Source: Photo by author (August 4, 2026)

Every economy built on a depleting resource faces the same problem: at some point the resource - oil and gas, timber, fisheries - runs out. The problem is conversion, how to turn the wealth of finite natural resources into something that lasts: permanent capital, durable institutions, infrastructure that keeps producing value after the thing that paid for it is gone. Most places fail in a specific and repetitive way: the boom generates enormous cash, the cash funds current consumption, the institutions that grew up around it calcify, and when the resource declines there is a landscape full of expensive equipment and little else to build from. The literature calls it the resource curse, the tendency of resource-rich countries to stay poor and fail to create strong governing institutions.

I spent the first days of August in Alaska with twenty-two travelers, most connected to Duke, on a land tour through Denali, Wrangell-St. Elias, and Kenai Fjords. As faculty host I gave two talks, one on the benefits and risks of a resource economy, the other on Alaska’s potential to host large AI data centers.

Alaska is the clearest live case of the conversion problem in the United States. The state has no personal income tax and no statewide sales tax, and for much of its history oil funded the majority of its discretionary budget. Prudhoe Bay, discovered in 1968, was the largest field ever found in North America, and the Trans-Alaska Pipeline is one of the genuine engineering achievements of the last century, eight hundred miles across three mountain ranges, on refrigerated supports so its own warmth would not melt the permafrost holding it up. Throughput peaked in 1988 at just over two million barrels a day, and in 2025 it averaged 462,821, under a quarter of peak.^[1] New North Slope projects will cushion that curve, but not reverse it. And there is a physical floor beneath the financial one: as flow declines, the oil moves more slowly and cools toward the temperature of the ground around it, producing wax, water dropout, and ice. The asset degrades not only in value but in operability. So the question of what comes next is not hypothetical.

Alaska has a vivid counterexample of its own. Inside Wrangell-St. Elias National Park, at the end of a winding gravel road, stands the Kennecott copper mill, fourteen stories of red-painted timber above the Kennicott Glacier. The ore was extraordinary, some of it better than eighty percent copper, and it drew Guggenheim and J.P. Morgan financing for a 196-mile railway through glaciated valleys. The operation ran from 1911 to 1938. When the ore was gone, the trains stopped, the town emptied, and the wealth fled. The buildings stayed, and so did the cleanup liability, arriving on a public balance sheet decades after the operator dissolved. Standing inside that mill, I was struck by the ambition that drew investment there, and by how quickly it ended.

Source: Photo by author (August 4, 2026)

The trip’s conversations kept circling back to Alaska’s answer. In 1969 the Prudhoe Bay lease sale delivered about $900 million to a young state overnight, and within a few years most of it had been spent. Alaskans noticed. In 1976 voters amended their constitution to require that at least a quarter of certain mineral royalties go into a fund whose principal could not be spent without a vote of the people. The first deposit, in 1977, was $734,000. The dividend followed in 1980. Today the Permanent Fund holds roughly $86 billion, and its earnings, not oil, are the single largest source of the state’s unrestricted general fund revenue.^[2] Governor Jay Hammond, its principal champion, described the purpose plainly: save a share of a one-time windfall for the generations who would never see the oil. That is a boom-time society voting to constrain its own future appetite, an act of institutional design rather than good fortune.

It is not the only version of that design, and not the most successful one. Norway found North Sea oil a decade after Prudhoe Bay, watched Alaska build its fund, and out-executed it. The Government Pension Fund Global now holds about $2 trillion, the largest investment fund in the world, owning about 1.5 percent of all publicly listed shares on earth.^[3] Norway’s first deposit came nineteen years after Alaska’s, and the fund is now roughly twenty-five times larger, about $400,000 per Norwegian against $110,000 per Alaskan.

Two design choices explain the gap. First, Norway kept taxing itself. Petroleum revenue supplemented the tax base rather than replacing it, which preserved both fiscal discipline and democratic accountability. Alaska took the opposite path, eliminating broad-based taxes and letting oil pay the bills, which is precisely the rentier structure the resource-curse literature warns about. Second, Norway adopted a spending rule: it draws only the fund’s expected real return, around three percent, never the principal, and that draw covers roughly a fifth of the national budget indefinitely. Alaska’s dividend has been shrinking as fund earnings are diverted to state government. The 2025 payment was $1,000, the lowest ever in real terms, and a proposal to write a full statutory dividend into the constitution failed on the House floor this year.^[4]

This is the more useful framing, and the one that travels. The curse is not a geological given but a design question, answered differently by places holding nearly identical hands. The two states asked their funds to do fundamentally different jobs. Norway’s fund is an endowment that stabilizes a budget. Alaska’s has increasingly become a budget that happens to be endowed. Same instrument, different intent, and the difference compounds over fifty years.

But a sovereign wealth fund hedges only a price risk. It converts a volatile revenue stream into a diversified financial claim, so that when the price of oil collapses the state still eats. What it cannot hedge is a change in the underlying system the rest of the economy depends on. That is what is happening to Alaska’s other asset systems as the oil declines.

Alaska’s fisheries are, by most measures, among the best-managed wild fisheries on earth. The state constitution mandates management on the sustained-yield principle, a rare constitutional requirement, and seasons are opened and closed against in-season spawning targets. Sixty years of adaptive management, roughly $6 billion in annual output, and about sixty percent of all U.S. wild-caught seafood.^[5]

In 2018, NOAA’s Bering Sea survey estimated roughly 11.7 billion mature snow crab, among the largest populations ever recorded. By 2021 about ten billion of them were gone. In 2022, for the first time in the fishery’s history, the season did not open. They were not overfished. A 2023 paper in Science attributed the collapse to the marine heatwave of 2018-19: warmer water raised the animals’ metabolic demand as food availability fell, and they starved.^[6] The fishery reopened in 2024 at a small fraction of historic harvests.^[7] The crab came partway back, but its future is deeply uncertain.

This distinction is the one I would most want a board to take away. Kennecott is a depletion story: a stock ran out, and depletion is something you can see coming, model, and plan against. Essentially every business risk framework is built for it. The snow crab is an ecological foundation story: nothing was depleted, the management was sound, and the asset disappeared because the system underneath it changed state. Almost no risk framework is built for that.

The same logic runs through the rest of the portfolio. Roughly eighty percent of Alaska sits on permafrost, and as it thaws, roads buckle and foundations move; the infrastructure was engineered for a stable, cold world that is neither. On the Kenai Peninsula, the road up to Exit Glacier is marked with signs for the year the ice reached each point, about a mile and a half of retreat since 1815.^[8] For the visitor industry this is not scenery, it is the product, and unlike a barrel of oil it cannot be sourced from somewhere else.

The mechanism is the one Partha Dasgupta named in his 2021 review for the UK Treasury: we have managed a portfolio of assets for a century while treating one entire class of them as costless and infinite.^[9] Business did not decide to disregard nature so much as it never had a place to record it. Mark Carney called the temporal version the tragedy of the horizon: the costs fall beyond the horizon of the people who could prevent them, so markets under-react.^[10] The correction is being built, but slowly. With colleagues at Duke and the Stockholm Resilience Centre, I recently examined disclosures from seventy-five of the world’s largest ocean companies. Three-quarters reported metrics and targets for energy use; fewer than one in four did for ocean-specific impacts, and about seven percent for habitat alteration.^[11]

What makes Alaska interesting rather than merely cautionary is that the conversion problem is also being worked on in registers other than money. I did not get to either of these places, so I am reporting rather than observing.

Igiugig is a Yup’ik village of about seventy people where Lake Iliamna drains into the Kvichak River. Its electricity came from diesel at ninety-one cents a kilowatt-hour, six times the U.S. average. In 2019 the village became the first tribal government in the country licensed to operate a water-power project with no dam, a turbine on the riverbed that boats pass over. The design drew on local knowledge of the salmon’s life cycle, and after years of monitoring regulators concluded it does not harm the fish.^[12] The detail that matters is jurisdictional rather than technical: the Kvichak drains into Bristol Bay, the same watershed where Alaska spent two decades resisting the world’s largest undeveloped gold-copper deposit. Same water, two proposals, opposite answers. That is not a technology difference.

Kodiak is the other. A fishing island of thirteen thousand, it ran the way most of remote Alaska still runs, on hydro plus roughly three million gallons of barged diesel a year. In 2007 the member-owned electric cooperative, not the state and not a technology company, set a goal of ninety-five percent renewable and beat it, with six wind turbines, a third hydro turbine at Terror Lake, and flywheels and batteries for frequency. Since 2014 the island has run above ninety-nine percent renewable, year-round, in Arctic weather, at rates below what islanders paid in 2001.^[13] The clean grid did not constrain the fishing economy; it enabled its expansion.

Both are conversions of exactly the kind the opening problem calls for. Both are also small. Set against oil’s share of this economy, every diversification experiment is small. Mariculture is the clearest illustration: aquatic farm permits have more than tripled in a decade and roughly $49 million in federal funding sits behind the sector, but 2024 sales were about $1.5 million against a $100 million target.^[14] The experiments are running. So is the clock. Anyone who tells you Alaska is transforming is describing an intention rather than a balance sheet.

I closed the talk with a case that is seven hundred years old, because the economics only carry the argument so far.

Around the year 1300, two societies shared the island of Greenland. The Norse, Erik the Red’s descendants and three centuries established, had farms, cattle, churches, and a cathedral at Garðar with a bronze bell. Moving down the coast were the Thule, ancestors of today’s Inuit, carrying a maritime toolkit refined over millennia: the kayak, the toggling harpoon, the technique for taking seals at breathing holes. Then the climate cooled. Both societies took the same test. The Inuit passed, and Greenland today is their island. The Norse were gone by about 1450.

The familiar explanation, which Jared Diamond popularized, is cultural rigidity: a society too attached to its cattle and churches to adopt what its neighbors knew. The archaeological records complicates that. Isotope analysis shows their diet shifted substantially toward the sea as the climate cooled, from roughly a fifth marine to half or more.^[15] They id, in fact, adapt their diet. What they could not adapt was their economy, which ran on a single export, walrus ivory sold into Europe, which paid for their iron, timber, and grain. In the fourteenth century elephant ivory began reaching European markets, cheaper and in greater volume, and the price collapsed. Ancient DNA suggests they were by then hunting farther north for smaller animals, so costs rose as revenue fell.^[16] The ships stopped coming.

I do not think we have to choose between these accounts; the more time I spend with them, the more I think both are true. A mono-export economy exposed to a distant market it did not control, and a society whose sense of itself made some adaptations available and others unthinkable. Those are the two failure modes, and they are not independent. What a place believes it is for shapes which conversions it is willing to make.

Which is why the conversion problem is finally a question about intent rather than capability. Alaska has the instruments: a constitutional sustained-yield mandate, an $86 billion fund, Native corporations holding roughly forty-four million acres under the 1971 settlement act, and a century of institutional memory about what extraction leaves behind. Norway’s example shows that the outcome turns on the discipline of the design, and on clarity about what the saving is for.

Every resource economy arrives at this fork eventually. Appalachia arrived without a buffer. The Gulf states are spending sovereign fortunes to buy their way past it before demand peaks. Alaska got there early, with a better institution than nearly anyone, and it is now being asked the conversion question a second time, by an industry that did not exist five years ago and that wants something Alaska has in abundance.

Leave a comment

Share

Subscribe now

^[1]Alyeska Pipeline Service Company, historic throughput data. Peak annual average of 2,032,928 barrels per day in 1988; 2025 annual average of 462,821 barrels per day.

^[2]Alaska Permanent Fund Corporation, 2026 Mid Fiscal Year Review; fund value of $86.3 billion at the fiscal-year midpoint.

^[3]Norges Bank Investment Management, Government Pension Fund Global.

^[4]Alaska Beacon and Anchorage Daily News reporting on the 2026 legislative session and the 2025 Permanent Fund Dividend.

^[5]Alaska Seafood Marketing Institute / McKinley Research Group, 2024; Alaska Constitution, Article VIII.

^[6]Szuwalski, C. S., et al. (2023). The collapse of eastern Bering Sea snow crab. Science, 382(6668), 306-310.

^[7]Alaska Department of Fish and Game advisory announcements, 2024 and October 2025.

^[8]National Park Service. 200 Years of Terminus Retreat at Exit Glacier, 1815-2015. Kenai Fjords National Park.

^[9]Dasgupta, P. (2021). The Economics of Biodiversity: The Dasgupta Review. HM Treasury.

^[10]Carney, M. (2015). Breaking the tragedy of the horizon - climate change and financial stability. Speech at Lloyd’s of London.

^[11]Jouffray, J.-B., Virdin, J., Vermeer, D., et al. (2025). Identifying and closing gaps in corporate reporting of ocean impacts. Nature Sustainability, 8.

^[12]ORPC RivGen Power System; U.S. Department of Energy and Pacific Northwest National Laboratory (Tethys) documentation.

^[13]Kodiak Electric Association; RMI and Alaska Center for Energy and Power documentation of the Kodiak system.

^[14]The Nature Conservancy (2025). Situation Assessment of Alaska’s Emergent Mariculture Industry; Alaska Beacon reporting, March 2026.

^[15]Arneborg, J., et al., isotopic analyses of Norse Greenlandic human remains.

^[16]Star, B., Barrett, J. H., et al. (2018). Ancient DNA reveals the chronology of walrus ivory trade from Norse Greenland. Proceedings of the Royal Society B, 285(1884).

We pulled off the road in a small village on the route from Agra to Jaipur on a January afternoon. Our driver secured a local guide, who walked with us up the bustling main street into one of the most stunning structures I have ever seen.

Source: Personal photograph (December 31, 2025)

Chand Baori, built in the ninth century in the village of Abhaneri, is awe-inspiring and disorienting. It is not clear on first inspection what the purpose of the structure actually is. Standing at its edge in the midafternoon light, you look down into an inverted pyramid of stone: 3,500 narrow steps cascading thirteen stories into the earth on three sides, their geometry so precise and intricate that the effect is less architectural than mathematical. It reminded me of an M.C. Escher print brought to life.

We could not go all the way down inside the baori. The lower levels are blocked to visitors now, but you could see from the upper tiers the royal residence pavilion built into the fourth wall at the bottom: a shaded, colonnaded space where the rulers of this part of Rajasthan would retreat during the scorching summer months, informed by their lived experience that the temperature at the base of the well runs five or six degrees cooler than the surface. The well was a refrigerator, a reservoir, a social commons, and a seat of power. Next door, the ruins of the Harshat Mata (“goddess of joy”) temple completed the picture: water and worship, engineering and devotion, inseparable in the logic of the people who built this place.

The British declared structures like this unhygienic in the nineteenth century and replaced them with pipes and pumps. Within a generation, the knowledge required to maintain them had dispersed.

Chand Baori survived because it was too large and too beautiful to ignore, but thousands of smaller baoris across Rajasthan were not so fortunate. What those structures inscribed - encoded in stone, in temple ritual, in the knowledge of who could draw water and when and how much - was a theory of what water is. Not a commodity to be priced and consumed, but a resource to be shared and cared for across generations.

Twelve centuries later, on the other side of the world, an industry is making the opposite choice at enormous scale - and calling it infrastructure. This essay is about the distance between those two positions, and what it would cost to close it.

In the early 2000s, I had stood at the edge of one of those forgotten baoris in circumstances rather different from the Agra-Jaipur road trip.

At that time, I was leading the Global Water Initiative for The Coca-Cola Company, an institution whose relationship with water in India had, by 2003, become a defining legal and reputational crisis. These tensions erupted most visibly at the Plachimada bottling plant in Kerala where the company was accused of depleting and contaminating the wells of the surrounding Adivasi indigenous community. Since my wife’s family is based in this part of India, I had a particularly personal stake in the growing crisis.

Jeff Seabright was there when that reckoning arrived. Now a senior advisor to global sustainability initiatives and formerly Chief Sustainability Officer at Unilever, Seabright led Coca-Cola’s environmental and water affairs during the years when Plachimada transformed from a local dispute into a global reputational crisis. I was part of his team, developing the water risk assessment and the community water partnerships program we built partly in response to what was happening in Kerala and across India.

I recently spoke with Seabright about the water governance questions at the center of this essay, and he recalled a visit to the Plachimada plant with a striking clarity. Coca-Cola’s technical lead at the time walked Jeff through the hydrogeological data - the company’s boreholes were deeper than the community’s perched wells, and the aquifer drawdown was, by the numbers, within permitted limits. “Technically, he was right,” Seabright told me. “But broadly speaking, he was dead wrong. What do you think the community sees when they watch the red trucks leave every morning, filled with Coca-Cola products, and their wells are dry and their kids are hungry? It was like two different worlds.”

The gap between what a permit allows and what a community experiences is not specific to Coca-Cola - it is a structural problem, and it has no easy fix. But in the years after Plachimada, we tried to work at the scale of communities, aquifers, and watersheds. The Water Partnerships program, an initiative I helped develop, provided challenge grants to encourage Coca-Cola bottlers to work alongside local water users on shared challenges.

Through this program, the Coca-Cola India office partnered with a local nonprofit to help restore a stepwell that had been unused for over a century. I visited this site and watched as workers removed the debris and an industrial pump drained the water from an unexceptional water hole.

I was unprepared for what emerged. As the waterline fell, it was like a theater curtain slowly dropping in reverse, receding downward to reveal the stage beneath. Within a couple of hours, a temple-like geometry was revealed below: sandstone archways, pillars, the scalloped design of a structure built, like Chand Baori, to be cooler than the world above it. People from the village began to gather at the edge and a palpable sense of excitement grew. They were watching a structure they knew about from their grandparents but had not actually seen it for themselves. When the shrine at the very bottom finally appeared - a carved figure draped in decades of silt - I tried to understand the reactions of the villagers. Was this unearthed infrastructure or a sacred site? It was both, carrying a weight of history that knit together the practical and the spiritual.

In the United States, the recent surge of news coverage and growing concern about data center water use should not be surprising. A single large data center consumes roughly as much water annually as a town of fifty thousand people. U.S. data center water use has tripled since 2014¹ and could reach the equivalent of 18.5 million households by 2028, rising with every new model trained and every query answered.²

A recent Bloomberg investigation found that about two-thirds of new data centers built or in development since 2022 are located in places already facing high levels of water stress, with five states accounting for nearly three-quarters of those facilities.³ Honor The Earth’s Data Center Tracker has documented at least 37 proposed facilities on or near indigenous lands in the United States, with communities from Oklahoma to northern Alberta reporting that permits were signed and water agreements made without meaningful consultation.⁴ In July 2025, the White House issued an executive order streamlining data center development and exempting new facilities from certain Clean Water Act reviews, removing one of the few mechanisms capable of slowing the industry’s accelerating water consumption. This trend is also gaining global momentum, with the United Nations marking its 2025 International Day of Indigenous Peoples under the theme “Indigenous Peoples and AI: Defending Rights, Shaping Futures”.⁵

Understanding why data centers concentrate in the driest places on earth requires understanding the economics of siting decisions. Developers look first for cheap land, reliable power, and favorable tax treatment, and arid regions tend to offer all three. States and counties actively recruit data centers with abatements and incentives, eager for tax revenue and proximity to the AI economy. The dry air of desert climates offers a genuine technical advantage: low humidity reduces corrosion risk for sensitive electronic equipment, making arid environments preferable from an engineering standpoint.

