Advocates of nuclear power are having a bit of a moment. It is, be sure, a confected moment, as I'll explain, but it is indisputably a moment. And therefore problematic. This is an industry that never gives up and to which the normal rules don't apply. Even as market forces have made it abundantly clear that nuclear for new generating capacity is very far down the priority list, in the normal course of events, the Bitcoin/Crypto boom threw the industry its first lifeline in a long time. Huge excitement about so much new electricity needed to sustain one of the biggest generic con tricks of the early 21st century. High fives all around. But it never really happened.

And then it was green hydrogen. With every hard-to-abate sector queuing up for its share of vanishingly small volumes of green hydrogen (today and for the next 10 years), the Knights of Nuclear were up into their saddles just as fast as enough hobby horses could be corralled together, riding straight into that quagmire of technological, financial, and regulatory problems that has sucked down so many nuclear dreams in the past. Again, nothing happened.

But, as I say, this is an industry that never gives up. It doesn't need to, with nuclear governments around the world holding open their chequebooks (to give away our money) for whatever nuclear boondoggle next materializes. And that's exactly where we are today, in what I can only describe as “mid-boondoggle,” this time with AI filling the breach in a way that neither crypto nor green hydrogen ever managed to do.

I'll return to this in a moment, but we have to stand back first to take stock of the industry's unique status. Nuclear power inhabits two parallel universes. The first is performance-driven, measured in the hard-edged, empirical metrics of output, levelized cost of energy, cost of capital, capacity factors, and so on. The second is hype-driven, measured in the soft metrics of PR column inches, declaratory enthusiasm, big-name wet dreams (with Bill Gates still top of that particular league table, though OpenAI’s Sam Altman is now pressing hard for that top slot), and, most importantly of all, the steady flow of taxpayer dollars moving reassuringly onto companies’ P&L. This is a set of metrics unique to the nuclear industry, shaped by them, lovingly embellished as times change and boondoggling opportunities evolve. Which is all that's needed, apparently, to persuade governments to go on flinging billions of our taxpayer dollars at them, with no serious performance review deemed necessary or relevant.

You really have got to hand it to them. So let's start with the hard metrics of performance—and then move on to the AI-enhanced soft metrics of hype.

My go-to authority for distinguishing between performance and hype is the annual World Nuclear Industry Status Report (WNISR). Over the years, it has built up a formidable reputation both for factual accuracy and editorial insight. Its principal authors are all nuclear experts, with a distinctly skeptical take on information emanating from the industry itself.

According to this year's WNISR, the industry today is all about Russia (the dominant vendor in terms of international sales) and China. Of the six “new starts” in 2023, five were in China. Five new reactors came online in 2023; five closed down, resulting in a “net decline” in generating capacity of 1 GW. Compare that with a “net gain” of 460 GW of new renewable capacity. Investment in renewables in 2023 was 27 times as much as in nuclear. Nuclear costs are still surging – both solar and wind are already significantly cheaper, and their costs continue to fall. Delays on new builds are chronic, and the average age of the 408 reactors around the world is 32 years.

WNISR summarizes the situation as follows: “Contrary to widespread perception, nuclear power remains irrelevant in the international market for electricity generation systems.”

Moving over into the parallel universe of hype—incomparably enhanced as it now is by the all-encompassing allure/threat of AI, compared to which Bitcoin is a pissy little bit-part player in the corner. We've all read the growth projections for AI-enabled markets—from billions today to trillions tomorrow. We've probably all clocked that none of this comes for free in energy terms while ignoring the geeky little granular details—for instance, every single AI-assisted search requires 10 times as much electricity as today's search engines, with 10 times the carbon footprint.

I won't weary you with the extrapolated increases in electricity consumption for all the new data centers that this entails—but it's going to be a lot. On par with the electricity consumption of small countries. There’s a countervailing view, to be fair (which points to the huge potential gains in energy efficiency in buildings, transport, and manufacturing, which AI could make possible), but that all sounds very speculative at the moment. By contrast, the data centers are being built right now, ever bigger, already gobbling up more and more electricity. Nor, to get to the point, will I invite you to ask why this AI boom must not—ever, on any terms—be subjected to much deeper scrutiny as to the balance of costs and benefits. AI represents the apogee of latter-day technological determinism: if it can be done, it will be done. So suck it up.

Shift a bit further now into the thinking of those Titans that run the big tech corporations. Once you've determined what that growth curve looks like and calculated the concomitant electricity demand curve, and you know you can't use fossil fuels, and you sussed out the green hydrogen hype long ago, and you're just not sure that renewables + storage + efficiency can do the heavy lifting, where do you go? Well, straight back into that self-same nuclear quagmire!

In the last few months, we've had turbo-hyped nuclear announcements from most of the tech titans. Google has done a deal with Kairos Power for a “fleet of advanced nuclear power stations, providing us with 500 MW by 2035.” Amazon has pledged $500 million towards a 300 MW small modular reactor by the end of the decade. And Bill Gates has started construction on TerraPower’s Natrium fast reactor on the site of a retired coal power plant in Wyoming – courtesy of $2 billion from the Department of Energy, coming on top of all the billions of taxpayers’ dollars he’s already received. Microsoft has signed up as the first customer for electricity from Sam Altman’s fusion startup with Helion Energy—promised by 2028! Altman’s $375 million will deliver in a few years what tens of billions of dollars of public money have failed to deliver over 75 years!

Microsoft has also signed a 20-year Power Purchase Agreement with Constellation Energy (with $1.6 billion as table stakes) to restart one of the units at the Three Mile Island reactor—that caused a bit of a stir back in 1979 with what was then one of the world's worst-ever nuclear accidents. Microsoft has already committed to paying twice as much for each unit of nuclear electricity as it would have to pay for renewable electricity—and that’s before the inevitable delays and setbacks kick in. However, “Federal tax credits available to restart closed nuclear power plants will sweeten the deal.”

Just as they already have for the global energy giant Holtec, which secured a $1.5 billion loan from the Department of Energy (through the Inflation Reduction Act) to reopen the Palisades nuclear reactor, which closed in 2022. (The loan was originally to decommission the reactor, but now there's a much more lucrative AI-enabled deal to be done!). You can see what I mean by “a moment!”

Don't hold your breath on any of this. Not one small modular reactor or advanced modular reactor has a design certificate from regulators, let alone a construction and operating license. Projects are constantly delayed or canceled. And precisely because these smaller reactors can't achieve the economies of scale of today's standard reactors, whatever electricity they eventually produce (if any) will be eye-wateringly expensive. A boondoggling bonanza the nuclear industry itself can barely believe.