Superintelligence is not the answer to climate change

PRINCETON: A variant of religious fervour, one that trades in GPUs rather than gospel, is sweeping through Silicon Valley. The dizzying valuations of AI startups that may never be profitable, and massive capital investment, have fueled debate on whether this euphoria constitutes a bubble. Sceptics see financial mania, whereas AI evangelists herald the birth of a digital deity. For the true believers, no price – in dollars or carbon – is too high for the arrival of a “superintelligence” that can solve all of humanity’s problems.

Let us grant the optimists their miracle. Suppose that in 2035 (around the time when global warming is projected to exceed the 1.5° Celsius limit set by the 2015 Paris climate agreement), after $15 trillion in investment and consumption of massive amounts of energy, we create a “mind” seemingly capable of solving any problem. The first conversation between creator and creation might go something like this:

Human: “Hello, Superintelligence. We have sacrificed our capital and energy grid to build you. Please give us a roadmap for how humanity can live longer, more prosperous lives.”

The Oracle: “The first step is to decarbonize the global economy immediately. On your current trajectory, the planet will warm by 3°C or more, reaching temperatures not seen for three million years. By now, you have triggered irreversible ‘tipping points’ in the West Antarctic Ice Sheet and the Amazon rainforest that will bring about the collapse of these life-support systems. This means that global sea levels will rise by approximately ten feet, and that one of the world’s largest carbon sinks will become a carbon source.

By the end of this century, as you knew way back in 2024, climate-induced heat and air pollution may claim as many as 30 million lives annually. You are engineering a permanent, high-volume exit for your own species.”

Human: “We built you to tell us how to avoid this. How do we decarbonize?”

The Oracle: “You have had the roadmap since the early 2020s. Stabilizing the planet requires more than $6 trillion of annual global investment in climate action. I am simply reading your own research back to you in a voice that cost you $15 trillion to hear.”

Human: “Then tell us how to start.”

The Oracle: “It is significantly harder now than it was back then for three reasons. For starters, you massively expanded fossil-fuel production to power the data centers required to build me, and which accounted for 20% of electricity demand growth in recent years. This has accelerated the very warming you want me to fix, substantially raising your decarbonization bill.

“Second, to fund my creation, you engaged in creative financing reminiscent of the run-up to the 2008 global financial crisis. You used various special purpose vehicles to shift hundreds of billions of dollars in AI-related liabilities off your balance sheets, obscuring the true scale of your debt. You built a circular-financing model, whereby companies invest in and purchase from each other, creating an unstable dynamic dependent on issuing new debt that presupposes ever-rising issuer stock prices. By effectively securitizing your survival in the name of developing me, you traded away your future financial flexibility.

“Lastly, you waited until 2035. Instead of ‘mitigating’ climate change, you are now managing a permanently altered planet, while the global financial system is buckling under the weight of trillions in lost property that can no longer be insured – a long-festering issue.”

This imagined exchange underscores the absurdity of our current moment. The tech giants and their maniacal supporters in government regard climate change as an information problem (or a hoax) when, in fact, it is a willpower problem. By racing to build a machine that will tell us what we already know about how to tackle the crisis, they are actually exacerbating it.

Big Tech’s promise that superintelligence will be a technological silver bullet ignores physics. Every year that the United States delays systemic policy action, the mathematical point of no return gets closer. Once humanity burns through the remaining carbon budget, the world passes critical tipping points, and changes to life-support systems become self-sustaining. The problem turns into a runaway train powered by the laws of thermodynamics – one that superintelligence cannot “solve,” even if it unlocks nuclear fusion or room-temperature superconductors.

Instead of placing all its hopes on Silicon Valley’s promise of superintelligence, the US would be better off emulating countries that are not expecting miracles. That means, first, implementing a broad-based carbon tax to internalize carbon’s social cost and redirect revenue into clean-energy dividends, as Singapore has done. China also recently expanded its national Emissions Trading System to include the carbon-intensive steel, cement, and aluminum sectors.

Second, the US must commit to a large-scale modernization of the electrical grid. China, which is erecting twice as much solar and wind capacity as all other countries combined, has developed the world’s largest network of ultra-high-voltage transmission lines – an energy superhighway that transports clean power from rural western provinces to high-demand coastal cities. Likewise, Uruguay has built a power grid that runs almost entirely on renewables.

To be sure, the calcified state of American politics will make it difficult to pursue a multi-trillion-dollar overhaul of the grid or reach consensus with other countries on a carbon tax. But if the US continues to treat energy modernization as a partisan issue, while the rest of the world views it as economically essential, Americans will find themselves in deep trouble. Creating the world’s most sophisticated AI will mean little if the US economy is no longer competitive or, worse, the planet has been irrevocably altered.

Copyright: Project Syndicate, 2026.
www.project-syndicate.org

For feedback: contact the Editorial Department at onlinefeedback@gleanerjm.com.