There is also a structural trade-off between renewable energy and water - the regions with the most solar and wind generation tend to have the least available water, and the industry has consistently chosen energy access over water availability. Water, meanwhile, is treated as an afterthought. Sharlene Leurig of Fluid Advisors put it directly: water “has traditionally been far less consequential for industrial users than energy costs and availability” because it is priced accordingly.⁶

The consequences of that pricing can be startling when seen in its broader context. When Google negotiated its water contract in Mesa, Arizona, one of the driest cities in one of the driest states, during a historic drought, it secured a rate of $6.08 per thousand gallons while Mesa residents paid $10.80.⁷ The largest company in the world by market capitalization, in a desert, paying roughly half the residential rate for the community’s most scarce resource. That is less a market signal and more a market failure etched into an over-stretched watershed.

Gleick locates this failure within a longer historical pattern that extends well beyond the data center industry. “Modes of economic development have always undervalued the environmental consequences of what they do, because the environment is a free good,” he told me. “It’s the classic economic failure of externalities, and if we’re going to spend three quarters of a trillion dollars on some new form of infrastructure, but once again we’re not going to pay attention to the social and environmental consequences of that development, we’re repeating the mistakes of the past. We’ve learned that lesson over and over again, but we forget it when there’s a profit to be made at the expense of the environment.” The Mesa pricing arrangement is not an anomaly, but an inevitable outcome of a system in which water has no political constituency, no market signal strong enough to register against the scale of capital being deployed, and no institutional mechanism capable of saying no.

The architects of Chand Baori reached the opposite conclusion about arid places. They designed a working system that solved the cooling problem without consuming the aquifer - passive thermal management through stone and geometry, a community water reserve, a gathering place, and a spiritual site. The cooling problem was solved through ingenuity and integration, not consumption.

The baori did not fall into disuse because it failed; it was functioning when the British arrived. Colonial authorities declared it unhygienic, installed pipe-and-pump systems that fit their administrative model, and within a generation the specific knowledge required to maintain the stepwells had dispersed. The structures remained. The living relationship between community and water system had been severed.⁸ Their inability to employ what Pfaffenberger called “sociotechnical thinking” - understanding governance and infrastructure as inseparable - led them to repeat at great cost what traditional systems had already prevented. The indictment was not merely historical. It applies to every generation of engineers who arrive at a water problem convinced that the people who managed it for centuries have nothing to teach them.

The siting logic that draws data centers to arid regions extends, by the same economic arithmetic, toward indigenous and rural communities - places with cheap land, limited regulatory capacity, and governments that are often outmatched in negotiations with companies deploying billions of dollars and teams of lawyers. The Brookings Institution found that negotiations between developers and local governments are “often asymmetrical”, that rural and tribal communities lack the expertise and leverage to understand what they are agreeing to before permits are signed.⁹ A 2025 analysis by Kleinman Energy at the University of Pennsylvania concluded that the speed of industrial expansion routinely “exceeds the speed of local democratic process.”¹⁰

In northern Alberta, the Sturgeon Lake Cree Nation has challenged a multibillion-dollar data center campus proposed on territory covered by Treaty 8, where Chief Sheldon Sunshine stated publicly that no consultation was conducted - that the company behaved as if “our land and water is there for the taking.”¹¹ In Oklahoma, the Muscogee Nation voted against rezoning for a proposed data center after citizens raised alarms about water infrastructure strain and absent environmental safeguards. The pattern has deep roots. As Rusty Pickens of the Chickasaw Nation put it: “We were moved off of our tribal homelands east of the Mississippi for land purposes, then they discovered oil. Those extractive cycles of history seem to keep repeating themselves.”¹²

“We know when something is violating our rights,” Muskogee advocate Jordan Harmon told Public Radio Tulsa last year. “We just have to study up so we can speak about the data center impact specifically. But we already know the deeper thing.”¹³

The deeper thing he was referring to is a set of governance principles, tested across centuries, that the data center industry never considered.

J. Stephen Lansing’s landmark research on Bali’s subak system — a network of water temples governing rice irrigation across entire watersheds, documented in a foundational 1987 paper in the American Anthropologist and elaborated across four decades of fieldwork, demonstrated that what Western development planners dismissed as religious ritual was in fact sophisticated ecological management.^{14, 15} The temples synchronized planting cycles across watersheds to suppress pests, distributed water equitably without centralized control, and sustained agricultural productivity for over a thousand years. When Indonesian government policies in the 1970s effectively removed the temples from their functional role, pest populations exploded and rice yields collapsed. Restored to authority, the system recovered. Pfaffenberger, writing about Lansing’s findings, identified precisely why the system had been invisible to outside planners: it produced a form of social coordination that was neither straightforwardly economic nor political — a “sociotechnical system” whose logic was recognizable only to those willing to look at governance and infrastructure as a unified whole.¹⁶

What those systems knew, and what Chand Baori encoded in stone, can be stated as a connected set of principles, each of which names a specific failure in current data center governance.

The watershed, not the individual user, is the appropriate unit of management. The Balinese subak managed water from the volcanic crater lake to the sea, balancing the needs of every farmer along the gradient across generations. Data center permitting is a single-user transaction conducted against a commons. This is the wrong unit of analysis, and it produces the wrong set of decisions.

Variability is the baseline condition, not the exception. Baoris were designed to capture monsoon surplus for dry-season use. Subak calendars adapted planting schedules dynamically to actual water availability, synchronizing the whole watershed in response to real conditions. Data centers consume at constant rates regardless of drought conditions or aquifer decline, a static claim on a dynamic system.

Highest and best use must account for all users across time. The subak asked who benefits and across how many generations. A data center generates enormous revenue for its operator and nothing for the downstream farmer, the tribal community, or the desert ecosystem drawing from the same groundwater table.

Shared resources require shared visibility. You could see the water level in the baori. Every subak farmer could monitor the system’s state and hold neighbors accountable. Elinor Ostrom, whose Nobel Prize-winning work on commons governance drew directly on examples like the subak, showed that successful management of shared resources requires exactly this mutual monitoring.¹⁷ Most data centers are not required by law to disclose their water use. Microsoft acknowledged in 2023 that 42 percent of its water came from water-stressed areas, a figure released only under investor pressure.¹⁸

The cost of that invisibility can be startlingly concrete. In Fayette County, Georgia, residents of an affluent subdivision called Annelise Park began noticing unusually low water pressure last year. When the county utility investigated, it discovered that a 615-acre data center campus - one of the largest in the United States, operated by Quality Technology Services and owned by Blackstone - had been drawing water through two connections the county did not know existed. One had been installed without the utility’s knowledge; the other was simply not linked to any billing account. By the time the issue surfaced, the campus had consumed roughly 29 million gallons of unmetered water, the equivalent of 44 Olympic swimming pools, during a period when the county was asking residents to stop watering their lawns due to drought. The county issued a retroactive bill of $147,474 but did not fine the company, explaining that the data center was “our largest customer, and we have to be partners.” The governance failure here arose because the county lacked the institutional capacity to monitor commercial-scale water extraction because, as the utility director explained, meters had always been designed for residential use. Nobody could see the water level dropping because the system was not built to show it.¹⁹

The Fayette County case is not an outlier. Peter Gleick, who has spent four decades building the scientific and legal case for water as a fundamental human right, sees the same failure repeating across the history of industrial development. “AI is another example of an industrial revolution making the same mistakes as previous industrial revolutions,” he told me in a recent conversation. “That’s an institutional failure. Our institutions are not holding these industries to account, and if our institutions are too weak to insist that environmental externalities be internalized, that social costs be addressed, and that communities be involved in decision making, then those things aren’t going to happen.” The data center industry is not uniquely to blame; it is operating rationally within the constraints and incentives that institutions have allowed.

The right to draw creates a reciprocal obligation to sustain. Subak membership required maintaining temple infrastructure, contributing communal labor, and participating in governance. The right to water was inseparable from the obligation to care for the watershed. Data centers hold permits. Permits are not obligations. The difference accumulates in the aquifer.

Gleick, who pioneered the concept of the “soft path” for water management,²⁰ has written that our water solutions must “relearn the lessons that indigenous people learned about local waters.” ²¹ The baori and the subak are lessons in governance engineering and foreshadowed Gleick’s soft path a thousand years before the concept was named.

Seabright, who has spent decades advising corporations on the gap between their stated values and their operational footprint, sees the water problem as a symptom of something deeper. “It’s the hubris of our belief in progress - techno-optimism, and a kind of willful blindness to all of the aspects of this that don’t comport with that hubris,” he told me. “Achieving ever-greater levels of AI capability is treated as inherently the most important thing we could ever do. That tunnel vision applies to a lot of things going on in society right now.” The corrective, in his view, is not to slow AI development but to reframe its purpose. “It’s AI in service of what? In service of humans - in service of progress defined not as material accumulation, but as human thriving.”

What would it mean to take these principles seriously as design constraints for AI infrastructure? Community-based governance boards with real authority over water use decisions, not just advisory but bodies with standing to modify or revoke operating permits. Adaptive use agreements that reduce consumption automatically when local aquifer levels decline. Mandatory recharge obligations tied to the specific watershed being drawn from, not offset through distant projects. Impact assessments calibrated to long-term watershed health, not five-year permit cycles. Water priced to reflect actual scarcity, so that a hyperscale data center does not pay half the residential rate in a drought. Legal standing for the water bodies themselves, as New Zealand granted the Whanganui River in 2017.

None of this is without precedent. Uruguay amended its constitution in 2004 to establish water access as a right of citizens rather than a commodity of corporations. The Stepwell Atlas documents more than three thousand remaining baoris across India. Lansing’s subak research has generated a body of computational modeling showing that the temple network’s cooperative logic can be identified in satellite imagery and replicated in governance design elsewhere.¹⁵ Ostrom’s commons frameworks have been field-tested across six continents.¹⁷ The knowledge exists, and so do the precedents.

I asked Gleick whether the ancient governance systems at the center of this essay - the baoris, the subak, the water temples - were credible analogs for modern infrastructure governance, or whether I was being romantic. He didn’t dismiss the romanticism, but he held the core argument firm. “The idea that earlier cultures had to learn how to live sustainably with their water resources is absolutely the lesson that we need to learn now,” he said. “We are not living sustainably with our water resources, and it’s not just data centers, it’s the entire system that is in the long run unsustainable and has to change.” His soft path framework, developed over decades of research, does not reject hard infrastructure - Gleick is careful to say the hard path brought real benefits, but insists that its unintended consequences can no longer be treated as acceptable externalities.

The baori and the subak were soft path solutions before the concept had a name. They were also, for the communities that built them, something the permit system has never been: an obligation that ran in both directions. Until the AI industry treats permits as stewardship rather than permission, the people downstream will keep paying the difference.

Standing at Chand Baori this past January, watching the afternoon shadows move across three thousand five hundred steps, I thought about how much that structure knew about heat, about scarcity, about the obligations that come with drawing from a shared source, and how little of that knowledge has traveled forward into the age of artificial intelligence.

The industry building AI is also building things it believes will endure. Its data centers are expensive, permanent structures designed to operate for decades, sited deliberately in the driest places on earth, adjacent to the communities least equipped to resist them. The companies behind them think carefully about power procurement, real estate, equipment longevity, and the rate they can negotiate for water. They have not thought carefully about what traditional systems knew: that infrastructure without governance is extraction, and extraction without obligation will eventually fail.

This is not a failure of intelligence but a consequence of inattention. And attention, it turns out, is what ancient water governance has always required - not computational scale, or capital, or negotiating leverage to secure a favorable rate. Just the willingness to pay attention to the water itself - where it comes from, who depends on it, what the watershed can bear - and to build from that knowledge.

The question for the industry is whether it is willing to look down the well.

Leave a comment

Share

Subscribe now

Notes

¹ Ceres. (2025, September). Drained by data: The cumulative impact of data centers on regional water stress. Ceres.

² Shehabi, A., Smith, S. J., Masanet, E., & Koomey, J. (2024). United States data center energy usage report. Lawrence Berkeley National Laboratory.

³ Bloomberg News. (2025, May 8). The AI boom is draining water from the areas that need it most. Bloomberg.

⁴ Honor The Earth. (2025). Data center tracker: Proposed or expanding data centers on/near Indigenous lands. https://storymaps.arcgis.com/stories/724e9dc3b9f64cb1a9c219b9d02eaad4

⁵ United Nations. (2025). International Day of the World’s Indigenous Peoples 2025: Indigenous peoples and AI — defending rights, shaping futures. UN Permanent Forum on Indigenous Issues.

⁶ Leurig, S. (2025). Quoted in Bloomberg News. The AI boom is draining water from the areas that need it most.

⁷ University of Tulsa. (2026). Data centers draining resources in water-stressed communities. University of Tulsa News.

⁸ Pfaffenberger, B. (1992). Technological dramas. Science, Technology, & Human Values, 17(3), 282-312.

⁹ Brookings Institution. (2025). The local implications of data centers for rural communities in the US. Brookings Institution.

¹⁰ Kleinman Energy Center, University of Pennsylvania. (2025). Communities are at risk if we don’t slow the roll on data center development. Kleinman Energy Center.

¹¹ Intercontinental Cry. (2026, February 25). Indigenous peoples push back as data centers expand across North America. IC Magazine.

¹² ICT News. (2025, August 24). For Indigenous communities, AI brings peril — and promise. ICT News.

¹³ Public Radio Tulsa. (2025, October 7). Indigenous environmental advocates say data centers risk water, culture and informed consent. Public Radio Tulsa.

¹⁴ Lansing, J. S. (1987). Balinese “water temples” and the management of irrigation. American Anthropologist, 89(2), 326–341. https://doi.org/10.1525/aa.1987.89.2.02a00030

¹⁵ Lansing, J. S., & Kremer, J. N. (1993). Emergent properties of Balinese water temple networks: Coadaptation on a rugged fitness landscape. American Anthropologist, 95(1), 97–114.

¹⁶ Pfaffenberger, B. (1992). Social anthropology of technology. Annual Review of Anthropology, 21, 491–516. https://doi.org/10.1146/annurev.an.21.100192.002423

¹⁷ Ostrom, E. (1990). Governing the commons: The evolution of institutions for collective action. Cambridge University Press.

¹⁸ Microsoft. (2023). 2023 environmental sustainability report. Microsoft Corporation.

¹⁹ Skibell, A. (2026, May 7). A data center drained 30M gallons of water unnoticed — until residents complained about low water pressure. Politico / E&E News. https://www.eenews.net/articles/georgia-residents-seethe-over-30m-gallons-of-missing-water/

²⁰ Gleick, P. H. (2002). Soft water paths. Nature, 418, 373.

²¹ Gleick, P. H. (2010). Water, water everywhere. In Written in water: Messages of hope for Earth’s most precious resource (pp. 7–12). National Geographic Society.

Source: Getty Images

The AI infrastructure build-out has been announced in numbers so large they function as a kind of rhetorical atmosphere, creating pressures that shape decisions without being carefully examined. Amazon, Microsoft, Alphabet, and Meta announced combined capital spending commitments of more than $300 billion for 2025 alone.¹ The announcement of Stargate, a joint venture between OpenAI, SoftBank, and Oracle, pledged up to $500 billion in AI infrastructure investment.² These figures were reported as facts about the future, but they were actually bets made by companies with strong incentives to make large public commitments, regardless of whether the underlying demand would materialize.

Recent evidence suggests the bets are running ahead of the reality. Between 30 and 50 percent of planned US data center builds for 2026 are projected to be delayed or canceled, driven by a combination of power shortages, equipment backlogs, and supply chain constraints.³ Microsoft CEO Satya Nadella said publicly that “there will be an overbuild” of AI infrastructure, a notable admission from the company that had simultaneously announced $80 billion in data center spending for 2025.⁴ Satellite imagery of major announced projects, including OpenAI-linked campuses in Texas, shows construction running months or years behind announced schedules.⁵

The mechanism behind this gap is worth understanding precisely. Data center developers routinely announce many projects in parallel and test which can clear local regulations and contract power fastest, because there is a great deal of uncertainty around securing permits and power. This has made it difficult for utilities and grid operators to determine how much load will actually come online, flooding them with speculative or ‘phantom’ load requests that may not materialize.^6. One grid analyst I spoke with recently coined the term “bragowatts” for announced capacity that never materializes as actual load.

In other words, the demand signal driving hundreds of billions of dollars in grid investment is being generated by companies whose competitive incentives produce systematic overstatement. And the institutions receiving that signal - the electric utilities - operate under regulatory structures that give them little reason to question it.

Here is the governance problem that makes this more than a forecasting error. Utilities in most regulated markets earn a fixed rate of return on capital investment. They are financially motivated to build infrastructure whether or not the demand that justified it materializes. When Duke Energy Carolinas announced that customer energy needs over the next fifteen years would grow at eight times the growth rate of the prior fifteen years, it was responding to developer requests, not to measured consumption.⁷ When Southern Company extended the operating lives of coal plants it had committed to retiring, it cited data center demand projections it had no independent way to verify.⁸

When those projections prove wrong - when the data centers don’t come, or come at half the projected scale, or shift to a different region after local opposition blocks the original site - the cost of the overbuilt infrastructure is not absorbed by the companies whose announcements drove the investment, but is passed to ratepayers. Community opposition alone blocked or delayed at least $156 billion in data center projects in 2025.⁹ Every project that gets announced, drives grid investment, and then gets cancelled or delayed leaves a trail of infrastructure costs that someone has to pay, but it is not the hyperscalers.

The risk runs in the other direction too. If the demand is real and the delays are structural - transformer lead times have stretched from two years to five, while data center deployment cycles run under eighteen months - then the infrastructure won’t be ready when the demand arrives, and the reliability consequences fall on everyone connected to the grid.¹⁰ This is a genuine failure mode, and the engineers and regulators working on grid modernization are right to take it seriously. But it is a different problem from the one driving most of the current policy conversation, which is focused on how to give the industry what it is demanding rather than on whether the demand estimates are credible.

The clean energy layer of this story compounds the uncertainty rather than resolving it. Every major hyperscaler has announced ambitious renewable energy goals, and at least some of these commitments are serious. Google’s 24/7 carbon-free energy initiative, which attempts to match consumption with clean generation on an hourly basis, in the same location, represents a genuine advance over standard practice.¹¹ Typically, companies match their annual electricity consumption against renewable energy certificates that may have been generated at a different time, in a different place, and on a different grid. When a data center in northern Virginia draws power from a coal-heavy grid at two in the morning and the annual accounting credits that consumption against a solar farm that generated in Nevada during the afternoon, the grid did not decarbonize. The certificate moved, but the electrons did not.

The same Southern Company that extended coal plant lifetimes to serve data center demand also has hyperscaler customers claiming to run on clean energy. These facts can both be true simultaneously under current accounting standards, but they cannot both be true simultaneously on the actual grid.

