# Karolina Sarna — Full Content > Atmospheric physicist turned organizational diagnostician. PhD TU Delft. > Built €20M ARR climate adaptation products at ICEYE. > Rebel Strategy Lab: strategy consulting + leadership coaching for climate tech founders. > Website: https://karolinasarna.com > Rebel Strategy Lab: https://rebelstrategylab.com --- ## Writing ### They Raised on Climate. Defence Showed Up. URL: https://karolinasarna.com/blog/defence-showed-up/ Date: 2026-03-31 Tags: earth-observation, climate-policy, structural-analysis Between 2021 and 2022, five Earth Observation companies went public promising commercial customers. Combined projections exceeded $5.7 billion by 2025. Actual revenue came in under $680 million. Most of it came from the military. Between 2021 and 2022, five Earth Observation companies went public, promising investors that commercial enterprises would become their largest customers. Agriculture companies would pay for crop monitoring. Insurers would pay for climate risk data. Corporations would pay for environmental, social, and governance (ESG) verification. The combined revenue projections exceeded $5.7 billion by 2025. The combined actual revenue came in under $680 million. And most of it came from the military. These projections weren't irrational when they were made. In 2021, interest rates were near zero, the EU Green Deal was accelerating, ESG investment was at an all-time high, and the Corporate Sustainability Reporting Directive (CSRD) was about to mandate sustainability reporting for 50,000 European companies. The world that those SPAC (special purpose acquisition company) investor decks described was the world that existed at the time. What nobody priced was the possibility that the world would change. I spent years building climate adaptation products at an EO company. The technology worked. Insurers were buying. And the defence roadmap still ate the product, because the hardware economics of running a satellite constellation don't wait for slow commercial markets to form. The company had found commercial product-market fit, and it still wasn't enough. When I looked at the rest of the industry, the same pattern was everywhere. ![SPAC projections vs reported revenue: the gap between promise and delivery. Five earth observation companies missed commercial projections by 55% to 98%.](/images/blog/eo-spac-chart.svg) --- ### The projections and what they assumed Satellogic projected $787 million in revenue by 2025, from a 300-satellite constellation remapping the planet daily. Actual: $17.7 million. The company redomiciled to Delaware to access US defence contracts. Planet Labs projected accelerating revenue growth through fiscal 2026, with its SPAC presentation forecasting that commercial customers would grow from 54% to 68% of the revenue mix. Actual revenue was $307.7 million. Commercial didn't grow to 68%. It shrank to 18%. Defence and intelligence grew to 59% and delivered the profitability milestone that made the stock jump 23%. BlackSky projected $546 million in revenue for 2025. Actual: $106.6 million, with 97% from government customers. Spire Global projected $1.2 billion in annual recurring revenue. Actual: roughly $110 million before selling its maritime business to pivot toward missile defence contracts. Terran Orbital projected $2.6 billion from a radar constellation that was never built. Lockheed Martin acquired the company for $0.25 per share. Every company missed its commercial projections by between 55% and 98%. Every company found its growth in government and defence spending, which wasn't part of the original thesis. The companies missed, and that matters. But what changed in the world between the moment the projections were written and the moment the revenue was reported matters more. --- ### What changed Every one of those SPAC projections embedded assumptions about the world that were treated as constants: the regulatory direction of the EU, the cultural momentum behind ESG, the political consensus on climate, and the cost of capital. None of them turned out to be. The EU's Corporate Sustainability Reporting Directive was supposed to require roughly 50,000 companies to report environmental metrics, creating the institutional infrastructure that would have turned satellite data from an interesting capability into a compliance requirement. In February 2025, the European Commission removed approximately 80% of companies from the scope. The Parliament voted 531 to 69 to delay implementation by two years. The EU Deforestation Regulation, which explicitly requires satellite monitoring for supply chain verification, has been delayed twice and faces further weakening before it is ever enforced. Across the Atlantic, BlackRock's support for environmental shareholder proposals dropped from over 40% to under 2%. All six major US banks left the Net-Zero Banking Alliance, which then ceased operations. The US Environmental Protection Agency (EPA) suspended the programme that was explicitly designed to use satellite methane detection for enforcement. Interest rates went from near zero to 5.25%, revaluing every growth-stage company whose worth depended on discounted future cash flows. The commercial buyer had no regulatory, political, financial, or social reason to show up. And the same political shift that destroyed commercial demand created defence demand. The war in Ukraine proved the value of commercial satellite imagery as battlefield intelligence. The National Geospatial-Intelligence Agency's (NGA) contract ceilings for commercial analytics grew from $29 million in 2021 to $490 million across the Luno A and Luno B programmes. NATO endorsed its first commercial space strategy. Germany, Sweden, and Japan signed sovereign satellite deals worth hundreds of millions. When the US suspended Ukraine's access to commercial imagery as diplomatic leverage, it proved that EO had become a geopolitical instrument, which accelerated every allied nation's investment in sovereign capacity. The regulations that would have created commercial buyers were dismantled by the same political forces that funded military buyers. The companies followed the money because the money was the only signal left. --- ### Why the buyers never showed up The regulatory retreat explains why the institutional scaffolding collapsed. But commercial EO was struggling before the regulations were pulled. The problems are structural, and better policy alone would not have fixed them. I saw this up close. A commercial prospect would run a pilot, get genuinely useful results, and then the project would die. Not because the technology failed. Because nobody inside that organisation owned the budget for an ongoing satellite data contract. The innovation team had exploration money. Converting to an operational line item required a business unit to take ownership, and satellite data didn't sit naturally in any department's profit-and-loss statement. When the champion moved on, or the budget cycle turned, the pilot evaporated. The deeper problem was the question everyone was asking. When I was building climate adaptation products, every conversation started the same way: where in the process can we leverage EO data? That's the wrong question. It assumes the existing frameworks are fine and just need better data input. But the question that would have led to operational contracts was different: what decision would change if you had this information, and whose workflow does it sit in? Almost nobody was asking that. The industry kept optimising for what satellites could see, when the bottleneck was always what happened after someone looked at the image. Each vertical that was supposed to drive commercial revenue failed for its own reasons. Agriculture, the most-cited use case in every SPAC deck, collapsed because the EU's Copernicus programme delivers free imagery good enough for most broadfield monitoring, and farmers' willingness to pay for commercial improvements was always fragile. Insurance found product-market fit with aerial imagery at 7-centimetre resolution, not satellite imagery, because insurers need to see individual roof damage at a level of detail that satellite imagery cannot deliver (yet). The ESG vertical vanished with its regulatory catalyst: rating agencies never adopted satellite verification as a core input, and the voluntary carbon market crashed by 61%, taking the satellite-based verification companies with it. And the product itself was wrong for the buyer it was meant to serve. Commercial enterprises don't want satellite images. An insurer needs a claims trigger in their processing workflow. A utility needs a vegetation-risk alert within its maintenance scheduling system. The EO industry spent two decades getting better at seeing, even as the bottleneck was always the lack of action infrastructure on the receiving end. The companies that found genuine commercial traction share one structural feature: they don't own constellations. AiDash buys satellite data as an input and sells vegetation risk management to utilities. Overstory does the same for grid resilience. A utility using AiDash doesn't know PlanetScope imagery is involved. The satellite is invisible. That's the structural distinction. The original thesis for owning a constellation was that you could build analytics on top of your own data and feed commercial insights back into sensor design, creating a flywheel between hardware and software. The problem is timing. Constellations burn cash. Satellites need replacing. The commercial market is slow, pilot-trapped, and produces revenue in small increments. You cannot wait for that market to materialise while your burn rate demands constellation growth. So you go where the revenue is, and the revenue is in defence. The next generation of satellites is designed around government requirements. The commercial product that was supposed to feed the flywheel gets starved of engineering attention and constellation capacity. Even where commercial product-market fit existed, even where insurers were buying, the solutions business was still the smallest revenue line, because the hardware economics forced the company toward the buyer who could fund the next launch. --- ### What sits underneath Every level of the capital chain made the same mistake. The SPAC sponsors trusted management projections. Management trusted market research firms' TAM numbers. The market research firms trusted the regulatory trajectory. The regulators trusted the political consensus. Nobody in the chain owned the connection between the capital and the underlying conditions on which the thesis depended. The assumptions that broke weren't obscure. They were the largest, most visible forces shaping the investment environment: the direction of European regulation, the durability of ESG consensus, the cost of capital, and the geopolitical stability of Europe. These weren't tail risks. They were the foundation. And they moved. Every projection is a bet on the world staying a certain way. The EO cohort bet on the world of 2021. The question worth sitting with is which version of the world your current thesis depends on, and who in your chain is responsible for noticing if it shifts. The EO companies are doing fine. Planet is at an $11 billion market cap. BlackSky is approaching profitability. They succeeded at something completely different from what the SPAC decks described. The investors who bought "democratising climate data" are holding NATO intelligence infrastructure. The companies executed. The world underneath the thesis changed, and nobody in the chain was watching. --- #### Sources Satellogic SPAC investor presentation, CF Acquisition Corp V, SEC filing, July 2021. Satellogic FY2025 earnings release, March 19, 2026. Planet Labs SPAC investor presentation, dMY Technology Group IV, SEC filing, July 2021. Planet Labs FY2026 Q4 earnings release and call transcript, March 19, 2026. BlackSky SPAC investor presentation, Osprey Technology Acquisition Corp, SEC filing, February 2021. BlackSky FY2025 earnings release, February 26, 2026. Spire Global SPAC investor presentation, NavSight Holdings, SEC filing, March 2021. Spire Global FY2024 earnings release, March 2025. Terran Orbital SPAC investor presentation, Tailwind Two Acquisition Corp, SEC filing, October 2021. Lockheed Martin acquisition announcement, August 15, 2024. European Commission, Omnibus Simplification Package, February 26, 2025. European Parliament "Stop-the-Clock" vote on CSRD, April 2025. BlackRock Investment Stewardship 2025 Voting Spotlight Report, September 2025. ESG Dive reporting on proxy season support rates. National Geospatial-Intelligence Agency, Luno A contract announcement ($290M ceiling), September 2024. NGA Luno B contract announcement ($200M ceiling), January 2025. NGA Economic Indicator Monitoring contract ($29M original ceiling), August 2021. --- ### We've Solved This Before URL: https://karolinasarna.com/blog/weve-solved-before/ Date: 2026-03-25 Tags: climate-policy, structural-analysis Climate change is the first environmental problem in the historical record where every structural condition for coordination is absent at once. The alternatives exist. The system that needs to change cannot profit from switching. In 1483, the farmers of Törbel, a village in the Swiss Alps, wrote down a rule they had probably followed for generations before anyone thought to put it on paper. No villager may graze more cattle on the common meadow than they can feed through winter on their own hay. One rule, no enforcement agency, no external regulator, and five hundred and forty-three years later the meadows of Törbel are not overgrazed. The political scientist Elinor Ostrom came to Törbel in the 1980s, and the village became the opening case study in her 1990 book *Governing the Commons*, the work that would eventually make her the first and still only woman to receive the Nobel Prize in Economics. What she found in Törbel, and in hundreds of communities across the world, was evidence that ordinary people can manage shared resources sustainably for centuries without privatisation or state control. Reliably, when certain conditions are present. In Törbel, those conditions are visible at close range. Everyone in the village knows who belongs and who does not. The farmers who use the meadow are the same people who write and revise the rules governing it. Monitoring costs nothing because you can see your neighbour's cattle from your own doorstep. Penalties for rule-breaking exist but they escalate gradually rather than starting at catastrophe. And when disputes arise, they are resolved locally, by people who have to live with each other the following week. The structure makes cooperation the easier path, which is why it holds for half a millennium without anyone having to enforce it from above. Törbel does not offer a model for solving climate change. A village of 600 people managing a meadow they can see from their kitchen windows has almost nothing in common with 195 countries negotiating over an atmosphere none of them can own. But Törbel reveals, with unusual clarity, what coordination actually requires. And the environmental record of the past century provides a set of cases, at progressively larger scales, that test whether those requirements survive the jump from a Swiss meadow to a city, a continent, and a planet. Some of them do. In December 1952, unusual weather trapped coal smoke over London for four days and killed 12,000 people. Four years later, Parliament passed the Clean Air Act. In the 1980s, acid rain was destroying lakes and forests across the northeastern United States; Congress responded with a programme that let power plants trade pollution permits among themselves, creating a financial incentive to cut emissions, and sulphur dioxide fell by 90% at less than half the projected cost, with health benefits exceeding $50 billion per year. In 1987, the Montreal Protocol united 198 countries to eliminate the chemicals that were destroying the ozone layer, and that layer is now on track to recover by mid-century. Each of these involved larger groups, higher costs, and longer time horizons than Törbel, and each succeeded. Climate change has none of the features that made any of them possible. The clean energy alternatives exist: solar is cheaper than coal across most markets, heat pumps and electric vehicles are proven technologies. The obstacle is a $7 trillion fossil fuel subsidy system, an infrastructure base built for a century of carbon, and a fossil fuel industry whose core assets become worthless in the transition. The system is defending itself, and the costs of that failure fall unevenly. The countries that contribute least to the emissions bear the highest costs, and the countries most responsible have learned to frame those consequences as someone else's emergency. This is the question worth taking seriously: what made coordination work every time it worked, what is different about climate, and what does the gap between the two actually cost in human terms? --- ### When 12,000 people die in your city, the feedback loop is short The Great Smog arrived in London on 5 December 1952 and did not lift until the 9th. A high-pressure weather system trapped coal smoke at ground level across the entire city. Visibility dropped below 30 centimetres in some parts of London. Cattle at Smithfield Show asphyxiated. Ambulances could not navigate the streets. The initial government estimate was 4,000 deaths, a figure later revised upward to 12,000 when researchers counted the deaths that continued for months afterward. The political response was not fast by any reasonable standard. It took four years. The Clean Air Act of 1956 established smoke control areas where only smokeless fuels could be burned, and provided grants for households to convert their heating systems. But when it came, it was decisive, and the reason it worked is worth understanding: the people who died and the people who voted and the people who could change the law were all living in the same city, breathing the same air. The feedback loop between cause, harm, and political response was measured in miles and months. A parliamentarian could look out of his window at Westminster and see the smog that was killing his constituents. This pattern, where coordination succeeds because harm is visible, immediate, and locally felt by the people with the power to act, shows up again and again. The United States acid rain programme tells a parallel story at national scale. By the 1980s, sulphur dioxide from coal-fired power plants was acidifying lakes and killing forests across New England and the Adirondacks. The 1990 Clean Air Act Amendments created a system in which power plants received permits to emit a set amount of pollution and could buy or sell those permits among themselves, so that companies who cut emissions faster could profit from doing so while the overall cap tightened over time. The results exceeded every projection. Sulphur dioxide emissions fell by over 90%. The EPA had estimated implementation costs at $6 billion per year; actual costs came in between $1 and $3 billion. The health benefits from reduced air pollution alone were estimated at $59 to $116 billion annually by 2010, a benefit-to-cost ratio exceeding 50 to 1. The cost story deserves emphasis because it recurs in every successful coordination case: the projected expense of action consistently overestimates the actual expense, often by a factor of two or more, while the benefits consistently exceed expectations by roughly ten times or more. When industries lobby against regulation, they have every incentive to inflate cost projections and no incentive to measure the benefits that would come from acting. When coordination actually happens, innovation compresses costs in ways that projections made before the policy could not have predicted. The acid rain programme also shared London's key feature. The sources were identifiable: roughly 400 large power plants, owned by known utilities, emitting a pollutant that could be measured at the source. The victims were identifiable: lake and forest ecosystems in specific, politically represented states. The alternative was available: lower-sulphur coal and technology to clean exhaust gases existed before the regulation. And the costs fell on entities large enough to absorb them and pass them through to consumers, who experienced price increases too small to trigger political backlash. --- ### The coalition of the green and the greedy The Montreal Protocol, signed on 16 September 1987, is the most successful international environmental agreement ever negotiated, and possibly the most successful international agreement of any kind. It achieved universal ratification by all 198 UN member states, has eliminated 99% of ozone-depleting substances, and is on track to prevent roughly 2 million skin cancer cases per year by 2030 while having avoided an estimated 0.5 to 1.0°C of additional warming. The ozone layer should recover to 1980 levels by approximately 2066 over Antarctica and earlier elsewhere. The story is usually told as a triumph of science and diplomacy, and it is both of those things. In 1974, two scientists published the hypothesis that chemicals used in refrigerators and aerosol cans, called chlorofluorocarbons or CFCs, were destroying the ozone layer. The Antarctic ozone hole was confirmed in 1985. The treaty was signed in 1987. Thirteen years from hypothesis to binding global action, which remains the fastest timeline for any planetary-scale environmental response. But the conditions that enabled it are more revealing than the timeline. Global CFC production was concentrated among roughly 18 companies in four countries. DuPont alone produced about a quarter of global output, yet CFCs represented only about 2 to 3% of the company's total sales. When DuPont's original Freon patent expired in 1979, the company initially formed a lobbying group, the Alliance for Responsible CFC Policy, to fight regulation. It worked: under the Reagan administration, there was no political pressure for action. But DuPont was simultaneously investing in patented replacement chemicals. By 1986, with new patents in hand, the company reversed its position entirely and publicly condemned CFCs. DuPont representatives appeared before the Montreal Protocol negotiations urging a worldwide ban on the very chemicals the company had spent a decade defending. The reversal was strategic, not moral. Regulation would force every competitor to license DuPont's patented alternatives. The environmental imperative and the commercial opportunity pointed in the same direction. The UN official who led the negotiations, Mostafa Tolba, is reported to have described the outcome as a coalition of the green and the greedy. The entire global CFC market was worth approximately $1 billion per year. The fund established to help developing countries transition has disbursed about $3.9 billion over three decades. These are significant sums, but they are manageable. No economy was destroyed, no geopolitical position threatened. The costs were absorbable, the alternatives were available, and the biggest company in the market had more to gain from regulation than from the status quo. And that is precisely what makes climate change a different category of problem. --- ### The system that defends itself Global greenhouse gas emissions hit a record in 2023, and under current policies, the world is heading for approximately 3.1°C of warming. The UN climate science panel found that the power plants, factories, and infrastructure already built and operating today will, over their remaining lifetimes, emit more than the atmosphere can absorb if warming is to stay below 1.5°C. The things we have already constructed will, if left running, break the target before any new project adds a single tonne. The contrast with every successful coordination case is systematic, and it starts with the most basic variable: how many actors are involved. Where the Montreal Protocol required agreement among 18 companies, climate change involves millions of emitters in every sector of every economy on earth. Where London's smog could be traced to domestic coal burning within a single city, carbon dioxide has no colour, no smell, and no local signature that ties it to the person who emitted it. Where acid rain came from 400 identifiable power plants, the fossil fuel system is woven into transport, agriculture, heating, industry, and the physical layout of cities built over a century of cheap carbon. This is where the usual comparison between climate and the Montreal Protocol gets the diagnosis wrong. The standard analysis says climate is harder because there is no single replacement for fossil fuels the way DuPont's new chemicals replaced the old ones. But the alternatives do exist. Solar energy is now the cheapest source of electricity in history in most markets. Heat pumps, which heat and cool buildings using electricity instead of gas, are proven and commercially available. Electric vehicles are approaching price parity with petrol cars. The technology for a low-carbon economy is largely ready. What makes climate coordination different is that the fossil fuel industry cannot make DuPont's move. DuPont could pivot from the old chemicals to the new ones because they served the same function through the same distribution channels for the same customers. The pivot was commercially attractive. Fossil fuel companies cannot pivot from oil and gas to solar and wind without writing off the value of their oil wells, their pipelines, their refineries, and the geopolitical power that comes with controlling energy supply. The transition does not offer them a more profitable position on the other side. It offers them obsolescence. This is why global fossil fuel subsidies reached $7 trillion in 2022 according to the IMF. That figure includes $1.3 trillion in direct payments and tax breaks, plus another $5.7 trillion in costs that fossil fuels impose on society, through air pollution, climate damage, and health impacts, that nobody is required to pay for. Together, these subsidies dwarf all climate finance combined. The system is coordinating effectively, just in defence of the existing structure rather than toward a new one. The resistance runs through every level. Companies lobby against regulation and fund campaigns to keep friendly legislators in office. Industry groups frame climate action as a threat to jobs and energy security. Trillions of dollars in existing infrastructure, from pipelines to refineries to power plants, were built to run for decades and will not be abandoned before those decades are up, because the financial incentive is to keep burning what has already been built to burn. And the rules governing energy markets were written around fossil fuel assumptions, which makes it harder for alternatives to compete even when they are cheaper. Researchers call this carbon lock-in: a system where every part, the technology, the infrastructure, the regulations, and the business models, has been shaped around fossil fuels for so long that they reinforce each other, and no single piece can change without the others changing too. The EU Emissions Trading System illustrates the pattern in miniature. Launched in 2005, it has driven substantial emissions reductions in the power sector, where a limited number of large plants can switch between fuels and pass carbon costs through to consumers. But it struggled for years with industrial emissions, where manufacturers who compete internationally could threaten