What If AI Can’t Solve Climate Change?

Today’s internet is inundated with “AI slop,” and mocking the often bizarre outputs produced of language models (recall the Google AI that recommended making pizza with glue) has become an online pastime. Yet artificial intelligence advocates have not been deterred from their claims of a utopian future made possible by AI. Whatever problems we face — including climate change — one day, we are told, they will be solved by the magical power of computing. The breathless headlines have been around for years: “How artificial intelligence can tackle climate change”; “How AI could power the climate breakthrough the world needs”; “Here are 10 ways AI could help fight climate change”; “9 ways AI is helping tackle climate change.”

Like much of the hype around AI, the specifics aren’t necessarily wrong. AI could help us understand the impacts of climate change more comprehensively, and can be used to locate solutions to particular challenges in technology and manufacturing. But as the extraordinary energy demands AI will impose on our system are coming into focus, and as some of the most important corporate AI leaders join hands with what could be the most anti-environment administration in history, the big picture problem becomes even clearer. Artificial intelligence can’t solve climate change because doing so will always require passing through the bottleneck of politics.

For those hoping to bring us to a glorious future guided by superintelligent computers, claiming that AI will solve climate change has become more urgent as the energy demands of the technology increase. Google reported last summer that since 2019, its emissions have increased by 48% because of its use of AI. The International Energy Agency projects that by 2026, AI will consume 1,000 terawatt-hours of electricity, as much as the entire nation of Japan, the world’s fourth-largest economy. Countries around the world are rushing to develop their own AI systems (the surprising capabilities of a new Chinese system called DeepSeek just sent the stock market tumbling), any of which could entail the same scale of energy demand as the ones created by American tech giants.

But imagine if we could snap our fingers and make that problem disappear? That’s what OpenAI CEO Sam Altman suggested in a recent interview with Bloomberg. “Fusion’s going to work,” he said when asked about AI’s energy demands, going on to say that “quickly permitting fusion reactors” is the answer — particularly those made by Helion Energy, a company whose executive chairman is, you guessed it, Sam Altman.

Of course, Helion has no fusion reactors to permit yet because no one does. Fusion’s promise of essentially limitless clean energy at low cost is tantalizing, which is why billions of public and private dollars have been invested in fusion research. But while technological gains are being made, there is still a great deal of uncertainty about how long it will take until fusion can reach commercial scale. It might be 10 years, or 20, or 50 or 100 — no one knows for sure.

But blithely insisting that incredibly complex problems will be solved easily and quickly is a specialty of tech barons. And if AI itself finds the solution to our energy problems? Even better.

There’s no question that AI is improving at a rapid pace, even if there are some things it’s still terrible at. And when it comes to climate, over time it will probably help produce incremental gains across a wide number of areas, from manufacturing efficiency to urban planning. But the more dramatic and consequential any idea is — whether it comes from AI or not — the more likely it is that it will have to move through the political process in order to be implemented.

And that’s where AI can’t help. A machine learning system can’t tell you the precise formula to please a recalcitrant senator or navigate a hundred city councils with different ideas about what kinds of clean energy projects they’ll allow in their towns. Politics is about people — their goals, their incentives, their fears, their prejudices — and it’s far too messy to be solved with numeric calculation, even by the most powerful AI system imaginable.

Let’s say that a year from now, an AI came up with both an entirely new way to design a fusion reactor and a revolutionary battery design that offered longer and denser storage, together solving so many of the problems scientists and engineers struggle with today. How would the fossil fuel industry react to this development? Would it say, “Oh well, oil and gas had a pretty good run, but now the world can move on”?

Of course not. It would use its extraordinary resources to battle against their competition, just as they always have. That’s what it did in the last election cycle, when it spent $450 million on campaigns and lobbying to preserve the industry and the riches it generates.

And while many hoped that the Republican Party would moderate its views on climate, at the moment it looks more like it is going backward — not just looking to undo every bit of progress made under the Biden administration, but also undermining renewable energy wherever it can. President Trump seems determined to destroy wind energy in America, which has been growing rapidly in recent years. Whether he succeeds will be up to the political system, not the inherent usefulness of a millennium-old technology.

In politics, good ideas don’t always win out. Who has power and what they are after will always matter a great deal, as Trump and the people he is bringing into the federal government are showing us right now.

AI can be a tool that helps us reduce emissions and mitigate the effects of climate change; the fact that its boosters regularly offer absurdly optimistic timelines for societal transformation doesn’t mean the underlying technology isn’t remarkable. But “solving” climate change isn’t merely a technological problem. It will always be a political one as well, and even the smartest piece of software won’t solve it for us.