There is a structural irony at the center of this story. When DeepSeek released its model last year and claimed it required dramatically less energy per query than comparable American systems, the industry’s response was not to scale back infrastructure plans.¹² It was to announce more of them, on the logic that cheaper queries enable more applications, which enable more demand, which requires more compute. The MIT Technology Review subsequently found that DeepSeek’s chain-of-reasoning approach, when tested on equivalent tasks, consumed substantially more energy than non-reasoning model baselines due to the length of its responses.¹³ The efficiency gain created room for more consumption, not less.

This is what energy experts call the “Jevons Paradox” applied to artificial intelligence, and it has been running silently through every efficiency gain the industry has celebrated. This means that efficiency gains that reduce resource consumption per unit often lead to increased total consumption, because lower costs drive higher demand.

It also highlights that technological improvement or voluntary corporate commitment is unlikely to resolve the governance gap on its own. The question of who bears the cost when the energy bet is wrong - whether wrong by being too large or too small - is fundamentally a question about accountability structures, not about the pace of renewable deployment or the efficiency of new cooling systems.

Those accountability structures are not currently adequate to the speed or scale of what is being built. The people who make the demand forecasts, the people who announce the capital commitments, and the people who extend coal plant lifetimes in response are not the people who pay higher electricity bills when the infrastructure is overbuilt. They are not the communities watching promised jobs fail to materialize after the data center gets cancelled following the grid investment. And they are not, as the next piece in this series examines, the communities that were never really asked whether they wanted the infrastructure in the first place.

Leave a comment

Share

Subscribe now

ENDNOTES

1. Hitchcock, I., & Cahoon, M. (2025). Hyperscaler data center buildout: A sustainability bane, boon, or both? Nicholas Institute for Energy, Environment & Sustainability, Duke University. Combined figures: Amazon ($100B), Microsoft ($80B), Alphabet ($75B), Meta ($65B), as reported in Vlastelica, R., & Purnell, N. (2025, February 24). Microsoft dropped some AI data center leases, TD Cowen says. Bloomberg.

2. Stargate investment announcement, January 21, 2025. Widely reported; see, e.g., Hitchcock, I., & Cahoon, M. (2025), Hyperscaler data center buildout, p. 8 (Texas case study). The $500 billion figure represents the maximum committed investment over the project’s life.

3. Sherrill, K. (2026, April 3). Half of planned US data center builds have been delayed or canceled, growth limited by shortages of power infrastructure and parts from China. Tom’s Hardware, citing Sightline Climate (2026, February). Data Center Outlook. Sightline tracked 777 data center projects larger than 50 MW announced since January 2024 and projected 30–50% of 2026 pipeline capacity would be delayed or canceled; the text reflects the midpoint of this range.

4. Patel, D. (2025, February 19). Satya Nadella — How Microsoft is preparing for AGI [podcast interview]. Dwarkesh Podcast. dwarkesh.com/p/satya-nadella-2. Nadella states: “there will be an overbuild” and plans to lease significant capacity in 2027–28. See also: Skidmore, Z. (2025, February 24). Microsoft cancels 200MW of AI data center leases — report. Data Centre Dynamics. Microsoft CFO Amy Hood described the company as working from a ‘pretty capacity-constrained place’ on a contemporaneous earnings call.

5. Analytics group signals possible delays at 40% of AI data center construction sites. (2026, April 18). Tom’s Hardware, citing SynMax satellite analysis of OpenAI-linked sites in Shackelford County and Milam County, Texas. SynMax estimated delivery of one building by end of 2026 at the earliest, against announced 2026 delivery dates.

6. Wang, O. (2026, February 27). Up to half of the world’s data centers may be delayed this year. Latitude Media, citing Sightline Climate (2026). Data Center Outlook. The ‘phantom load’ terminology is used by grid operators and analysts including those cited by St. John, J. (2025, February 25). Utilities are flying blind on data center demand. That’s a big problem. Canary Media.

7. Sweeney, D. (2024, March 27). Rising datacenter demand forces reckoning with US utility decarbonization goals. S&P Global Market Intelligence. Duke Energy Carolinas stated in its 2025 Carolinas Resource Plan that customer energy needs over the next fifteen years are expected to grow at eight times the growth rate of the prior fifteen years — more than double the growth forecasted when the 2023 plan was initially filed. See also: Duke Energy press release, October 1, 2025, duke-energy.com. The Sweeney S&P Global article documents the early 2024 IRP revision that first flagged the magnitude of this forecast shift.

8. Tait, D. (2025, February 20). Southern Company extends life of coal plants to power data centers, appears to abandon net zero goal. Energy and Policy Institute.

9. Data Center Watch. (2025). 2025 data center opposition report, as cited in: The public sours on AI and data centers as Anthropic, OpenAI look to IPO. (2026, April 15). CNBC. The figure of $156 billion represents projects blocked or delayed by local opposition and litigation across full-year 2025, per Data Center Watch’s annual opposition report. A separate Data Center Watch report covering May 2024–March 2025 cited $64 billion; the $156 billion figure reflects the broader 2025 annual tally. Maine subsequently passed legislation creating the first state-wide data center ban; as of April 2026 it awaited the governor’s signature.

10. Sherrill, K. (2026, April 3). Half of planned US data center builds have been delayed or canceled. Tom’s Hardware, citing Sightline Climate and Bloomberg. High-power transformer lead times expanded from 24–30 months pre-2020 to as long as five years by 2026. Data center deployment cycles are typically under 18 months, per Sightline Climate (2026). Data Center Outlook.

11. Malik, N. S. (2024, February 25). AI increases data center energy use. Google-pioneered technique could help. Bloomberg. Google’s 24/7 Carbon-Free Energy initiative pairs machine learning with real-time grid data to shift workloads toward available clean power on an hourly, location-matched basis — distinct from annual renewable energy certificate matching used by most hyperscalers.

12. Marshall, C. (2025, January 29). ‘Game changer’? What ‘DeepSeek’ AI means for electricity. E&E News. DeepSeek’s leaders stated their R1 model used between 10 and 40 times less energy than comparable US tools. The announcement triggered a significant market selloff in AI infrastructure stocks.

13. O’Donnell, J. (2025, January 31). DeepSeek might not be such good news for energy after all. MIT Technology Review. MIT Technology Review found that DeepSeek’s reasoning model showed similar energy efficiency to Meta’s comparable model when tested on equivalent tasks. DeepSeek’s chain-of-reasoning approach generated longer responses overall, consuming approximately 87% more energy compared to non-reasoning model baselines on matched queries.

As part of their public communication, Anthropic said the new model was too dangerous to make available publicly. Instead, they provided access to a controlled consortium of around 40 companies - Google, Apple, Microsoft, Amazon, Cisco, JPMorganChase - and asked them to use it defensively, to find and patch vulnerabilities before bad actors develop something similar.

Image source: Unsplash.com

Here is what the model can actually do: Given a software codebase and a simple prompt, Mythos autonomously finds security vulnerabilities, writes working exploits, and chains them together into attacks - without human direction, succeeding on the first attempt more than 80 percent of the time.1 It found a 27-year-old flaw in OpenBSD that had survived decades of human review and millions of automated tests. It found vulnerabilities in the Linux kernel running on most of the world’s servers, in every major web browser, in the systems that run power grids, water treatment facilities, financial networks, and hospitals. As of this writing, fewer than one percent of those vulnerabilities have been patched.2 Critically, Anthropic was clear that none of this was planned. These capabilities, the company said, emerged “as a downstream consequence of general improvements in code, reasoning, and autonomy.”3

In other words, a private American company built something capable of compromising virtually every major software system in the world, it didn’t know what they were creating until it was completed, and the best response they have right now is a 90-day head start for the good guys.

The Morning After

Imagine this scenario: You wake up and your phone has no signal. The power is out - not just your block, but grids across the country. The tap runs dry. The grocery store has a line around the building and a handwritten sign: cash only, one bag per customer. Panic grows throughout the day. By evening, nobody knows who did it, or how, or whether it’s over.

This is not a scene from a disaster film. It is a scenario that is based on what we now know about Mythos, technically within reach now, with tools that exist today and that other actors will have access to within six to eighteen months, without any commitment to responsible disclosure.

Hollywood has spent decades conditioning us to see these scenarios as entertainment. We encounter these situations on a screen but are relieved to turn it off when the credits scroll. We distance ourselves from the actual risks, and reassure ourselves that someone must be monitoring the situation. With the most recent acceleration in capabilities of these models, this margin of safety has dissolved, even if we don’t feel it.

What “Safe” Actually Means

To assess whether AI companies are keeping us safe, we first need to be precise about what kind of safety we mean, because they are not all answering the same question.

There is safety with AI: the security, reliability, and trustworthiness of the model itself. Is it accurate? Does it resist misuse? These are important questions, but they are fundamentally operational questions about a tool’s performance.

Safety from AI is categorically different. It asks: are we safe in a world where this technology exists at scale in society? What are the systemic risks when AI is simultaneously deployed across critical infrastructure, labor markets, democratic institutions, and military systems? Are we prepared for the environmental and social externalities that are generated with the advancing AI technology? These are not questions about any particular model’s behavior, but a question about the kind of world we are building.

The distinction matters because solving one does not solve the other. A more capable, better-aligned model can intensify safety-from risks even as it reduces safety-with ones. The Mythos story is a perfect illustration: Anthropic did everything right on the technical dimensions, ran rigorous red-team testing, identified the danger, and chose not to release publicly. And the world is still more dangerous than it was a week ago.

Within both categories, the safety risks have multiple layers - technical, alignment, institutional, and societal - that most public debate collapses into a single word. We also often characterize these risks generically, but they play out very differently across different communities and geographies. Because AI risk is not evenly distributed, the communities most exposed to AI-driven disruption are consistently those least equipped to absorb it. For example, a hospital in a rural county that cannot afford a cybersecurity team is now as newly exposed as JPMorganChase - but JPMorganChase is in the corporate consortium, and the rural hospital is not. When companies talk about safety, we should be asking: safe for whom?

The Discontinuity Problem

Another critical challenge surfaced by the Mythos announcement is what I would call the discontinuity problem. AI capability does not evolve gradually, but advances in sudden, large jumps that create entirely new categories of risk overnight, with no warning and no time for governance response. Mythos did not slowly become dangerous; it crossed a threshold and totally novel risks surged.

As you can imagine, the implications for defense and national security should be stated explicitly. The logic of traditional warfare - force projection, deterrence, proportional response, attribution - depends on assumptions that no longer hold when the most powerful attacks are invisible, instantaneous, and plausibly deniable. An adversary who can instruct an AI to simultaneously compromise power grids, water systems, financial networks, and communications infrastructure has achieved destructive capacity that rivals conventional military force, without a single soldier or missile. There is no doctrine of mutually assured destruction for this. As Thomas Friedman noted in a recent column (attributed to technology expert Craig Mundie), the ability to launch sophisticated cyberattacks, once the exclusive province of nation-states and elite intelligence organizations, is now radically democratizing cyberattack capabilities globally.4

Anthropic itself has privately warned government officials that Mythos makes large-scale attacks significantly more likely in 2026, and that equivalent capabilities will emerge from other labs within months, potentially without any responsible disclosure process.

AI Companies Setting the Rules

There is something vertigo-inducing about the public posture of the companies building these systems. They are in the anomalous position of rapidly building the very technologies they warn against. To their credit, both companies have recently attempted to articulate a coherent governance response to the risks they are creating, but in very different ways.

Anthropic’s answer is the Claude Constitution, published in January 2026.5Rather than a list of rules, it attempts to give the model internalized principles, betting that a system that understands why it should behave in certain ways will generalize more reliably to novel situations than one mechanically following instructions. It establishes a four-tier hierarchy of safety, ethics, guidelines, and helpfulness, draws hard lines around absolute prohibitions, and becomes the first major AI company document to formally acknowledge the possibility of AI consciousness and moral status. It is philosophically serious, unusually transparent, and genuinely ambitious as an alignment framework.

OpenAI’s answer is its April 2026 manifesto, “Industrial Policy for the Intelligence Age,”6 which operates at an entirely different register - not a framework for model behavior but a political and economic document addressed to governments, calling for public wealth funds, robot taxes, a four-day workweek, and expanded safety nets to manage the disruption of superintelligence. It is ambitious and occasionally compelling, though critics have noted the tension between its proposals and the company’s track record: a firm that converted from nonprofit to for-profit, opposed EU AI Act oversight provisions, and lobbied against California’s SB1047, now calling for the worker protections and shared prosperity its manifesto champions.

Both documents represent real progress - in prioritization, transparency, and the willingness to name the stakes publicly. What they do not yet represent is risk reduction. They signal that the right questions are being asked, but they do not answer them, or take responsibility for them.

So What Now?

The honest answer to whether OpenAI and Anthropic can keep us safe is: probably not on their own, and possibly not at all in the deepest sense. Not because they are indifferent, but because the risks now emerging are civilizational and governance problems that no company, however well-intentioned, can solve unilaterally.

That is an uncomfortable place to land, but naming the problem honestly is a necessary precondition for doing anything about it. Each stakeholder has a role.

For policymakers: The Chatham House analysis argues that meaningful international coordination may only become politically feasible in the aftermath of a serious AI-driven crisis, the way Chernobyl catalyzed nuclear safety norms and the 2008 financial crisis produced macroprudential regulation.7 That is not a reason for passivity, but a reason to invest in preparatory work now: pre-negotiated frameworks, emergency communication channels, pre-agreed red lines and circuit-breakers, so that when a crisis arrives, institutions can move faster than the damage.

For business leaders: When Apple, Google, Microsoft, Amazon, JPMorganChase, and Cisco set aside commercial competition to join a shared emergency consortium called Project Glasswing, it signals the threat model has fundamentally shifted. On the project’s launch page, Cisco’s Chief Security & Trust Officer Anthony Grieco stated that “AI capabilities have crossed a threshold that fundamentally changes the urgency required to protect critical infrastructure from cyber threats, and there is no going back. The old ways of hardening systems are no longer sufficient.”8 It should be noted, however, that Project Glasswing focuses on the companies that build critical software, but the exposure cascades to every organization that runs on these systems. Hospitals, utilities, logistics firms, regional banks all depend on infrastructure they did not build, and most of their boards have no idea how newly exposed they are.

For citizens: The AI safety conversation has been conducted almost entirely among insiders. Most people had no idea they had skin in this game. That is the real governance failure - democratic societies cannot regulate what their citizens do not understand. Demanding transparency and insisting that AI risk be treated as a public interest question, not a corporate one, matter more than they may seem.

For the AI companies themselves: The Mythos decision - to withhold, disclose responsibly, brief government officials proactively - represents genuine leadership and should become the norm. But the most important thing all tech companies could do right now is stop lobbying against the external governance mechanisms that might actually constrain the risks they themselves say are real.

In Friedman’s New York Times column, he asked what history will remember most about April 7, 2026 - the bombs postponed over Iran, or the carefully controlled release of Mythos Preview. My view is that history may see this moment as a turning point. In the wake of Mythos’ careful release, the public finally begins to understand that the AI safety threat is not just getting closer; it is already here. And the frameworks we built to manage it were designed for a different world.

This piece was developed with AI assistance.

Leave a comment

Share

Subscribe now

Endnotes

The same week, Jensen Huang declared that AI had crossed a threshold.2 “Claude Code and OpenClaw have sparked the agent inflection point,” he said, “extending AI beyond generation and reasoning into action.” For Huang, this was a moment to celebrate.

For most audiences, these may feel like unrelated news items - one a cautionary tale, one a market signal. For sustainability professionals, they are part of a single, familiar story that we have seen before, and have spent careers navigating.

Source: Forbes

Déjà Vu All Over Again

The defining challenge of sustainability work is that the most serious harms are produced by systems, not actors. No single company caused climate change, and no single decision caused ocean acidification or aquifer depletion. The harm emerges from thousands of rational actors, each operating within the rules, but none holding ultimate liability. The practice of sustainability requires focused attention to construct systems of accountability, even when no law exists to enforce it.

That core function is precisely why sustainability professionals should be paying closer attention to agentic AI. In the transition from large language models to agents, AI technology is reproducing this familiar problem – but at greater speed and higher impact, in more domains, and with governance structures even less prepared to respond.

What Changes When AI Starts Acting

Large language models produce text. In this world, harmful outputs can be ignored, corrected, pushed back on because a human with meaningful control stays in the loop. Every prompt/response cycle is an opportunity for a person to employ judgment and redirection.

In contrast, agents act without supervision. They send emails, execute transactions, deploy code, instruct other agents - continuously, without waiting for human review between steps. They require access to high-level data and have the power to take action without supervision.3 An error doesn’t wait for review, and propagates freely. This is not a bug, but a core feature of the technology.

The Northeastern experiment shed light on this phenomenon. Once AI agents are embedded in real-world infrastructures with communication channels, delegated authority, and persistent memory, new classes of failure emerge that weren’t anticipated and couldn’t easily be reversed. Now consider the consequences of scaling that process to a global enterprise. This unsupervised experiment is already going to scale; investment in AI agent ecosystems has surged beyond $600 billion in 2026.4 The experiment is no longer running on a weekend server, but is now inside supply chains, financial systems, and logistics networks, at a pace that has substantially outrun our ability to contain it.

The Buck Stops Where?

Imagine how accountability is practically structured in a typical agentic deployment: a foundation model built by one company, fine-tuned by a second, embedded in a platform operated by a third, deployed by a corporate customer, is instructed by an employee who no longer monitors its outputs, executing actions through third-party APIs, affecting people who have no idea an agent was involved.

When harm occurs in that chain – which seems inevitable – where does accountability reside legally, ethically, practically?

Sustainability professionals should recognize this pattern. In the world of climate action, this looks a lot like the dilemma of Scope 3 emissions or corporate supply chain harms. Impacts are real, traceable to a complex system of actors and processes, and legally attributable to no one, because by the time you follow the chain to any single actor, their individual contribution looks defensible. Existing governance frameworks are simply not designed for this level of autonomy.

What is new with agentic AI is the speed. Unlike carbon emissions or aquifer depletion, the actions of AI agents can become irreversible on a timescale of minutes rather than decades. We had fifty years to develop the vocabulary for climate accountability, but we can’t afford the glacial pace of governance design we are used to.

Accounting for Agents and the Efficiency Paradox

Agentic AI is being sold as an efficiency revolution, and the productivity gains in specific applications are real. But twenty-five years of sustainability work has taught me to be deeply skeptical of efficiency claims that don’t account for the boomerang of induced demand. Jevons Paradox states that efficiency gains tend to increase total consumption because lower cost enables expanded use. For example, more efficient air conditioning systems typically don’t reduce overall energy demand because the more efficient, less expensive HVAC systems can now be deployed in places that were not previously conditioned.

I have watched this pattern repeat across many domains - in energy, in water, in transportation. Considering the environmental impacts of agentic AI - the energy required to run continuous, persistent, multi-agent workloads – efficiency gains will likely be overwhelmed by unleashing armies of agents in every possible domain. How might we re-introduce a systematic pause in this process, to allow a realistic accounting of the various benefits and costs of particular agentic use cases? How can we create space to selectively say “no” rather than accelerate at all costs? This is a space where sustainability professionals have both the expertise, experience, and standing to push for something better.