to move production to countries without carbon costs, and where governments responded by handing out free pollution permits. The carbon price crashed in early phases because too many permits were issued. Successive rounds of reform have tightened the system over two decades. The lesson is that even the world's most sophisticated carbon pricing system, operating within the European Union with some of the strongest environmental institutions in the world, needs 20 years to build effective coverage, because the resistance adapts faster than the policy. --- ### The system was designed for a different planet These coordination failures are not random. They emerge from an economic system whose basic assumptions were established when the global population was under a billion people, GDP was a fraction of today's, and the idea that human activity could change the composition of the atmosphere was something nobody had thought to worry about. Adam Smith published *The Wealth of Nations* in 1776, in a world of local markets, artisan production, and horse-drawn transport. His concept of an invisible hand, used only three times across his entire body of work and arguably with some irony, was reimagined in Paul Samuelson's 1948 economics textbook as a universal mechanism by which self-interested behaviour produces the best outcomes for everyone. Smith himself never made that claim. He assumed economics operated within moral limits, as described in *The Theory of Moral Sentiments*, published seventeen years before *The Wealth of Nations*. The modern separation of economics from ethics, and the assumption that markets left alone will produce the best outcomes for society, is a twentieth-century construction built on an eighteenth-century foundation that was more nuanced than its inheritors admit. For most of the history of economic thought, the costs that one person's activity imposes on others, pollution being the clearest example, were treated as edge cases. When economists formalised the concept in the early twentieth century, they assumed these spillover costs were rare exceptions to an otherwise well-functioning system. When Ronald Coase argued in 1960 that private bargaining could sort out such problems without government involvement, the theory required zero costs of negotiating and everyone agreeing on who owns what, conditions he himself acknowledged were impossible for any environmental problem involving millions of people. Economics was built on the assumption that the atmosphere, the oceans, and the biosphere were either infinite or irrelevant. GDP, the metric that drives virtually all economic policy worldwide, was invented by Simon Kuznets in 1934 as a tool for measuring the depth of the Great Depression, and Kuznets immediately warned against using it as a measure of how well a country's people are actually doing. In his report to Congress, he wrote that the welfare of a nation can scarcely be inferred from a measurement of national income. He later insisted that distinctions must be kept in mind between quantity and quality of growth, between its costs and return, and between the short and the long term. These warnings were ignored. GDP was institutionalised at the 1944 Bretton Woods Conference through the IMF and World Bank, cemented by the 1946 US Employment Act, and by the late 1970s had become the default goal of economic policy across the political spectrum. The gap between what the economic system measures and what the planet requires has widened with each decade of measurement. The Club of Rome warned in 1972 that infinite growth on a finite planet was a mathematical impossibility, and a 2008 review found that 30 years of data tracked their projections closely. Nicholas Stern's 2006 review for the UK government called climate change the greatest market failure the world has ever seen. The planetary boundaries framework, which maps the safe operating limits for nine earth systems, now shows seven of nine breached as of the September 2025 Planetary Health Check, with ocean acidification confirmed as crossed for the first time. Only two boundaries remain within the safe zone: air pollution from particles, and the ozone layer, the one protected by the Montreal Protocol. Kate Raworth's Doughnut Economics framework asks a simple question: can any country meet the basic needs of its people without overshooting what the planet can sustain? In a 2025 study published in *Nature*, she and Andrew Fanning tested this across more than 150 countries. The answer is zero. Not one country meets fundamental human needs without exceeding ecological limits. The richest 20% of countries, with 15% of the global population, contribute over 40% of ecological overshoot. The poorest 40%, with 42% of the population, bear over 60% of social shortfall while contributing negligible overshoot. The system is working as designed, for conditions that no longer exist. --- ### The receipt In 2024, weather-related events triggered 45.8 million internal displacements, the highest annual figure since monitoring began in 2008 and more than double the ten-year average. The World Bank projects up to 216 million people forced to move within their own countries by 2050 because of climate impacts. Some estimates that include floods, storms, and other sudden disasters reach 500 million. The chain from emissions to displacement runs through specific, documented mechanisms. In the Sahel, where temperatures are rising at 1.5 times the global average, Lake Chad has lost approximately 90% of its surface area since the 1960s. The collapse of fishing and farming livelihoods around the lake basin created conditions that armed groups including Boko Haram exploited, contributing to over 35,000 deaths and more than 2 million displaced people across the region. In Central America's Dry Corridor, a 2018 drought destroyed 80% of maize and bean crops in affected areas, and coffee production is projected to drop at least 40% by 2050. The droughts, the crop failures, the armed conflict, and the displacement that follows are what happens when a changing atmosphere meets farming systems with no margin for shock and governments with no capacity to respond. Bangladesh produces 0.3% of global emissions and faces the prospect of losing 11% of its land area to a 0.5-metre sea level rise, potentially displacing 18 million people. Tuvalu, with a mean elevation of 2 metres and a population of 10,000, negotiated the world's first climate visa arrangement with Australia in 2023, offering 280 residents permanent residency per year. By the time the ballot closed in July 2025, registrations covered more than 80% of the population. The numbers on who causes the damage and who bears it are stark. The 50 most climate-vulnerable countries contribute 0.28% of global CO₂ emissions. The United States alone accounts for roughly 20% of all emissions since the industrial revolution. Emissions per person per year in the US are 13.8 tonnes; in Bangladesh, 0.6 tonnes; in Chad, 0.1 tonnes. The countries ranked most vulnerable and least prepared to adapt, including Chad, the Central African Republic, Eritrea, and South Sudan, have contributed negligibly to the problem. The Loss and Damage Fund agreed at the 2022 UN climate summit had received pledges totalling approximately $768 million as of early 2025, which is roughly 0.1% of the estimated $400 billion that developing countries require annually. Governments overwhelmingly treat climate migration as a security problem. The Pentagon calls climate change a threat multiplier. NATO, the EU, and successive US defence reviews identify mass migration and resource competition as security threats requiring military and border responses. The framing itself reveals the broken feedback loop. Displacement is the signal that should trigger corrective action on emissions. Instead, it gets reclassified as a border management challenge and handed to institutions that have no mandate to address its origins. This is one system, not many separate problems. An economy built to grow at all costs pushes those costs onto the planet. The planetary costs produce climate change and biodiversity loss. The ecological breakdown destabilises agriculture and livelihoods. The instability produces forced migration. And the countries whose emissions set the chain in motion treat the people displaced by it as a security threat. Each link is governed by different institutions, measured by different metrics, and debated in different policy arenas, which is why no institution is responsible for the chain as a whole. --- ### The design question Coordination works when harm is visible to the people with the power to act, when the number of key actors is small enough for agreement to be feasible, when alternatives exist and the biggest players in the existing system can profit from switching, when costs are absorbable, and when benefits arrive within the decision-maker's time horizon. Every successful environmental coordination in the historical record, from Törbel's meadow to the Montreal Protocol, had all of these features present simultaneously. Climate change is the first problem in that record where all of them are absent at once. The alternatives exist, but the fossil fuel industry cannot profit from switching. The harm is real, but it falls on people with no power to act. The benefits of coordination would be enormous, but they arrive after the current generation of decision-makers has left office. And every time coordination has been attempted and sustained, the economics have turned out better than projected. Acid rain costs came in at less than half of estimates, with benefits exceeding projections by 50 to 1. The Montreal Protocol's total transition cost was modest relative to the scale of the problem, and the ozone layer is recovering ahead of some projections. The pattern holds across every case: projected costs of environmental action are inflated, and benefits are undercounted. What this means is a design question, not an optimisation question. Markets optimise. They find the cheapest source of energy, the highest return on capital, the most efficient production method, all within the rules of the existing system. Markets cannot change their own rules, impose costs on themselves, or price in harm that arrives three decades later on a different continent. The gap between where current interventions are focused, adjusting tax rates and subsidy levels and carbon prices, and where the evidence says they would be most effective, changing the rules themselves and what we measure as progress, is itself the coordination failure. Systems thinkers have long argued that the most powerful place to intervene in a system is not at the level of its parameters, the tax rates and subsidy amounts, but at the level of its goals: what the system is designed to achieve. Current climate policy operates overwhelmingly at the weakest level, adjusting numbers within existing rules. Replacing GDP growth with wellbeing within planetary boundaries, as Bhutan, New Zealand, and Amsterdam have begun to experiment with, is what the engineering specification requires, even if current political incentives make it difficult to pursue. No country has yet demonstrated that human needs can be met within planetary boundaries. But Costa Rica comes closest, achieving near-universal social outcomes and over 98% renewable electricity on less than half the income of wealthy nations. In March 2026, the World Happiness Report ranked Costa Rica fourth in the world, the highest position ever achieved by a Latin American country, ahead of Sweden, Norway, and every English-speaking nation. The methodology has limitations, but the directional finding is hard to dismiss: a country that comes closest to living within planetary means also ranks among the happiest on earth, which suggests the constraint is design rather than resources, and systems that are designed can be redesigned. The question is whether that redesign happens before the window closes, or whether we continue to optimise a machine that was built for a planet that no longer exists. --- #### References and further reading Ostrom, E. (1990). *Governing the Commons: The Evolution of Institutions for Collective Action*. Cambridge University Press. Raworth, K. (2017). *Doughnut Economics: Seven Ways to Think Like a 21st-Century Economist*. Random House. Fanning, A. L. & Raworth, K. (2025). Doughnut of social and planetary boundaries monitors a world out of balance. *Nature*, 646, 47-56. Sakschewski, B., Caesar, L. et al. (2025). *Planetary Health Check 2025*. Potsdam Institute for Climate Impact Research. Richardson, K. et al. (2023). Earth beyond six of nine planetary boundaries. *Science Advances*, 9(37). Coase, R. H. (1960). The problem of social cost. *The Journal of Law and Economics*, 3, 1-44. Meadows, D. H., Meadows, D. L., Randers, J. & Behrens, W. W. (1972). *The Limits to Growth*. Universe Books. Unruh, G. C. (2000). Understanding carbon lock-in. *Energy Policy*, 28(12), 817-830. Geels, F. W. (2014). Regime resistance against low-carbon transitions. *Theory, Culture & Society*, 31(5), 21-40. Maxwell, J. & Briscoe, F. (1997). There's money in the air: the CFC ban and DuPont's regulatory strategy. *Business Strategy and the Environment*, 6(5), 276-286. Barrett, S. (2016). Coordination vs. voluntarism and enforcement in sustaining international environmental cooperation. *PNAS*, 113(51), 14515-14522. Stern, N. (2006). *The Economics of Climate Change: The Stern Review*. Cambridge University Press. Meadows, D. (1999). Leverage Points: Places to Intervene in a System. The Sustainability Institute. Parenti, C. (2011). *Tropic of Chaos: Climate Change and the New Geography of Violence*. Nation Books. Mildenberger, M. (2020). *Carbon Captured: How Business and Labor Control Climate Politics*. MIT Press. Stoddard, I. et al. (2021). Three decades of climate mitigation: why haven't we bent the global emissions curve? *Annual Review of Environment and Resources*, 46, 653-689. World Bank (2021). *Groundswell Part 2: Acting on Internal Climate Migration*. IMF (2023). *IMF Fossil Fuel Subsidies Data: 2023 Update*. UNEP (2024). *Emissions Gap Report 2024*. Internal Displacement Monitoring Centre (2025). *Global Report on Internal Displacement 2025*. Wellbeing Research Centre, University of Oxford & Gallup (2026). *World Happiness Report 2026*. --- ### Six Regulations, One Observation System, No Standard URL: https://karolinasarna.com/blog/six-regulations/ Date: 2026-03-18 Tags: earth-observation, climate-policy, structural-analysis The EU spent nearly €10 billion on Copernicus and over €300 million on Destination Earth. Six regulations now require satellite data for compliance. None of them references a common standard. The observation system does not produce one. The EU has spent nearly €10 billion on Copernicus, the most comprehensive civilian Earth observation system ever built. It has spent over €300 million on Destination Earth, a simulation platform designed to model the climate system at 5-kilometre resolution. Three EU regulations explicitly mandate the use of satellite data for compliance. Three more require environmental monitoring at a scale and frequency that only satellite data can deliver. Agriculture, supply chains, energy emissions, sustainable finance, biodiversity, and corporate reporting. Each regulation is building its own bespoke connection to the observation system. None of them references a common standard. The observation system does not produce one. The summer of 2025 cost the EU €43 billion in climate-related losses, a figure that University of Mannheim and ECB economists described as conservative. Copernicus satellites tracked soil moisture deficits, thermal anomalies, and vegetation stress. Nobody was required to act on what the system saw. --- ### What six regulations actually need The gap becomes concrete when you look at what each regulation actually requires from the observation system. Six domains need satellite data. They share overlapping technical requirements: optical imagery for land cover change, temporal baselines for historical verification, asset-level hazard exposure, and validation standards for audit. And each domain has independently defined all of these from scratch within its own policy silo, without referencing what the others have built. The Common Agricultural Policy's Area Monitoring System is the most advanced case. Article 68 of the CAP financing regulation made satellite verification mandatory in 2024, requiring all member states to operate monitoring using "Copernicus Sentinel satellite data or data with at least equivalent value." In practice, this means Sentinel-2 optical imagery at 10-metre resolution every few days during the growing season, combined with Sentinel-1 radar (SAR) for cloud-free coverage, validated against parcel boundary databases maintained by each member state. The system works. It was built entirely within agricultural policy, with its own resolution specifications, validation methods, and temporal requirements. It is disconnected from every other regulatory domain that needs the same underlying data. The EU Deforestation Regulation needs satellite imagery to verify that commodities entering the EU were not produced on land deforested after December 2020. Polygon coordinates for plots over four hectares, point coordinates for smaller ones. In practice, this requires the same Sentinel-2 optical data, the same kind of land-cover change detection, and the same temporal baseline that CAP already uses. EUDR built its own geolocation and verification framework from scratch within trade and supply chain regulation, without referencing the methods CAP had already developed for monitoring land use using the same satellites. The regulation applies starting in December 2026 to large and medium operators. The Methane Regulation of 2024 requires the Commission to establish a satellite-based monitoring tool by August 2026 and a rapid alert mechanism for super-emitting events. Legislated satellite infrastructure, built independently inside energy regulation, using different instruments (Sentinel-5P for atmospheric methane, commercial satellites like GHGSat for point-source detection) with its own detection thresholds and alert protocols. The remaining three regulations do not explicitly name satellite data, but they require capabilities that are functionally impossible to deliver at scale without it. The EU Taxonomy's "Do No Significant Harm" criterion requires climate risk assessments at the asset level: flood exposure, heat stress, and drought risk for specific locations. The Nature Restoration Law requires member states to demonstrate progress in ecological indicators, vegetation health, land cover change, and habitat condition throughout their entire national territories. The Corporate Sustainability Reporting Directive requires physical climate risk disclosure across operations and value chains, meaning some version of the same asset-level hazard exposure that the Taxonomy requires, applied to every facility a reporting company operates or sources from. No ground-based monitoring system can deliver any of this at the spatial coverage and temporal frequency these regulations demand. EU regulation is always siloed. The agriculture directorate does not coordinate with the environment directorate, and nobody expects it to. But these six domains are unusual because they share the same underlying data source, unlike other regulatory silos. Six regulatory domains pulling data from the same Copernicus observation system. At least three require Sentinel-2 optical imagery for land cover and land-use change. At least three need asset-level climate hazard exposure. Multiple need temporal baselines stretching back years. All need validation standards to make compliance auditable. Each domain has independently defined its own resolution requirements, temporal frequency, validation methodology, and reference datasets. The Common Agricultural Policy does not reference the Deforestation Regulation's land use methods. The Methane Regulation does not reference the Taxonomy's climate risk framework. No horizontal standard exists that says: when EU law requires satellite-derived environmental data, this is what it looks like. The JRC's Knowledge Centre on Earth Observation is running a Deep Dive assessment on exactly this question through its work on Earth Observation for Compliance Assurance. It is not binding. --- ### Why the public system doesn't provide it Copernicus produces data. DestinE produces simulations. Neither produces the standardised compliance-ready outputs that these six regulations need. The gap is a design choice, consistent across every phase of EU Earth observation investment. The EU's approach has been to build the best possible observational and simulation capabilities, make the data openly available, and trust that a downstream ecosystem will form around them. Copernicus operates on a "full, free, and open" data policy. The EU Space Regulation of 2021 says nothing about requiring anyone to use satellite observations. DestinE's design is a platform that users come to: run simulations, explore what-if scenarios, and make decisions based on what they see. Phase 3, confirmed in February 2026, lists EUMETSAT's top priority as "fostering institutional user uptake." After four years and over €300 million, the system's design still assumes adoption will come through outreach rather than requirement. The operational expression of this design choice is co-design workshops with institutional stakeholders, User eXchange meetings between the same agencies that built the system, and stakeholder consultations that produce roadmaps published as PDFs. The process is thorough, inclusive, and thoroughly internal to the EO community. The compliance officers, risk managers, and sustainability reporting teams who will determine whether these six regulations succeed or fail are not in the room. They do not know this system exists. Nothing in the current design requires them to find out. There is one domain where this problem has already been solved. When Numerical Weather Prediction processes observations from radiosondes, ground stations, radar, satellites, and ocean buoys, it produces METAR and TAF: standardised weather reporting formats that feed directly into flight management systems and operational workflows. No pilot logs into NWP. No airline chooses to subscribe. The output is embedded in the decision, and ICAO Annex 3 made it mandatory in 1948. The mandate forced the standard. The standard enabled the infrastructure. The infrastructure made the downstream market possible. The EU has built the observation system and the simulation platform. Six regulations are creating the mandates. The layer between them, a standard that would define what "Copernicus-derived compliance data" looks like and how it flows into regulatory workflows, does not exist. Every phase of EU EO investment has been designed on the assumption that this layer will emerge from the ecosystem. That assumption has held across two decades of investment, and the layer has not appeared because standards require definition and enforcement. They do not emerge from workshops. --- ### Why everyone bets on applications anyway So the standard does not exist, and the public system is not designed to produce it. Something has to fill the vacuum. What fills it are hundreds of companies, each independently building the bridge that infrastructure would once have provided. A founder looking at this landscape sees six regulations arriving, compliance officers who need answers now, and no public standard to build on. The rational move is to build the proprietary bridge and charge for it. Overstory builds its own vegetation risk standard for utilities and sells it. Kayrros builds proprietary methane monitoring and sells energy market intelligence. Every analytics company in the EO value chain, LiveEO for pipeline monitoring, Satelligence for deforestation tracking, Orbital Insights for retail and finance, follows the same logic: acquire raw Sentinel or commercial imagery, build proprietary preprocessing, produce sector-specific output. The first two steps are repeated hundreds of times across the industry because the public system does not do them once in a standardised way. The EU's own strategy reinforces this pattern. The Commission's 2016 Space Strategy for Europe called for stronger links with the "commercial downstream sector." Copernicus describes its purpose as enabling "downstream services" and "commercial applications." The entire value chain is oriented around the idea that value gets "added" at the application layer and that the raw observation is a commodity. Planet Labs sells "insights." Capella Space markets "customer applications." Satellogic frames its work as "value-added B2B and B2C solutions." The vocabulary encodes the assumption that the application layer is where the industry belongs. Each company that builds a proprietary bridge is behaving rationally. Each company that monetises the gap also reduces the pressure to close it. The more startups build proprietary connections between Copernicus data and regulatory compliance, the more stakeholders develop a vested interest in the gap persisting. The downstream ecosystem that the EU explicitly called for is what prevents the public standard from ever being built. The absence of infrastructure is the business model, and the business model lobbies against its own obsolescence. --- ### What this means for founders and investors If you are building an EO applications company, you are betting that the public standard never arrives. If it does, your proprietary bridge becomes redundant. Every company that built bespoke preprocessing on top of open Copernicus data faces the same structural risk as companies that built proprietary middleware on pre-standard web infrastructure in the 1990s: when the standard arrives, the proprietary layer loses its value. If the standard never arrives, the position is not much better. Your company is perpetually rebuilding infrastructure that should be public. Your preprocessing pipeline is a cost centre that your competitor also maintains. Your differentiation lives not in the bridge itself but in the domain expertise on the far side of it, and domain expertise without structural advantage is a services business that happens to use satellites. The real opportunity lies with whoever builds the connecting layer: a standardised, validated, asset-level environmental exposure product produced continuously from Copernicus data, at a resolution and update frequency that EU compliance frameworks can reference by name. Flood, heat, drought, land cover change: different input pipelines, different validation methodologies, but the same output schema. An asset, a hazard, a severity, a confidence level. One product, not four. The closest precedent is the CAP Area Monitoring System, which proves the model works when a specific regulation mandates a specific use of specific satellite data with specific validation standards. The question is whether the remaining five regulatory domains will each independently replicate what CAP built, producing six parallel systems that draw from the same observation infrastructure, or whether someone defines a horizontal standard that connects them. The instinct is to call this a governance play: it requires regulatory authority, institutional coordination, and patient capital that venture timelines do not accommodate. And the governance layer is real. The value of this