AI's electricity hunger is reviving nuclear power

Climate & Energy 4 min 80 sources

Surging demand — much of it from data centers — is pulling nuclear back into the energy picture and straining grids, even as the oil market stays tight, an Arctic drilling auction flops, and the instruments that watch the climate face cuts.

Key takeaways

Fast-rising electricity demand, much of it from AI data centers, is reviving nuclear power — a major US utility merger and a reactor reaching "criticality" both landed this week — while the same demand is already straining grids like Texas's.
The fossil side stays tight: Iran's Hormuz blockade keeps oil prices high, yet a hard government push to open Alaska's Arctic refuge to drilling drew only $3.7 million and two bidders — proof that opening land doesn't force companies to drill.
Clean power's next limit is land, not cost (solar needs ~417 acres per 100 megawatts), and cuts to US ocean-monitoring instruments threaten the early warning systems that track El Niño and Atlantic circulation.

Nuclear gets pulled back in

Nuclear power, written off for years as too slow and too costly, is suddenly back in the conversation. Two signs landed this week. NextEra and Dominion — both large US power companies — are merging in a deal set to reshape nuclear power across New England [1]. And the US Energy Department said an advanced reactor became the first of its kind to reach “criticality” [2].

Criticality is the moment a reactor sustains its own nuclear chain reaction — the core switches on and keeps itself going. It is a real milestone. It is also not the same as power flowing to homes: a first criticality is a test, and the road from there to electricity on the grid runs for years.

Why the renewed interest? After a decade of flat US electricity demand, demand is climbing fast. A big driver is the surge of AI data centers — warehouses of computers that draw power around the clock. Grids need “firm” supply: electricity available on demand, day or night, wind or no wind. Solar and wind are cheap but come and go with the weather. Nuclear runs nonstop. That round-the-clock reliability is what’s pulling the money and the policy back toward it. The thing to watch is simple: whether the new reactors actually deliver electricity, or just milestones.

The load that’s bending the grid

You can already see the strain. In Texas, the grid operator flagged that several large data centers and crypto-mining sites failed its reliability tests ahead of the summer peak, when demand runs highest [4]. The regulator is reviewing the failures and drawing up fixes [4].

The problem is the shape of the new demand. An AI data center is an enormous, concentrated load that switches on fast — the grid was built for slower, steadier growth spread across many users. When a single site pulls as much power as a town, it can drag down the voltage that keeps the local network stable. This is the demand pulling nuclear back into favour, and it is becoming a reliability-and-bills story for everyone sharing that grid, not just the tech firms building on it.

The fossil side stays tight — and a drilling push falls flat

The old energy system is under its own strain. Global oil inventories — the world’s stored buffer — are running dangerously low, and analysts warn the next price spike could come within weeks, hard enough to rattle markets [3]. The squeeze traces back to Iran’s blockade of the Strait of Hormuz, the channel that carries about a fifth of the world’s oil, which has kept prices high [5].

You might expect that tight market to make new drilling irresistible. It didn’t. The Trump administration’s Friday auction of oil leases in Alaska’s Arctic National Wildlife Refuge drew almost no interest [16]. The refuge is a remote, pristine habitat for polar bears, caribou and migratory birds. It offered 58 tracts; companies bid on five [16]. Winning bids totaled just $3.7 million, from only two bidders, one of them an Alaskan state agency [16]. The administration had pushed hard to “unleash Alaska’s resource potential,” sending three cabinet officials north to make the case [51]. The lesson is plain: a government can open the land, but it cannot make oil companies spend on remote, costly, reputationally fraught barrels they aren’t sure will pay off.

Clean power’s next limit isn’t cost — it’s land

Solar is now the fastest-growing source of electricity in the US [6]. Nearly 397 gigawatts of projects are waiting to connect to the grid, with about 70 gigawatts due to come online in 2026 and 2027 [6]. Cost stopped being the obstacle a while ago. The new one is space.

Solar farms are land-hungry. A single 100-megawatt project needs roughly 417 acres — about 316 American football fields [6]. The 70 gigawatts arriving over the next two years will need more than 320,000 acres of land [6]. A lot of that is farmland, which sets up a real contest between growing food and generating power. One answer researchers point to is agrivoltaics — putting panels above crops or grazing animals so the same acre does both jobs. The fight over where clean power goes, not whether it’s affordable, is the next phase of the build-out.