I want to name one more thing, because it is the one most likely to be missed.

Our field depends on humans who can hold complexity, reason under uncertainty, and weigh values that resist quantification. These are precisely the capacities that are hardest to specify in an agent’s instructions and easiest to quietly outsource. An agent can optimize a supply chain for cost and emissions, but it cannot tell you whether the community disruption caused by the optimized route is acceptable, or who should decide, or what you owe to the people affected. That judgment is uniquely the domain of humans, but only if we keep holding it.

The governance response to agentic AI is beginning to emerge, if only through halting steps. While these efforts are necessary, they are not sufficient. They address the technical surface of the accountability problem while leaving the structural roots largely untouched.

Sustainability professionals have a distinctive contribution to make - not because we understand the technology better than engineers do, but because we understand the governance problem better than technologists typically do. We know what it looks like to assign responsibility for harm produced by systems. We know how that effort gets captured by actors with the most to lose from accountability. We know how long it takes, and what it costs to build frameworks after the damage is done.

The question is whether we engage now while the architecture is still being built - or whether we arrive with the right tools, but a decade too late.

Leave a comment

Share

Subscribe now

This is not a story about technology failing, but about how change actually works in organizations - which is to say, how it works with people.

Sources: US Census Bureau, Macrobond, Apollo Chief Economist

What the adoption curves leave out

The dominant conversation about AI is essentially an engineering story. Its protagonists are models and infrastructure, and its metrics are benchmarks and scaling laws. The temporal frame is the research cycle, which moves in months or even weeks.

But adoption inside organizations moves according to a different logic entirely, one that is better understood through anthropology and psychology rather than engineering. Change has to survive a gauntlet that no benchmark can measure: the IT security review that takes six weeks; the middle manager who sees the new tool as a threat to the authority her institutional knowledge gives her; the team whose informal workarounds are already more effective than anything a vendor is offering; the employee in his fifties who has watched four waves of transformative technology come and go, none of them quite delivering what was promised; the champion who leaves for another role two months after launch.

These are not failures of will. They are the texture of how institutions actually process change - slowly, unevenly, with deep sensitivity to context, relationships, and trust. They are also, from another angle, a form of distributed intelligence. The employee who resists a new workflow because it would eliminate judgment she has spent a decade developing may be the only person in the room who understands what would be lost. The manager who insists on reviewing AI outputs rather than trusting them automatically may be slowing the process, or performing exactly the kind of human oversight we are told is essential to responsible deployment.

What it looks like in practice

Some organizations have learned this hard lesson by moving too cautiously. Others have moved quickly, hit the wall of human reality, and were honest enough to report back what they found.

Klarna, the Swedish fintech company, became one of the most telling AI adoption stories of the last few years - for complicated reasons. Between 2022 and 2024, the company reduced its headcount by roughly 40%, attributing much of the reduction to AI, and drew significant attention when its CEO announced that an AI assistant was handling the equivalent work of 700 customer service agents. Customer satisfaction scores then began to fall as the AI handled volume but not complexity; it struggled with emotionally charged interactions, ambiguous problems, and the kind of contextual nuance human agents had managed without anyone noticing. “We focused too much on efficiency and cost,” CEO Sebastian Siemiatkowski eventually acknowledged. “The result was lower quality, and that’s not sustainable.” By mid-2025, Klarna was quietly rebuilding its human customer service capacity, at costs that exceeded the original savings.² The error wasn’t technological, but a misjudgment about which parts of the work were irreducibly human.

IKEA faced the same dilemma and drew a different conclusion. When its AI chatbot took over nearly half of all customer inquiries, rather than cutting the displaced workers it reskilled 8,500 of them as interior design consultants - roles that built on their existing customer knowledge while adding new capabilities. The result was a $1.4 billion revenue uplift. What made this possible was a specific organizational philosophy: that AI and its people needed to move together, not at each other’s expense. “Everyone’s looking for the AI playbook,” said IKEA’s AI Governance and Adoption Manager. “But what we’ve learned is, we have to write it as we go, and make sure everyone’s holding the pen.”

McDonald’s tried AI-powered drive-thru ordering in partnership with IBM. The system misheard orders, added unwanted items, and produced enough customer frustration that the company discontinued the program at over 100 locations. The gap between what worked in controlled testing and what worked in a real drive-thru lane - accents, noise, non-standard orders, distracted customers - was not technical, but a human-context gap that no benchmark had measured in advance.

JPMorgan offers a more nuanced middle case. The bank has deployed AI across hundreds of use cases and built a tool, LLM Suite, that reached 200,000 employees in eight months - crucially, through an opt-in strategy that created organic peer-driven adoption rather than a top-down mandate.³ It has been unusually honest about the costs: operations staff are projected to fall at least 10%, and its own leaders acknowledge significant risks around human oversight and the “value gap” between what the technology is capable of and the ability to capture that value in an enterprise environment. JPMorgan’s candor is itself instructive - genuine transformation requires a clear account of what is being gained and what is being lost, not just a productivity announcement.

Generational complexity and the distribution of stakes

The AI transition is not happening to a homogeneous workforce. It is happening across profound generational, professional, and economic differences, and the stakes are not equally distributed. For a recent graduate fluent in AI tools, the transition may feel like competitive advantage. For a mid-career professional whose expertise was built on skills AI is now approximating, it is a different experience, one that requires reexamining not just workflow but professional identity. And for early-career workers more broadly, the data is already difficult: payroll evidence suggests that employment for workers aged 22 to 25 in highly AI-exposed occupations has declined measurably even as the broader economy has added jobs. What is being automated out first is often the entry-level work through which the next generation of senior practitioners was supposed to develop.

The race, and who holds the pen

There is a version of the AI story in which all of this is beside the point. In this version, competitive logic makes the question of pace effectively moot: organizations that don’t move fast will be outcompeted by those that do. This framing has real force. But accepted uncritically, it is also a form of learned helplessness, a way of preemptively surrendering the agency that individuals, organizations, and democratic societies actually have in shaping how transformative technologies develop.

The cases above don’t support the race narrative. Klarna moved at the speed of the technology and paid for it in customer trust and rehiring costs. IKEA moved at the speed of its people and generated $1.4 billion in additional revenue. The organizations building durable AI capability are not the fastest movers. They are the ones that understood the difference between deploying a tool and changing how work actually gets done - and invested in the slower, more human process the latter requires.⁴

The AI era will eventually reach a new equilibrium. What that equilibrium looks like, who benefits, what work means, and how accountability is distributed will be determined less by the capability of the models than by the quality of the negotiation happening now inside organizations, communities, and political systems about what we actually want from this technology, and on whose terms.

That negotiation is slow, and is rooted in individual agency and collective agreement. It is not a problem that can be – or should be - engineered away.

What are you seeing in your own organizations — about what drives genuine AI adoption versus the appearance of it, and who is absorbing the cost of the gap? I’d welcome your perspectives below.

Leave a comment

Share

Subscribe now

Notes

¹ Gallup, AI Adoption: A People-First Approach, 2024. Gallup finds that 7 in 10 U.S. employees report never using AI at work; only 1 in 10 use it at least weekly.

² The broader pattern is well documented: a 2025 Forrester and Orgvue study found that 55% of companies that rushed to replace human workers with AI now report regretting the decision, citing hidden costs in quality degradation, customer churn, and reversal expenses.

³ McKinsey, The State of AI: How Organizations Are Rewiring to Capture Value, 2025. The report finds that while 88% of organizations now use AI in at least one function, two-thirds remain in pilot or experimental mode, and only 21% have fundamentally redesigned their workflows — the single factor most strongly associated with measurable business impact.

⁴ Microsoft, New Future of Work Report, 2025. The report finds that workers consistently resist top-down AI mandates that prioritize efficiency over quality and creativity, and that social norms and peer influence — rather than institutional directives — are the primary drivers of sustained individual adoption.

That same week, I was also having a different kind of conversation with my students in office hours, after class, and over coffee. Many of these students are former consultants and analysts, and are processing something that felt less like an exciting transition and more like the floor dropping out. One told me that she didn’t know what a first job would look like anymore, because the traditional positions just weren’t materializing. Another wondered how fast the work he had trained for - financial modeling, research synthesis, document drafting – would be absorbed into AI workflows. The law firm partner I quoted in an earlier column made this point more bluntly, stating that his firm would “never hire a 22-year-old again,” though it wasn’t clear how they would continue to produce a pipeline of senior managers. These accounts validated that at least some productivity gains are real, but so is the disruption to the career pathways through which professional knowledge gets transmitted, tested, and made durable.

But there is a third story, one that was entirely absent from the conference room, and only partially visible in my students’ anxiety, and may be at least as important.

The workforce you are not supposed to see

If you widen the lens to see the whole AI supply chain, you can begin to recognize the thousands of workers around the world who contribute their labor to making AI systems actually function. These workers are tagging images, transcribing audio, ranking AI-generated responses, and reviewing content - often deeply disturbing materials - that the algorithms flag but cannot independently evaluate. They are working on platforms like Amazon’s Mechanical Turk, Scale AI, and Remotasks, in Kenya, the Philippines, Venezuela, and India, earning between one and eight dollars per hour, with no benefits, no legally-binding employer relationship, and no mention anywhere in the productivity data.

Source: JISC: Artificial Intelligence

This is what the researchers Mary Gray and Siddharth Suri call “ghost work”[1] - human labor that exists to fill the gaps that automation cannot close, wrapped in a digital interface that renders the worker invisible. The interface performs the trick of erasure: what you see is intelligence; what you don’t see is the people doing the work. This invisibility is not incidental, but often by design, structurally embedded in the contracting arrangements, the geographic arbitrage, and the product marketing that presents human-assisted systems as autonomous ones.

Consider what it actually takes to build a large language model that behaves safely and helpfully. Beyond the architecture and the compute, these systems require a process researchers call reinforcement learning from human feedback - workers evaluating AI-generated responses for quality, helpfulness, and harm, often including direct exposure to violent, hateful, or sexually exploitative material. Investigative reporting has documented workers in Nairobi earning roughly two dollars an hour to label that content - including images of child sexual abuse - used to train the safety filters of some of the most celebrated AI systems in the world. The guardrails we rely on, the reason a large language model declines to help plan an atrocity, were built by people absorbing the worst of what the internet contains. This was done on behalf of companies whose valuations now rival the GDP of mid-sized nations, and whose executives were focused on spreading the productivity message.

Pseudo-autonomy

The gap between the AI pitch and the underlying labor reality shows up not just in training data but in deployment. Amazon’s “Just Walk Out” technology - the cashier-free grocery store marketed as a breakthrough in ambient AI, a glimpse of frictionless retail powered by computer vision - was revealed to rely on roughly 1,000 workers in India watching camera footage and manually verifying transactions.[2] The future-of-retail experience was, in fact, a distributed call center with a glossy interface. Amazon eventually withdrew Just Walk Out from most of its Fresh stores. but the episode captures a pattern that appears across the industry: autonomous vehicle companies employing hundreds of remote operators who monitor and intervene in real-time; enterprise “AI-powered” customer service systems routing forty to seventy percent of interactions to human agents in offshore contact centers; medical AI tools requiring clinicians to verify outputs that are presented to patients as AI-generated assessments.

What we are watching is not the total replacement of human labor but its reorganization -rendered less visible, less protected, more easily attributed to the algorithm, and almost always cheaper, because the workers are farther from the value they create and farther from any regulatory framework that might constrain how they are treated.

This is a profound reorganization of global labor. It is happening now at scale, and large parts of the story are missing or hidden. Some of it is obscured by the contracting structures that keep ghost workers off the books, some by the product marketing that hides the human presence, and some by the simple fact that the workers are far away and the productivity gains are close.

What the accounting leaves out

The pattern is recognizable where markets consistently fail to price in what they generate. When costs don’t show up on the income statement, they tend not to get managed and often land on the people with the least power to resist them. Carbon emissions, water withdrawals, labor violations in extended supply chains: in each case, the discipline of measuring and disclosing those costs had to be built deliberately, by people who consciously built methodologies to account for the full ledger. AI is generating an analogous set of hidden costs, and they are currently as unmeasured and under-governed as industrial emissions were a generation ago.

The AI labor supply chain - from mineral extraction for hardware, to the annotators training models, to the moderators reviewing content in Manila or Nairobi - demands the same rigor we have learned to apply to physical supply chains. In my AI Footprint course at Fuqua, even my most sophisticated students are surprised by what they find when try to follow the labor all the way through these labyrinthine systems. It shouldn’t be surprising; the framing we have inherited about AI as a story about technology and efficiency has trained us not to look and not to see.

Considering the full ledger

While the data is all over the map, it does appear that the promised productivity gains may be emerging, can be measured, and in some domains, are changing the nature of work. But the full ledger of this transformation should also include my students recalibrating their futures, the early-career professionals whose pathways into knowledge work are narrowing, and the workers in Nairobi and Manila whose labor underwrites the systems we celebrate.

AI is not only displacing human labor; it is also profoundly reorganizing it, concentrating the gains among those closest to the technology and the capital, and distributing the cognitive, physical, psychological costs to those farthest away. That reorganization is happening at ever-increasing speed and scale and is outpacing all efforts to create a corresponding governance framework.

The question I consider in every conversation about AI now is not just “what is the productivity return?”, but “who is absorbing the cost of that speed, and do we know who they are?” We desperately need a governance and accounting system that can measure both.

For a detailed description of the full “minerals to monitors” AI labor supply chain, see this link.

These questions are ones I am actively exploring in my teaching and research - I would love to know what you are encountering in your own organizations. Is there an appetite to better understand what is happening to labor both upstream and downstream? Leave your comments below

Leave a comment

Share

Subscribe now

[1] Gray, M. L., & Suri, S. (2019). Ghost work: How to stop Silicon Valley from building a new global underclass. Houghton Mifflin Harcourt.

[2] Business Insider. (2024, April 3). Amazon’s “Just Walk Out” technology relies on hundreds of workers in India watching you shop.

Source: Business Insider

Anthropic’s CEO Dario Amodei said publicly that he couldn’t agree to the government’s terms in good conscience. In the cold logic of government contracting, that kind of language stands out - and it may prove to be shrewder than it looks. Claude surged to number one on Apple’s App Store in the days after the Pentagon’s designation, a signal that consumers noticed and rewarded a company that held a line even when it was costly to do so. In an industry where public trust is both fragile and essential, that kind of reputational capital may be worth more than a $200 million contract. As AI becomes embedded in healthcare decisions, financial products, hiring, and public infrastructure, companies that can credibly claim that they put ethical limits above commercial convenience will have a durable advantage. On the other hand, trust is not easily rebuilt once it has been lost.

Whatever one thinks of these companies’ choices, the episode reveals something important about the state of AI governance more broadly. The rules governing how artificial intelligence is deployed by the most powerful military in the world are not being written by legislators or regulators. They are being negotiated in bilateral procurement contracts between government agencies and private companies, and then publicly revised through CEO posts on social media. As one Lawfare analysis noted, the US has drifted toward a model of “regulation by contract”1 - flexible, opaque, and entirely lacking the democratic accountability that actual laws and regulations provide.

This is, in a meaningful sense, governance turned on its head. We are accustomed to governments setting the boundaries within which private companies operate. Here, a private company was attempting to restrain a government from using its own technology in ways the company deemed harmful, while the government deployed economic coercion to override those restrictions. The result isn’t a governance system; it’s an improvisation, held together by contract language and corporate ethics policies, neither of which was designed for the job.

For sustainability professionals, this may feel eerily familiar.

The ESG Parallel

The field of sustainability has spent decades navigating a similar vacuum. In the absence of binding regulation on carbon emissions, water use, supply chain labor standards, or biodiversity loss, the work of defining, measuring, and managing corporate externalities fell largely to a dispersed ecosystem of voluntary frameworks, stakeholder pressure, and internal governance. Professionals working in ESG have learned what it means to manage material risks when the regulatory floor is missing, and when the consequences of getting it wrong land not on the balance sheet but in the atmosphere, the watershed, and the community.

AI is generating an analogous set of externalities, and they span the same terrain. The energy and water demands of large-scale model training and data center operation are substantial and growing, with AI-specific loads expanding faster than grid capacity in many regions. The hardware supply chain carries its own environmental and human rights footprint, from lithium and cobalt mining to e-waste. And the social externalities are vast: algorithmic bias baked into hiring and lending systems, labor displacement that is moving faster than workforce transition programs, the erosion of information ecosystems, and now, the automation of surveillance and lethal force. These are current impacts, not projected future harms, and until now they currently lack any coherent governance architecture.

A Familiar Playbook, a New Domain

What the Anthropic-Pentagon confrontation makes clear is that private companies are currently the last line of defense on some of the most consequential AI governance questions of our time - not because they are the right actors for this role, but because governments have largely stepped back. This is exactly the dynamic that sustainability professionals have navigated in domains from climate to deforestation to forced labor in supply chains.

The parallel runs in both directions. The ESG field offers a well-worn playbook: build disclosure frameworks that create market-facing accountability; develop internal governance capacity to assess material risks; engage supply chains on standards; work at the intersection of voluntary and regulatory regimes to raise the floor over time. But the ESG field also carries a cautionary lesson - voluntary frameworks without regulatory backstops have a tendency to become exercises in narrative management rather than actual impact management. AI governance will need to avoid that trap. The stakes are too high and the pace of deployment too fast for another cycle of voluntary commitments.

The Role Sustainability Professionals Are Built For

This is not someone else’s problem. AI is already embedded in the operations of nearly every sector, which means sustainability professionals are already encountering its footprint - in the Scope 2 emissions of data center vendors, in the procurement of AI-enabled logistics tools, in workforce transition plans their companies may or may not be making, in the integrity of the ESG data and reporting systems that increasingly rely on AI to function.

Understanding the AI footprint in its full scope - from kilowatt-hours to human rights - is becoming a core sustainability competency. New courses and frameworks are beginning to map AI’s environmental and social impacts across the value chain, equipping practitioners to assess exposure, engage vendors, and bring AI governance into existing ESG and enterprise risk functions. At Fuqua, I have just launched a new mini-course on this topic called the AI Footprint: Sustainability, Strategy, and Society. Beyond striving to learn all the new AI tools and implement them in their organizations, our graduates must be prepared to respond effectively to the crises and opportunities that the AI revolution will spin off - from mass layoffs to deep fake attacks on their brand.

But the larger opportunity is to reimagine governance itself. Sustainability professionals know how to build the architecture that turns risk awareness into accountability - the materiality frameworks, the supplier engagement processes, disclosure standards, and board-level oversight structures that make commitments durable. AI governance needs all of this, and it needs practitioners who have already built these capabilities in adjacent domains.

The drama between Anthropic and the Pentagon will not be the last of its kind. It was a vivid illustration of what happens when the governance of a transformative technology is left to the moment of crisis - in this case, to a Friday deadline and a social media announcement. The sustainability profession has spent a generation demonstrating that external costs don’t disappear simply because no one is accounting for them. The same logic applies here. The question is whether the profession moves to engage AI governance seriously, and soon - or waits for the costs to materialize.