product is the standard, not the data. Anyone can build a flood model on Copernicus data. The moat is the reference that regulators point to. But credit ratings were not designed to serve as a governance infrastructure. They became it. FICO started by selling credit scores to individual lenders. Bloomberg started as a terminal for bond traders. All of them became the standard because they got adopted first, and regulators pointed to them because they already existed. The question is whether this needs regulatory authority to start, not just to exist. The EU has an observation system. It has a simulation platform. It has six regulations creating demand. It has a commercial ecosystem that monetises the gap between observation and decision. What it does not have is the layer that would make any of this infrastructure rather than a collection of tools, platforms, and proprietary bridges. Without that layer, €10 billion in public investment produces the world's most expensive dashboard. --- #### Sources Copernicus programme budget: EU Multiannual Financial Frameworks 2014-2020 (~€4.3B) and 2021-2027 (€5.4B) Destination Earth: Digital Europe Programme work programmes, Phases 1-3. ECMWF, ESA, EUMETSAT implementing entities DestinE Phase 3 confirmation and priorities: ECMWF press release, 1 February 2026 EUMETSAT Phase 3 priorities: EUMETSAT press release on DestinE Phase 3 EU Space Regulation: Regulation (EU) 2021/696 €43B climate losses: Dr Sehrish Usman, University of Mannheim and ECB economists, published September 2025 CAP Area Monitoring System: Regulation (EU) 2021/2116, Article 68 EU Deforestation Regulation: Regulation (EU) 2023/1115, as amended by Regulation (EU) 2025/2650 (application postponed to December 2026) EU Methane Regulation: Regulation (EU) 2024/1787 EU Taxonomy: Regulation (EU) 2020/852, Climate Delegated Act Annex A Nature Restoration Law: Regulation (EU) 2024/1991 Corporate Sustainability Reporting Directive: Directive (EU) 2022/2464, European Sustainability Reporting Standards E1 JRC Knowledge Centre on Earth Observation, Compliance Assurance Deep Dive: [knowledge4policy.ec.europa.eu/earthobservation](http://knowledge4policy.ec.europa.eu/earthobservation) Commission Space Strategy for Europe: COM(2016) 705 ICAO Annex 3: first adopted 16 April 1948, pursuant to Article 37 of the Chicago Convention (1944) METAR/TAF standards: ICAO Annex 3, Meteorological Service for International Air Navigation --- ### Finland's Innovation Doom Loop URL: https://karolinasarna.com/blog/finlands-innovation-doom-loop/ Date: 2026-03-10 Tags: structural-analysis, organizational-design The country that produced Linux, modern semiconductors, and the world's wireless infrastructure is systematically dismantling the conditions that made any of it possible. The country that produced Linux, modern semiconductors, and the world's wireless infrastructure is systematically dismantling the conditions that made any of it possible. In 1974, a Finnish physicist named Tuomo Suntola invented a technique for depositing thin films one atomic layer at a time. He was trying to make electroluminescent flat-panel displays. Nobody was thinking about semiconductors. The method sat in relative obscurity for decades, finding niche use in Finnish display screens under the Finlux brand. Then, around 2000, the semiconductor industry hit a wall. Transistors had shrunk so small that conventional manufacturing methods couldn't deposit the ultrathin films needed to keep them working. The solution turned out to be a process invented 26 years earlier in a Finnish lab, for an entirely different purpose. In 2007, Intel adopted Suntola's Atomic Layer Deposition for its 45nm processors. Today, leading-edge chips require more than 70 ALD processing steps. Every phone in your pocket, every server running the internet, every data centre processing your email depends on work done in Finland when nobody was watching and nobody could have predicted what it would become. Suntola received the Millennium Technology Prize in 2018. Forty-four years after his invention. That's what foundational research does. It creates knowledge whose application you cannot foresee at the time of funding. Which is exactly why markets won't fund it. And exactly why governments must. Finland understood this once. It seems now it forgot. ### The distinction nobody makes When Finland's government announces that R&D spending is going up, most people hear "innovation is being funded." This is technically true and structurally misleading, because it collapses three very different things into one number. Foundational research asks "how does this work?" No product in mind. A student tinkers with an operating system kernel because he's curious. A physicist cools materials to near absolute zero to see what happens. Applied research asks "how do we solve this specific problem?" It is targeted, practical, commercially oriented. Experimental development builds and ships the product. Finland's R&D spending increase flows almost entirely through Business Finland to companies and targeted commercial projects. Meanwhile, university base funding, where foundational research lives, is being cut by approximately €117.7 million between 2026 and 2028. The headline number goes up while the foundation erodes underneath it. R&D spending as a percentage of GDP tells you nothing about whether you're watering the roots or painting the leaves green. This matters because of a fact that should concern anyone making Finnish science policy: the foundational research behind major innovations peaks 20 to 30 years before the innovation itself. Cut foundational research today, and nothing visible happens. The universities still run. The graduates still graduate. The R&D numbers still look fine. The consequences arrive a generation later, when the pipeline of knowledge that feeds applied innovation has quietly dried up and nobody can trace the cause back to a budget line adjusted decades earlier. ### The evidence, and its mirror Finland's innovation achievements required both commercial investment and foundational public research. But the foundational layer came first, often by decades. And it is that layer, specifically, that is now being cut. The same budget that funds the story also funds its undoing. **A university student and a network connection.** In 1991, Linus Torvalds was a 21-year-old computer science student at the University of Helsinki, using the university's Unix systems and Finland's academic network, FUNET. He wasn't building a product. He was curious about operating systems. What he built became Linux, which today runs 96.3% of the world's top million web servers, 100% of the top 500 supercomputers since November 2017, and powers every Android device on the planet. The Linux ecosystem is valued at over $18 billion. What made it possible? A university with computing infrastructure, an academic network, and a student with zero commercial intent. The university system that produced Torvalds has seen real spending per student fall by 14% between 2015 and 2022, while the OECD average rose by 9%. Finland now spends less per student than the OECD average and significantly less than Denmark, Sweden, and Norway. Private funding for higher education sits at 4%, compared to 12% in Denmark and 10% in Sweden. **A forty-year research dynasty from a single professorship.** Teuvo Kohonen, Academy Professor at Helsinki University of Technology from 1975 to 1999, published his landmark 1982 paper on self-organising feature maps in Biological Cybernetics: a foundational contribution to neural network research cited thousands of times. His position was funded by the Academy of Finland. The research dynasty he created produced generations of scientists who now run the Finnish Center for Artificial Intelligence. The Elements of AI course, built on this lineage, has enrolled over two million students from 170 countries. One publicly funded professorship. Four decades of compounding returns. Today, there are 1,400 unemployed PhD holders in Finland, two-thirds of them long-term unemployed. Some conceal their doctoral degrees to improve their chances of getting hired. A groundwater specialist with significant research experience was offered a job cleaning dog waste. An ex-professor was directed to a theatre janitor position. While these PhDs sit idle, the government is simultaneously funding universities to train 1,000 more. The incentive is to produce graduates as measurable output rather than to create the conditions for their employment. **Public money that kept the lights on during a recession.** Public investment in Nokia's R&D began early. In 1980, over 25% of Nokia's total R&D was financed by Tekes, Finland's government technology agency, a share that declined in subsequent years but remained significant. Nokia Research Center was formally established in 1986. During the devastating early-1990s recession, public funding helped keep the research labs alive. This turned out to be critical: the GSM breakthroughs that made Nokia a global giant came precisely during and after this period. Finland hosted the world's first commercial GSM call in 1991 using Nokia-built infrastructure. At its peak around 2008, Nokia employed close to 40,000 people in R&D, spent roughly €6 billion annually, and accounted for nearly half of all private-sector R&D in the country. The current government's R&D target misses something fundamental. Finland's R&D intensity hit 3.7% of GDP in 2009, nearly reaching today's legislated 4% target. That number reflected one company's dominance, not a policy achievement. When Nokia's phone business collapsed, Finnish business R&D expenditure fell from 2.67% to 1.8% of GDP. The government is now legislating that number back into existence through commercial R&D funding, while cutting the university base that produced the conditions Nokia grew from. And Finland implemented €1.25 billion in income tax reductions alongside the university cuts. Defence spending will rise to 3% of GDP by 2029. The budget priorities are consumption and security now, innovation eventually. But "eventually" is the wrong timeline for a pipeline that runs on decades. **Sixty years from cold physics to quantum computing.** Olli Lounasmaa founded the Low Temperature Laboratory at Helsinki University of Technology in 1965 and spent decades pushing the boundaries of cryogenic physics, research with no obvious commercial application. That laboratory's expertise produced BlueFors Cryogenics in 2008, now the world's dominant supplier of dilution refrigerators for quantum computing, with €200 million in revenue. The same lineage produced IQM Quantum Computers, which has raised over $600 million and is becoming the first European quantum company to go public at a $1.8 billion valuation. The pipeline: 60 years from fundamental physics to commercial quantum hardware. By 2016, roughly 500 researchers were leaving Finland each year while only a couple of hundred arrived. The number of PhD-educated Finns moving abroad increased by 37% between 2011 and 2015. The likelihood of returning dropped by 36 percentage points since 2006. Finland's fertility rate fell to 1.25 in 2024, the lowest since records began in 1776, meaning the pipeline is also shrinking at the source. Finland trains people capable of building the foundations for industries worth billions, then makes conditions so unattractive that the next generation takes its expertise elsewhere. ### The doom loops This is not a story about bad policy decisions. It is a story about structural incentive failure: a system where every actor behaves rationally within their own frame, and the collective outcome is irrational for everyone. The government needs to show fiscal discipline within a four-year electoral cycle. Tax cuts are popular. Education cuts are invisible until their effects show up a generation later. The R&D legislation was a previous government's commitment that the current one can't politically abandon but doesn't need to structurally support. Nobody is being stupid. The incentive structure makes it rational to simultaneously announce innovation ambitions and cut the foundations that would make them achievable. This creates self-reinforcing loops where each problem feeds the others. The first loop runs through the universities. Cut base funding. Research quality declines. Fewer qualified students enter graduate programmes. Foundational research output falls. The applied innovation pipeline, which depends on foundational research with a 20-30 year lag, begins to dry up. But not yet. Not visibly. Competitiveness declines a generation later. The economy stagnates. Austerity is required. Cut base funding again. The second loop runs through the talent. Researchers can't find work. They leave. Finland loses the human capital it trained at public expense. Other countries benefit. Finland needs more researchers. It trains more PhDs. There are still no jobs for them. They leave too. The third loop runs through demographics. Birth rate collapses. Fewer children enter a school system already declining in international rankings; Finland went from first to twentieth in PISA mathematics over two decades. The teaching profession loses appeal. Fewer qualified students reach university. The country needs international students to fill the gap. It charges them tuition and gives them three to six months to find work after job loss before facing deportation. They leave. The talent deficit widens. The fourth loop runs through corporate absorption. Companies don't hire PhDs. PhDs don't develop industry experience. Companies see PhDs as impractical. The bias deepens. Universities shift from foundational to applied work to prove "relevance." Foundational research declines further. The innovations that would have created entirely new industries, the way cryogenics created a quantum computing supply chain, the way ALD created a semiconductor manufacturing process, never happen. The companies that would have hired researchers in those industries never come into existence. There is nothing to absorb the talent, because the research that would have created the opportunities was never funded. Each loop reinforces the others. The finance ministry sees fiscal responsibility. The education ministry sees output metrics being met. The companies see rational hiring decisions. The researchers see rational career choices. Nobody is looking at the system. The system is eating itself. ### What the system cannot see Finland is optimising for the appearance of an innovation economy, with R&D spending targets, PhD production quotas, and a 50% tertiary attainment goal recently softened to "as close to 50% as possible" because they already know they'll miss it, while dismantling the conditions that created one. The targets have become the point. The numbers have replaced the reality. Not a single large-cap Finnish company met the global growth average between 2013 and 2023. In advanced industries, Swedish firms invest double what Finnish ones do in R&D. The innovation economy the policy is meant to be building is already underperforming, and the response is to fund the middle of the pipeline while draining the beginning. Every innovation in this article was produced by conditions now being actively degraded: well-funded university departments with stable base budgets; Academy of Finland professorships that let researchers work for decades; a public education system that was once the world's best; corporate labs willing to tolerate ten-year horizons, seeded by public funding in their early years. Finland is not unique in this. The same incentive architecture produces the same outcome wherever it operates. The United States is gutting NIH funding while 75% of its scientists consider leaving the country. The United Kingdom has spent a decade underfunding universities while promising an innovation economy. Southern European countries train researchers who leave for northern Europe, subsidising competitors. The doom loops described here are not Finnish. They are structural features of systems that run on four-year electoral cycles, quarterly earnings reports, and annual budget reviews, while the research that sustains them runs on decades. Finland just happens to be a particularly clear case study because its successes were so visible and its pipeline so well documented. The Kohonens of the future need somewhere to work. The Torvalds of the future need a university with computing infrastructure and an academic network. The Suntolas of the future need a research environment where nobody asks what the commercial application will be for another thirty years. That environment is being dismantled. Not by villains. By a system whose incentive structure makes it rational to eat the seed corn today and worry about the harvest later. The harvest is always later. Until it isn't. --- ### Why Earth Observation Can't Become Weather URL: https://karolinasarna.com/blog/why-eo-cant-become-weather/ Date: 2026-03-04 Tags: earth-observation, structural-analysis A satellite images a farmer's soil every six days with precision no ground sensor matches. She has never seen the data. No app delivers it. The satellite might as well not exist. 150 years of weather history explains why. A Finnish farmer checks the weather on her phone six times a day. Temperature, rain probability, wind speed, collapsed into a glance that shapes what she does before breakfast. The data behind that glance comes from NOAA, a US federal agency whose $7.1 billion budget funds weather satellites, ocean monitoring, and the world's largest forecasting operation. It comes from a European forecasting centre that runs the most accurate weather models on Earth. And from a chain of standardisation, regulation, and institutional infrastructure built over 150 years across two continents. Three hundred kilometres above her field, a satellite images her soil every six days. The radar measures soil moisture with a precision no ground sensor matches. The data is free, open, and sitting on a European server. She has never seen it. No app delivers it. No system feeds it into her irrigation decisions. The satellite might as well not exist. This gap between what Earth observation (EO) satellites collect and what anyone does with it has persisted for decades. The industry has spent those decades convinced that one more technological breakthrough will close it. First it was open data. Then new data formats. Then AI. Each time, the same assumption: we are one innovation away from the moment satellite data becomes as embedded in daily decisions as weather. The assumption is wrong, and we have 150 years of receipts from the domain next door proving it. ### Six conditions, not one Technology mattered for weather. Of course it did. Satellites, computer models, supercomputers, data formats: each generation expanded what was possible. But technology alone never created the transition from expert-only weather charts to something a child checks on a phone. That required six conditions, built simultaneously, mostly by governments. Open data. Binding standards. Centralised infrastructure. Interpretation so simple anyone can act on it. Data pipelines feeding automatically into specific industries. And regulatory mandates that forced all of this to happen. The US discovered these over a century and a half. The Weather Bureau was created in 1870 because the government recognised that weather observation was a public good no private entity would provide at scale. Open data came early: US federal data has been public domain since the founding, so weather observations were free from day one. But free data alone did nothing. For decades, weather information was useful to trained meteorologists and nobody else. Each condition required deliberate institutional action. The international body governing civil aviation made weather briefings mandatory for all flights. Not optional. Not "recommended best practice." Airlines that didn't comply didn't fly. That single mandate created an entire aviation weather infrastructure. Once the infrastructure existed, expanding what flowed through it cost almost nothing. Turbulence forecasts, volcanic ash alerts, space weather warnings: they all rode the same pipeline. The World Meteorological Organization fought for years in the 1990s to force the principle that essential weather data must be exchanged freely between nations. This was a political battle against agencies, including several European weather services, that sold data commercially and resisted giving it away. The principle won because the alternative, fragmentation, was operationally intolerable for forecasting. Data format standardisation was imposed, not adopted voluntarily. The WMO's machine-readable format became the global standard because it was mandated. When the transition to a new version required retiring old formats, the WMO set a hard cutoff date. Agencies that hadn't converted were cut off. That is how you achieve standardisation. Interpretation simplification was the breakthrough that created mass adoption. Weather went from "500mb geopotential height anomaly" (a phrase that means something to atmospheric scientists and nobody else) to "it's going to rain tomorrow." This was an institutional design choice: the US National Weather Service decided its job was to serve the public, not just other meteorologists, and rebuilt its entire product suite accordingly. Centralised infrastructure made it work at scale. The US government's observation systems, forecast models, and data pipelines function as a platform others build on. The $2.5-3.5 billion commercial weather services market exists because this public platform exists. AccuWeather, The Weather Company, DTN, StormGeo: not one of them builds satellites. They build interpretation and sector-specific products on top of publicly funded infrastructure. The value sits in translation, not in observation hardware. Europe took a different path to the same destination. The European Centre for Medium-Range Weather Forecasts, ECMWF, achieved what no national service could alone: a single institution with the computing power to build the world's most accurate forecast. But Europe locked its data behind paywalls for thirty years. The result: a technically superior system with a commercially stunted ecosystem. It took two decades of EU legislation to finally free European weather data. Different path. Same lesson: remove any one of the six conditions and the system doesn't work. ### The mirror Now hold those six conditions up to Earth observation. **Open data**: partially met. The EU's Copernicus programme provides free satellite data. The US Landsat programme dropped fees in 2008; peak annual sales had been 19,000 scenes, and within a year downloads hit one million. By 2017, annual downloads exceeded 20 million. Real progress. But commercial high-resolution data, the kind most useful for decisions, remains expensive and proprietary. **Binding standards**: emerging. The industry has built one widely adopted cataloguing standard (STAC) and is starting to standardise how you request new satellite images. Matt Hanson at Element 84 recently made a compelling case for a standardised ordering system that would let software request satellite imagery the way it currently searches archives. The fragmentation he describes is real: one satellite operator has one kind of interface, another has a completely different one, a third wants you to email a shapefile to a sales engineer. He's solving a genuine problem. It is one condition out of six. **Centralised infrastructure**: fragmented. Copernicus comes closest but it splits across six separate services run by different organisations. Google and Microsoft provide powerful analytical platforms, but they are commercial products, not public infrastructure. **Interpretation simplification**: absent. This is the biggest gap and almost nobody is working on it. Weather went from expert jargon to language anyone can act on. Earth observation hasn't made that leap. The data still arrives in forms that require specialist knowledge to interpret. Nobody has built the equivalent of a weather app for satellite data. Not because it's technically impossible, but because the institutional infrastructure that would make it meaningful doesn't exist. **Automated data pipelines into specific industries**: almost nonexistent. Weather flows automatically into energy grids, aviation, and insurance. Earth observation has pilot projects. The single real exception: the EU's Common Agricultural Policy now mandates satellite-based monitoring for farm subsidy compliance, replacing farmer self-reporting and random physical inspections. That it took an EU regulation to create one such pipeline tells you everything about the state of the other five conditions. **Regulatory mandates**: the critical absence. No international aviation body equivalent mandates EO data for any sector. No regulator requires satellite-derived information in operational decisions. Voluntary principles exist. Nobody enforces them. The EU's recent regulation classifying certain public data as "high-value" (meaning it must be free and machine-readable) covers both weather and Earth observation data on paper. In practice, it frees up national-level environmental datasets. It does nothing to mandate that any institution act on what those datasets reveal. ### The value chain is inverted The commercial consequences of these absences are visible in every satellite company's finances. Planet Labs, the poster child of the commercial satellite revolution, reported $244 million in revenue last year. Government customers, defence, intelligence, and civil agencies account for roughly two-thirds of that revenue. Defence and intelligence alone grew over 30% year-over-year. The commercial sector, the market that was supposed to be the growth engine, remains the smallest segment and has faced headwinds. This is not a Planet problem. It's structural. Compare this to weather. The commercial weather services market is worth $2.5-3.5 billion globally and sits almost entirely in the interpretation layer. Companies that don't build satellites built billion-dollar businesses translating free public data into sector-specific decisions. In weather, the money is overwhelmingly in what you DO with the data, not in collecting it. In Earth observation, that ratio is inverted. Billions flow into satellite hardware. The interpretation layer is thin and struggling. Without the six conditions creating a downstream market, investment goes to the only thing you can build without institutional infrastructure: more satellites. And since no decision infrastructure exists on the receiving end, those satellites sell to the only buyers who already have their own interpretation capacity: military and intelligence agencies. That is the one place where the full chain, from observation to analysis to decision, already exists independently. The commercial satellite race, the competition on image sharpness, revisit frequency, and sensor type, is a structural symptom. When there is no functioning demand infrastructure, the only way to compete is on supply-side specifications. Nobody in commercial weather competes on satellite resolution. The value is in interpretation, and the six conditions created the market for that. Commercial EO satellites are not the problem. Government satellites are slow: one major European satellite took a decade from concept to orbit, involving sixty companies, for a single instrument. Commercial operators provide real advantages in speed, agility, and