The instruments that watch the climate face the axe

End on the quiet story. The administration is moving to dismantle a network of ocean-sensing instruments [40]. Scientists say it will hinder the monitoring of two systems that shape our weather [40]. One is El Niño in the Pacific. The other is the AMOC — the Atlantic Ocean circulation that carries warmth around the Northern Hemisphere [40].

Why it matters is best shown by example. In the winter of 2013-14, seabed instruments off the US west coast caught a vast mass of warm water, later nicknamed “the Blob” [40]. It was up to 4°C hotter than normal and spread across more than 1,500 kilometres of the north Pacific [40]. Those sensors gave scientists and the fishing industry early warning [40]. You cannot forecast, or even understand, a system you have stopped measuring. Cutting the instruments doesn’t change the climate. It just means seeing less of it coming.

02 · Lesson · why it matters

Fix the bottleneck, and it just moves

Every system has one true limit — and the moment you widen it, the limit jumps somewhere else, usually somewhere quieter than where you were just looking.

The wall that didn’t fall — it moved

For years, the case against clean energy was simple: too expensive. Solar and wind cost more than coal and gas, so they’d stay niche. Then the price of solar collapsed. Today it’s the cheapest electricity most of the world has ever had.

Problem solved? Not even close. The complaints just changed shape. Now it’s the grid that can’t carry the power. Now it’s the land the panels need. Now it’s what happens after dark, when the sun is down and demand is up. The wall everyone fought against — cost — fell. And a new wall was standing right behind it.

This is not a story of failure. It is how every system works, and once you see it, you can’t unsee it.

What a bottleneck actually is

In any system with many parts, one part is the real limit on the whole. Engineers call it the bottleneck — the narrowest point of the pipe, the slowest step on the line. A factory runs only as fast as its slowest machine. It doesn’t matter how fast the other machines go; the slow one sets the pace for everything.

The energy system’s bottleneck used to be cost. So that’s where everyone pushed — researchers, manufacturers, policymakers, all working to make clean power cheap. They succeeded. And the moment they did, cost stopped being the thing holding the system back.

Why fixing it doesn’t set you free

Here’s the part people miss. When you widen the bottleneck, the system speeds up — until something else becomes the slowest step. The limit doesn’t disappear. It gets handed to the next-narrowest part.

Watch it move in real time. Solar got cheap, so much of it got built that the grid became the new limit. Nearly 400 gigawatts of projects are now waiting in line just to connect to wires that can’t carry them. Build the wires, and land becomes the limit: all those panels need space, and that space is often farmland. Solve land, and the limit becomes firm power — keeping the lights on at night, which is why nuclear is suddenly back in the conversation. Each victory crowns the next constraint. The race never ends; the finish line just keeps walking.

The new limit hides in the boring places

And there’s a trap inside the trap. The next bottleneck is almost always quieter and less photogenic than the one you just beat.

Cost is a headline number everyone can argue about. Transmission lines, land-use permits, and ocean sensors are not. So we keep celebrating the visible win — “solar is cheap now!” — while the real limit slips into the dull infrastructure nobody puts on a poster. The Arctic auction this week was the same lesson, inside out. The government pulled the visible lever, opening protected land to drilling. But the real constraint was never permission. It was whether companies would actually spend the money — and they mostly wouldn’t. The loud lever moved. The quiet constraint didn’t.

Where the leverage really is

Once you can see the bottleneck moving, two things change about how you read the world.

First, “we solved it” almost never means “we’re done.” It means: go find where the limit just went. A solved problem is a signal to look for the next one, not to declare victory.

Second, the place where real progress is made is hardly ever the thing everyone is shouting about. That one is usually already being relieved. The leverage is in the next constraint — the quiet one no one has noticed yet. The people who actually move systems are the ones who can spot the bottleneck before it becomes the headline.

It always goes somewhere

This is not only about energy. It’s true of a project, a company, a body, a life. Hire more builders and the bottleneck becomes the review that approves their work. Speed up the reviews and it becomes the one person who makes the final call. Relieve them and it becomes the cash to pay for any of it. The limit keeps walking, room to room.

Progress doesn’t feel like crossing a finish line, because in a living system there isn’t one. There is only a constraint, somewhere, right now — and your real job is to keep finding where it moved. Stop asking “did we fix the problem?” Start asking “so where did the problem go?” In a system that’s actually alive, the answer is never nowhere.

03 · Lab · your turn

The Moving Bottleneck

Upgrade one stage of a clean-power chain each round and feel that only widening the current slowest step moves the system — and that each fix relocates the limit to the next stage.