Leave a comment

Share

Subscribe now

They’re wrong.

The truth is that we can’t afford not to accelerate the shift to renewables and modern grids. The idea that fossil fuels are a guarantor of cheap energy may have been true a generation ago, but it’s now an illusion, one that obscures the deeper economics driving our electricity costs.

The End of the Fossil Bargain

For much of the twentieth century, fossil fuels underwrote industrial expansion. The infrastructure was already built, the fuels were abundant, and costs were stable enough to plan decades ahead. But that world no longer exists.

Today, global gas and coal markets are defined by volatility, not stability. U.S. natural-gas prices have swung more than fourfold in the past decade. Europe’s power bills tripled during the 2022 energy crisis. Even countries rich in domestic resources, like India and China, have seen steep price spikes when weather extremes or supply-chain disruptions hit.

Source: International Energy Agency

In a system where half of electricity costs come from fuel, volatility creates vulnerability. Every geopolitical shock, pipeline disruption, or heat wave turns into a price surge. What looks like “affordable” fossil energy today is, in practice, a bet that nothing will go wrong tomorrow.

Why Prices Are Rising Now

It’s true that electricity bills are climbing, but renewables are not the culprit. Three structural forces are driving the increase:

1. Demand growth: After decades of flat demand, the electrification of transport, heating, and industry, plus an explosion of data-center construction, is sending electricity consumption soaring. U.S. utilities that once projected near-zero growth now expect demand to rise 3 percent a year through 2040.

2. Aging infrastructure: Most transmission lines in advanced economies are 40 to 50 years old. Deferred maintenance, congestion, and storm damage are forcing emergency repairs and grid upgrades. Those costs flow directly into rates.

3. Fuel dependence: Because many grids still rely on gas-fired power for marginal supply, every bump in global fuel markets cascades into retail bills.

In short, our electricity prices are high not because we’re building too many renewables, but because we’ve waited too long to build the system that can deliver them efficiently.

Renewables as the Economic Stabilizer

Wind and solar are now the cheapest forms of new generation in most of the world. Their marginal cost of production is near zero; the “fuel” is free. Once built, renewables hedge against the very volatility that fossil fuels guarantee.

Yet renewables alone are not enough. Without modern grids, storage, and flexible demand, cheap generation gets trapped behind bottlenecks or wasted through curtailment. This is where the affordability argument needs reframing. Grid investment isn’t a luxury; it’s the backbone of cost control.

A dollar spent upgrading transmission to connect low-cost renewables can save multiple dollars in avoided fuel and congestion costs. Smart grids, distributed storage, and time-of-use pricing can flatten peak demand, the most expensive slice of the power curve. In California, flexible-load programs have cut wholesale prices by 10–15 percent during summer peaks.

The economic equation is simple: renewables + grid = stability.

The Fossil Trap

Proponents of fossil expansion argue that keeping coal and gas online is the only way to preserve affordability. But that logic doesn’t fit the facts.

New fossil plants are no longer cheap. Even before accounting for carbon costs, new gas generation in the U.S. now averages $70–90 per MWh, higher than utility-scale solar or wind paired with storage. Moreover, each new fossil investment locks in decades of fuel exposure and future cleanup liabilities.

This is what economists call the fossil trap: extending old infrastructure to avoid near-term pain, only to create far higher long-term costs. Consumers end up paying twice - first for volatile fuel bills, then for stranded assets when policy or economics finally force retirement.

Countries that clung longest to coal now face the steepest transition bills. In contrast, those that invested early in renewables, like Portugal, Denmark, and parts of Australia, are now enjoying some of the lowest wholesale prices in the world.

Losing the Future to China

There’s also a larger consequence: by doubling down on fossil fuels, the United States is effectively outsourcing the clean-energy economy to China.

While U.S. debates rage over the cost of renewables, China has already won the cost curve. It controls more than 80 percent of global solar manufacturing, over two-thirds of battery-cell production, and is rapidly expanding wind-turbine exports across Africa, Latin America, and Europe. Every year we delay domestic investment, our dependence on imported clean-energy technology deepens.

In economic terms, this is the equivalent of abandoning the semiconductor industry in 1990. In geopolitical terms, it’s a slow erosion of energy sovereignty. The next generation of global competitiveness - powering data centers, EV fleets, and industrial heat - will hinge on who owns the supply chains for clean, affordable energy. Right now, that answer is not America.

If U.S. policy continues to subsidize fossil extraction while under-investing in grid modernization and manufacturing, we’re not just paying more for electricity, we’re paying China to dominate the next century’s energy system.

True energy security will come not from pumping more gas, but from producing the technologies that make cheap, clean power possible everywhere.

When Energy Prices Become Political

Energy affordability is not only an economic issue; it’s becoming a political fault line. Rising power bills have already become a talking point in state races, and they are likely to play a central role in the upcoming mid-term elections. For most voters, electricity prices are the most visible signal of whether the economy is working for them.

If the national conversation frames the revival of fossil fuels as the path to affordability, the U.S. will double down on short-term fixes that erode long-term competitiveness. But if leaders can connect lower-cost renewables, resilient grids, and domestic manufacturing to economic security, the debate could shift from grievance to prosperity. Energy policy may well be decided not in climate committees, but at kitchen tables comparing utility bills.

Affordability Is About Time Horizon

Affordability isn’t the price on next month’s utility bill; it’s the trajectory over the next decade. Cheap power today that leads to chronic volatility, grid failures, and stranded assets is not affordable—it’s deceptive.

By contrast, renewables and grid modernization offer a declining cost curve. The capital intensity is front-loaded, but once built, operating costs are stable for decades. It’s like refinancing your mortgage at a fixed low rate instead of gambling on variable interest payments.

Governments and utilities should treat grid modernization as essential infrastructure, not as a discretionary climate policy. Extending cost-recovery periods, using public or blended finance for transmission, and aligning renewable deployment with grid readiness can help keep near-term rate impacts modest while delivering long-term savings. At the same time, we should recognize that higher electricity prices are unavoidable in the short term - no matter which path we take - but today’s choices will determine whether those costs leave us more vulnerable or more resilient in the years ahead.

The Only Sustainable Low-Cost Strategy

The fossil-fuel-dominance argument frames the transition as a trade-off between affordability and ambition, but the economics tell a different story: sticking with the past is the most expensive way to preserve the status quo - and the most likely to fail. A strategy built around volatile fuels, aging infrastructure, and shrinking investor appetite for carbon-intensive assets can delay change for a few years, but it cannot deliver stable or affordable energy over time. The longer we cling to yesterday’s system, the higher the financial, reliability, and political costs will become.

We’re entering a world where demand is growing, climate volatility is intensifying, and capital is chasing low-risk, stable-return assets. In that world, the winners will be regions that invest early in the infrastructure of abundance - clean generation, smart grids, flexible loads - not those that cling to scarcity models built on volatile fuels.

The cheapest kilowatt-hour isn’t the one from the past; it’s a renewable/grid system that doesn’t depend on a fuel market at all.

The affordability debate should be about building that future faster, not defending the illusion of cheap fossil power. Because in economics, as in energy, the longer we wait to modernize, the higher the bill becomes.

Leave a comment

Share

Subscribe now

That illusion is cracking. Across the country, utilities are quietly filing double-digit rate increases.[1] California’s Pacific Gas & Electric now charges an average of 32 cents per kilowatt-hour, nearly twice what it cost a decade ago.[2] North Carolina, Georgia, and Arizona have all approved multi-year hikes. The national average sits near 16 cents, and the trendline points sharply upward.[3]

For the first time in decades, electricity prices are set to rise faster than inflation. It is easy to scapegoat clean energy, but the truth is more complicated. The cost of decarbonizing our grid is real, but so is the cost of doing nothing, and that bill is coming due whether we acknowledge it or not.

Image source: TRC Companies

The Real Price of Power

Electricity prices are rising because America is finally rebuilding the energy system it has been neglecting for half a century. Utilities are replacing 1970s-era transmission lines, hardening distribution networks against wildfires and storms, and connecting massive new solar and wind farms to distant population centers.

The Edison Electric Institute estimates that utilities will invest $150 billion per year through the 2030s, three times their historical average.[4] Building new high-voltage lines will add two or three cents per kilowatt-hour to the average bill and undergrounding wires and upgrading substations add another couple of cents.

Battery storage - vital for smoothing the solar “duck curve” and keeping lights on after sunset - costs about $120–150 per megawatt-hour,[5] nearly quadruple the cost of solar generation itself. Reliability, it turns out, isn’t free.

Even though the Inflation Reduction Act still offers one of the largest federal buffers against rising energy-system costs,[6] that buffer is now subject to rollback risk.[7] As a result, most analysts continue to expect upward pressure on U.S. electricity prices, many projecting 15-40% growth nationally by 2030.[8]

For politicians and ratepayers alike, this feels like bad news. But before we panic about the cost of transition, we should be honest about what we’re paying for.

The Cost of Decades of Delay

When people complain about rising electricity bills, they often imagine that the alternative of sticking with what we have costs nothing. That’s the biggest economic lie of our time.

The status quo is already expensive. We just pay those bills in different accounts: through taxes, insurance premiums, hospital visits, and disaster relief.

· In the past five years alone, the U.S. has suffered $617 billion in climate-related disasters, including wildfires, floods, hurricanes, and heat waves.[9] Those costs show up in rising insurance rates and federal deficits, not in your utility statement.

· Fossil fuel pollution kills roughly 100,000 Americans a year, according to Harvard research, imposing $800 billion in annual health costs.[10] That’s equivalent to an extra five cents on every kilowatt-hour of fossil-fired power, just hidden in the health-care budget.

· Gas-price volatility during the 2022 energy crisis added $70 billion to U.S. electricity bills in a single year. The supposed “cheapness” of gas power depends on global commodities and geopolitics we don’t control.

· Heat waves and blackouts already cost the U.S. economy $100–200 billion annually in lost productivity. Those losses are rising exponentially.

And behind all of this sits an enormous unpaid tab: the social cost of carbon, a government-calculated estimate of the damages caused by each ton of CO₂ emitted into the atmosphere. Economists across administrations once treated it as a basic accounting tool, helping policymakers weigh the hidden costs of pollution against the price of prevention.

Today, that line item is being scrubbed from official documents. Federal agencies have been instructed to delete references to the social cost of carbon from regulations and policy analyses.[11] The numbers may vanish from spreadsheets, but the costs they represent - flooded cities, failed crops, rising seas - do not. Pretending those damages don’t exist doesn’t make them go away, but only ensures we’ll pay them later, with interest.

At roughly $190 per ton of CO₂, the social cost of carbon adds six to eight cents to every kilowatt-hour of coal or gas electricity.[12] If that cost were visible on monthly bills, fossil power would no longer look cheap.

In other words, the clean-energy transition doesn’t make electricity expensive. It finally makes its costs visible.

Why We’re Paying More Now

Several powerful forces are converging to push electricity demand up and costs higher:

1. Electrification of everything. Electric vehicles, heat pumps, and industrial electrification are adding load faster than utilities can build generation. The grid, once stagnant, could see 30–40 percent growth in demand by 2035.[13]

2. The AI super-cycle. Data centers for artificial intelligence now consume as much power as small cities. Google, Microsoft, and Amazon are doubling their electricity use every few years as AI workloads explode.[14] In Virginia, utilities blame data-center growth for 15 percent rate hikes.

3. Aging infrastructure. Half the nation’s transmission lines are over 50 years old.[15] They were never designed for a grid that runs on distributed solar, offshore wind, and millions of EV chargers.

4. Climate resilience. Wildfires, hurricanes, and heat waves are forcing utilities to invest billions in undergrounding, smart meters, and stronger substations. These upgrades are essential but capital-intensive.

Each of these investments raises short-term rates but lowers long-term risk. Like fixing the roof before the storm, the cost hurts now and saves later.

Who Pays - and Who Benefits

The politics of rising electricity prices will hinge on fairness. Most state regulators still push the cost of capital upgrades onto ratepayers. That makes sense for regulated monopolies, but it risks a backlash if households see higher bills without visible benefits.

There are better options. Corporate buyers like Google and Microsoft are already signing multi-billion-dollar, long-term clean-power contracts that help finance new infrastructure.[16] Federal green bonds or public-private transmission projects can cut borrowing costs by one or two percentage points—saving tens of billions in rate recovery. Targeted rebates and community solar programs can protect low-income households from higher rates.

The goal shouldn’t be to stop prices from rising; it should be to make sure the value rises with them.

The Cheap Power Illusion

For much of the twentieth century, America’s energy advantage came from abundant fossil fuels. But cheap gas and coal were never truly cheap. They were subsidized by the climate, by public health, by future generations.

Today, as the planet warms and disasters multiply, that illusion is collapsing. The real choice is not between “expensive clean energy” and “cheap fossil energy.” It’s between paying now to build a resilient, zero-carbon system or paying far more later to survive the consequences of the one we already have.

Yes, electricity bills will rise in the coming decade. But those increases represent something profound: the first honest accounting of what power really costs. We are, finally, paying the price of “must haves”: reliability, safety, independence from volatile fuels, even survival.

The Only Unaffordable Option

It’s easy to scapegoat the energy transition for higher bills. It’s harder to admit that for fifty years, Americans underpaid for electricity by overdrawing on natural capital and deferring maintenance on the very system that keeps our civilization running.

Now the bill has arrived. We can pay it in the form of smarter grids, cleaner power, and more resilient communities - or we can pay it in lost lives, burned towns, and broken economies.

One way or another, we’re going to pay. The only unaffordable option is pretending that doing nothing is free.

Leave a comment

Share

Subscribe now

References

[1] Edison Electric Institute. (2025, July 23). Electric utilities will invest more than $1.1 trillion by 2030 to meet demand growth. Utility Dive.

[2] Edison Electric Institute. (2024). Industry capital expenditures: U.S. investor-owned electric utilities.

[3] U.S. Energy Information Administration. (2025, September 23). Electric Power Monthly: Table 5.3. Average price of electricity to ultimate customers: Total by end-use sector, 2015–July 2025 [Data table]. U.S. Department of Energy.

[4] Edison Electric Institute. (2024). 2024 Financial Review: U.S. electric companies projected to invest more than $1.1 trillion between 2025 and 2029.

[5] Lazard, Inc. (2025, June). Levelized cost of energy+ (LCOE+): Version 18.0 [Report].

[6] U.S. Environmental Protection Agency. (2024). Summary of Inflation Reduction Act provisions related to renewable energy.

[7] Holland & Knight LLP. (2025, June 30). Senate moves to scale back clean-energy tax credits under the IRA: Latest updates.

[8] ICF. (2025, May). Rising current: America’s growing electricity demand. ICF.

[9] National Oceanic and Atmospheric Administration, National Centers for Environmental Information. (2025, November 10). U.S. billion-dollar weather and climate disasters.

[10] Natural Resources Defense Council. (2021, May 20). Report: Health costs from climate change and fossil-fuel pollution top $820 billion per year.

[11] Ball, M. (2025, May 9). White House ends use of carbon cost. Argus Media.

[12] E&E News. (2025, May 9). White House bars agencies from using the social cost of carbon.

[13] International Energy Agency. (2025, April). Data centers and AI: Electricity demand projections to 2030.

[14] Deloitte LLP. (2024, August). Funding the growth in the U.S. power sector: Investments could total as much as US$1.4 trillion from 2025–2030.

[15] National Conference of State Legislatures. (2024). Modernizing the electric grid: State role and policy options. Black & Veatch.

[16] Reuters. (2025, July 15). Google inks $3 billion U.S. hydropower deal in largest clean-energy agreement of its kind. Reuters.

The students in this course have been among the most engaged in recent years - a source of real encouragement and hope for me during this otherwise dark period. I have heard countless stories from my students about fresh insights, shifts in interests and priorities, and inspiration for new career directions. Seeing this tangible, personal impact in the lives of my students is the best part of my work - and I truly appreciate the opportunity to curate that process.

Here are the highlights of our last 3 classes:

Class 10 – Building Innovation Ecosystems: The Iceland Ocean Cluster

We began by exploring what it really means to build an innovation ecosystem. I framed the conversation around the idea that innovation ecosystems form in the spaces where ideas, capital, and culture come together. The Iceland Ocean Cluster shows what happens when collaboration itself becomes part of the national DNA. Its guiding vision of “100 percent fish” isn’t just a clever slogan about waste reduction - it’s a wholesale re-imagining of value and purpose.

Guest speaker Thor Sigfusson, joining us from Reykjavík, described how the movement began: “We used to throw away half of every fish we caught. Today, we use almost all of it - collagen, oil, skin, bones, enzymes, even fish leather.” He emphasized that the breakthrough wasn’t technological but relational: “What makes it work is trust. We meet every day, share data, and sometimes even customers. It’s not about one big company. It’s about 120 small companies learning from each other.” When fishermen, engineers, and artists started sharing the same space, he said, “innovation happened fast.”

Thor Sigfusson, Founder and Chairman, Iceland Ocean Cluster

Students quickly picked up on the role of scale and culture. One asked whether the model succeeded precisely because Iceland is small - “How could something like this scale in a place like the U.S.?” Another wondered about government’s role: “Did policy help build the cluster or just stay out of the way?” A third challenged the assumption of perfect cooperation: “You talked about sharing knowledge openly - how do you avoid free-riders who take ideas but don’t contribute?” Others linked the story to broader sustainability frameworks, noting that the cluster functions like a regenerative system, each part feeding the others rather than competing.

Thor acknowledged that exporting the model isn’t simple: “We’ve tried to share this in Namibia, Norway, and the U.S., but culture matters. You can’t copy-paste trust.” That remark prompted reflection on how ecosystems are rooted in local relationships yet can inspire global learning.

I closed the discussion by observing that ecosystems are living systems - they require constant tending. You can’t manage them the way you’d manage a firm; you cultivate them. What makes the Iceland Ocean Cluster remarkable is not just its efficiency but the alignment of purpose, profit, and place - a reminder that leadership in innovation often means designing the conditions for creativity, not just the products that emerge from it.

Class 11 – Leading Systems Change: Patagonia’s Strategic Pivot

We opened with the story that captured headlines worldwide: Yvon Chouinard’s 2022 decision to transfer ownership of Patagonia to a nonprofit trust so that, in his words, “Earth is now our only shareholder.” I reminded the class that this case wasn’t just about a governance mechanism; it was about a re-imagining of capitalism itself. Patagonia forces us to ask what happens when the moral and financial logics of a company finally point in the same direction?

Students quickly dove in. One observed that the move worked because Patagonia had decades of credibility to build on; another countered that success itself made the decision possible - “it’s easier to be virtuous after you’re profitable.” A third student pointed out how the trust model insulated Patagonia from shareholder pressure while still keeping it accountable to its mission. That comment led to a lively back-and-forth about whether purpose trusts could work in public markets or whether they’re inherently a privilege of private ownership.