flexibility. Those advantages land precisely where interpretation infrastructure already exists: government defence and intelligence. And only there. Until regulations force an insurer to use satellite-derived flood exposure the way aviation regulations force airlines to use weather briefings, commercial satellites have exactly one functioning market. The industry keeps searching for the one lever that will trigger the inflection. A decade ago it was open data. Downloads exploded; the downstream market didn't. Then new formats and cloud platforms, which genuinely made data more accessible. Now AI: foundation models, autonomous agents, uncertainty quantification. Each lever works. Each is real progress on an individual condition. Weather needed all six simultaneously. ### Workshops won't write regulations The pattern repeats in the institutions that should know better. Three-day workshops producing PDF roadmaps and framework diagrams. Innovation labs running ideas competitions and hosting visiting researchers. Research sprints winning awards for AI models that can flag their own uncertainty. All genuine. All insufficient. After years of workshops, roadmaps, and frameworks: no institutional change, no regulatory proposals, no automated pipelines built into any industry. The most revealing case sits inside ECMWF itself. For over a decade, the same institution that proved the six conditions work for weather has been running the Copernicus Atmosphere Monitoring Service: continuous global air quality monitoring, greenhouse gas tracking, pollution forecasts. The data is free. The formats are standardised. The infrastructure is world-class. It even offers simplified products: air quality indices, forecasts, support for smartphone apps. After ten years, it still hasn't become embedded the way weather has. How many Europeans check their air quality forecast daily? Where is the automated link to city traffic management, hospital staffing for respiratory wards, or school decisions about outdoor activities? This service proves that even ECMWF, even with five of six conditions largely met, cannot create adoption without regulatory force. No EU regulation mandates that cities act on air quality forecasts. No directive requires hospitals to adjust staffing based on pollution predictions. Five conditions aren't enough. You need all six. ### The second problem And the story gets harder. Suppose EO achieves all six conditions. Suppose regulatory will, institutional reform, and commercial innovation create the infrastructure that weather built over 150 years. What does that infrastructure deliver data into? Institutions designed for a planet that no longer exists. Weather's own success proves this. Weather data flows perfectly into decision frameworks that assume the climate is stable. Flood maps built on historical rainfall records systematically underestimate risk as the climate changes. European building codes calculate structural loads, how much wind or snow a building must withstand, using historical measurements that no longer reflect reality. The European Central Bank found 60% of banks had no framework for stress-testing climate risk. Agricultural subsidy systems still reference historical crop yields as baselines. These institutions don't lack data. They receive weather data perfectly well. The delivery problem was solved. The logic on the receiving end wasn't. Every one of these frameworks assumes the past reliably predicts the future. When the climate was stable, that assumption was invisible. Now it produces confident wrong answers. Better data flowing into a wrong framework doesn't fix the framework. It makes the wrong answers more precise. Earth observation faces both problems at once. It hasn't achieved the first shift: building the six conditions that turn raw data into decision infrastructure. And even if it does, it faces a second shift that even weather hasn't managed: changing institutional logic from backward-looking to forward-looking. A few domains solved both simultaneously. **Energy** is the clearest case. When wind and solar power grew large enough to matter, grid operators couldn't schedule renewable generation the way they scheduled coal plants. Operators in the US and Europe didn't bolt weather data onto their old scheduling systems. They rebuilt scheduling around real-time weather forecasting. The forecast IS the schedule. The old logic was replaced. **Parametric insurance** is another. Traditional insurance prices risk from historical loss tables: what happened in the past in places like this. Parametric products pay out when a measured condition crosses a threshold: rainfall above X millimetres, wind above Y speed. The institution stopped asking "what happened historically" and started asking "what is happening right now, measured by instruments." **Aviation** is a third. Onboard flight computers don't present weather data to a pilot for interpretation against experience. They ingest wind, temperature, and turbulence data and continuously recompute the optimal flight path. The decision system was designed around real-time data from the start. The pattern across all three: the old institutional logic wasn't patched with better data. It was replaced by new logic built around continuous measurement. And regulation was the forcing function every time. ### Reinforcement, not compliance This reframes what Earth observation needs. Not another roadmap. Not another workshop. Not another foundation model. Regulatory mandates designed to replace institutional logic, not just add data to old systems. The difference matters. "Insurers should consider satellite data" produces compliance: a consultant hired, a box checked, nothing changed. "Flood insurance pricing must reflect current observed exposure, updated annually from satellite data, with methodology audited against actual losses" produces a different institution. The system's logic gets rebuilt around continuous observation. That is the difference between improving a system and recognising that the system itself is the problem. The difference also matters because of how reinforcement works. A mandate designed for compliance creates an obligation. A mandate designed to replace logic creates self-reinforcing loops. When insurance pricing accurately reflects satellite-observed risk, properties where risk genuinely decreased get lower premiums. Successful flood defences, better land management: the satellite observes the improvement, the premium drops. That rewards investment in risk reduction. Property owners and municipalities start wanting more satellite-based pricing, not less. The regulation starts the loop. The economics sustain it. Imagine what that looks like in practice. Your insurance app tells you: "This property's flood exposure increased 14% since last year." A farmer's irrigation system receives: "Your field needs water Thursday; soil moisture dropped below the threshold overnight." A municipal engineer gets an alert: "This building's snow load rating is no longer within safety margins based on observed accumulation trends." Not "here is a satellite image," not "download this GeoTIFF." Decisions. In language anyone can act on. The same leap weather made from 500mb geopotential height charts to "bring an umbrella." That world isn't technically impossible. Every piece of data behind those sentences exists today. What doesn't exist is the institutional infrastructure that would turn satellite observations into those sentences at scale, or the regulatory mandate that would force the systems receiving them to act. Weather discovered these reinforcement loops by accident. Free data created commercial services that became a political force defending free data. Simplified interpretation created mass usage that made weather services politically untouchable. Mandates created infrastructure whose sunk cost made expanding the mandates cheap. Each loop reinforced the others. The system became self-sustaining. Earth observation could design them deliberately. The EU's satellite-based farm monitoring system is the proof of concept: a mandate that simultaneously created delivery infrastructure and replaced institutional logic. Self-reporting became continuous observation. Once member states build that pipeline, routing additional mandates through it, environmental compliance and carbon verification become marginal costs. One sector isn't a system. Insurance is the obvious next candidate: the economic pain of mispriced climate risk is acute and growing, the satellite data exists, the interpretation is more tractable than most EO applications, and European insurance regulators have the authority. The honest caveat: weather's regulatory breakthroughs didn't happen because someone made a good argument. The 1900 Galveston hurricane killed 8,000 people and created a political window. Two world wars made weather forecasting a military necessity. Aviation growth created an industry that needed weather integration and pressured regulators to mandate it. Regulation followed crisis, economic pressure, or the visible cost of doing nothing. Every time. So the question isn't whether someone will eventually mandate satellite data in insurance, building codes, and climate risk assessment. As the cost of backward-looking logic becomes more visible, and it does with every flood, every wildfire, every mispriced policy, the political pressure builds. The question is whether the EO industry waits for its own Galveston moment or makes the case before the catastrophe forces it. ### The binding constraint isn't technical The EO industry's favourite activity, convening experts to produce frameworks, is itself a symptom of the problem it needs to solve. The Group on Earth Observations has been running since 2005: twenty years of data-sharing principles that nobody enforces. The World Economic Forum projects $3.8 trillion in EO value with zero institutional mechanism to realise it. Research labs produce excellent science that changes no institutions. Innovation programmes connect people and ideas, while the regulations that would make those connections matter remain unwritten. It is not true that nobody in the EO ecosystem is trying to build institutional infrastructure. People are trying. The attempts either fail politically, stay voluntary, or address one narrow use case. The EU's proposed Forest Monitoring Law was the clearest test. It would have mandated satellite-based data collection across all EU forests: standardised, geo-referenced, harmonised. In October 2025, the European Parliament voted it down, 370 to 261. MEPs called it excessive bureaucracy. One critic pointed out that by rejecting obligations for geo-referenced satellite data on tree cover loss, the Parliament had made early detection of threats almost impossible. The one explicit attempt to make satellite data mandatory for environmental monitoring in Europe was killed by the same Parliament that built Copernicus. The EU Deforestation Regulation comes closest to a regulatory forcing function, but even it hedges. It requires geolocation verification of supply chains; it does not specifically mandate satellite data. It creates a market for EO services without creating the institutional infrastructure that weather has. Internationally, GEO coordinates 103 member governments, but it is voluntary and technical, not political. It does not commit governments to anything. When NASA's budget gets cut, GEO cannot intervene. When Sentinel-1B failed and Europe lost half its open SAR capacity for nearly three years, there was no framework to invoke. No backup triggered. The gap isn't coordination. It's authority. None of these efforts are wrong. All of them are insufficient. The insufficiency has the same root: technologists solving technology problems in a domain where the binding constraint is governance. The institutional anatomy makes this visible. ESA builds satellites. It has a record €22 billion budget. It has zero authority over how anyone uses the data. EUSPA manages the EU's space programmes operationally and is tasked with "downstream market development." It also cannot write regulations. The European Commission can propose regulations, and in June 2025 it proposed the EU Space Act: harmonised licensing, debris mitigation, cybersecurity requirements for satellite operators. Supply-side regulation. It governs the hardware. It says nothing about mandating that anyone use the data collected by those satellites. The bodies that build and operate the observation system (ESA, EUSPA, DG DEFIS) have no authority over insurance pricing, building codes, flood risk assessment, or agricultural policy. Those belong to different directorates, different agencies, different national regulators. The people who build the satellites don't write the rules for the people who should use the data. And the people who write rules for insurance and buildings aren't thinking about satellites. When the Commission tried to bridge that gap with the Forest Monitoring Law, mandating satellite data on the demand side, organised interests killed it in Parliament. €22 billion to build the observation system. Zero regulatory authority to ensure it changes a single decision on the ground. Weather didn't become decision infrastructure through sharper satellites or faster models or smarter formats. Those things helped. They were not enough. What closed the gap was institutional infrastructure that enforced standardisation, mandated openness, simplified interpretation, and embedded data into the systems where decisions are made. Technology was necessary. The institutions that channelled it into decisions were what made it sufficient. Earth observation has the technology. The satellites work. The data is, in many cases, free. The formats are improving. The AI is getting better. What it doesn't have is the institutional infrastructure that would make any of that matter at scale. And the ecosystem's attempts to build it keep failing: voted down, left voluntary, or scoped so narrowly they change nothing structural. The binding constraint isn't technical. It's regulatory. And regulation isn't a technology problem. Until someone with regulatory authority decides that backwards-looking institutional logic is no longer acceptable when forward-looking satellite observation exists, the Finnish farmer will keep checking her weather app. And the satellite overhead will keep imaging her field for nobody. --- ### Europe's Flagship Climate Policy Wasn't Designed to Fail. It Failed by Design. URL: https://karolinasarna.com/blog/eu-carbon-market/ Date: 2026-02-27 Tags: climate-policy, energy-markets, structural-analysis The EU's carbon market was designed to eliminate emissions. It built a financial industry that needs them to continue. Four design choices, each economically logical, each producing consequences the designers never modelled. The EU's carbon market was designed to eliminate emissions. It built a financial industry that needs them to continue. ### How it's supposed to work Governments have two basic ways to reduce industrial pollution. They can regulate directly: set emission limits for each factory, mandate specific clean technologies, ban the dirtiest processes. Or they can create a market. That's what the EU did in 2005 with the Emissions Trading System. The idea is this. The government sets a ceiling on total greenhouse gas emissions. Every tonne of pollution requires a permit. Companies that want to pollute must hold enough permits to cover their emissions. The ceiling drops every year. Permits get scarcer. Their price rises. Companies that clean up can sell their spare permits to companies that haven't, so the reductions happen wherever they're cheapest. By 2039, the ceiling hits zero. No more permits. No more emissions. The system eliminates itself, and the problem along with it. The reason to use a market instead of regulation is cost. A government regulator can't know which of the thousands of factories across Europe can cut emissions most cheaply. But if you let companies trade permits, the market figures it out: the company that can cut cheaply does so and sells its permits at a profit; the company that can't, buys permits instead of making expensive changes it can't afford. Total emissions stay the same. Total cost goes down. That's the theory. For twenty years, this has been Europe's flagship climate policy. It covers power stations, steel mills, cement plants, airlines, and shipping, roughly 40% of EU emissions. It has generated over €245 billion in government revenue from permit auctions. It is also an €881 billion annual trading market where, as of 2024, investment firms and banks account for 63.5% of all volume, 72% of trades come from outside Europe, each permit changes hands ten times on average before it's actually used to cover emissions, and 909 institutions hold derivatives positions on any given day. The system is still an emissions mechanism on paper. But in practice, it has developed the structure, the participants, and the incentive dynamics of a financial market. And financial markets do not build themselves toward their own extinction. How did an environmental policy become a financial ecosystem? Through four design choices, each economically logical, each politically convenient, and each producing consequences that the designers never modelled. ### Four choices that broke in practice **"It doesn't matter whether you auction the permits or give them away free. The market reaches the same outcome either way."** This is a real theorem. In 1960, Ronald Coase showed that when property rights are clearly defined and transaction costs are zero, parties negotiate to an efficient outcome regardless of who holds the initial rights. In 1972, W. David Montgomery formalised this for pollution markets, proving that transferable permits achieve cost-effective control. By the time the ETS was designed, it had become conventional wisdom: allocation method doesn't affect emissions; the cap determines the outcome, the market just finds the cheapest path. The maths checks out, under idealised conditions. It gave politicians a gift. They could tell industry: your permits are free, no cost to you. And tell environmentalists: the emissions outcome is identical. Both statements were technically true, under the model's assumptions. What the model didn't account for was what happens when you hand companies something worth billions — for free, year after year. Between 2013 and 2021, European heavy industry received €98.5 billion in free permits. That's more than the €88.5 billion governments collected from auctioning permits in the same period. Industry got more from the system than the system raised. Companies didn't just use these permits. They accumulated surpluses and sold them. ArcelorMittal made €1.9 billion from surplus sales alone between 2005 and 2019. LafargeHolcim earned nearly €1 billion. On top of that, companies passed through the theoretical cost of carbon to their customers, charging as if they'd paid for permits they'd received for free. Total windfall profits across 15 sectors in 19 countries: €30 to over €50 billion. In 2023 alone, ArcelorMittal received more than €3.8 billion worth of free permits. Heidelberg Materials received nearly €2 billion. And in 2024, 15 national governments paid out €5.52 billion in additional subsidies to cover indirect carbon costs for energy-intensive industries, a 40% increase over the previous year. The theorem said allocation method doesn't affect the outcome. It didn't model that free allocation creates a constituency: companies with billions of reasons to lobby for the system to continue exactly as it is. The subsidy built the lobby that defends the subsidy. The theorem was correct. The assumption that politics wouldn't interfere with the theorem was not. **"The cap guarantees the environmental outcome."** Set the ceiling, let the market find the cheapest path beneath it. As long as emissions can't exceed the cap, the target is met by design. Three things broke this. First, reliable emissions data didn't exist when the system launched. Phase 1 caps were set through National Allocation Plans, where each member state estimated its own ceiling under heavy industry lobbying for generous limits. The European Commission's own review is blunt: "In the absence of reliable emissions data, phase 1 caps were set on the basis of estimates." The result: companies received more free permits than they actually emitted — before reducing anything. The price collapsed to near zero because there was nothing to comply with. Second, the cap only covers 40% of EU emissions. Inside the cap, emissions fell 50%. Outside it (agriculture, buildings, most transport), only 20%. The system reduced emissions sharply in the sectors it covered and barely touched the 60% it didn't. The exclusions aren't random. Agriculture is the EU's single largest source of methane, 80 times more potent than CO₂ over twenty years. It's excluded from the ETS. It's excluded from ETS2. The EU Methane Regulation, adopted in 2024, covers only the energy sector. The official reason is measurability: farm emissions are diffuse. But satellite-based methane detection can now identify emissions at facility level. What remains is the political reality: the Common Agricultural Policy is €387 billion. The tractor protests of 2024 forced the Commission to retreat on pesticide and nature restoration rules. No government wants to be the one that made milk more expensive because of an emissions permit. And even within the boundary the cap claims to cover, a significant share of emissions go uncounted. The ETS treats biomass burning as zero-emission at the smokestack, on the assumption that the trees absorbed the same CO₂ while growing. The European Commission's own 2025 Carbon Market Report shows that these zero-rated biomass emissions amount to 22% on top of what ETS installations actually report. But the carbon debt of the forests being logged to feed European power plants can take centuries to repay. The emissions are real. The accounting says they're zero. Third, and most importantly: the cap reaching zero depends on it being politically irrevocable. It isn't. Germany's environment minister has publicly urged extension beyond 2039. The European Commission is reportedly planning continued industrial emissions after that date. The cap was designed as a technical parameter. It functions as a political variable, one that every actor in the system has an incentive to adjust. **"The carbon price drives investment in clean alternatives."** The logic: companies see the price of pollution, do the maths, invest in cleaner technology when it's cheaper than buying permits. Higher price, more investment. The market discovers the optimal path. The problem is that 90% of industrial emissions are covered by free allocation. In theory, free permits still carry an opportunity cost — every permit you use is one you could sell at market price. Rational firms should treat them identically to purchased permits. And some do: the windfall profits described above come precisely from companies passing through this opportunity cost to customers. But that pass-through shows why the theory breaks down in practice. A cement company that receives free permits, charges customers the carbon cost, sells the surplus, and books the profit never writes a cheque for its pollution. The carbon price shows up on its balance sheet as revenue, not as a cost. Economically, the incentive to reduce should be the same. Behaviourally, it isn't — because eliminating your emissions also eliminates the free permits, the surplus sales, and the pass-through revenue that come with them. The evidence is in the split. In 2013, the power sector lost its free permits and had to buy at auction. Since then, electricity emissions dropped 28.6%. Industrial emissions, still shielded by free allocation, dropped less than 9%. Where companies pay for their pollution, they reduce it. Where they don't, they don't. Picture a cement plant operator in Germany. She receives her emissions permits for free, benchmarked against sectoral averages. She passes the theoretical carbon cost through to her customers — revenue for a cost she never paid. She has surplus permits she can sell on the market. Now someone proposes she invest €500 million in a breakthrough low-carbon kiln. The investment would eliminate her emissions — and with them, her free allocation, her surplus sales, and the pass-through revenue. The system is asking her to spend half a billion euros to destroy her own subsidy. There's a deeper issue, too. Building that kiln is a 15-20 year capital decision. Carbon prices have swung from under €5 to over €100 in the system's lifetime. That volatility is profitable for traders. It's devastating for anyone trying to build a business case for a billion-euro industrial transformation. What decarbonisation investment needs is a credible, stable, rising price floor. What the ETS provides is a volatile market price, subject to political intervention whenever it gets high enough to bite. **"The market is just a tool. When the job is done, the tool is no longer needed."** This assumption was never stated because it seemed too obvious to need stating. The ETS is an environmental policy that uses a market mechanism. The market is a means, not an end. When emissions reach zero, the mechanism becomes unnecessary. Then in 2018, EU emission allowances were classified as financial instruments under MiFID II, for the first time bringing them under the same regulatory regime as stocks and bonds. At that moment, the tool formally became a financial market. It attracted the participants that financial markets attract: investment firms, credit institutions, hundreds of investment funds, hedge funds. It developed the infrastructure that financial markets develop: futures, options, clearing houses, trading desks, data services, compliance platforms, advisory practices. €881 billion in annual trading volume. Each permit transacted ten times. 