AI’s power crisis is funding the post-fossil-fuel future
A few days ago when I posted an essay about the AI bubble I promised to write a separate post addressing the issues of power consumption and data centers – because they are two totally separate issues.
I’ve been closely following the pros and cons of AI for over 3 years now. ChatGPT says I’ve had my account for over 900 days. I also have 51 years of experience living through end of the world situations – and yet – here we all are.
From my research, the existential dread people are experiencing about the power used by AI is a vast overreaction.
“But what about the jobs, Rachel??? What about the jobs?????”
That’s a legitimate concern – but not one I’ll be addressing in this particular post – that’s a different issue – which I will write about separately.
Anyway, back to power consumption. When I started writing this essay, I was under the impression that Google planned to be carbon neutral by 2030 – but after doing some more research – those plans seem to have changed. Google is currently moving away from that goal, not toward it. You may have seen a recent quote from Google’s AI infrastructure boss Amin Vahdat about needing to double compute capacity every six months to keep up – and that is exactly why.
As it turns out, for many years, Google claimed to be carbon neutral by buying cheap carbon offsets – paying to plant trees or protect forests to cancel out their emissions. It was the easy version of looking green. But in 2023, they stopped buying those offsets and stopped making that claim. They admitted offsets weren’t a real solution and pivoted to a much harder goal called Net Zero, which requires actually reducing their own pollution.
Here’s the problem. Since making that switch, their emissions haven’t dropped. They’ve gone up.
Google’s greenhouse gas emissions have increased roughly 48 percent since 2019. To compete in AI, Google has to build massive, power-hungry data centers. That exponential growth in hardware requires massive amounts of electricity, and much of it still comes from fossil fuels.
So while Google set a new, harder target – Net Zero by 2030 – the AI arms race is actively sabotaging their ability to hit it.
But here’s where it gets interesting. The worry has shifted. A few years ago, the concern was largely reputational – it that’s shifted.
Companies wanted to look green for the public. Now, in late 2025, the worry is existential. They need to find enough actual electricity to keep the servers running. Wind and solar aren’t scaling fast enough to meet the 24/7 demands of AI.
So the major players did something that should make every free-market conservative in America weep with joy. They opened their checkbooks and started funding the future of energy themselves.
Microsoft is arguably the most aggressive. They signed a deal to restart the Three Mile Island nuclear plant in Pennsylvania. Yes, that Three Mile Island – the site of the famous 1979 partial meltdown. Unit 1, which was shut down for economic reasons, is being brought back online to provide dedicated, round-the-clock carbon-free energy exclusively for their data centers. No government mandate. No taxpayer subsidy. A private company decided it needed clean power, so it went and bought a nuclear plant.
Google is betting on something different. In late 2025, they signed the world’s first corporate agreement to purchase power from Small Modular Reactors developed by a company called Kairos Power. These are mini nuclear reactors that can be built faster and placed closer to demand. The first one isn’t expected online until roughly 2030, but the point is that Google is funding the development of next-generation nuclear technology because they need it to exist.
Amazon took a direct investment route. They led a 500 million dollar investment into X-energy, another SMR developer, and signed an agreement with Energy Northwest to build four advanced nuclear reactors in Washington state. Private capital flowing into advanced nuclear at a scale that government energy programs haven’t matched in decades.
Then there’s OpenAI. Sam Altman is personally betting on nuclear fusion – the process that powers the sun – which is theoretically limitless and clean but has never been commercially viable. Altman has invested hundreds of millions into Helion Energy, which claims they can build a working commercial fusion plant by 2028. Microsoft has even pre-agreed to buy electricity from Helion, essentially pre-ordering power from a star in a jar that doesn’t exist yet.
The AI power crisis has actually turned Silicon Valley into the biggest private funders of next-generation energy on the planet. And yes – they’re not doing this because they love the environment. They’re doing it because they literally cannot build enough data centers without inventing new ways to generate electricity. Profit motive is dragging us toward a post-fossil-fuel future faster than any carbon tax ever could.
But, let’s be real, the difference between what Microsoft is doing and what OpenAI is betting on is the difference between reusing a 1970s engine and trying to build a star in a jar. Microsoft is buying fission – proven, messy, 70-year-old technology. You split heavy atoms like uranium, release heat, boil water, spin turbines. It works. The downside is radioactive waste that stays dangerous for millennia and the theoretical risk of meltdown, though modern plants are extremely safe.
Read that again – all experts say that modern nuclear power plants are extremely safe.
Microsoft isn’t building anything new. They’re paying to restart an old reactor to get guaranteed power by 2027.
Fusion, on the other hand, is how the sun works. You take two light atoms like hydrogen and smash them together so hard they fuse into helium. It creates way more energy than fission.
Zero long-term radioactive waste. And it cannot melt down – if something goes wrong, the reaction just stops. But you have to contain plasma hotter than the sun, around 100 million degrees. Humanity has mostly spent more energy starting the reaction than we’ve gotten out of it and most physicists think Helion’s 2028 deadline is fantasy. But private money is flowing into solving the problem anyway.
This is, frankly, the kind of private-sector technology investment that Republicans have wet dreams about. No government picking winners and losers. No subsidies distorting markets. Just companies that need a thing to exist pouring billions into making it exist. The AI arms race might accidentally solve climate change as a side effect of the profit motive.
Does this need regulation and close oversight? Hell yes.
These are nuclear reactors we’re talking about. Some of these designs have never been built at scale. The gap between now and nuclear means they’re still burning natural gas and extending the life of coal plants to keep the servers running today. Their emissions are spiking right now despite their clean plans for tomorrow. Someone needs to be watching.
But that brings us to the actual problem. We need government officials and regulators who understand what’s happening. People who can tell the difference between a Small Modular Reactor and a cold fusion scam. Officials who grasp that the current power grid cannot handle what’s coming and that the solutions being funded are genuinely novel. Not dinosaurs who think “the cyber” is a series of tubes.
The same members of Congress who can’t figure out how to unmute themselves on Zoom are supposed to be overseeing the largest private investment in nuclear energy in American history. The same regulators who spent decades blocking nuclear permits for environmental reasons now need to fast-track approvals for technology that could eliminate fossil fuel dependence entirely.
This is another reason to vote out the politicians who still think coal is coming back. The future is being built by companies that need power so badly they’re willing to restart Three Mile Island and fund fusion research. If our government can’t keep up, we’re going to miss the biggest energy transition in a century – one that’s happening not because of regulation but despite decades of it.
Sure, the whole setup isn’t really about saving the planet. It’s about securing enough power to win the AI race. But the side effect might be a world that runs on nuclear instead of fossil fuels. Private greed accidentally building the clean energy future that public policy couldn’t deliver. That’s either deeply inspiring or deeply depressing, depending on how you decide to view the world.
Me? I’m an optimist because that’s what the facts tell me to be.
SOURCES
https://www.reuters.com/technology/microsoft-deal-resurrect-three-mile-island-nuclear-plant-2024-09-20/
https://www.cnbc.com/2024/10/14/google-signs-deal-to-get-power-from-small-modular-nuclear-reactors.html
https://www.aboutamazon.com/news/aws/amazon-nuclear-small-modular-reactor-net-zero-carbon
https://www.helionenergy.com/
https://sustainability.google/operating-sustainably/net-zero-carbon/