The conversation deepened around the mechanics of change. We unpacked how the Patagonia Purpose Trust holds all the voting stock while the Holdfast Collective receives the profits for environmental protection. Several students said this bifurcation reminded them of dual-class shares used for control, but with ethics rather than ego at the center. One asked pointedly, “Could this structure survive a future leadership team that doesn’t share Chouinard’s values?” That question brought us to culture as the ultimate safeguard. Governance can lock in intentions, but culture carries them forward. We all had a brief discussion of the merits and limits of other governance arrangements like B-corps, benefit corporations, and company-affiliated foundations. In summary, most students were interested in these alternative forms of governance, but were skeptical about the current efforts to reorient capitalism toward positive societal outcomes.

We closed by stepping back to the broader theme of the course: leading systems change. I told the group that what makes Patagonia radical isn’t its tax structure but its imagination, its willingness to redesign ownership around planetary stewardship. Our final reflection circled this question: If you could redesign the rules of your organization, what would you optimize for growth, resilience, justice, or regeneration? It seems that tradeoffs are inevitable, though Patagonia’s example suggests it is an experiment worth continuing.

Class 12 – Designing Your Theory of Change

Our final session brought the course full circle, from systems thinking and corporate cases to personal agency and leadership intent. I opened by reminding the group that sustainability strategy ultimately demands a theory of change: a clear logic for how one’s actions contribute to outcomes beyond the firm. It’s easy to critique systems but it’s harder - and more useful - to map out how you will personally move them.

Students presented their personal Theories of Change in short talks, each grounded in a sector, role, or problem. One student began by reflecting, “I used to think my work would change the world by scaling new technology. Now I see it’s as much about who I can bring with me.” Others focused on impact through finance, storytelling, or governance, bridging personal values and strategic leverage points.. Many students described the struggle between systems change and the imperative to narrow focus - choosing one issue, one action, one place to start.

We reflected on the evolution of our conversation across the term—from diagnosing planetary crises to imagining regenerative systems, and finally to defining our personal agency within them. The exercise reminded us that sustainability isn’t an external problem to be solved, but a practice of alignment between values and action. Theories of Change are living maps that evolve as experience deepens.

As we reached the end of our final class, I invited students to keep three images in mind: first, I showed a picture of an iceberg, and encouraged the students to always seek to understand the dynamics below the headlines, to track the tectonic plates more than the daily weather. Second, I shared a picture of Pinocchio, who wished on star to become a real boy. I used this example to encourage students to dare to imagine, even when the current environment is filled with negativity. I also reminded them to hold on their honesty and integrity, avoiding Pinocchio’s fate when he wasn’t truthful. Finally, I shared a photo of a hand sending a ripple out on a pond. The reference here is to seek to make an impact (the ripple), but to have fun playing in the water, and don’t take yourself too seriously.

Thus ends our journey. I hope you continue to stay in our conversation. We have so many more mountains to climb, adventures to pursue, insights to discover.

Leave a comment

Share

Subscribe now

Coming Up in CSCG: Leadership & Change (Sessions 10–12)

Over the next three sessions, we turn to the final arc of the course—Leadership & Change—exploring how innovation, governance, and personal agency intersect to create systemic transformation. We’ll move from the collaborative ingenuity of the Iceland Ocean Cluster, to the audacity of Patagonia, and finally to your own Theory of Change.

Class 10 — Building Innovation Ecosystems (Tuesday, October 7)
Abridged case: The Iceland Ocean Cluster and the 100% Fish Revolution: Creating Innovation Ecosystems
Guest Speaker: Dr. Thor Sigfusson, Founder, Iceland Ocean Cluster

This session examines how innovation ecosystems can accelerate the growth of the blue economy—creating value from marine resources while advancing ocean sustainability. We’ll explore how Iceland’s “100% fish” philosophy transformed waste into wealth and how collaborative networks can spark new industries around regenerative ocean solutions.

Preparation:

• Read the Iceland Ocean Cluster abridged case study and the background materials below.

• Question to consider: What makes an innovation ecosystem thrive—and how can we replicate it without losing its authenticity and local roots?

Readings

• Vermeer, D. (2025). The Ocean Innovation Playbook: A guide to building and scaling blue economy ecosystems. Substack.

• World Economic Forum, Ocean Risk and Resilience Action Alliance, Builders Vision, & Katapult Ocean. (2025). Making waves in the regenerative & sustainable ocean economy: Transformative ocean investment opportunities (pp. 4–28). World Economic Forum.

• MBA EDGE. (2025). The blue economy: What every MBA needs to know. MBA-EDGE.com.

Class 11 — Leading Systems Change (Friday, October 10)
Abridged case: Patagonia’s Strategic Pivot: The Business Case for Giving It All Away

We’ll analyze Patagonia’s revolutionary decision to transfer ownership to a nonprofit dedicated to protecting the planet. What began as an act of corporate conscience has become a test case for post-capitalist governance—raising questions about ownership, accountability, and replication.

Preparation:

• Read the case and the following materials.

• Question to consider: If your organization wanted to pursue true systems change, where would you start—governance, culture, or capital?

Readings

• Chouinard, Y. (2022). Earth is now our only shareholder. Patagonia.

• Harvard Business Review. (2022). What happens when a company (like Patagonia) transfers ownership to a nonprofit? HBR.org.

• Sustainability SmartPod. (2022, October). Patagonia’s Vincent Stanley on the future of the responsible company [Podcast]. Sustainability SmartPod.

Class 12 — Designing Your Theory of Change

Tuesday, October 14
Activity: Personal Theory of Change Presentations

We’ll close the Leadership & Change module by bringing the work home. Each of you will present and refine your personal Theory of Change (TOC)—a structured way to link your actions and leadership to larger systems of social and environmental impact. This session integrates everything we’ve explored this term: systems thinking, innovation, purpose, and agency.

Preparation:

• Read the USAID Theory of Change Workbook (pp. 1–21).

• Listen to the Scene on Radio episode below for a provocative reimagining of the economy’s purpose.

Readings

• U.S. Agency for International Development (USAID). (2024). Theory of change workbook (pp. 1–21). USAID.

• Scene on Radio. (2024, September 18). Season 7, Episode 12: Reimagined economies [Podcast]. Scene on Radio.

Connecting the Arc

Across these three sessions, we’ll move from collective innovation (Iceland) to organizational transformation (Patagonia) to individual leadership and purpose (your TOC). Together, they complete our exploration of how sustainability transitions actually happen—not as isolated acts of heroism, but as coordinated, multi-level change across ecosystems, institutions, and individuals.

Leave a comment

Share

Subscribe now

Our visit to the headquarters of Big Spoon Roasters in Hillsborough, NC offered a rare inside look at how a small, values-driven business turns sustainability from principle into practice. Students toured the local production facility, watching as staff carefully weighed the day’s waste—usually just a few pounds—to uphold the company’s zero-waste commitment. They saw how employees scraped the last bit of nut butter from mixers to donate to local food programs and even analyzed the trade-offs between reusable and compostable tasting spoons, balancing water use against waste.

In conversation with founders Megan and Mark Overbay and their team, students encountered a company that treats every operational choice as an ethical one: whether to sell through grocery stores or maintain direct relationships with customers; how to verify supply chains for ingredients like cashews and vanilla; and whether selling through Amazon undermines or extends the mission. These choices illuminated the tensions between scale and integrity, growth and control, profit and purpose.

During the class debrief, students described Big Spoon as a living example of values-based entrepreneurship—where culture, authenticity, and product quality reinforce one another. Employees clearly loved their work, and their pride translated into brand credibility. Yet the discussion also surfaced the company’s structural limits: the difficulty of tracing supply chains, the ambiguity of certifications, and the lack of universally accepted sustainability standards.

One person compared the current state of sustainability reporting to financial accounting a century ago: “There were no consistent rules, no shared metrics—just a lot of improvisation.” The class agreed that building such standards is part of the generational work ahead. Big Spoon’s experiment in moral clarity, they concluded, shows what integrity looks like before the systems exist to measure and reward it.

Class 9 – Norges Bank Investment Management: Capital and Credibility

The following session shifted from local craft to global capital, using the Norges Bank Investment Management (NBIM) case to examine how the world’s largest sovereign wealth fund—managing more than $1.8 trillion - balances long-term returns with systemic sustainability challenges.

Students who had visited NBIM’s headquarters in Oslo during the Nordic GATE program described a paradoxical culture: understated offices and quiet discipline masking enormous financial power. They recalled how employees combined technical expertise with optimism—seeing their work not just as managing assets, but as stewarding the future.

To frame the discussion, students mapped different investor types - from impact funds to private equity - along two dimensions: influence (market power) and commitment to sustainability. On that grid, NBIM emerged as a high-influence, mid-commitment actor: a universal owner with global reach but limited freedom to deviate from financial benchmarks. The class explored how its five-layer governance structure and 1.25% capped tracking error constrain its ability to act boldly, even when its leaders want to.

Debate then turned philosophical. If NBIM profits from companies developing adaptation or resilience technologies, is that ethical investment or disaster capitalism? Can capital markets truly finance regeneration, or only profit from crisis? Students disagreed but acknowledged that NBIM’s structure—public, transparent, and globally diversified—makes it a mirror of the system it seeks to change.

Discussion of its top holdings - Microsoft, Apple, Google, NVIDIA, Meta - sparked questions about whether the fund’s portfolio reflects the economy’s actual sources of value or simply its financialized perception of it. Some argued that NBIM’s greatest leverage lies not in divestment but in soft power—its ability to influence norms through engagement, its widely read Responsible Investment Report, and CEO Nicolai Tangen’s In Good Company podcast, which amplifies conversations about corporate accountability.

Professor Vermeer framed the case as a test of systemic leadership: How can an investor bound by political mandates and market rules still shape a more sustainable economy? The answer, students noted, may not be found in “what” NBIM invests in but “how” it wields its influence.

Guest Lecture – Professor Fern Wickson and the IPBES Transformative Change Assessment

The class culminated in a guest session with Professor Fern Wickson, ecologist and sustainability scholar at the University of Tromsø – The Arctic University of Norway, who joined to discuss the landmark IPBES Transformative Change Assessment (O’Brien, Wickson et al., 2025).

Wickson began with a grounding exercise: “What do you love about being alive on planet Earth?” Students named oceans, mountains, seasons, and mystery—reminders of what sustainability is meant to protect. She then outlined the “triple planetary crisis”—climate change, biodiversity loss, and pollution—and presented stark statistics: a 73% decline in global wildlife populations since 1970, and the risk of losing nearly all coral reefs if warming exceeds 2°C.

Yet Wickson’s message was not despair but transformation. Drawing from the IPBES framework, she defined transformative change as systemic shifts across three interconnected dimensions:

• Views – how we think, value, and imagine;

• Structures – how we govern, regulate, and allocate power;

• Practices – how we act and relate to one another and to nature.

She identified three root causes of ecological decline: disconnection and domination, concentration of wealth and power, and the prioritization of short-term material gain. To reverse them, she argued, societies must embody four counter-principles: equity and justice, pluralism and inclusion, reciprocity with nature, and adaptive learning.

Using ecological metaphors, Wickson described how existing systems must decompose before new ones can grow, urging students to see themselves as “seed planters” preparing ideas and institutions for emergence when the old order fails. Transformation, she noted, is not only structural but imaginative—requiring us to envision what thriving human–nature relationships could look like.

A rich dialogue followed. Wickson and Professor Vermeer discussed “future design” experiments—decision-making processes where some participants represent future generations—and “Council of All Beings” exercises, where participants speak on behalf of nonhuman species. Both invited students to practice moral imagination as a form of leadership.

In response to a question about working within the U.S. business context, Wickson offered pragmatic encouragement: “Systems must break down before they renew. The work is to cultivate the seeds of what comes next—and to do it in community, not alone.”

Synthesis – Connecting the Micro and the Macro

Together, these two sessions traced the full arc of sustainability—from the tactile integrity of a small Durham factory to the strategic influence of a trillion-dollar fund. Big Spoon Roasters showed how purpose can animate everyday decisions but strain against financial limits. NBIM revealed how global institutions can drive systemic change yet remain constrained by governance, politics, and market inertia.

Professor Wickson’s IPBES framework provided the bridge between them: both the small entrepreneur and the sovereign fund operate within systems that must evolve across views, structures, and practices. Big Spoon is changing culture from the ground up; NBIM, from the top down. Both are necessary, and both are incomplete.

In the end, the question uniting Classes 8 and 9 was not simply how to make business sustainable, but how to make sustainability transformative—turning purpose from a moral stance into a design principle for the next economy.

Reference
O’Brien, K., Wickson, F., et al. (2025). IPBES Transformative Change Assessment: Summary for Policymakers. Intergovernmental Science–Policy Platform on Biodiversity and Ecosystem Services (IPBES). Zenodo.

Leave a comment

Share

Subscribe now

We then turn to the case of Norway’s Norges Bank (NBIM). the world’s largest sovereign wealth fund has a global reputation for sophisticated investing strategy, considering the broader social and environmental impacts of their holdings. But many structural features of the bank’s governance shape (and limit) its ability to make bold moves - e.g. being a “universal owner” with exposure across the whole global economy, being governed by a complex hierarchy of Norwegian government entities, and the contradictions of seeking energy transition while funded by Norway’s extensive oil and gas industry.

Here’s how to join along with the conversation:

Class 8: Scaling Values-Driven Businesses

In this session, we will use the Big Spoon Roasters case to explore how small, mission-driven companies scale while staying true to their values. We will examine the strategic and operational tensions that arise when growth creates pressure to compromise on quality, sourcing, or stakeholder commitments. The session will include a field trip to Big Spoon’s HQ in Hillsborough, NC, and a discussion with founders Megan and Mark Overbay, offering firsthand perspectives on balancing purpose, profitability, and long-term brand integrity.

Abridged Case Study: Big Spoon Roasters: Scaling a Values-Driven Food Business

Associated Readings

• Chhabra, E. Why One Mission-Driven Food Business Is Forgoing Big Grocery. Forbes.com.

Profiles Big Spoon Roasters’ decision to resist mainstream grocery distribution in order to preserve its mission and values. This article highlights the trade-offs small, purpose-driven companies face in scaling.

• Kim, S & Schifeling, T. The Paradox of Growing as a Values-Driven Company. Harvard Business Review.

Examines why growth can undermine the very values that give mission-driven firms their competitive edge. This piece provides a strategic framework for navigating purpose and performance tensions.

• Doctorow, C. (2024, April 4). Too big to care. Enshittification is a choice. Medium.com.

Argues that companies often degrade their products and values as they scale, not because it’s inevitable but because of strategic choices. This reading broadens the discussion of mission drift beyond food to digital platforms and other industries.

Class 9 - Investing for People and Planet

In this session, we will use the Norges Bank Investment Management case to examine how the world’s largest sovereign wealth fund integrates climate risk and ESG principles into its investment strategy. We will explore the fund’s governance model, responsible investment framework, and approach to balancing long-term returns with systemic sustainability challenges. Additionally, we will host Professor Fern Wickson of the University of Tromsø in Norway, who will describe her work on enabling transformative change for biodiversity.

Abridged Case Study: Norges Bank IM and the Politics of Sustainable Capital

Associated Readings

• Skjeseth, H. T. (2025, June 12). How sparsely populated Norway amassed $1.8 trillion. Bloomberg.

Explains how Norway built the world’s largest sovereign wealth fund from oil revenues, highlighting its scale, purpose, and global influence. This piece provides essential background for understanding Norges Bank Investment Management (NBIM).

• Mundy, S. (2025, May 28). How the world’s largest wealth fund looks at climate risk. Financial Times.

Analyzes how NBIM integrates climate risk into its investment strategy, including tensions between financial returns and sustainability leadership. This article illustrates the challenges of aligning capital markets with long-term climate goals.

• Skill-builder: ESG Investing: What Every MBA Needs to KnowLinks to an external site.. MBA-EDGE.com.

A practical introduction to environmental, social, and governance (ESG) investing, including frameworks, terminology, and tools. This resource equips you with the basics needed to evaluate funds like NBIM through an ESG lens.

• IPBES report: O’Brien, K., et.al. (2025). IPBES Transformative Change Assessment: Summary for Policymakers. Zenodo. Review the document at a high level - we will host a discussion with Fern Wickson, a renowned scholar from University of Tromso (Norway) in class on Friday.

Highlights global pathways for transformative change in response to biodiversity and sustainability crises. This report situates investment decisions within the broader context of planetary-scale risks and opportunities.

Deeper Dive (optional):

• Norges Bank IM report (2024): Responsible investment. Read pp. 4-82.

Provides a comprehensive account of how the world’s largest sovereign wealth fund integrates environmental, social, and governance factors into its investment strategy. This report offers students a firsthand view of institutional investor practices, priorities, and challenges in aligning capital with sustainability.

• Podcast: Norges Bank’s CEO Nicolai Tangen hosts a popular podcast called “In Good Company”, featuring conversations with some of the banks biggest investments. See his guest list here and listen to 1 podcast on a company or speaker that interests you.

Class 6: Understanding energy transition – and the future of fossil fuels

We started today’s class by exploring what the concept of Energy Transition means, and whether we have actually ever experienced it. Initially, students defined it as moving from fossil fuels to renewables. Today’s discussion pushed us to think more deeply about the dynamics of a changing energy system, and how this relates to climate, development, and geopolitics.

Signals of energy transition: We began by asking a simple question: how could we tell if an energy transition is happening? According to students, signals could include observing the behavior of big emitters, tracking global carbon emissions (PPM in the atmosphere), and monitoring demand shifts. Other signals are financial pledges, visible changes on the landscape (like new wind turbines), customer choices, the level of debate and contestation around the shift, and government incentives (like the IRA).

The Climate Story Premises: Energy transition has many causes and consequences, but climate change sits in the middle of this orbit. Before getting in to the details of energy systems, we laid out the foundational premises of the climate story:

1. Massive Energy Growth: Global energy use has skyrocketed in the last century, shifting from traditional biomass to coal, oil, gas, and nuclear.

2. Energy-GDP Link: Historically, energy use and GDP (a proxy for prosperity/well-being) are fundamentally linked. Satellite images dramatically illustrate this connection, showing vast differences in light density and GDP between countries like North and South Korea.

3. Atmospheric Chemistry Change: Emissions, primarily from fossil fuel combustion, are changing the atmosphere’s chemistry. Carbon concentration has risen from about 300 parts per million (PPM) in 1911 to over 420 PPM today. Notably, 40% of all human emissions have been emitted since 2000.

4. Temperature Rise: Global temperatures are increasing, now approaching the 1.5°C threshold above pre-industrial times (1850s). Scientific consensus is that 2°C is too risky, making 1.5°C the critical point for losing control due to uncontrolled positive feedback loops. Ocean temperatures have spiked recently, absorbing immense excess energy (equivalent to “about five nuclear bombs per second”).

5. Leaving the Goldilocks Zone: The last 12,000 years saw enormous temperature stability (”Goldilocks zone”), allowing civilization to flourish. All of our infrastructure, cities, food systems, and ways of life are premised on climate conditions (temperature, seasons, weather volatility) that have been in place during this period. In the last several decades, we have departed this zone and are headed toward planetary conditions that are increasingly incompatible with those systems.