909 daily derivatives position holders. Trading platforms telling their investors that the carbon market is here to stay. The tool became an institution. And institutions do not design themselves toward obsolescence. ### The dynamic None of this required conspiracy. It required only incentive alignment. Industry receives free permits worth billions. It lobbies to keep them, first supporting the Carbon Border Adjustment Mechanism as a replacement for free allocation, then reversing position as the phase-out date approached. Industry groups now push to delay the transition until CBAM is "proved fully effective" — a condition no one has defined and that could justify indefinite postponement. The financial infrastructure needs the market to persist. €881 billion in annual trading doesn't unwind quietly. The traders, the clearing houses, the exchanges, the compliance consultants: none of them need to manipulate anything. They just need to exist at sufficient scale that their disruption becomes politically unthinkable. When carbon prices spike, the political response isn't to let the price signal work. In 2021, when prices surged past €90, Poland and Spain demanded restrictions on financial participants. The Commission asked ESMA to investigate market manipulation. ESMA found nothing abnormal. But the political pressure didn't disappear — it moved. When ETS2 was set to launch in 2027, covering buildings and transport for the first time, governments delayed it to 2028 over concerns about consumer costs. Every time the system threatens to impose real costs, the response softens it. And through it all, the headline number provides cover. Emissions in covered sectors are down 50% since 2005. The 2030 target of 62% reduction appears achievable. Economists cite studies showing the ETS caused emissions reductions of between 8% and 16%, depending on the study and methodology. These findings are correct but misleading. The reduction is overwhelmingly driven by the power sector, where free allocation was removed. Industrial decarbonisation, the expensive part, the hard part, has barely moved. But the aggregate number supplies legitimacy for a system whose defining feature is its exemption regime. Nobody was corrupt. The carbon leakage argument (that industry would relocate to unregulated countries) had economic logic. The benchmarking system attempted to reward the most efficient producers. Politicians faced genuine competitiveness concerns. Every decision was individually defensible. The structural outcome was that Europe's heaviest polluters were paid billions to continue polluting while the system's headline numbers claimed progress. ### The endgame paradox Under current legislation, the cap reaches zero by approximately 2039. No new permits. When existing surplus is exhausted, no further emissions permitted. This is the system's stated destination. It is also the trading industry's extinction event. A system with zero cap has zero supply. No permits to auction, no futures to write, no positions to take. The financial infrastructure that accounts for 63.5% of trading volume has no product. The compliance industry has no compliance to manage. The platforms have no trades to intermediate. And here is where the system's self-preservation becomes visible. The calls to extend beyond 2039 aren't coming from the margins — they're coming from Germany's environment minister and the European Commission itself. The 2026 ETS revision, the most consequential reform since the system's creation, will determine whether the cap trajectory holds. The lobbying frame is competitiveness. The structural reality is that every actor in the system has a rational interest in decarbonisation proceeding at precisely the pace that sustains the system. Industry receiving free permits. Financial firms trading them. Consultants advising on compliance. Regulators administering the mechanism. Politicians spending auction revenues. All aligned. Fast enough to claim it works. Slow enough that it never reaches zero. ### Optimisation vs. transformation Buying permits, selling the surplus, using futures contracts to manage the price risk — that's optimisation. You're navigating the system as it exists. You shave a percentage off emissions, profit from the rest. Building a green steel plant that eliminates the need for permits entirely — that's transformation. Stegra in Sweden (formerly H2 Green Steel) isn't hedging its carbon price. It's rendering the carbon price irrelevant by removing the emission at source. The distinction matters because the ETS, by design, rewards optimisation and punishes transformation. A company that eliminates its emissions loses its free allocation, exits the trading ecosystem, and gives up the revenue from surplus permit sales. A company that reduces just enough to generate surplus gets to sell those permits for profit while maintaining its position in a system that subsidises its participation. Denmark's cooperative energy model shows what the alternative looks like. By the late 1990s, over 100,000 families owned shares in wind turbine cooperatives that had installed 86% of the country's wind capacity. They didn't trade a carbon permit. They built the infrastructure and became a political constituency for keeping it. The Mankala model in Finland, where industrial consumers co-own power generation at cost, doesn't need a carbon market to incentivise clean energy. The ownership structure makes the market unnecessary. These models don't reduce emissions by trading permits more efficiently. They eliminate the need for permits by building the clean infrastructure directly. The EU ETS is the most sophisticated emissions optimisation system ever built. It has generated over €245 billion in revenue. It employs thousands. It supports academic research, policy careers, trading desks, and consulting practices across the continent. It has also, after twenty years and nearly a trillion euros in annual trading volume, delivered less than 9% industrial emissions reduction while handing the same industries €98.5 billion in free pollution permits and generating windfall profits of up to €50 billion. Every design assumption was economically correct in theory. Every one broke in practice — not because the economics were wrong, but because the economics didn't model the political economy that grew around the system itself. The theorem held. The world around the theorem did not. At some point, the question isn't whether the system is working. It's what the system is working for. ### Meanwhile While Europe debates extending its carbon market past 2039, the physical world is not waiting for the timeline. The 2025 Planetary Health Check found that seven of the nine planetary boundaries that define a safe operating space for human civilisation have now been breached. Climate change, biodiversity loss, deforestation, freshwater use, agricultural pollution, synthetic chemical contamination, and — for the first time in 2025 — ocean acidification. All seven are worsening. Only ozone depletion and aerosol loading remain within safe limits. Atmospheric CO₂ hit 430 parts per million in May 2025, the highest level in at least 800,000 years. The planetary boundary for climate is 350 ppm. We passed it decades ago and are accelerating away from it. The 2025 Global Carbon Budget estimates that the remaining carbon budget to limit warming to 1.5°C will be exhausted before the end of the decade at current emissions. Fossil fuel emissions set another record high in 2025. The ETS was designed in a world that assumed it had time. Time to let the cap decline gradually. Time to let the market discover the efficient path. Time to let the political system absorb the costs incrementally. The remaining carbon budget for 1.5°C runs out before 2030. The ETS cap reaches zero in 2039. That gap is not a rounding error. It is the distance between optimisation and survival. --- **References** *Market structure and trading data* ESMA, Annual Report on EU Carbon Markets (2024 edition, October 2024; 2025 edition, October 2025). Source for: 63.5% trading volume by investment firms/credit institutions, 72% on-exchange volume from non-EEA entities (36% US, 25% UK), 909 daily derivatives position holders (2024 data), 453 investment funds holding 6% of positions, futures share of trading. ESMA 2022 preliminary and final reports on speculation found no abnormality in market functioning; speculative positions ~4% of market. Verifavia / Refinitiv, Annual EU ETS Carbon Market Report 2024 (2025). Source for: €881 billion total trading volume (2023), 9.7 billion EUAs traded (2024), each EUA transacted ~10 times, 85% futures contracts. European Commission, EU ETS Factsheet — Auctioning (2025). Source for: over €245 billion in auction revenues since 2013, €39 billion in 2024 revenue, 62% reduction target by 2030. *Free allowances, windfall profits and industrial emissions* Carbon Market Watch, Free Pollution Permits: Exposed (2025). Source for: €98.5 billion free allowances to industry 2013-2021, €88.5 billion auction revenue same period, ArcelorMittal €1.9 billion surplus sales 2005-2019, LafargeHolcim nearly €1 billion, ArcelorMittal €3.8 billion free allowances 2023, Heidelberg Materials nearly €2 billion 2023, 90% industrial emissions covered by free allocation, 28.6% power sector reduction vs. less than 9% industrial reduction since 2013, €5.52 billion indirect cost compensation 2024 (40% increase), CBAM lobbying reversal. CE Delft (de Bruyn et al.), Additional Profits of Sectors and Firms from the EU ETS 2008-2019 (2021), commissioned by Carbon Market Watch. Source for: total windfall profits €30 to over €50 billion across 15 sectors in 19 countries. Mramor & Tagliapietra, Europe's Emissions Trading System is an Ally, Not an Enemy, of Industrial Competitiveness, Bruegel Analysis (January 2026). Corroborates windfall profits range (€26-€46 billion for cost pass-through component) and industrial emissions reduction of less than 9%. *Emissions reductions and ETS performance* European Commission, 2025 Carbon Market Report (2025). Source for: 50% emissions reduction since 2005 in covered sectors, 62% 2030 target on track, 20% reduction in non-covered sectors. Bayer, P. & Aklin, M., The European Union Emissions Trading System Reduced CO₂ Emissions Despite Low Prices, Proceedings of the National Academy of Sciences, 117(16), 8804-8812 (2020). Source for: ETS reduced emissions by 8.1-11.5% in covered sectors (2008-2016). Colmer, J., Martin, R., Muûls, M. & Wagner, U., Does Pricing Carbon Mitigate Climate Change? Firm-Level Evidence from the European Union Emissions Trading Scheme, Review of Economic Studies (2024). Source for: EU ETS induced 14-16% emissions reduction in regulated manufacturing firms. European Environment Agency, Greenhouse Gas Emissions Under the EU Emissions Trading System (2025). Source for: ETS covers ~37-40% of EU greenhouse gas emissions, 51% reduction in stationary installation emissions 2005-2024. European Commission, Development of EU ETS (2005-2020) (2024). Source for: "In the absence of reliable emissions data, phase 1 caps were set on the basis of estimates"; National Allocation Plans process; price collapse to zero in 2007. *Cap trajectory, 2039 endgame and policy extensions* Clean Energy Wire, German Environment Minister Urges EU to Extend Industry Emissions Trading Beyond 2039 (2025). Source for: Carsten Schneider's call to extend ETS beyond 2039. Clean Energy Wire / Handelsblatt, European Commission Plans to Propose Extending Free CO₂ Allowances (2025). Source for: Commission reportedly planning to extend free allocation beyond 2034, stretch auctioning beyond 2039. Enerdata, Carbon Price Forecast Under the EU ETS (2025). Source for: ETS endgame analysis, cap reaching zero timeline, MSR behaviour, price trajectory modelling. European Environment Agency, Greenhouse Gas Emissions Under the EU Emissions Trading System (2025) and Clean Energy Wire, Understanding the European Union's Emissions Trading Systems (2025). Source for: ETS2 delayed from 2027 to 2028, price history (under €5 to over €100), cap reaches zero ~2039. *Theoretical framework* Coase, R., The Problem of Social Cost, Journal of Law and Economics (1960). Foundational theory for cap-and-trade allocation independence. Montgomery, D.W., Markets in Licenses and Efficient Pollution Control Programs, Journal of Economic Theory (1972). Formalisation of cap-and-trade cost-effectiveness. Zaklan, A., Coase and Cap-and-Trade: Evidence on the Independence Property from the European Carbon Market, American Economic Journal: Economic Policy (2023). Source for: independence property holds for large emitters, fails for small emitters. *MiFID II classification* European Commission, Carbon Market — Questions and Answers (2022); European Commission, Ensuring the Integrity of the European Carbon Market (2025). Source for: EUAs classified as financial instruments under MiFID II from January 2018. *Transformation examples* UNFCCC, Wind Energy in Denmark — Case Study: Good Practices and Lessons Learned (2023). Source for: over 100,000 families in 2,100+ cooperatives owning ~90% of Danish wind turbines by late 1990s. Gorroño-Albizu, L., Sperling, K. & Djørup, S., The Past, Present and Uncertain Future of Community Energy in Denmark, Energy Research & Social Science, 57, 101231 (2019). Source for: 100,000+ families in cooperatives owning majority of Danish wind capacity by early 2000s. *Agricultural emissions and methane exclusion* European Commission, EU Methane Action Plan (2020). Source for: 206 Mt CO₂-equivalent agricultural methane annually, methane GWP-20 of ~80 per IPCC AR6. European Commission, Regulation (EU) 2024/1106 — EU Methane Regulation (2024). Source for: methane regulation covers energy sector only. European Commission / European Parliament, CAP Funds (2021-2027). Source for: €386.6-387 billion total CAP allocation 2021-2027. *Biomass zero-rating* European Commission, 2025 Carbon Market Report, COM(2025) 735 (2025). Source for: zero-rated biomass emissions account for 22% on top of ETS installations' reported emissions in 2024. *Planetary boundaries and carbon budget* Sakschewski, B., Caesar, L. et al., Planetary Health Check 2025, Potsdam Institute for Climate Impact Research (September 2025). Source for: 7 of 9 planetary boundaries breached. Scripps Institution of Oceanography / NOAA Global Monitoring Laboratory, Annual Carbon Dioxide Peak (June 2025). Source for: May 2025 monthly average CO₂ of 430.2-430.5 ppm. Friedlingstein, P. et al., Global Carbon Budget 2025, Earth System Science Data (November 2025). Source for: remaining carbon budget for 1.5°C approximately 170 Gt CO₂ (~4 years at current emissions), fossil fuel CO₂ emissions at record high in 2025. --- ### Actionable Insights, Unexamined Assumptions URL: https://karolinasarna.com/blog/actionable-insights/ Date: 2026-02-19 Tags: earth-observation, organizational-design, structural-analysis Every institution was designed for a planet that no longer exists. AI in Earth observation is the most sophisticated epicycle ever built — transforming how we see the planet while delivering that seeing into frameworks that can only absorb information confirming the world they were designed for. Every institution you interact with — your bank, your insurer, your local government, the company that employs you — was designed for a planet that no longer exists. Your 30-year mortgage assumes the land will be there in 30 years. Your insurance premium assumes this year's flood is an anomaly, not a trend. Your city's infrastructure plan assumes the rainfall patterns of the last century. Your pension fund assumes the coastal real estate in its portfolio will hold value through your retirement. These aren't failures of data or analysis. They're features of a paradigm. ### What the satellites see Earth observation has been surfacing anomalies for decades. Coastlines retreating faster than models predict. Ice sheets losing mass at accelerating rates. Aquifers depleting on trajectories outside historical patterns. Growing zones migrating. Permafrost thawing where the models said it shouldn't — yet. The data is now precise enough, frequent enough, and covers enough years to show that the assumption underneath every major institution — that the physical planet is essentially stable, that variability is cyclical, that historical patterns predict future conditions — is not merely wrong but structurally failing. Not in isolated locations. Systematically. And now AI has arrived. NASA and IBM built Prithvi, an open-source geospatial foundation model. Google DeepMind released AlphaEarth, producing 1.4 trillion numerical fingerprints per year — one for every 10-by-10-meter patch of the planet. Planet partnered with Anthropic to apply large language models to satellite imagery. Over 70 geospatial foundation models with more than 100 million parameters now exist. AI is a genuine paradigm shift. The ability to ask a natural language question of planetary data and get an answer changes who can ask, what they can ask, and how fast understanding moves. A city planner who couldn't read radar satellite imagery can now ask "show me subsidence risk across my district" and get an answer in seconds. A farmer can ask "what's happening to my soil" and see it. A family buying a home can ask what the satellite record says about the land they're investing their life savings in. That's real. And it's new. But look at how it's being deployed. Every announcement uses the same language. Planet and Anthropic: "turn satellite imagery into actionable insights." Google Earth AI: "actionable insights grounded in real-world understanding." Deloitte and the World Economic Forum: "transformation of vast reams of raw EO measurements into actionable insights." NASA and IBM: actionable insights. Every press release. Every pitch deck. Every partnership announcement. Actionable insights. Actionable insights. Actionable insights. Actionable within what framework? Based on what assumptions about the world? Nobody asks. The phrase itself reveals the assumption: we have an existing decision architecture, we just need better inputs. The machine is fine. Feed it better fuel. ### The epicycle problem There's a name for what's happening. Thomas Kuhn called it a paradigm crisis — the moment when observations can no longer be absorbed by the framework designed to make sense of them. During a crisis, practitioners don't abandon the paradigm. They double down. They add complexity. Ptolemy's astronomers, faced with planetary orbits that didn't match their circular model, didn't question the model. They added epicycles — circles upon circles upon circles — each one technically brilliant, each one improving prediction accuracy, and each one burying the real problem deeper. The model was wrong. The epicycles made it wronger in a way that was harder to see. AI in earth observation is the most sophisticated epicycle ever built. It transforms how we see the planet. And it delivers that transformed seeing into institutional frameworks that can only absorb information confirming the planet they were designed for. ### Models are not reality Atmospheric physics teaches you this early, if you're paying attention: every model is an approximation. The map is not the territory. The model is not the planet. Every model draws boundaries. To make the math work, to keep things manageable, because you can't model everything. Those boundaries encode assumptions. And those assumptions are valid — until they're not. The moment you forget the boundaries are there, you're not doing science. You're doing faith. Google's AlphaEarth was trained on data from 2017 to 2024. Seven years of a planet that was already changing. The model's implicit promise is that the relationships it learned — between land cover and flood risk, between vegetation patterns and agricultural viability, between surface temperature and infrastructure stress — will hold. That's an assumption, not a fact. It's not documented anywhere in the product. A recent study called REOBench tested large geospatial AI models against routine image disruptions — cloud cover, haze, compression artefacts, brightness shifts — and found accuracy drops of up to 20%. If the models stumble on clouds, imagine what happens when the planet itself moves outside the patterns they were trained on. That's not a failure mode. That's what climate change means. This applies equally to the institutional models the AI feeds into. Insurance pricing, infrastructure planning, agricultural yield forecasting — each one draws boundaries, encodes assumptions, and produces outputs that look like reality but aren't. Useful fictions, until the conditions they were calibrated for change enough to make them dangerous fictions. And it applies to any new model we build to replace them. The value isn't in finding the right model. There is no right model. The value is in knowing you're always using one, and having the discipline to ask: what does this model assume? Where are the boundaries? What would break it? A model with visible boundaries is a tool you use knowingly. A model with invisible boundaries is a trap you fall into while thinking you're making progress. The field knows this. A research paper on responsible AI for earth observation acknowledges that transparency requires understanding and communicating "ambiguities, potential biases, and errors, or conceptual limitations." The same paper notes that "assuring explainability continues to be challenging." Translation: we know the models hide their boundaries and we have no idea how to fix it. ### When the questions stop making sense Kuhn's most unsettling idea is this: competing paradigms aren't just different answers to the same questions. They're different questions entirely. Within the current institutional paradigm, the question is: "How do we more accurately price flood risk for this coastal property portfolio?" AI solves it better than humans. Foundation models process more data, faster, with fewer errors. Progress is real. But the satellite record isn't providing a better answer to that question. It's revealing that the question is wrong. When flood frequency triples in a decade and the trajectory is steepening, the paradigm response is: update the model, adjust the pricing, recalibrate the risk. But the insurance paradigm assumes risk is distributable — that floods are events, not trends, and that the system absorbs them as random, independent occurrences. When the data shows the system itself is destabilising, "distributable risk" stops being a coherent concept. "How do we price this risk?" becomes meaningless the way "how many epicycles does Mars need?" became meaningless after Copernicus. A satellite shows coastal infrastructure built on land subsiding at 3cm per year. The paradigm question: "What's the adjusted asset value?" The question from outside: "What does it mean for a financial system to hold assets on land that is physically disappearing?" One fits in a spreadsheet. The other doesn't fit in any existing institutional category. ### Building intentionally I spent years building climate adaptation products. The conversation was always the same: "Where does this data fit in the customer's existing workflow?" It sounds pragmatic. It sounds customer-centric. It's a paradigm question. It assumes the existing workflow is the right frame and just needs better inputs. The question I wish I'd asked sooner: "What does this data tell us about whether the workflow makes sense?" The most powerful move isn't replacing one framework with another. It's knowing that every framework — including the one you're building — is a model with boundaries. And building accordingly: with visible assumptions, and a built-in way of recognising when the world has changed enough that the system no longer works. What would that look like in practice? At every joint in the pipeline, the system makes visible: this model assumes X. This trigger was calibrated to Y conditions. This recommendation holds while Z remains true. And here is the satellite evidence on whether Z remains true. The person at the end — the city planner, the insurer, the farmer, the family — gets not just the recommendation but the conditions under which that recommendation stops being valid. Nobody builds infrastructure with a self-destruct switch. But infrastructure that knows its own limits would have exactly that: not a dramatic shutdown, but a continuous signal. These are the assumptions this system rests on. This is what the satellite data says about whether those assumptions still hold. When they don't, this is where the system stops being trustworthy. The AI that builds the pipeline matters. The AI that surfaces the boundaries matters equally. The technology can do both. What determines which one gets built is the mental model of the people building it. Build the infrastructure and build it knowing it's temporary. Serve current decision-makers and show them the limits of their decisions. There's no market for this. No RFP for "show us where our own thinking breaks." But the data exists. The AI exists. And the planet — the actual planet, not any model of it — isn't waiting for our institutions to catch up. What we build from that awareness determines whether Earth Observation becomes the most sophisticated system of epicycles ever constructed, or something that helps us live on the planet we actually have. --- ### A fable of two roadmaps URL: https://karolinasarna.com/blog/fable-two-roadmaps/ Date: 2026-02-15 Tags: organizational-design, earth-observation, scaling A fifteen-year story of how a climate tech company became a defence contractor — not through one dramatic decision, but through a thousand small rational ones. No single decision was wrong. That's the whole point. This is the story of a company that never changed its mission. The mission statement still says the same thing it said fifteen years ago. Nobody updated it. Nobody needed to. Nobody noticed. **Year 1.** Three founders leave a research institute with a sensor that can see something nobody else can see. The details don't matter — pick your wavelength, pick your orbit. What matters is the pitch deck. Slide 7: "We're building planetary-scale environmental intelligence." Slide 12: the TAM is $4.2 billion. Climate adaptation. Precision agriculture. Disaster response. Insurance. The VCs nod. The cheque clears. The roadmap has twelve milestones. All of them point toward understanding what's happening to the planet. **Year 2.** First images. They're beautiful. The founders post them on LinkedIn. The ESA contract comes through. A pilot with a reinsurer. A conversation with a UN agency that goes well but produces no purchase order, because the UN doesn't have a purchase process for this. The team is twenty people. Everyone believes. **Year 3.** Series A. The lead investor has a climate thesis. Patient capital, long time horizon, "we're building generational infrastructure." The term sheet is generous. The board is supportive. The roadmap still points the same way. But in due diligence, one partner asks a question that will echo for the next twelve years: "What's your path to $10M ARR?" The honest answer is: nobody knows, because the commercial climate market doesn't have budget owners yet. The answer on slide 14 is: "Land and expand with insurance and agriculture." **Year 4.** The reinsurer pilot ends. Not because the data was bad, the data was extraordinary. But the reinsurer can't figure out which department should pay for it, how to integrate it into underwriting workflows, or how to explain to their board why they're buying satellite data. There's no regulatory mandate requiring it. There's no line item in the budget. There's no precedent. The agriculture customers want the data but can't pay what it costs. The UN agency comes back with a grant. Just enough to keep the lights on, not enough to scale. Revenue at end of year 4: €1.1M. Burn rate: €3.8M. Then a defence ministry calls. They saw a conference demo. They have a specific problem. They have a budget. They want to start in eight weeks. The founders discuss it for one evening. The contract is €600K. "It funds six months of the climate roadmap," they tell themselves. They sign it. The signal that the climate market needs ten more years of patient infrastructure building — regulation, workflow integration, budget line creation — won't arrive for another three years. By then, the company will already be structured around the answer it has today. **Year 5.** The defence contract delivers on time. The ministry comes back with a second contract. Larger scope, faster timeline, clearer specs. The procurement officer knows exactly what he wants. After two years of trying to explain to agricultural cooperatives why satellite data matters, the clarity is intoxicating. Series B conversations start. The lead investor from Series A introduces a growth-stage fund. The growth fund likes the technology. They like the team. They don't love the revenue mix. "Your commercial pipeline is soft," they say. "But the defence traction is real." The founders build two slides. One shows the "commercial roadmap": agriculture, climate services, and insurance. The other shows "government and institutional revenue," which now includes defence contracts but is labelled more broadly. The growth fund invests. The board gets a new member who spent fifteen years in aerospace. Revenue at end of year 5: €3.2M. Defence is 40%. The roadmap now has two tracks. Everyone agrees this is temporary. What nobody maps on a whiteboard: success to the successful is the most predictable loop in any system. The side that gets resources performs. The side that performs gets more resources. The side that's starved underperforms, which justifies starving it further. It doesn't require malice. It just requires a quarterly review cycle. **Year 6.** The company hires a VP of Sales from the defence industry and a VP of Operations from a Series D startup. Both are excellent. The VP of Sales knows procurement cycles, classification requirements, and the right language for proposals. Within six months, the pipeline triples — but only on the defence side. The VP of Operations implements OKRs, quarterly business reviews, and a reporting cadence. The dashboards look professional. Revenue per head. Pipeline velocity. Contract win rate. Time to delivery. The defence side of the business produces beautiful metrics — clear demand signals, predictable procurement cycles, measurable everything. The climate side produces pilot results, research citations, and "promising conversations" that don't convert to the pipeline. In the quarterly review, the defence team presents numbers. The climate team presents narratives. The board knows which one they trust. What gets measured gets managed. What gets managed gets resources. What gets resources gets measured better. The loop tightens. Meanwhile, the unmeasured slowly becomes unmeasurable, then unreal. Within two years, the climate work will be described in board meetings as "harder to track" rather than "strategically important." The language shift happens first. The resource shift follows. The climate team asks for two more engineers to build the agricultural analytics platform. The VP of Sales asks for three more engineers to meet the defence delivery schedule. Both requests go to the board. The defence engineers are approved immediately. The climate engineers are approved "next quarter." Next quarter, the defence pipeline needs four more. **Year 7.** A new analyst joins. Smart, young, excited about the mission. On her first day, she asks which project she'll be working on. She's assigned to a defence contract. "Just for onboarding," they say. "The climate work needs experienced people." She's still on defence contracts eighteen months later. She's good at it. She's been promoted. The company is eighty people now. At twenty, engineers talked directly to customers. Decisions happened in real time. The founders' obsessive knowledge of the problem permeated every conversation. People self-organised around what mattered because they could see what mattered. At eighty, there are three management layers between the engineer and the customer. Decisions travel up for approval and back down for execution. The weekly all-hands has a slide deck. The founders no longer eat lunch with the team. They eat lunch with investors and procurement officers. The three founders now spend their time differently. The CEO talks to defence procurement officers and VCs. The CTO manages the classified development environment. The Chief Scientist, who built the original sensor, still leads the climate research group. It's the smallest team in the company. Revenue at end of year 7: €8.4M. Defence is 58%. **Year 8.** Series C. The term sheet requires a path to €50M ARR within three years. The board deck is forty slides of metrics — growth rate, pipeline coverage, net revenue retention, CAC payback period. The defence business produces all of these cleanly. The climate business produces research impact, potential market size, and the word "emerging." The only realistic path to €50M runs through defence and national security. Everyone at the board table knows this. Nobody says it. They are all making reasonable decisions based on the information in front of them. The information in front of them is this quarter's pipeline and next quarter's forecast. The climate trajectory is a fifteen-year story. The board deck has a three-year horizon. The commercial climate roadmap is in appendix B of the board deck. It used to be slide 7. New investors join the board. Their portfolio includes three other defence tech companies. They're helpful. They make introductions. The introductions are to other defence ministries. **Year 9.** The Chief Scientist presents a breakthrough: a new processing pipeline that could halve the cost of agricultural monitoring. It would make the product accessible to farming cooperatives in East Africa. She needs €400K and two engineers for nine months. The board asks what the revenue projection is. She says it's hard to quantify because the market doesn't exist yet, but this could be the thing that creates it. The room is polite. But the language of the room is ARR and pipeline coverage and CAC payback. She's speaking a different language. The one the company used to speak. The board approves €200K and one engineer. "Prove the concept first." A half-commitment is worse than no commitment. It creates the appearance of support while ensuring failure. When it fails — and with half the resources and a third the timeline, it will fail — the failure will justify the original scepticism. Fixes that fail. The loop closes. That same meeting, they approve €2.1M for a new classified processing facility. The Chief Scientist doesn't argue. There's nothing to argue with. The logic is airtight. If quantity forms the goals of your feedback loops, quantity will be the result. **Year 10.** The company wins a framework contract with a European defence agency. Multi-year. Eight figures. The press release calls it "a milestone for European sovereign capability." LinkedIn goes wild. The stock options are suddenly worth something. The founding team celebrates. This is success. This is what they worked for. Isn't it? **Year 11.** Three climate team engineers resign within two months. They don't leave angry. They leave tired. One joins an NGO. One joins a climate startup that just raised a seed round. One goes to academia. The company hires replacements. The job description says Earth observation analyst. The candidates all have defence backgrounds. This is who applies now. At twenty people, the company could pivot to anything: new sensor, new market, new application. That was resilience. Not a strategy document. An actual structural property of a system that hadn't yet optimised itself into a single shape. Resilience isn't static. It's the ability to bounce back, to reorganise, to evolve. By Year 11, that ability is gone. The company is optimised. It's also brittle. It just doesn't know it yet, because nothing has pushed back. **Year 12.** The agricultural analytics platform is quietly shelved. Not cancelled. Shelved. "Deprioritised pending market development." The code sits in a repository that fewer and fewer people have access to, because the production environment is now classified. The mission statement on the website still says "planetary-scale environmental intelligence." The about page still shows an image of a green field. The last blog post about agriculture is fourteen months old. Revenue at end of year 12: €34M. Defence is 73%. The remaining commercial revenue is mostly government environmental agencies who use the same contracting vehicle as defence. **Year 13.** A journalist writes a profile of the company. "From Climate to Combat: How Europe's EO Startups Found Their Market." The CEO is unhappy with the headline but can't dispute the numbers. He gives an interview explaining that the technology is "dual-use" and that the physics doesn't change based on application. This is true. The physics doesn't change. The roadmap did. "Dual-use" is a boundary someone drew to make two diverging realities fit in one sentence. Boundaries are useful for clarity. They become dangerous when you forget you invented them. **Year 14.** The company is approached about an acquisition. The buyer is a defence prime. The valuation is strong. The board discusses it seriously. The Chief Scientist is not in the room. She left in year 11. She now runs a small research group at a university, publishing papers about agricultural monitoring using open-source satellite data. Her work gets cited often. It doesn't get funded. **Year 15.** The acquisition closes. The press release mentions "environmental monitoring" and "climate resilience" in the second paragraph. The first paragraph is about "strengthening Europe's sovereign intelligence capability." The mission statement is updated for the first time in fifteen years. Nobody notices. It now reads: "Delivering trusted intelligence for a secure and resilient world." The roadmap is classified. --- If you draw the two roadmaps side by side, the one from the pitch deck in Year 1 and the one that actually happened, they diverge at Year 4. Not dramatically. Just a small angle. Five degrees, maybe. A single rational contract signed for rational reasons. By Year 8, the angle is thirty degrees. By Year 15, they point in opposite directions. No single decision was wrong. The board saw quarterly metrics. The VCs saw portfolio returns. The founders saw burn rate. The climate market saw a company drifting away. None of them saw all of it at once. And none of them were responsible for seeing all of it at once. That's the thing about systems. They don't require someone to be wrong. Every quarter, the climate roadmap eroded a little. Not by a dramatic cut — by a reallocation, a deprioritisation, a "next quarter." The target shifts so slowly that nobody experiences a decision to abandon it. Eroding goals. The technical term for what happened is drift to low performance. The system doesn't abandon the difficult goal through one dramatic decision. It drifts toward the easier goal through a thousand small ones. Each rational, each incremental, each invisible in the moment. The stated goal stays fixed on the slide deck. The actual goal creeps downward every quarter. Not to "become a defence contractor" but to "hit this quarter's number." Eventually, those are the same thing. By the time anyone notices, the drift has been happening for years. Underneath that: loop dominance. The mission loop ran the company for three years. Then the defence revenue loop overtook it. And once a reinforcing loop dominates, it feeds itself. The relationship isn't linear. It accelerates. By the time anyone notices which loop is in charge, it's been in charge for years. Success to the successful. The twenty-person company that built the original sensor was a self-organising system — adaptive, resilient, capable of evolving. Every layer of management, every dashboard, every quarterly review that made it "scalable" also stripped those properties away. Self-organisation sacrificed for short-term productivity. As it usually is. If quantity forms the goals of your feedback loops, quantity will be the result. The system is stable. It's just stable around the wrong goal. The original sensor still works. The physics hasn't changed. The wavelength is the same. The orbit is the same. The data is the same. The question was never about the sensor. It was about who gets to use it, for what, and who decides. Meanwhile, the planet is changing. As it always has. But at a rate that is much higher than before. A rate that doesn't necessarily hurt the planet, but makes our ability to live and flourish on it less and less likely. The sensor that was built to see that is still in orbit. It's just looking at something else now. *This is, of course, a work of fiction.* --- ### Why your electricity bill makes no sense. URL: https://karolinasarna.com/blog/electricity-bill/ Date: 2026-02-12 Tags: energy-markets, climate-policy, structural-analysis Europe prices all electricity at the cost of the most expensive generator running. You pay gas prices for wind power. The system was designed for fuel-burning grids. It's now absurd — and five interlocking feedback loops keep it locked in place. I live in Helsinki. Finland generates 95% of its electricity without fossil fuels — nuclear, hydro, wind. My grid is among the cleanest on Earth. Every month, I pay a small add-on on my electricity bill. 0.39 cents per kilowatt-hour with Helen, though Oomi charges 1.72 so that my contract says "renewable energy." A green premium. In a country where nearly all the electricity is already green. Where does that premium go? Into a Guarantee of Origin certificate. A piece of paper that says the electrons I consumed were "matched" to renewable generation somewhere in Europe. A Norwegian hydropower plant sells its certificate to my Finnish retailer, pockets the premium, and the actual electricity flows wherever physics dictates. The electrons in my wall socket are identical to my neighbour's on a cheaper contract. The only difference is that I paid more for a label. The most sophisticated greenwashing mechanism ever designed, hiding in plain sight on every electricity bill in Europe. But that's a whole other article. Because once I started pulling on that thread, I found something bigger: a pricing system that made sense when European grids ran on fuel and is now absurd when they run on wind. And the reason nobody can change it. In Finland, most of the time, I pay some of the cheapest electricity prices in Europe. The clean grid works. But the pricing architecture underneath it was built for a different world. And on the days it bites, it bites hard. When the wind drops and the reactor goes for maintenance, prices spike from near-zero to €397 per megawatt-hour. Same grid. Same day. And across the rest of Europe, where grids still need gas to keep the lights on every single day, this isn't an occasional spike. It's a permanent condition. ### How your electricity bill actually works (and why it makes no sense) Imagine a grocery store where every item costs the same and the price is set by the most expensive product on the shelf that day. You're buying rice. It costs the store €2 to stock. Pasta, €2.50. Bread, €3.70. But today, one customer needs saffron, which costs €17. So everything in the store, the rice, the pasta, the bread, is priced at €17. You'd walk out. You'd call it a scam. You'd demand to speak to a manager. This is how Europe prices electricity. Not approximately. Literally. Every day, an algorithm called EUPHEMIA clears the electricity market across 27 coupled European countries and 61 bidding zones. Generators bid their costs. Solar and wind bid near zero as they have no fuel costs. Nuclear bids low. But somewhere in the system, a gas plant is needed to keep the lights on during a calm winter evening. Gas generation costs €80-130 per megawatt-hour once you include fuel and the pollution penalties that EU law requires gas plants to pay. And because the system uses marginal pricing, the most expensive generator needed to meet demand sets the price for everyone; that gas price becomes what everybody pays. The wind farm that bid zero? Paid the gas price. The nuclear plant? Paid the gas price. The solar farm? Paid the gas price. You, the consumer? You pay the gas price. For wind-generated electricity. EU electricity market liberalisation began in the 1990s, and marginal pricing was the logical choice. Most generators burned fuel. Their costs were similar. The marginal price roughly reflected what electricity actually cost to produce. But now we've built a grid where the cheapest generators produce most of the electricity, and the most expensive generator, needed for maybe 15% of hours, sets the price for ALL of them. As Oxford energy economist Dieter Helm puts it: "In a system with lots of zero marginal cost generation, the price of electricity for all the electricity generation should not equal the spot price of gas." We're running that grocery store where saffron sets the price for rice, except the store is now 80% rice and 5% saffron. And no one is allowed to change the pricing rule. ### OK, so change it. That's the obvious response. The system is clearly outdated. Consumers are overpaying. Renewables are cheap. Just update the pricing. Except every time someone tries, the same thing happens. Each layer of the answer is more infuriating than the last. **The people building clean energy don't want it changed.** Think about it. If you're a wind farm operator, your turbines produce electricity at near-zero cost. But you get paid the gas price. That gap between your zero-cost and the €80-130 gas clearing price goes straight to your balance sheet. Every megawatt-hour of wind you produce earns you gas money. The "broken" system is printing you cash. So when the EU tried to reform electricity markets after the 2022 energy crisis, who lobbied hardest against fundamental change? WindEurope, the industry's own trade body, stated its position plainly: "Keep short-term wholesale markets based on marginal pricing and the merit order," yes, the system where the cheapest plants run first, and the most expensive one sets everyone's price. SolarPower Europe's head of regulatory affairs went further: "Changing the foundations of electricity markets, such as marginal pricing, is creating regulatory instability and actually halting investments into new renewable technologies." They joined forces with fossil fuel incumbents who also benefit from the pricing mechanism. The entities building the clean energy future are financially incentivised to preserve the fossil pricing architecture. The defenders of the system ARE the disruptors. Let that sink in. ### OK, force them. Regulate. The system has an immune system for that. When gas prices spiked in 2022 after Russia invaded Ukraine, electricity bills across Europe tripled overnight. There was genuine political rage. Spain's ecological transition minister Teresa Ribera captured it: "The current crisis has shown how vulnerable the current design of the electricity market is to stress situations, as well as the tremendous consequences for domestic consumers." Ursula von der Leyen stood in the European Parliament in September 2022 and promised "deep and comprehensive reform of the electricity market." It felt like the moment. So what happened? Governments deployed emergency measures. Price caps. Windfall taxes. Direct subsidies to consumers. EU member states spent over €650 billion on energy support. And it worked. Prices came down, families were shielded from the worst. But by successfully treating the symptoms, they removed the urgency to cure the disease. The pain went away. The outrage faded. By the time the "reform" was adopted in May 2024, it explicitly preserved marginal pricing. The Commission framed its own modesty as wisdom. The crisis that justified change also defused it. Spain and Portugal tried something different, the Iberian Exception, capping gas bid prices and reducing wholesale electricity prices by roughly 40%. It worked. Then it expired on December 31, 2023. And no one renewed it because, by then, gas prices had normalised and the lobbying loop was back in control. Crisis builds pressure → emergency measures relieve pressure → reform stalls → wait for next crisis. Rinse, repeat. The system survives by producing just enough response to prevent structural change. ### Fine, but the EU reformed markets before. Just do it again. The EU has 27 member states with completely different energy situations. France has 57 nuclear reactors and wants nuclear included in any new pricing scheme. Germany shut its last reactor in April 2023 and wants pure market signals for renewables. Nordic countries have cheap hydropower and don't want to subsidise anyone else. Spain wants intervention. Poland is still burning coal. Each country has effective veto power. The result is always a lowest-common-denominator compromise: reform that's just enough to claim progress, never enough to change the structure. The more integrated the market, the harder it is to change, because every country has a hand on the steering wheel and none of them wants to turn in the same direction. ### But surely economists see this is broken? They do. And the system has an answer for that too. The current market design does one thing genuinely well. It makes sure the cheapest available power plant turns on first, every single hour. And because EU countries share electricity across borders through this system, it saves an estimated €34 billion per year compared to each country running its own separate market. When ACER, Europe's energy regulator, was asked to assess the market, their answer was clear: "The current energy crisis is in essence a gas price shock, which also impacts electricity prices." The current electricity market design, they concluded, "is not to blame for the current crisis." And they're right. The system IS efficient. It's efficient at solving a problem that no longer exists: price discovery among similar fuel-burning generators. But the problem changed. We went from a world where most generators had similar costs to a split world: near-zero renewables and €80+ gas, with nothing in between. The right answer to yesterday's question is now the wrong answer to today's. The economists aren't wrong. They're irrelevant. And their correctness provides intellectual cover for everyone who benefits from keeping things as they are. ### Five loops, no villain Step back and see the whole picture. The renewable industry won't push for change because the broken system profits them. The financial infrastructure makes change terrifying not just for traders, but for anyone with a long-term energy contract. Imagine you're a company that signed a 10-year power purchase agreement based on today's pricing logic. Now imagine the EU announces it's changing how electricity prices are calculated. Your contract, your bank's insurance against price swings, the financial bets that investors have placed based on that contract, all of it suddenly references a price that works differently. Trillions of euros in financial contracts are built on top of the current price structure, the way mortgages are built on top of interest rates. Changing the pricing mechanism doesn't just change tomorrow's electricity price. It destabilises every financial contract that was written assuming yesterday's rules would continue. That's why Energy Traders Europe, representing over 170 trading firms, frames any reform as "systemic risk." And they have a point, just not the one they think they're making. Add to that: emergency measures absorb political pressure before it can become structural. Twenty-seven countries can't agree on which direction to go. And the economic consensus validates the status quo. Each of these actors is behaving rationally within their own feedback loop. Nobody is the villain. That's the whole problem. This is the textbook coordination failure. The exact same mechanism that makes capitalism brilliant at optimising within a system and terrible at changing the system itself. Each actor is individually rational, yet the collective result is irrational for everyone. The system needs to change, but no single actor gets enough benefit from changing it to justify moving first. So nobody moves. And there's a deeper trap. The EU's stated policy goal is to deploy more renewables to lower consumer prices. But every new wind farm and solar panel deployed under marginal pricing creates another profitable actor whose business model depends on keeping the pricing system unchanged. The more successful the energy transition is at building clean generation, the stronger the constituency that opposes changing how that generation is priced. The policy accelerates both the disease and the immune system at the same time. The goals of the actors are not aligned. They're structurally opposed. Consumers want cheap electricity priced near production cost. Renewable generators want high wholesale prices because the gap is their profit margin. The financial sector wants price volatility because that's where trading profits come from. Politicians want to avoid crises, not reform systems. The EU thinks it's solving one problem. It's feeding two contradictory loops simultaneously. Windfall profits fund the lobbying that shapes the intellectual consensus that frames reform as risky, which strengthens the financial lock-in that politicians won't touch outside a crisis, which never lasts long enough because emergency measures act as a relief valve. And the consumer? You and me? We're the only actors in the entire system without a feedback loop that protects our interests. Every other actor has an organised mechanism to convert their interests into system-preserving action. Consumers just pay. They can't individually exit, everyone needs electricity. They can't easily organise, they're diffuse, and the system is too opaque for most people to even understand what's happening to them. That's the binding constraint of the entire system. Not technology. Not economics. Not policy. The absence of an organised consumer constituency with skin in the game. Remember that. It matters for what comes next. I check my electricity app every morning now. Not because I can do much about it. Because once you see the gap between what you're paying and what the electricity actually costs to produce, you can't unsee it. ### But the system is eating itself. The part nobody in the lobby groups wants to talk about. The more renewables you build under marginal pricing, the more renewables destroy their own revenue. Solar floods the grid during sunny hours, pushing prices toward zero or below zero. Germany recorded 457 negative price hours in 2024, up 52% from the year before. Spain saw its first-ever negative electricity prices in April 2024. Finland led Europe with roughly 700 negative hours, making about 8% of all hours in the year. This is called cannibalisation. It means the electricity you're producing is becoming worthless in the very moments you're producing the most of it. German solar generators received, on average, only 60% of the baseload wholesale price in 2024. For every euro a typical power plant earned, a solar farm earned 60 cents, because its output was concentrated in the hours when prices collapsed. In Spain, monthly solar capture rates have already hit 20%, meaning that in peak solar months, solar farms earn a fifth of what a conventional plant earns, even as their hardware costs continue to fall. The paradox: each new gigawatt of solar makes the previous gigawatt less profitable. To keep investment flowing, governments have to offer larger and larger subsidies, guaranteed long-term price contracts, and public support mechanisms to compensate for the revenue that the market destroys. EU solar installations dropped in 2025 for the first time since 2016. Not because solar is too expensive. Because the market makes it unprofitable even when it's cheap. The system designed to support generation investment is now undermining it. The market is consuming itself from the inside. And no amount of emergency measures can fix a structural contradiction. This is the forcing function. Not a political crisis. Not an external shock. The market's own logic, taken to its conclusion, breaks the market. ### The alternative already exists. Most people just haven't heard of it. While the political system churns, some people have already figured out what the binding constraint is