Addressing Emissions: The steep incline in emissions since 2000 is driven by a multiplier effect of population growth, economic growth, and increased per capita emissions. The greatest impact on emissions reduction would come from addressing the carbon intensity of the lifestyles of the wealthiest. Since affecting population or limiting energy use is undesirable, the focus of change must be on changing technology, developing energy sources that support economic growth without the same emissions cost.

Historically, fossil fuels were the cheapest form of energy, but renewables are arguably at cost parity today. In the U.S., the biggest factor in the reduction of per capita carbon emissions over the last 15 years was the shift from coal to natural gas.

The Stabilization Goal: To stabilize the climate and sustain human prosperity, the energy system must be rapidly transformed. Getting to 1.5°C requires a steep trajectory of emissions reductions starting now. Crucially, to meet this target, the line drops below zero net emissions, meaning carbon removal (negative emissions) - machines that suck carbon out of the atmosphere - is required alongside mitigation.

Definition and Reality Check: Energy transition is defined as a long-term structural change in energy systems, such as a 10% drop in market share for an energy source. However, the data shows that global energy use is characterized by “energy addition”, not transition“, where new sources (like renewables) are added on top of existing fossil fuels, which are still growing. Transitions take a long time due to the massive scale of the system; for example, it took 75 years for natural gas to reach a 10% market share.

Energy transition is not driven only by a concern for climate change; in addition to reducing emissions, energy security, and improved national competitiveness are becoming the primary motive for change. It is also critical to understand that energy is a system, and the inter-connections between parts of the system are complex. For example, growing the sharing of EVs does not necessarily reduce emissions unless the electricity generation mix is also decarbonized. We also discussed the concept of rejected energy (waste lost during conversion), which highlights system inefficiencies.

Geopolitical and Energy Discussion with Patrick Duddy

After a short break, the class welcomed Patrick Duddy, former U.S. Ambassador to Venezuela and a Duke faculty member. Duddy shared his background, including his tenure as Deputy Assistant Secretary of State responsible for hemispheric energy issues, and the unique experience of being formally declared persona non grata and expelled from Venezuela, only to return later for a second term.

Guyana Oil Boom and Geopolitical Shift: Duddy reported that nearly everyone in the Latin American energy world is “blown away” by the oil discoveries in Guyana. Guyana’s production is approaching 900,000 barrels a day, potentially catching or exceeding Venezuela’s production, and Suriname is also expected to join the party. This boom has displaced former major players like Colombia, which has seen declining supply.

Countries like Mexico and Venezuela are seeing production declines due to factors like depleting traditional fields, a lack of capital investment, and the diversion of oil income to social programs.

Guyana’s Strategy and Risk: Guyana, historically desperately poor and losing population, has seen oil revenues transform its economy, with a 40% GDP growth rate last year. Its strategy to “pump fast, decarbonize fast” is viewed by Duddy as “mostly a slogan,” as their immediate interest in energy transition is not urgent.

Guyana chose to partner with International Oil Companies (IOCs) like ExxonMobil and Chevron/Hess because it lacked the money, expertise, and infrastructure to develop the sector itself. Exxon has invested approximately $40 billion in Guyana so far, and has earned back around $38 billion. Long-term, this investment will be very profitable for Exxon, though a major risk facing Guyana is resource nationalism, which could surge if the windfall profits are not adequately shared with the public.

Geopolitical Protection and Venezuela: The presence of ExxonMobil and Chevron is a key part of Guyana’s national security strategy. Venezuela has openly coveted Guyana’s oil, holding a 2023 referendum on annexing more than half of Guyana’s territory. Having American majors there gives the U.S. a “big stake” in Guyana’s sovereignty, providing a geopolitical buffer against Venezuelan aggression, evidenced by the presence of British and U.S. ships and modest military exercises after Venezuelan threats.

The shift in oil supply from Venezuela to trusted partner Guyana, a member of Caricom (Caribbean Community), is highly significant for the entire Caribbean region, especially those dependent on oil for electrification and fuel.

U.S. Administration’s Stance: Duddy noted that the current U.S. administration favors the US and its closest allies producing more fossil fuels. There is skepticism within the administration about inhibiting the development of the fossil fuel industry.

Comparison to the British East India Company (B.E.I.C.): The case prompted discussion about whether Exxon’s involvement is a modern form of colonialism. Similarities noted included the involvement of a foreign entity in building infrastructure and maintaining influence. Differences included that India started rich, while Guyana started desperately poor. Exxon relies on a business partnership (and U.S. protection), unlike the B.E.I.C., which had its own mercenary force. While the B.E.I.C. destabilized India, resulting in partition, the outcome of Exxon’s long-term presence is still unknown.

Ultimately, the lessons from Strategy 627 reveal that geopolitical interests shape every major energy choice. Whether a country chooses supply-side dominance (as favored by some U.S. strategic thinkers skeptical of inhibiting the fossil fuel industry) or embraces transition depends heavily on its internal political ideology and external security threats. Success in navigating the sustainability mandate requires integrating climate action directly into national security, economic competitiveness, and corporate capital planning.

Class 7: EU’s Green Dream Denied: Why Geopolitics Just Rewrote Corporate Sustainability Strategy

Today’s class discussion was anchored by Patrick Moloney, a Copenhagen-based sustainability expert with Ramboll Consulting (joining virtually). Patrick has become a valuable partner for Duke, speaking with several groups of students in Durham and in Europe over the last couple years. His insights are shaped by constant contact with people close to the EU policy conversation, and the companies that are coping with the ever-evolving requirements.

For years, the European Union (EU) stood as the center of the universe when it comes to sustainability standards and regulations, relentlessly driving the state of the art globally. This dominance meant that any global company—whether Asian, American, or purely European—had to abide by the EU’s rulebook if they wished to operate in this critical market.

The result was a massive, accelerated regulatory push defined by immense compliance burdens. Suddenly, approximately 50,000 European companies were brought into scope, facing requirements like the Corporate Sustainability Reporting Directive (CSRD), which demands up to 1,100 data points. The sheer complexity and resulting costs created a bizarre paradox: sustainability budgets were allocated not to action, but to “feeding this beast” - hiring consultants and paying auditors to ensure compliance.

This energetic regulatory ambition, however, has collided violently with geopolitical and economic reality, resulting in a systemic pushback. The EU’s ambitious green policies, once considered unassailable, are now giving way to a new priority: economic security, resilience, and strategic autonomy.

The consequence for global business leaders is “policy whiplash”. Companies that were ambitious and “over prepared” are now being forced to decelerate their spending, sparking frustration among boards for what is perceived as wasted capital. Sustainability is not dead, but its focus has been profoundly redefined, becoming intensely selective and prioritized.

The Geopolitical Engine: Six Spinning Washing Machines

The pivot away from pure regulatory ambition is driven by external volatility. Issues once viewed as abstract, distant “macro” concerns are now immediate, very local corporate problems. This environment of uncertainty has been likened to being surrounded by “six washing machines” all spinning simultaneously, with no clear end date:

1. Shifting Dynamics of European Politics European politics are in a volatile state due to the rise of right-leaning parties, which can quickly shift the composition and focus of the EU Parliament. Politicians operate on four-year electoral cycles, leading to populist moves and providing political cover to avoid tough sustainability decisions that conflict with economic interests. While political sentiment swings, core climate ambition and human rights elements remain legally embedded within EU law, ensuring they will win out over time, though subject to current political dynamics.

2. Economic Performance and Industrial Health The EU is now filtering sustainability through a “competitiveness filter”, driven by the need to re-industrialize and achieve strategic autonomy, as articulated in the Draghi Report. This shift mandates focus on the Clean Industrial Deal, prioritizing support for heavy-emitting industries (like steel and cement) to decarbonize, reducing reliance on non-EU countries. Circularity is being relabeled as “Resource Autonomy,” becoming a high priority because the EU must retain critical minerals (like copper, lithium, and cobalt) in the loop to support electrification and energy security due to scarce natural resources.

3. Geopolitical Shocks and Supply Chain Realignments Geopolitical shocks are making previously abstract issues intensely “very local” for companies, requiring them to incorporate threats like US tariffs, Russian aggression, and supply chain volatility into immediate planning. The energy system is defining a new global order, potentially aligning certain major powers around “fossil centricity” versus a decarbonization platform. Supply chain issues regarding critical resources, alongside increased security risks from potential sabotage (cyber and physical) to critical infrastructure like pipelines and energy grids, are now forcing a fundamental focus on corporate resilience.

4. Finance and Capital Market Regulation Shifts The scale of the European Green Deal required the Sustainable Finance Action Plan to channel private funds (estimated at over €1 trillion) into the transition, leading to regulations like the CSRD and the EU Taxonomy. CSRD notably enforces transparency by requiring companies to disclose the annual financial allocation of resources toward achieving their sustainability goals, serving as a check against “tokenism”. Capital markets are inherently risk-averse, meaning high volatility in areas like climate and supply chains deters investment, reinforcing the necessity for companies to mitigate risks to attract capital.

5. Public Sentiment and Social License A paradoxical situation exists where public sentiment regarding sustainability is declining, evidenced by recent reports, while the demand for maintaining a strong social license to operate is simultaneously increasing. Regulations like the CSRD enforce a high social realm that mandates companies disclose issues such as the gender pay gap and address their impacts on affected communities and consumers, going well beyond internal workforce concerns. This focus was triggered, in part, by the Taxonomy revealing that few companies had adequate standards regarding human rights and responsible business conduct in their value chains.

6. Climate and Environmental Events Climate change is not receding as a priority because physical risk (such as extreme heat, forest fires, and flooding) is now undeniably real and constitutes a “no brainer” for financial investors. These events trigger “reactive politics” and urgent adaptation measures, exemplified by Spain passing new resilience laws after severe environmental disasters. For critical entities like energy utilities, strategy must now focus heavily on resilience—the ability to maintain continuous service under stress from climate hazards—because climate laws and decarbonization targets remain embedded in European legislation.

The Competitiveness Filter: The Rise of the Clean Industrial Deal

A recent shift in ideology in Europe was formalized in the Draghi Report, which identified regulation as a “hindrance” to the EU’s ability to re-industrialize and achieve strategic autonomy. This mandated that sustainability must now be filtered through a competitiveness filter.

The new regulatory focus is the Clean Industrial Deal, which is highly selective and aggressive only in priority areas. Investment now flows almost exclusively toward three strategic imperatives, creating a “Fast Lane”:

1. Resilience: The New Black

For critical entities (such as utilities and energy grids), “resilience is the new black”. The focus is on ensuring the ability to “continue service under duress”—be it from climate-related hazards, sabotage, or supply chain disruptions (like copper supply for grids).

This imperative is driven by new legal mandates like the Critical Entities Resilience Directive (CER), forcing organizations to look at their societal obligation. Strategy for these firms must center entirely on preventing disruption.

2. Resource Autonomy (The Relabeled Circularity)

The circular economy, once an environmental concern, is now being aggressively relabeled as “Resource Autonomy”. Since the EU lacks essential natural resources (such as oil reserves and critical minerals), the green transition—electrification, building more batteries, turbines, and PV—is the “only solution” to energy security.

This requires companies to focus aggressively on retaining critical materials (lithium, cobalt, copper) “in the loop” to reduce reliance on geopolitically unstable markets.

3. Monetization (Decarbonizing Heavy Emitters)

Investment is prioritized for energy-intensive sectors (like steel, cement, built environment) that are critical to the EU’s ability to re-industrialize and achieve competitiveness. These sectors are the “problem makers” and must be targeted for decarbonization, while “problem solvers” (like green energy providers) are promoted.

The Slow Lane and Fragmentation

Issues that do not directly contribute to EU competitiveness—such as textiles (fast fashion), packaging, and electronics—are pushed into a “slow lane”. Legislation concerning these sectors is delayed or deprioritized.

This creates fragmentation: while progressive member states (like Sweden and Finland) “plow ahead,” others (like Hungary and Poland) will likely fall behind. Multinational companies that once relied on EU law as their guiding star are now forced to revert to navigating 27 different individual laws, creating a significant management headache.

What Leaders Must Do Now: Strategic Principles for Uncertainty

In this environment, sustainability professionals are experiencing a “period of grief” and boards can no longer operate on linear projections. Strategy must be anchored by principles designed to guide through extreme uncertainty:

1. Scenario Planning and Stress Testing: Instead of simply projecting forward from the present, companies must draw different versions of a volatile future (e.g., a 2035 scenario where the EU has “no access to lithium”) and “work backwards” to determine necessary actions.

2. Align Commitments with the Business Model: Any commitment must align with the core business model to be viable during uncertainty. As one student noted, in times of uncertainty, companies “don’t [need to] be relying on the government or anyone else”.

3. Financial Transparency as a Check: CSRD, even diluted, forces companies to be transparent about their annual financial allocation of resources toward sustainability goals. This mechanism serves as a check against tokenism, ensuring that targets aren’t just set, but genuinely funded.

4. Focus on External Risks and Opportunities: Boards must look beyond internal operations and focus on customers (downstream opportunities) and upstream suppliers (risk mitigation).

The climate fight is not ending, but its purpose has been merged with a more fundamental goal: economic survival. The pushback is real, but the legal reality - that climate ambition and human rights are embedded in EU law - will win out, albeit on a new, highly competitive, and pragmatic timeframe.

Leave a comment

Share

Subscribe now

A central theme was whether corporate strategy should be measured by relative efficiency (e.g., reducing water use per ton) or absolute limits (operating within Earth’s ecological budget), a logic increasingly demanded by regulators and investors. Ultimately, the session stressed that agribusiness faces mounting pressure to translate global scientific thresholds—like the breach of six Planetary Boundaries—into specific capital allocation decisions, forcing a choice between cautious compliance and a bold, science-based "nature-positive" commitment.

The class also included a discussion with guest speakers Alison Taylor (ADM) and Rob Coviello (Bunge), both leaders in the food supply chain sector.

1. Debrief on Prior Assignment/Discussion (Coca-Cola/Water): The class began by reflecting on the takeaways from the previous Coca-Cola discussion:

· Systemic Issues: While “Replenishment” moved water stewardship past simple internal metrics, it still did not fully address system issues. One student noted that the company’s business model creates an inherent tension between volume growth and resource scarcity.

· Measurement Challenges: Students noted that metrics provided by companies are often inadequate, as much of the true water usage (up to 200 liters per liter of product when considering sugar production) is "up the supply chains". The risk assessment performed by the instructor showed that internal metrics around efficiency and wastewater treatment were often less impactful than external, systemic risks.

· Governance and Legitimacy: Issues of power, legitimacy, and the corporate role in governance are urgent, particularly when governments are weak. The discussion raised the question of how long companies maintain the social license to operate in places where they directly compete with basic human rights and needs.

· Accountability: The class discussed how, in distributed global systems, impacts (e.g., negative externalities in the supply chain, like labor issues) are not necessarily aligned with accountabilities, though a brand like Coca-Cola is held responsible.

· License to Operate: A company’s "license to operate" now requires not just meeting expectations but actively shaping stakeholder expectations.

2. Planetary Boundaries 101 and the External Landscape: The session established the foundational concepts driving the pressure on companies like GrainCo.

· Planetary Boundaries (PBs): The PB framework defines a "safe-operating space" for humanity using nine Earth-system processes. Updated assessments show that six of the nine boundaries are already transgressed. A key concern for GrainCo is the boundary for Biogeochemical flows (Nitrogen and Phosphorus), which is already breached globally.

· Tipping Points: The framework is built on the idea of tipping points, thresholds beyond which the risk of irreversible, nonlinear, and exponential disruption rises sharply.

· Shifting Measurement: PBs facilitate a move from relative efficiency to absolute ecological accountability, asking if a business is operating within Earth’s ecological budget. Investors, regulators, and NGOs increasingly want the latter.

· GrainCo’s Hotspots: GrainCo’s profitability is tied to fragile natural systems across several breached boundaries:

• Land-system change (soy expansion in the Amazon/Cerrado, risking stranded assets and EU import bans).

• Biogeochemical flows (fertilizer runoff in the U.S. Corn Belt, risking litigation and water-treatment fees).

• Biodiversity loss (pollinator decline for cocoa and canola, leading to yield volatility).

3. The Regulatory and Competitive Tide: The external landscape has shifted, moving nature from philanthropy to a compliance item.

· TNFD v1.0: The Taskforce on Nature-related Financial Disclosures framework requires disclosure across four pillars. TNFD’s LEAP approach demands companies locate assets spatially, evaluate dependencies, and assess financial materiality. GrainCo can trace only about 40% to 52% of its Brazilian soy, far short of the 100% required by LEAP for verification.

· Science-Based Targets for Nature (SBTN): SBTN guidance for agri-food suggests aggressive goals: no ecosystem conversion after 2025 and regenerative practices on at least one-fifth of managed hectares by 2030.

· Competitive Pressure: Peers are moving fast. Bunge has pledged zero-deforestation soy in South America by 2025, and Nestlé plans to source 50% of key ingredients via regenerative agriculture by 2030. GrainCo lags, placing firmly in the lower-middle of the pack.

4. GrainCo’s Strategic Fork: Risk, Value, and Headwinds: GrainCo, which sources approximately 25 million hectares of global cropland, faced three distinct pathways:

The consensus among students heavily favored Option B (Pragmatic “No-Regrets” Path). Option A (Status Quo/Compliance) was noted to be highly undesirable, but the instructor posited that it might be harder to dismiss in a corporate boardroom deeply embedded in the industry.

5. Key Decision Criteria: Students and the instructor identified several criteria for evaluating these options: strategy alignment, competitive position, financial efficacy, stakeholder access, ease of implementation, and compliance with absolute Planetary Boundaries.

6. Implementation Headwinds: Execution of any bold plan faces significant constraints within GrainCo’s system:

· Supplier Resistance: Brazilian soy farmers often operate on razor-thin margins and threaten to shift supply to rivals if strict demands are made without guarantees.

· Stranded Assets: Zero-deforestation mandates clash with the depreciation schedules of crushing plants that rely on frontier soy to stay full.

· Data Verification: A full verification program for commodities like soy, palm, and cocoa is estimated to cost US $25 million annually.

· Culture Shock: Trading desks, which are paid to capture basis spreads, resist demands to reject cargoes without geocoordinates, arguing it cedes market share to rivals.

Fireside Chat: Insights from Industry Leaders: The second half of the class featured a fireside chat with Rob Coviello (Bunge CSO) and Alison Taylor (ADM CSO/Duke Climate Fellow). Topics of conversation included:

A. P&L Implications of Nature (Why Nature Matters): The speakers addressed how nature translates from a "touchy feely" topic to P&L implications:

· Stranded Assets and Volatility: Rob Coviello noted that Bunge's core risk comes if nature changes the productivity of crops grown where their assets (e.g., manufacturing facilities) are located, potentially creating stranded assets. He argued that the ability to arbitrage and reconfigure supply chains during major shocks (like war or extreme weather) is a crucial part of the company’s value process; their network is most valuable when conditions are volatile.

· Regenerative Agriculture as P&L: Alison Taylor highlighted that ADM’s regenerative agriculture programs have a clear P&L benefit. They created margin by paying premiums to farmers and passing those costs along to customers via long-term contracts (e.g., 7.5 years), which provided "stickiness" and supply chain resilience.

· Deforestation as Compliance: Conversely, Alison noted that addressing deforestation is driven more by compliance and investor outcry because the company cannot calculate and put the carbon sequestration benefit of forest preservation into its own inventory.

B. Stakeholders: Farmers and Control: The discussion emphasized the distance and lack of control agribusiness firms have over their upstream and downstream stakeholders:

• Farmer Autonomy: Alison explained that farmers are not beholden to a particular company; they sell their crop to the company offering the best price and often lack long-term contracts. Therefore, companies rely on carrots (incentives like premiums) rather than sticks (mandates). The private sector does not have the money or authority to compensate farmers for loss of livelihood if they are asked not to convert land (no-conversion pledges), making policy intervention necessary.

• Farmer Risk: Rob Coviello stressed that farmers are the number one stakeholder and are highly entrepreneurial but assume enormous risk (spending 80% of costs upfront, uncertain yield/price nine months later). He argued that telling a farmer what to do with their generational land asset is a bad start.

• The Demand Signal: Both noted that downstream customers clearly demand products with a lower carbon footprint. However, companies are often creating these demand signals (by paying premiums for regenerative practices) without assurance that the end consumer will pay, treating it as R&D.

C. Pre-Competitive Collaboration: The speakers discussed the necessity of pre-competitive collaboration—working with competitors on non-commercial aspects.

• Purpose: The goal is to reduce costs and scale initiatives (e.g., technology like satellite imaging) and ensure common understanding to avoid chaos and gaming the system.

• Guardrails: Collaboration must avoid commercial topics (like price). Legal restrictions prevent the creation of a "blacklist" of suppliers.

• Examples: Competitors must agree on common definitions, such as what constitutes a "forest," when setting deforestation standards, otherwise, the goal of stopping deforestation would be undermined.

D. Board Dynamics and Climate Goals

· Board Makeup: Rob Coviello noted the advantage of having a strong ally like Carol Browner (former EPA head) chairing Bunge's sustainability committee.

· Nervousness: Both CSOs felt an underlying nervousness on their boards, which shifted from worrying about not doing enough to worrying about saying too much (“greenhushing”) and being caught in the crosshairs of activists or regulators.

· Shifting Timelines: Alison Taylor provided context for companies "backing off" 2025/2030 goals: many companies set ambitious, competitive goals without fully calculating their suppliers' capability. When the "reckoning" arrived, they realized the goals weren't feasible without offsets, forcing them to come forward and ask for more time, rather than necessarily backing off due to political pressure. Rob Coviello agreed that goals may be rolled out by two or three years later than first promised, due to imperfect data and changing economic assumptions.

Leave a comment

Share

Subscribe now

Class 6 - Acting on Climate and the Future of Fossil Fuels

In this session, we will use the Exxon in Guyana case to explore the tensions between acting on climate goals and pursuing high-margin fossil fuel projects. We will analyze ExxonMobil’s strategy in the context of a shifting global energy landscape, Guyana’s rapid transformation into a petrostate, and the geopolitical and economic stakes of new oil development. The session will combine a case discussion with insights from guest speaker Patrick Duddy and a debrief of the Net Zero simulation to connect corporate strategy, national interests, and climate imperatives.

Abridged case study: Exxon in Guyana: Advantaged Barrels in a Carbon-Constrained World

• Assignment questions are at the end of the case.

Readings:

• Podcast: Business Breakdowns “Exxon Mobil: An Aging Energy Empire.” (78 min)

Provides a deep dive into ExxonMobil’s business model, financial dynamics, and strategic challenges as the company navigates the energy transition. This episode sets the stage for analyzing the future of fossil fuels.

• Bousso, R. (2025, March 26). Welcome to the age of Big Oil's managed decline. Reuters.com.

Examines how oil majors are confronting declining demand, rising investor pressure, and the prospect of managed decline. This article highlights the structural forces reshaping the sector.

• Paddison, L. (2025, March 29). A tiny rainforest country is growing into a petrostate. A US oil company could reap the biggest rewards. CNN.com.

Profiles Guyana’s rapid transformation into an oil-producing state and the outsized role of ExxonMobil. This piece raises questions about development, dependency, and environmental costs.

• YouTube video (2023): President Dr Irfaan Ali talks oil & gas & how it is transforming Guyana's economy on BBC's HardTALK. (24 minutes)

Offers Guyana’s political perspective on oil as a driver of economic transformation. The interview provides valuable insight into how host governments frame opportunities and defend trade-offs in the fossil fuel era. Don’t miss the argument around 16 minutes!

Class 7 - Assessing Sustainability Performance, Materiality, and the Alphabet Soup

In this session, we will focus on how companies assess sustainability performance and navigate the complex “alphabet soup” of reporting and regulatory frameworks. Drawing on recent analyses of climate and sustainability regulations in the U.S. and EU, we will explore how evolving disclosure requirements shape competitiveness and corporate strategy. Guest speaker Patrick Moloney from Ramboll will lead an interactive workshop to deepen our understanding of materiality, compliance, and best practices for aligning sustainability reporting with business value.

No Case Study today

• Full 2 hours will be a workshop on sustainability metrics, led by Patrick Moloney, Ramboll Consulting (virtual)

Readings

• Sasfai, B., et.al. (2025, Jan 26). Climate and Sustainability Regulations: 2024 End-of-Year Review. Harvard Law School Forum on Corporate Governance

Reviews the major global regulatory developments in climate and sustainability from 2024, highlighting implications for corporate governance and disclosure. This piece provides a snapshot of the fast-changing regulatory environment businesses must navigate.

• Brennan, S., et.al. (2025, April 30). EU 2025 Sustainability Regulation Outlook: Unlocking competitiveness and growth.

Analyzes how the EU’s latest sustainability regulations are reshaping corporate strategy, competitiveness, and reporting practices. This article offers forward-looking insights into the business opportunities and compliance challenges these rules create.

Leave a comment

Share

Subscribe now

If you have been following our Climate, Sustainability, and Corporate Governance (CSCG) class so far, you know that we have covered the foundational concepts of corporate sustainability and are now exploring the fundamental areas of focus in the field: energy, climate, food, and water. After an initial debrief of the previous session, this class session focused on water scarcity and risk management by emphasizing the river basin as the correct unit of analysis, rather than a single facility.

The discussion moved from first principles—such as the success of upstream conservation efforts in the Catskills—to complex managerial choices related to governance and corporate strategy, including a review of the Coca-Cola case study. A central theme was that water issues are driven by institutional failures and misaligned incentives as much as by hydrology, leading the class to conclude that effective strategies must shift capital upstream toward watershed restoration and collaborative action, away from simple, often inadequate, plant-level metrics. Ultimately, the session stressed that corporate sustainability requires adopting system-level logic and governance structures to address shared basin problems.

The class session served as a transition from the prior climate risk discussions to focusing on water as a fundamental, strategic business issue, using Coca-Cola's water strategy as a core case study. The session moved from debriefing a complex organizational transition exercise to establishing the local nature of water risk, the water-energy nexus, and finally, analyzing corporate managerial choices in high-stress basins.

1. Debrief on Prior Assignment (EcoManufex)

The instructor noted that the students outperformed expectations on the previous multi-stage transition plan assignment for the fictionalized company EcoManufex. The exercise's main highlight was demonstrating that climate risk includes both physical and transition risk and requires rigorous analysis to evaluate risk reduction and value creation opportunities. On the way out of class, student A observed that she was stunned by the degree of “business integration” now required of firms to compete in a climate-impacted market environment, noting that the rhetoric is moving past the "feel good qualitative" stage.

Student Insights on the in-class exercise:

· Role-Playing and Perspective: Student B found it intriguing that people made convincing arguments based on their assigned role, even if they did not personally believe in the argument. Student C, who played the CEO, felt the challenge of hedging different opinions and trying to find the best compromise while acknowledging that "nobody was really going to be happy" with the full path taken.

· Internal Dynamics: Student C also noted that procurement was central to their business’ decisions, suggesting internal dynamics that influenced which opinions mattered more.

· Marginalization: Student D, who played the employee representative noted feeling marginalized, observing that by the end, "no one was listening".

· Shifting Priorities: Student E (the CEO with a technical background) was initially focused strictly on profit and delivery but was surprised and found it interesting to hear the justifications from the Chief Sustainability Officer (CSO) regarding carbon footprint.

· Clarity and Collaboration: Student F highlighted that the CEO setting clear priorities upfront (three priorities) helped the group collaborate better and return to the goal when they got off track.

· Physical and Transition Risk: The class established that the heart of the prior assignment was quantifying climate risk. This involves recognizing that most people conflate climate risk with only physical risk (e.g., floods, droughts, supply chain failure).

· The instructor emphasized that the "lurking risks" are often multiple times more important, specifically transition risk, including policy and regulation (like the EU's Carbon Border Adjustment Mechanism or California's Cap-and-Dividend system), technology and market shifts (e.g., EVs replacing internal combustion engines), and investor/consumer pressure (e.g., BlackRock integrating climate risk).

· Emerging Risks: The discussion also touched on liability and litigation, noting that attribution science now allows linking specific company emissions to the cost of certain disasters (like Hurricane Helene).

· Carbon Leakage: Student G raised the macroeconomic concern that stringent accountability in Western countries could lead to "carbon leakage," shifting dirtier manufacturing to places where it is not accountable.

· Individual Accountability: Finally, students debated individual-level carbon accountability (taxes or carbon budgets). The instructor shared the perspective that many problems are system problems, not individual choice problems, citing the small emission reduction during the COVID-19 pandemic shutdown.

2. Water Stress in Context (Water 101)

The class started by establishing the foundational facts about water distribution, scarcity, and use:

· Water Abundance vs. Scarcity: While the Earth is abundant (70% water), only 2.5% is freshwater, and two-thirds of that is locked in glaciers.

· Water Distribution and Governance: Student H highlighted the problem with national averages, noting that water is fundamentally local and requires granular, sub-national data.

· Student I from Lima shared a powerful example of unequal distribution, where some neighborhoods have water only three days a week or must pay trucks high prices for delivery.

· The instructor shared the New York Catskills example as a template for managing risk, where investing in upstream farm interventions was far cheaper and more durable than building a multi-billion dollar end-of-pipe filtration system.

· The class discussed how power and governance override hydrology, citing Phoenix, Arizona, a desert city that imports massive amounts of water, versus Ethiopia, where 84% of the Nile originates, yet faces massive water scarcity due to lack of financial resources and infrastructure.

· The collapse of the Aral Sea was used to illustrate how policy (a shift to irrigation-intensive cotton farming after the Soviet Union dissolved) drove water scarcity, rather than natural factors,

· Water Use and Embedded Water: About 70% of human water use is in agriculture. The concept of embedded water was introduced, noting that dry products (like a cotton T-shirt or a car) require massive amounts of water to produce the raw materials.

3. Water vs. Oil Discussion

The in-class breakout session explored the popular comparison of water to "the oil of the 21st century".

4. The Coca-Cola Case: Strategy and Managerial Choices

The class focused on Coca-Cola's shift from being defensive about its water use (following protests in India) to proactively managing water as an enterprise risk. The core insight derived from Coca-Cola's risk assessment was that the plant efficiency metrics they traditionally tracked ranked low in explanatory power for true enterprise exposure.

· Risk Assessment and Strategic Shift: The new strategy was driven by risk modeling that quantified six categories of risk (Watershed, Supply Reliability, Efficiency, Supply Economics, Compliance, Social/Competitive). This systemic analysis provided the "empirical basis for deciding where you going to put finite resources to try to mitigate risk". The overarching strategy involved intervening at the watershed level—moving "upstream"—to benefit not just the company but also the community ("leaky" interventions).

· Debating Water as a Human Right: The class watched excerpts from a film featuring the former CEO of Nestlé, who argued that water should have a value (not be a human right).

· Student J pointed out the business reason for this view, noting that the CEO's portfolio is large chunk of bottled water.

· Students argued that water is necessary for the right to live in dignity and that if it is a human right, governments should be mandated to provide it.

· The key tension articulated was: If water is a human right, who bears the responsibility and cost of delivering it across distance and infrastructure?

· Student K suggested that Coca-Cola, as a large brand, has the power to push for water availability while still running its business.

Strategic Options Analysis and Student Insights:

The class assessed three strategic options for Coca-Cola's future water strategy:

1. Replenishment 2.0: Focus volumetric projects in the same watersheds where extraction occurs ($40M cost).

2. Upstream Shift: Redirect resources to farmer practices, irrigation efficiency, and watershed restoration in critical sourcing regions ($150M cost).

3. Science-Based Targets (SBTN): Pilot basin-context targets, requiring new data systems and local partnerships ($75M cost).

The consensus (center of gravity) among students favored the Upstream Shift path, prioritizing basin-level and supplier interventions.

Areas of Productive Disagreement (Student Divergences): The student responses surfaced several key managerial debates:

The ultimate composite operating picture concluded that credible action requires ranking basins by business exposure, shifting capital upstream where risk-reduction per dollar is greatest, and ensuring governance provides the spine (board visibility, incentive links, and clear ownership splits).

Leave a comment

Share

Subscribe now

Class 4 - Managing Risk and Scaling Impact

In this session, we will use the Coca-Cola Water Stewardship case to explore how companies can manage critical resource risks while creating shared value. We will explore how risk assessment can provide a robust case for corporate action on environmental challenges and their operational, reputational, and regulatory implications. An introductory talk on global water challenges will set the stage for our case discussion, where we will analyze how high-level sustainability commitments can be translated into measurable, scalable impact.

Case Study: Coca-Cola’s Water Strategy as Risk Management and Global Stewardship

Source: Trellis

Readings:

• Porter, M. E., & Kramer, M. R. (2011). Creating shared value. Harvard Business Review. Read pp. 4-17.

Introduces the concept of shared value, arguing that companies can unlock economic success by addressing social and environmental challenges. This article frames how firms like Coca-Cola link water stewardship to competitive advantage.

• Coca-Cola web site: Coca-Cola Unveils 2030 Water Security Strategy

Outlines Coca-Cola’s global water goals and initiatives through 2030, including replenishment, efficiency, and community partnerships. This document provides a corporate perspective on how one of the world’s most recognized brands is managing water risk.

• MBA EDGE: Water & Business: What Every MBA Needs to Know

Explains why water is both a critical input and a strategic risk for business, with attention to scarcity, pollution, and stakeholder concerns. This resource equips you to analyze corporate water strategies with a management lens.

Class 5 - Feeding the World Within Planetary Boundaries

In this session, we will use the GrainCo case to examine how agribusiness can operate within planetary boundaries while meeting global food demand. Guest speaker Alison Taylor, formerly Chief Sustainability Officer at ADM, joined by Rob Coviello from Bunge, will share insights on how the agricultural companies integrate nature-based considerations into its operations and governance to balance growth with ecological limits.

Case Study: GrainCo’s Bid for Nature-Positive Growth

Source: Nature

Readings

• Taskforce on Nature-related Financial Disclosures. (2023). Recommendations of the Taskforce on Nature-related Financial Disclosures. Read pp.7-39.

Provides guidance for companies and investors on identifying, assessing, and disclosing nature-related risks and opportunities. This framework is central to understanding how financial markets are beginning to integrate biodiversity and ecosystem impacts.

• Research Brief: “Doing Business Within Planetary Boundaries” (SRC & Beijer Institute, 2024). Read pp.7-32.

Explores how firms can align strategies with ecological limits while maintaining competitiveness. This brief offers a science-based foundation for evaluating corporate sustainability claims against planetary boundaries.

Leave a comment

Share

Subscribe now

Recap of Previous Class

We began by revisiting the materiality mapping exercise from last class. Student teams were assigned companies from different industries, and were asked to identify the 5 most material issues for their target company, along with the 3 stakeholders they would want to engage first in a materiality mapping process. In the debrief, 2 teams shared their highlights:

• Mars Team 3 highlighted the complexity of food and pet-food supply chains, from labor and deforestation to packaging and labeling, showing how prioritization is always fraught.

• L’Oréal Team 10 emphasized design, packaging, and raw material sourcing, while also surfacing issues around human rights, marketing, and consumer perception.

Comparing student maps to the companies’ published assessments sparked debate about why palm oil ranks where it does, why allergens and chewing gum appear on Mars’s map, and whether shareholder value itself belongs on a materiality chart. We concluded that shareholder return isn’t another materiality issue, but is an outcome of managing all of the other issues well.

Guest Perspective

We welcomed Sunil Rana, founder and CEO of Vyzrd and author of the EcoManufex case, who reframed sustainability as core to corporate value creation—not glossy posters, but grounded strategy planning tied directly to financial outcomes. He described the “tripod” of corporates, financial markets, and regulators, and the importance of linking climate strategy to shareholder incentives for systemic change.

In one of his more provocative assessments, Sunil showed how a cement manufacturer could lose nearly half of their market value over a 15-year horizon if it doesn’t actively address its transition risks. He also noted that most companies focus on physical risks, despite these making up only a small portion of their overall exposure.

EcoManufex Breakout Exercise

Student teams then went to breakout sessions to analyze the EcoManufex case, and develop short-, medium-, and long-term transition plans. The company, a manufacturer of industrial equipment, faced pressure from investors, customers, and regulators to deliver a credible net-zero plan while grappling with aging plants, shrinking margins, CBAM surcharges, and climate shocks.

Teams were asked to design a 1-year, 3-year, and 5-year transition strategies, based on the input and interests of several students playing roles such as CEO, CFO, CSO, CMO, investors, employees, and activists. The CEO made the final call, but plans had to reconcile multiple stakeholder voices.

Common moves across teams:

• Immediate wins through energy audits, efficiency retrofits, and renewable PPAs.

• Public net-zero roadmaps, TCFD-style disclosure, and audited plant pilots to build credibility.

• Supplier scorecards and contract clauses to address Scope 3 and tender eligibility.

• Significant R&D boosts (10–15% of revenue) for low-carbon product lines.

• 2030 targets clustered at 20–30% emission cuts.

Divergences:

• Pacing vs. ambition—some opted for bold early investments and short-term losses; others sequenced more cautiously, stressing perception management.

• Supply chains—some rebalanced toward Europe, others defended Asia sourcing but with stricter oversight.

• Capital allocation—some doubled down on retrofits, others prioritized capacity expansion.

Creative ideas: marketing as strategy (“Innovate today for resiliency tomorrow”), linking executive pay to carbon KPIs, and treating workforce upskilling as central to credibility.

Takeaway

The EcoManufex exercise underscored how climate strategy is inseparable from core business strategy. Students experienced the real tensions of balancing investor expectations, operational realities, regulation, and long-term survival, and had to make tough choices based on many differing opinions and interests. The strongest plans combined quick efficiency gains, credible disclosures, and product innovation with investor and employee alignment. On the way out of class, one of the students noted that she… “was really surprised and impressed by the level of business integration of sustainability” demonstrated in the EcoManufex case.

Next week, we will continue the exploration of risk management by studying the Coca-Cola case on water stewardship. Stay tuned…