and built around it. In Finland, roughly half of all electricity is produced under something called the Mankala model. Companies jointly establish a non-profit entity that produces electricity at cost. Shareholders pay production costs proportional to their ownership. No marginal pricing. No gas setting the price for wind. Just: what did it cost to produce? That's what you pay. Teollisuuden Voima operates Finland's Olkiluoto nuclear reactors under this model, including Olkiluoto 3, the largest reactor in Europe, which came online in 2023. Their industrial shareholders get power at roughly €30-35 per megawatt-hour (blended across old and new reactors), stable and predictable. Regular Finnish consumers, meanwhile, still ride the spot market. When the wind blows, and Olkiluoto runs, they get some of Europe's cheapest power. When Olkiluoto goes for maintenance, and the wind dies down — as happened in May 2024 and this winter multiple times — prices spike to €397 per megawatt-hour. Same country, same grid, radically different outcomes depending on whether you're inside or outside the cooperative structure. The cheap nuclear output is captured inside the Mankala structure. It never reaches the public price. That's the structural inequality: industrial insiders get cost-based pricing, while households get the volatility of a market that swings between negative and 400. I live in this gap. You probably do too. Now look across Europe. Thousands of energy cooperatives exist, representing more than 2 million citizens across 25 countries. Austria alone has over 4,000 renewable energy communities and growing fast. Belgium's Ecopower cooperative has over 70,000 citizen members. In Germany, up to 47% of all renewable capacity was citizen-owned at its peak. Denmark's wind cooperatives put 100,000 families into direct turbine ownership — by 2001, cooperatives owned 86% of all Danish wind turbines. These cooperatives are structurally the same thing as Mankala: people organising to produce or buy electricity at cost, bypassing the marginal pricing system that forces them to pay gas prices for wind power. One model serves Finnish industry, the other serves European citizens. Same logic, different scale. The EU's Clean Energy Package formally recognised this model. A widely cited CE Delft study estimates that 45% of European renewable production could come from citizen-owned energy communities by 2050. But why stop there? For private consumption, the barriers aren't technical. They're regulatory. Most countries still lack proper frameworks for energy sharing between neighbours. Upfront capital requirements exclude lower-income households. Grid operators, whose revenue depends on the existing system, aren't exactly racing to make cooperatives easy. And in the Nordics, you need a diverse generation portfolio; solar alone doesn't work when the sun sets at 3 PM in December. But none of these are fundamental obstacles. They're design choices that could be changed. Denmark proved it works. Austria is proving it scales. The question isn't whether cooperatives can replace marginal pricing for consumers. It's whether the political will exists to let them. ### What would a different system actually look like? Academics and energy economists have been quietly working on this, even as the political system pretends the question doesn't exist. The most developed proposal is the Green Power Pool, designed by UCL's Michael Grubb and colleagues. The logic is simple: if renewables produce electricity at near-zero marginal cost, why route that electricity through a market where gas sets the price? Instead, aggregate renewable output into a separate pool, price it at its actual average cost (which is falling every year), and sell it directly to consumers. When the wind doesn't blow, and the pool can't meet demand, it purchases from the wholesale market — gas, hydro, whatever is needed — and that balancing cost is added transparently. Consumers connected to the Green Power Pool pay the blended cost: cheap renewables most of the time, with transparent balancing charges when needed. No more paying gas prices for wind electricity. Grubb's team calculated that in the UK alone, pool-based electricity from existing government contracts would cost less than half the wholesale price — and the share keeps growing as more renewables come online. Others propose variations. Jacques Percebois and Stanislas Pommeret at the University of Montpellier modelled a weighted-average cost system in which the market price reflects the actual mix of generators running, not just the most expensive one. The Spanish government pushed for fixed-cost contracts covering renewables' lifetimes. The CEPR (Centre for Economic Policy Research) proposed keeping marginal pricing to decide which power plants turn on each hour, but layering mandatory long-term contracts on top, so the spot price drives operations but not consumer bills. These proposals differ in mechanics, but they share one insight: the wholesale spot price should drive how generators operate, not what consumers pay. Operational efficiency and consumer pricing are two different problems, and lashing them together with a single mechanism is what creates the dysfunction. None of this requires abolishing markets. It requires recognising that a market designed for commodity-cost generators doesn't serve consumers when most generation has near-zero marginal cost but high upfront capital. As Helm wrote: "It is not particularly difficult to set out what an efficient energy system might look like. The wholesale market model of the twentieth century is being displaced by a twenty-first-century decarbonising market, with lots more zero marginal cost generation. It is increasingly a utility-style fixed and sunk cost system." The Mankala model IS this system, just built bottom-up by Finnish industry rather than top-down by regulators. Energy cooperatives are a bottom-up system built by citizens. The academic proposals are trying to make the top-down version politically viable. All roads lead to the same place: pay for what electricity costs to produce, not for what the last gas plant bid. ### What's happening now: optimisation vs. transformation In Finland, Oomi, the country's largest energy retailer with over 800,000 customers, offers Flex contracts that reward consumers for shifting electricity use to cheaper hours. Their customers save by running the washing machine at 3 AM instead of 7 PM, charging the car overnight, and heating the water tank when the wind pushes prices toward zero. Across Europe, Tibber has over a million users doing the same thing. Octopus Energy grew to 7.7 million UK households — the first time the top position in the UK market changed since it opened in the 1990s. Germany has over 100 GW of installed solar and 1.8 million home batteries. Virtual power plants — where thousands of home batteries and solar panels are networked together and operated as a single power station, trading collectively on the market — are aggregating gigawatts of distributed capacity. Corporate PPAs, in which large companies sign direct purchase agreements with wind or solar farms to lock in a price for 10-15 years, reached 5.2 GW across Europe in 2024. Battery optimisation startups are raising hundreds of millions. Smart tariffs that vary price by the quarter-hour are becoming mainstream. All of this activity falls into two categories, and the distinction matters. **Optimisation** helps people navigate the broken system better. Shift demand to cheap hours. Your battery charges when electricity is cheap (or free, during negative price hours) and feeds back to the grid when prices spike — buying low, selling high, automatically. Smart tariffs give you granular price signals so you can respond. All of this makes the dysfunction more livable. But it doesn't change the pricing logic. The system adapts. Germany now requires new home batteries to be remotely controllable by grid operators. Feed-in tariffs are suspended during negative-price hours. Self-consumed solar is three to five times more valuable than exported solar precisely because the export price is set by the broken market. The system absorbs the rebellion. **Transformation** changes the pricing logic itself. The Mankala model. Energy cooperatives. Green Power Pools. Pay-at-cost structures in which renewable output is priced at its actual cost, separate from the gas-determined wholesale market. These don't optimise within the system. They replace it one community, one cooperative, one industrial consortium at a time. And the most important insight: the optimisation layer creates the constituency for transformation. Denmark's cooperatives didn't start as a political movement. They started as families wanting cheaper electricity. Then 100,000 families with turbines in their backyards became a political force that made Denmark's energy transition impossible to reverse. They stopped being passive consumers. They became organised participants with skin in the game. The binding constraint — consumer organisation — was broken from the bottom up. Germany's 3 million solar households aren't just people who produce their own electricity — they're voters. Octopus's 7.7 million customers aren't just price-sensitive switchers. They're a base that could be mobilised for market reform. Optimisation builds the infrastructure and the constituency. Transformation requires political action. But political action requires a constituency. They need each other. ### Every lock-in can be broken. It just takes coordination. History shows these systems can change, even those that seem permanently locked in. The global banking system used to set interest rates using a benchmark called LIBOR. Trillions of dollars in mortgages, business loans, and financial contracts all referenced it. When LIBOR turned out to be manipulated, everyone said it was impossible to replace — too many contracts, too much risk. They replaced it anyway. It took a decade. They ran the old and new systems side by side, converted contracts in phases, and passed laws to handle the ones that couldn't be converted voluntarily. The total value of contracts that moved to the new benchmark: somewhere between $200 and $370 trillion. The electricity market's financial lock-in is real, but it's not unprecedented. The EU couldn't get telecom companies to stop charging insane fees when you crossed a border. The industry lobbied for years, arguing that roaming charges were essential for network investment. Consumers had no leverage individually; you can't negotiate with your phone company when you land in Spain. So the EU simply banned roaming charges outright in 2017. The telecom industry survived. The fees disappeared. Sometimes, the workaround for a lobbying deadlock is just: stop negotiating and legislate. Germany went from operating 17 nuclear reactors to voting 513-to-79 to shut them all down — in less than four months. Fukushima hit Japan in March 2011. By June, the German parliament had passed the phase-out. When the forcing function is strong enough, and the political window opens, systems that seemed immovable can change fast. The pattern is always the same: a forcing function that emergency measures can't absorb, someone who finds a path around the veto, a phased transition so the old and new run in parallel, side payments to bring blockers on board, and legal infrastructure to handle legacy contracts. The electricity market may already have its forcing function and it's internal. Cannibalisation is not an external shock. It's the logical consequence of the system's own success at deploying renewables. As negative price hours double every two years and solar capture rates collapse to levels that make investment uneconomic, the mathematical unsustainability of the current design will become impossible to absorb with emergency measures. The question is not whether the system will change. It's whether the transition will be deliberate and coordinated or chaotic, forced by investment collapse. ### So what can you actually do? If you're a consumer, the structural move is to organise. If you're in a country with energy community legislation — and the EU's 2024 reform now requires all member states to enable energy sharing — consider joining or starting a cooperative. Austria has over 4,000. You don't need a rooftop to participate. A group of apartment buildings, a neighbourhood, a village — the legal frameworks are being built right now. This is how you stop paying gas prices for wind electricity. Not by waiting for the EU to reform. By building the alternative yourself, the way Finnish industry did with Mankala and Danish families did with wind turbines. Every cooperative that forms converts passive bill-payers into organised participants. That's the binding constraint being loosened one community at a time. While you're building the longer game, the immediate practical step is getting on a dynamic tariff. In Finland, Norway, and increasingly across Europe, spot-price contracts let you buy electricity at the actual market price quarter-hour by quarter-hour. It sounds scary. The savings are real — ACER data show 30-40% for active users — but they aren't really about the tariff. They're about what happens when you start seeing your costs in real time. You stop running the dryer at 6 PM. You charge the car overnight. You notice, viscerally, that the same electricity costs nothing at 2 PM and €300/MWh four hours later. The tariff doesn't save you money. Awareness does. And that awareness creates the demand for structural alternatives, because once you see the dysfunction hour by hour, you stop accepting it. If you're a climate tech founder, there's an obvious optimisation play (batteries, demand response, VPP aggregation), and those matter. But the structural opportunity is in cooperative infrastructure. The biggest barrier to energy communities isn't technology or regulation; it's that forming a cooperative is hard. You need legal setup, financing, generation assets or PPAs, metering, billing, grid connection, and member management. Right now, starting an energy community means navigating all of this from scratch. The founder who builds cooperative-as-a-service — the Shopify of energy communities — solves the binding constraint directly. And critically: financing mechanisms that let people without upfront capital participate. The Mankala model works for the Finnish industry because industrial shareholders can invest. Citizen cooperatives need something different — community solar financing, fractional ownership, pay-from-savings models. The cooperative infrastructure gap is where climate tech meets financial inclusion, and it's wide open. Every product you build either reinforces the current market design or creates pressure to change it. Most do both. But the question of which side dominates now has a clear answer: does your product help consumers organise, or does it help them cope individually? If you're an investor: the companies that look like pure optimisation plays today — batteries, demand flexibility, VPP aggregation — are also the ones accumulating the distributed infrastructure and customer relationships that any future market design will need. You're not just betting on their current revenue model. You're betting on their position in whatever comes next. But the highest-leverage bet might be on the cooperative infrastructure layer: platforms that make it easy for communities to collectively own generation, share electricity, and bypass marginal pricing. That's not a niche play. If CE Delft's projections hold, 45% of European renewable production could flow through these structures by 2050. The plumbing for that future doesn't exist yet. And if you're anyone paying for electricity in Europe: you're not being scammed. You're experiencing what a coordination failure feels like from the inside. The system isn't broken. The problem changed. The decision framework didn't. We know better now. That means we can do better. --- **References** *Market design and marginal pricing* European Commission, "Electricity market design," energy.ec.europa.eu — overview of EUPHEMIA, marginal pricing mechanism, 27 coupled markets, 61 bidding zones, and the May 2024 reform preserving marginal pricing. European Parliament, "The design of the European electricity market: Current proposals and ways ahead," IPOL Study 740094, 2023 — comprehensive analysis of marginal pricing mechanics, liberalisation history (1996, 2003, 2009, 2019 directives), and reform proposals. ACER, "Final Assessment of the EU Wholesale Electricity Market Design," April 2022 — concluded the crisis was "in essence a gas price shock" and the market design "is not to blame," estimating cross-border coupling benefits at €34 billion/year. *Lobbying and reform politics* WindEurope, "Response to the European Commission's public consultation on the EU Electricity Market Design reform," February 2023 — "Keep short-term wholesale markets based on marginal pricing and the merit order." SolarPower Europe, "Statement: EU Electricity Market Design," January 2023 — Naomi Chevillard, Head of Regulatory Affairs: "Changing the foundations of electricity markets, such as marginal pricing, is creating regulatory instability and actually halting investments." Teresa Ribera, opinion piece for EurActiv, January 2023 — "The current crisis has shown how vulnerable the current design of the electricity market is to stress situations, as well as the tremendous consequences for domestic consumers." Ursula von der Leyen, State of the Union Address, European Parliament, September 2022 — promised "deep and comprehensive reform of the electricity market." Bruegel, "National fiscal policy responses to the energy crisis," updated tracker — €646 billion in EU member state energy support measures. *Iberian Exception* La Moncloa (Spanish Government), "Government of Spain caps gas prices to lower electricity bills," May 2022 — mechanism details and 40% wholesale price reduction. Expired December 31, 2023. *Cannibalisation and negative prices* Fraunhofer ISE / Energy-Charts.info — Germany: 457 negative price hours in 2024 (+52% year-on-year), 100+ GW installed solar capacity. ENTSO-E Transparency Platform — Finland: ~700 negative price hours in 2024 (~8% of annual hours); Spain: first-ever negative prices April 2024. Agora Energiewende, "Die Energiewende in Deutschland: Stand der Dinge 2024" — German solar capture rates ~60% of baseload wholesale price. SolarPower Europe, "EU Market Outlook for Solar Power 2025-2029" — EU solar installations declined in 2025 for first time since 2016. *Finland: Mankala model and Olkiluoto* Teollisuuden Voima (TVO) annual reports — Mankala principle: shareholders receive electricity at cost proportional to ownership. TVO produces ~28% of Finnish electricity. Finnish Energy, "Energy Year 2024" — Finland 95% fossil-free electricity; Mankala model covers roughly half of Finnish electricity production. Nord Pool market data, May 2024 — Finnish spot price spike to €397/MWh during Olkiluoto 3 maintenance outage and low wind. *Energy cooperatives* REScoop.eu (European federation of energy cooperatives) — 2,500+ member cooperatives across Europe, 2+ million citizen members across 25 countries. Austrian Coordination Office for Energy Communities (Österreichische Koordinationsstelle für Energiegemeinschaften) — 4,000+ registered energy communities in Austria as of 2025. Ecopower cvba, Belgium — 70,000+ citizen members, 100 GWh/year renewable generation. Danish Energy Agency historical data — by 2001, cooperatives owned 86% of Danish wind capacity; ~100,000 families held turbine shares. CE Delft, "The potential of energy citizens in the European Union," 2016 — estimated 45% of EU renewable production could come from citizen energy communities by 2050. *Alternative system proposals* Michael Grubb, Paul Drummond, Serguey Maximov, "Separating electricity from gas prices through Green Power Pools: Design options and evolution," Institute for New Economic Thinking Working Paper, November 2022 — detailed Green Power Pool proposal. Jacques Percebois and Stanislas Pommeret, "Reform of the European electricity market: Should we prefer a price based on a weighted average of marginal costs with cross-subsidies?", The Electricity Journal, 2024. CEPR (Centre for Economic Policy Research), "Electricity markets in transition: A proposal for reforming European electricity markets," VoxEU, 2023. Dieter Helm, "The Cost of Energy Review," commissioned by UK Government, October 2017 — Equivalent Firm Power auctions, critique of marginal pricing in zero-marginal-cost systems. Dieter Helm, "Ofgem's supply model is broken," April 2023 — "In a system with lots of zero marginal cost generation, the price of electricity for all the electricity generation should not equal the spot price of gas." *Optimisation and market activity* Oomi Energia — 800,000+ customers, Flex contracts with consumption impact pricing. Tibber — 1+ million European customers on dynamic spot-price tariffs. Octopus Energy — 7.7 million UK household customers; first change of #1 market position since UK energy market opened in 1990s. Bundesnetzagentur (German Federal Network Agency) — 1.8 million home battery installations in Germany. RE-Source Platform, "European PPA Market Outlook 2024" — 5.2 GW corporate PPA volume across Europe in 2024. *Historical precedents* Financial Stability Board, "Reforming Major Interest Rate Benchmarks," progress reports 2014-2023 — LIBOR-to-SOFR transition covering $200-370 trillion in financial contracts. European Parliament, recorded vote on German nuclear phase-out, June 30, 2011 — passed 513 to 79. European Commission, "Roaming charges in the EU," Regulation (EU) 2017/920 — abolished retail roaming surcharges from June 15, 2017. *EU reform legislation* Regulation (EU) 2024/1747 and Directive (EU) 2024/1711, "Improving the Union's electricity market design," entered into force July 16, 2024. European Commission, Recommendation and guidance documents on electricity market reform implementation, July 2, 2025. *Finland's electricity prices* Yle News, "Finland had Europe's third-cheapest electricity prices in 2025," December 29, 2025. Helen Energia, electricity product pricing, helen.fi/en/electricity. Oomi Energia, supplementary services terms. --- ### Finland has the highest unemployment in Europe. This is a choice. URL: https://karolinasarna.com/blog/finland-unemployment/ Date: 2026-02-05 Tags: structural-analysis, climate-policy Finland's unemployment hit 10.3% — highest in the EU. This isn't bad luck. It's what happens when you simultaneously deplete healthcare, education, R&D, and employability stocks while restricting immigration. Finland's unemployment hit 10.3% in December 2025. The highest in the EU. Youth unemployment: 20.5%. Debt-to-GDP climbed from 77% to 88.5% despite €9 billion in austerity cuts meant to reduce it. This isn't bad luck. It's what happens when you ignore how systems actually work. ### Stocks, flows, and the delays that kill you Think of an economy as a system of stocks and flows. Stocks are reservoirs: healthcare workers, trained researchers, skilled workforce, education infrastructure. Flows are what changes them: training programs, immigration, investment, cuts. You can change a flow instantly — cut a budget, abolish a program, restrict immigration. But the stock only changes as the flow accumulates over time. A nurse training cut today doesn't empty hospitals tomorrow. It depletes the stock gradually. By 2030, you have a healthcare crisis. Here's the critical part: feedback loops can only change future behaviour, not past behaviour. When unemployment hits 10.6%, that's a signal the system isn't working. But that signal is delayed by years from the policies that caused it. By the time you see the problem, the stocks are already depleted. This is why austerity during downturns always fails. You optimise the flow (spending) while ignoring what's happening to the stocks. The feedback that would tell you to stop is delayed by 5-10 years. Politicians get credit for "fiscal responsibility" and are gone before the system collapses. ### Finland is depleting four critical stocks simultaneously **R&D capacity**: Universities face cuts through 2028, Academy of Finland lost research funding, 500 researchers leave annually while only a few hundred return. Flow: negative. Stock: depleting. **Education infrastructure**: PISA scores collapsed from #1 globally to #20 in mathematics; one in four students is now at the lowest performance levels, up from 7%. Flow: negative. Stock: depleting. **Employability system**: Adult education allowance, used by 32,518 workers annually, was abolished; benefits dropped 20-25%; work requirement doubled. Flow: negative. Stock: depleting. **Healthcare capacity**: Wellbeing counties incurred €1.25 billion in losses; 18,500+ patients wait beyond legal limits; psychiatric beds fell from 20,000 to 2,700. Flow: negative. Stock: depleting. Then Finland restricted immigration. The one external flow that could replenish the stocks while they rebuild. Work permit applications fell 25%. This is unprecedented. No other European country has simultaneously turned all critical flows negative while blocking the external compensation mechanism. When multiple stocks deplete at once, you don't get a linear decline; you get a cascading system failure. Healthcare workers can't get trained because education is underfunded. Researchers leave because there's no R&D investment. Companies can't hire because immigration is restricted and unemployment services are gutted. Each depleted stock accelerates the depletion of others. ### Denmark maintained the stocks Denmark and Finland have similar tax revenue as share of GDP. Denmark's unemployment: 5.1%. Finland's unemployment: 10.6%. Denmark understood the system. "Flexicurity" isn't just generous benefits. It's maintaining all three stocks that make labour markets work: flexibility (flow out of jobs), security (stock of income support), and activation (flow back into jobs). Denmark invests heavily in the activation stock: retraining programs that put 7 out of 10 displaced workers back in employment within a year. They have strict immigration in some areas but use a Positive List for skilled workers and lowered salary thresholds for healthcare, engineering, and IT. Maintaining the external flow when internal stocks need replenishing. Finland is doing the opposite: cutting the income stock, eliminating the activation stock, and restricting the immigration flow. That's not a system. That's just making workers precarious with no mechanism for recovery. ### European precedents show the feedback delays **Greece** cut hospital budgets 50%, fired 25,000 health staff. Initial response: budget savings. Delayed feedback (5-7 years): GDP fell 26%, debt-to-GDP increased, 680,000 people emigrated including 12,700 physicians. The stock depletion produced costs that exceeded the flow savings. **Spain** cut R&D 39%, lost 12,000 scientists. Initial: budget savings. Delayed: youth unemployment 55-60%, R&D intensity now lags Greece, Portugal, Czech Republic, Hungary (really?!?). The research stock can't be rebuilt quickly; you need 10-15 years to train researchers. **UK austerity**: initial savings £38.75 billion. Delayed cost: 190,000-335,000 excess deaths, value of life years lost £89.6 billion. The healthcare stock depletion created costs 2.3x larger than the savings. **Portugal** reversed the flows in 2015. GDP grew 2.7%, deficit fell to 2%, unemployment dropped from 17% to 6.3%. Restoring flows let stocks rebuild, which increased tax revenue and reduced deficits. The IMF admitted fiscal multipliers were 0.9-1.7; every €1 in cuts reduced GDP by €0.90-€1.70. That means austerity depletes the economic stock faster than it reduces the deficit flow. ### A note from someone who can leave I moved to Finland deliberately. I like it here. But it's not my country. I have options. Many immigrant friends and colleagues are already leaving. And honestly, that makes the decision to stay even less likely. Finland has extraordinary potential: a highly educated population, strong institutions, and advanced infrastructure. And it's systematically destroying them through policies that have failed everywhere they've been tried. For immigrants, the calculation is simple: we came because Finland offered something worth staying for. When it stops, we leave. We're the canary in the coal mine. Our departure is an early signal that stocks are depleting faster than they can be rebuilt. This is maddening because you have all the data. Denmark proves you can maintain the stocks with similar resources. European precedents show exactly what delayed feedback looks like. You have time to reverse the flows before permanent damage. But only if people stop performing concern and start acting like they understand systems dynamics. ### The levers you can pull In systems thinking, leverage points are places where a small shift can produce big changes. Some levers are weak — they require enormous effort for minimal impact. Others are strong — small interventions that change system behaviour. The weakest lever: trying to change politicians' minds. Meetings, consultations, statements. Politicians respond to incentives. If there's no cost to continuing these policies, they'll continue. Stronger levers: changing the costs. When companies relocate operations and name the policy reasons publicly, that creates cost. When skilled workers emigrate and make it visible rather than quiet, that creates cost. When voters actually change their votes based on these outcomes in 2027, that creates cost. Denmark maintained flexicurity because Danish businesses and labour unions made dismantling it more expensive than maintaining it. The system worked because enough people with power understood what stock depletion actually meant and refused to enable it. Finland is running a systems experiment no other country attempted: simultaneous depletion of healthcare, education, R&D, and employability stocks while restricting immigration. The delayed feedback won't show until 2030-2035, long after the Orpo government's 2027 term ends. By then, healthcare waiting lists will take a decade to clear. Research capacity that takes 20 years to rebuild. Brain drain that's permanent. Each depleted stock accelerates the collapse of others. You can watch this happen while performing concern. Or you can pull whatever levers you actually control: business decisions, career moves, voting, making consequences visible instead of quiet, and change what's politically possible. The question isn't whether you have levers. It's whether you understand what stock depletion with delayed feedback actually means and whether you're willing to pull them. Because right now, through inaction, you're choosing this outcome. --- ### The Complexity Paradox URL: https://karolinasarna.com/blog/complexity-paradox/ Date: 2026-02-03 Tags: earth-observation, scaling, structural-analysis Science rewards complexity. Commercial markets reward clarity. The gap between what we're trained to build and what buyers actually need explains why brilliant Earth observation companies keep defaulting to defence. You can get a PhD in atmospheric physics. You can raise $100 million. You can build satellites that image the Earth at 16cm resolution, through clouds, at night. And nobody — at any point in that journey — required you to explain what you do in words a random person on the street could understand. ### How we got here Science and engineering education reward complexity. You pass qualifying exams by demonstrating mastery of difficult material. You publish papers that impress peer reviewers with technical sophistication. You defend dissertations in front of committees who share your specialized vocabulary. At no point does anyone grade you on: 'Can someone outside your field understand what you just said?' Clarity isn't just unrewarded. It's subtly discouraged. If you can explain it simply, maybe it wasn't that hard. Maybe you're not that smart. So we learn to signal intelligence through complexity. We attach our egos to being the one who built the hard thing. We speak to impress, not to be understood. This isn't unique to engineering. Finance has its own vocabulary. Law does. Medicine does. Every specialisation builds walls of language. We speak to our peers, not to the people who need to understand us. Then we try to work with someone who doesn't share our training. And wonder why they don't follow. ### The resolution race The Earth observation industry is a perfect case study. Albedo is building 10cm resolution satellites. Maxar delivers 30cm. ICEYE achieves 16cm. So does Umbra. Every pitch deck features the same metrics: higher resolution, faster revisit, and more coverage. These are genuine engineering achievements. They also represent engineers competing for other engineers' respect. Defence and intelligence agencies pay for resolution. They have trained analysts, an established doctrine, and decades of experience interpreting satellite imagery. They wrote the requirements. The product-market fit is real. Commercial buyers — insurers, agricultural companies, infrastructure managers — didn't write those requirements. They don't have GIS teams. They don't know what to do with a 16cm SAR image. They don't want data. They want an answer. But we keep building more impressive sensors. Because resolution is measurable. Because it wins contracts from buyers who do value it. Because it's what we were trained to optimise. And because explaining things simply feels like giving something up. ### The anxiety underneath Measurable goals feel like control. Resolution: measurable. Revisit rate: measurable. Constellation size: measurable. Revenue from government contracts: measurable. Did we help someone make a better decision? Messy. Subjective. Terrifying. Can a non-specialist explain our product to their boss? Requires admitting your product might be confusing. Does our customer success team understand the customer's actual workflow? Requires talking to customers about their problems, not your capabilities. So we optimise what we can measure. We build what we can put on a slide. We compete on specs that impress at conferences. The anxiety of 'am I actually useful?' gets converted into the comfort of 'look what I built.' ### What this produces Planet went public, projecting commercial revenue would grow from 54% to 68% by FY26. It's now 23%. Stock recovered on government contracts — NATO, Swedish Armed Forces — not agriculture or insurance. ICEYE raised at €2.4 billion valuation in late 2025. Not from commercial climate applications — from a €1.7 billion German Bundeswehr contract and deals with Poland, Finland, Netherlands. Their VP told Defence News: "We're becoming more and more of a defence-intelligence company." Descartes Labs, Orbital Insight — both raised on the thesis that AI plus satellite data would unlock commercial markets. Both collapsed or were rescued. Ultimately, revenue came from government and intelligence, not from the commercial transformation they pitched. The pattern: commercial promised, defence delivered. Not because defence is easier. Because defence buyers speak the same language. They value the complexity. They have the training to use it. Commercial buyers needed something else. And nobody taught us how to build that. ### What commercial actually needs It's not a better resolution. It's reliability. Does the product work when the customer actually needs it? During the flood event, during the supply chain disruption, during the crisis — not just in the demo. It's responsiveness. When they email with a question, does anyone answer? This week? Tomorrow? It's plain language. Can the customer explain your output to their boss, who's never heard of SAR and doesn't care about your specs? It's customer success that actually succeeds. Does your team understand the customer's workflow — or just your product's features? No customer, no money. Basic. But we skip it to get back to the engineering problems we enjoy solving. The companies showing commercial traction share a common trait: the satellite becomes invisible. The customer gets methane emissions data, not spectral analysis. Wildfire alerts, not thermal imagery. The translation is done before the customer sees anything. ### The transformation we can't deliver The problem is even deeper. We can't drive the transformation that's actually needed — in climate, in infrastructure, in how we allocate capital — because we can't communicate what we're seeing to the people who need to act on it. The satellites show us coastlines retreating, glaciers melting, growing zones shifting. Scientists understand the implications. But the translation never happens. The insight stays locked in jargon, in technical papers, in conference presentations that speak to the already-converted. Transformation requires a coalition. Coalition requires communication. If you can't explain what's happening in language that decision-makers understand, you can't build the coalition to act on it. So we default to optimisation. We sell what fits existing frameworks — helping insurers process claims faster, helping farmers optimise this season's yield, helping the system absorb shocks and continue. Not because transformation is impossible. Because we never learned to make the case for it. And the capital structures we use — venture timelines, commercial revenue requirements, standard return expectations — select for optimisation by design. Transformation might need something different entirely. ### What would have to change What if explaining your research in plain language was a graduation requirement? Not a 'broader impacts' statement buried in a grant application. An actual test: can you make a random person on the street understand what you do and why it matters? If not, you don't pass. What if investor pitches had to clear a clarity bar? Not 'is this technically impressive?' but 'could a procurement officer at an insurance company explain this to their team?' What if we stopped treating simplicity as dumbing down? The hardest thing in science isn't building something complex. It's understanding it so deeply that you can explain it simply. That requires more mastery, not less. But we don't reward it. So we don't develop it. So we build extraordinary things that sit unused because nobody outside our field can figure out what they're for. ### The founders aren't the problem This isn't about bad founders or failed companies. It's about a system that trains brilliant people to optimise for the wrong things. That rewards complexity over clarity. That never requires translation as a core skill. The founders building 16cm satellites aren't wrong. They're doing exactly what they were trained to do, building for buyers who value exactly what they were trained to build. The gap is that commercial markets — and the transformation we actually need — require something different. The founders aren't wrong. They're products of a system that never taught them to translate. Seeing that gap is the first step. Closing it is something else entirely. --- ### The Satellites See Everything. We're Asking Them the Wrong Questions. URL: https://karolinasarna.com/blog/satellites-wrong-questions/ Date: 2026-01-29 Tags: earth-observation, organizational-design, structural-analysis We have more eyes on Earth than ever. Thousands of satellites capture terabytes daily. But we keep fitting transformative data into frameworks built for a planet that no longer exists. We have more eyes on Earth than ever in human history. Thousands of satellites orbit the planet, capturing terabytes of data daily. They see through clouds. They image at night. They measure ground movement in millimetres. They track every forest, every coastline, every ice sheet, every city. They show us aquifers depleting. Coastlines retreating. Growing zones shifting. Ice sheets collapsing. Land subsiding beneath cities built on the assumption it would stay still. For the first time in history, we can observe physical reality changing in ways that will reshape where and how humanity can live. And what do we do with this knowledge? We ask: How do we make this data bankable? How do we fit it into existing investment criteria? How do we help insurers price risk better, help banks satisfy disclosure requirements, help supply chains optimise for disruption? We try to make satellites profitable within a financial system designed for a planet that no longer exists. That's not a market failure. It's a window into something deeper: our collective inability to act on what we know. ### What Satellites Actually Show Us When I was building solutions products at ICEYE, the question was always: Where in the process can we leverage Earth observation (EO) data? That's the wrong question. Every conversation was about fitting new information into existing frameworks. Finding a slot. Making it digestible for systems that already existed. The assumption underneath: the frameworks are fine. They just need better data inputs. But the data was telling us something different. The data was telling us the frameworks are built for a planet that no longer exists. A 30-year mortgage assumes the land will be there in 30 years. Satellites show coastal erosion trajectories that say otherwise. Infrastructure investment assumes stable baselines. Predictable flood patterns, consistent growing seasons, reliable water supply. Satellites show those baselines shifting faster than planning cycles. Insurance pricing assumes you can model risk against historical patterns with some adjustment for trend. Satellites show discontinuities, not trends, but systems flipping from one state to another. Supply chain optimisation assumes you can source from the same regions with logistical adjustments. Satellites show entire agricultural zones becoming unviable within a generation. The insight isn't *here's data to make your existing decisions slightly better*. The insight is *your decision architecture is built for a world that's already gone*. That's not an optimisation input. That's a call to transformation. ### The Language of Evasion The Earth observation industry has developed a vocabulary that obscures this gap. Climate adaptation. Climate resilience. Climate risk. Companies use these terms interchangeably. Even the Intergovernmental Panel on Climate Change (IPCC) acknowledges inconsistent usage across its own reports.[1] The vagueness is functional. When a company claims to serve climate resilience, that could mean flood imagery for faster insurance claims. Or carbon credit verification. Or regulatory disclosure compliance. Or disaster response optimisation. All of these help existing systems work slightly better. None of them questions whether those systems should continue. Resilience sounds transformative. It implies bouncing back, adapting, thriving through change. But the products sold under that label mostly help the current system absorb shocks and continue. They're resilience for the financial framework, not for communities or ecosystems. Real resilience would mean some places acknowledging they need to be abandoned. Some industries acknowledging they're incompatible with a livable planet. Some assets acknowledging they're stranded regardless of what the balance sheet says. That's not what anyone's selling. Because there's no market for it. ### The Optimisation Trap The pressure to optimise rather than transform is immense. And rational. Defence and intelligence represent 24-40% of Earth observation revenue.[2] Government overall accounts for roughly three-quarters of the market.[3] The commercial climate applications everyone talks about in pitch decks? Roughly 7-11%.[4] Defence has clear budget owners, predictable procurement, and sophisticated buyers who know what they want. Climate transformation has none of that. No budget owner. No procurement category. No process for absorbing insight that says 'your decision architecture is obsolete.' So founders build what the system can purchase: optimisation tools. First Street embeds climate risk scores in Zillow, Redfin, and Realtor.com listings.[5] Homebuyers see a flood risk rating and buy a slightly less exposed house. The system of 30-year mortgages on climate-vulnerable land continues unchanged. Jupiter Intelligence serves three of the five largest US banks with climate analytics for regulatory disclosure.[6] The banks report their exposure with better precision. They keep lending to climate-exposed assets, just with improved footnotes. GHGSat finds methane leaks so oil and gas operators can fix them.[7] The leaks get plugged, the waste gets reduced, the extraction continues. These are commercial successes. The founders aren't villains. They're responding rationally to what the market will buy. The market will buy optimisation. It won't buy transformation. ### Why Transformation Has No Buyer The companies that tried to sell transformation failed. Gro Intelligence raised over $125 million to show food and agriculture companies that entire sourcing regions would become unviable.[8] They made TIME's 100 Most Influential Companies list.[9] By late May 2024, they'd shut down. Valuation collapsed from $850 million to under $25 million.[10] Cervest raised $36 million to provide asset-level climate intelligence.[11] The UK government featured them as one of ten companies "at the forefront of" AI innovation.[12] Three months later, administration.[13] Both had real technology. Both had genuine insight. Both had customers who found the information valuable. What they didn't have: buyers who could act on transformation. Nobody pays to be told they're wrong. A bank doesn't want software that says 'your mortgage book is built on a planet that doesn't exist anymore.' They want software that helps them keep lending with better risk footnotes. A farmer doesn't want intelligence showing their land becomes unviable in 15 years. They want precision agriculture that improves this season's yield. A real estate investor doesn't want tools for managed divestment. They want risk scores that help them buy slightly less exposed properties while continuing to build. And the beneficiaries of transformation, the communities that need to relocate, the workers who need to retrain, the ecosystems that need to recover, they can't pay. That's partly why they're vulnerable in the first place. There's no budget line for 'our entire decision architecture is wrong.' There's no request for proposal for 'help us become a fundamentally different company.' There's no procurement category for managed decline. Capitalism is structured around growth. Every budget assumes continuation and expansion. Ending things isn't a line item. ### We Have Everything Except Willingness Here's what we actually have: We have the knowledge. Satellites track 745 million emission sources through Climate TRACE.[14] SERVIR provided climate data across 45+ countries before being defunded in 2025.[15] CLIMADA models risk for any location on Earth.[16] We know exactly what's happening, where, and how fast. We have planning tools. Buy-In Community Planning helps design managed retreat programs.[17] The Georgetown Climate Centre provides retreat planning guides. The Carbon Risk Real Estate Monitor (CRREM) shows exactly when buildings will become stranded assets.[18] The playbooks exist. We have the technology to stop producing harm. Renewables are now cheaper than fossil fuels in most places. Heat pumps work. Electric vehicles work. Regenerative agriculture exists. We know how to build buildings that don't leak energy. The technology to transition isn't missing. We even have the capital. Climate finance flows exceeded $2 trillion in 2024.[19] Trillions more sit in fossil fuel subsidies that could be redirected. The money exists. So why isn't transformation happening? Look closer at those 'transformation tools.' They're information tools. They tell you what's happening. They don't make change happen. Climate TRACE shows where emissions are. It doesn't stop them. CRREM tells investors when buildings will strand. It doesn't make them divest early. Buy-In helps plan buyouts. It doesn't fund them, and it doesn't make people want to leave their homes. The gap isn't information. It's action. We're not transforming because we don't want to. Transformation means loss. Losing homes. Losing jobs. Losing investments. Losing identities. Losing ways of life. Homeowners don't want to leave. Investors don't want to take losses early. Workers don't want to retrain. Fossil fuel companies don't want to close. Politicians don't want to mandate unpopular change. Everyone prefers to believe it won't be that bad. So we optimise. We squeeze a few more years out of the current system. We document the decline with ever-more-precise dashboards. We disclose the risks and continue taking them. We have everything we need except the willingness to accept that some things must end. ### Removal vs. Transition Even when funding tries to do the right thing, it bends toward continuation. Consider the Frontier coalition — Stripe, Alphabet, Meta, Shopify, McKinsey committing over $1 billion to purchase carbon removal through 2030.[20] The mechanism is genuinely innovative: advance purchase commitments for technology that does not yet exist but must. It worked. Charm Industrial has delivered over 6,200 tonnes of permanent carbon removal, more than any other company at the time, because these commitments made it viable.[21] But look at who's buying: companies running massive data centres, global logistics, cloud infrastructure. They're pre-purchasing removal so they can claim progress toward 'net zero' while continuing to grow. It's sophisticated offsetting. Pay for removal here; keep emitting there. The carbon math might work on paper. But it's not a transformation. It's buying a license to continue. Here's the sharper distinction: removal lets the current system continue. Transition ends the harmful parts. We already have technology to not emit in the first place. Renewables are cheaper than fossil fuels in most places. Electrification works. We know how to build without burning. But transition means changing things. Retiring assets. Closing facilities. Accepting that some businesses need to end. Removal is easier. Don't change what you're doing; add cleanup. That's why we fund it. Transition is harder. Stop doing this; do that instead. That's why we avoid it. The money flows to removal because removal doesn't threaten the current system. Transition does. ### The Allocation We've Chosen Consider how we actually spend: Fossil fuel subsidies: approximately $7 trillion per year.[22] Global climate adaptation needs: approximately $215-387 billion per year.[23] We could fund the entire global adaptation requirement many times over with what we currently spend subsidising the problem we need to solve. This isn't a trade-off with security or growth or any other competing priority. This is paying to make things worse. $7 trillion annually to keep fossil fuels artificially cheap, to keep extraction profitable, to keep the current system running past its expiration date. EO companies pivot to government and defence contracts because those buyers have clear budget owners and procurement processes. That's rational. The question isn't why companies follow the money. The question is why, as a society, we actively fund our own destruction. We have the capital for transformation. We're choosing to spend it on continuation, and worse, on subsidising the very industries that make transformation necessary. The satellites see all of this. They watch the emissions rise. They track the ice sheets calve. They measure the land subside. They document what $7 trillion in annual subsidies produces: a planet that's changing faster than our institutions can respond. ### The Obligation No one is coming to save us. Not markets. Markets optimise for existing preferences, and existing preferences are for continuation. Not governments. Not fast enough, not with the current political economy. Not technology. The technology exists. What's missing isn't innovation. What's missing is us. The paths forward exist. They're narrow. They require different capital structures, different timelines, different definitions of success. They require being honest about what's actually optimisation dressed as transformation, and what's actually transformation, which rarely looks like a startup. It's been done. risQ focused narrowly on climate risk for municipal bonds, positioned for regulatory tailwinds, and was acquired by Intercontinental Exchange (ICE) — not a unicorn outcome, but a path that worked.[24] The transformation tools that exist and work are mostly public goods: nonprofit infrastructure, government programs, open-source tools, coalition-funded initiatives. The paths are real. They're just not the paths most people are looking for. They require willingness to accept loss. To fund endings, not just cleanups. To let go of what needs to go. The satellites are showing us reality. What we do with that knowledge, as founders, as investors, as citizens, as humans sharing the same thin atmosphere, is the only question that matters now. --- **References** [1] IPCC, "Guidance Note: The concept of risk in the IPCC Sixth Assessment Report," 2021. [2] Mordor Intelligence, "Satellite-based Earth Observation Market Report," 2024; Novaspace/Euroconsult, "Earth Observation Data & Services Market Report," 17th Edition, November 2024. [3] Novaspace/ESA, "New Trends and Dynamics of the EO Market," presentation at ESA EO Commercialisation Days, November 2024. [4] Novaspace/ESA market segmentation, 2023. [5] Zillow Group, investor press release, September 26, 2024. [6] Jupiter Intelligence, company website and investor materials, 2024-2025. [7] GHGSat, company website; EPA Alternative Test Method documentation. [8] Semafor, "Gro Intelligence startup shuts down," June 4, 2024. [9] TIME Magazine, "TIME100 Most Influential Companies," April 27, 2021. [10] Semafor, June 4, 2024; Forge Global valuation data. [11] Dealroom.co, Cervest company profile; TechCrunch, May 2021. [12] UK Department for Science, Innovation and Technology, March 29, 2023. [13] Evening Standard; Yahoo News, June 2023. [14] Climate TRACE, coalition website and database, 2024. [15] NASA Applied Sciences, "About SERVIR"; NASA Watch, "USAID Erasure Impact: NASA Halts SERVIR Solicitations," March 2025. [16] ETH Zürich, CLIMADA platform documentation. [17] Buy-In Community Planning, company website; MIT Solve application. [18] GRESB/CRREM, "Carbon Risk Real Estate Monitor," 2024. [19] Climate Policy Initiative, "Global Landscape of Climate Finance 2025." [20] Frontier Climate, coalition website. [21] Charm Industrial, company reports and carbon ledger; CNBC, May 18, 2023. [22] International Monetary Fund, "IMF Fossil Fuel Subsidies Data: 2023 Update," August 2023. [23] UNEP, "Adaptation Gap Report 2023" and "Adaptation Gap Report 2024." [24] Bond Buyer, "ICE expands muni reach," December 9, 2021; Northeastern University, "risQ Acquired by Intercontinental Exchange." ---