In May 2024, Congress banned Russian uranium. Last year, American reactors still bought more enriched uranium from Russia than from any plant in America.
That is not a loophole, and it is not an oversight. It is what happens when a country bans something it has forgotten how to make.

In September 2024, Microsoft did something nobody had done before. It signed a twenty-year deal to restart a reactor at Three Mile Island and send the power to the data centres running its AI. Three Mile Island is the site of America’s worst nuclear accident. The reactor that melted down in 1979 is still sitting there, next door to the one Microsoft is paying to wake up, which had been switched off for five years because it could not compete on price.

That is what the AI boom is doing to the energy map. Amazon and Google have signed deals for small modular reactors. Old plants are being switched back on and new ones planned. The buyer behind nearly all of it is the data centre, which needs power that does not stop at sunset.

And yet two years on, Three Mile Island still is not running. It is due back in 2027. Ordering the reactor turns out to be the easy part.

You cannot just feed a reactor mined rock. The uranium has to be enriched first, its fissile content concentrated in a handful of highly specialised plants. That enrichment step is the hard part, and it is the one link in the chain the West cannot simply buy its way around.

Here is the trap. Russia runs about 43% of the world’s uranium enrichment capacity, more than the United States, Britain, Germany, the Netherlands and France put together (World Nuclear Association). Add China’s CNNC and 59% of world capacity sits inside the two countries the West is busy decoupling from. In America the dependence is more direct still: Russia supplied roughly a quarter of the enrichment services US reactors bought in 2025, a bigger share than American plants supplied themselves (US Energy Information Administration). The West let its own capacity wither for years, for the simplest reason there is. Russian fuel was cheap.

Mining is not the problem. Kazakhstan digs about 39% of the world’s uranium, Canada 24%, Namibia 12%. Three countries that are not Russia account for three-quarters of world mine output, and Russia itself digs only 5%. Supply there has its own wobbles, and Kazatomprom is cutting output by 10% in 2026, but ore is a market, not a chokehold. Conversion and fuel fabrication are specialised work, and they are not scarce either. Everything narrows at one step, enrichment, and that is precisely the step Russia dominates.

Put the same chain on a map and the asymmetry gets starker. The ore comes from everywhere. The plants that turn it into fuel do not.

The stages are one thing. The names that own them are another, and the list gets very short in the middle.

So Washington banned it. The order of events matters here, because it usually gets told backwards. Washington moved first. The Prohibiting Russian Uranium Imports Act was signed on 13 May 2024 and took effect that August. Moscow retaliated in November 2024, banning uranium exports to the United States. But the American ban shipped with a release valve. The Department of Energy can waive it through 2027, and it has been waiving it, because the supply cannot actually be replaced yet. The ban is real. So is the dependence.

A ban does not create capacity. It only creates a deadline, and the deadline and the capacity do not arrive in the same year.

America is scrambling, and the scramble is real money. Centrus produced the first new American-made HALEU, the high-assay fuel the new small reactors need, in more than seventy years, at Piketon, Ohio. The catch is the scale. It runs at about 900 kilograms a year, and a fleet of reactors needs tonnes.

Then January 2026 happened. The Department of Energy handed out $900 million each to two projects: Centrus, to take Piketon to commercial scale, and Orano, for Project Ike, a $5 billion enrichment plant at Oak Ridge, Tennessee whose licence application the Nuclear Regulatory Commission accepted in May 2026 on an accelerated schedule. In June, Urenco USA announced a 2.1 million SWU expansion in New Mexico, and broke ground in August.

Now read the dates. Urenco’s new plant starts construction in 2029, turns its first cascades in 2032, and is not fully installed until 2036. Centrus’s new capacity arrives in 2029. Orano’s lands in the early 2030s. Every one of those announcements is welcome, and not one of them enriches a gram of uranium this decade.
You can order a reactor and wait ten years. An enrichment plant is the thing you should have ordered ten years ago.

Uranium tells the story, but only if you look at the right price. Spot uranium is a thin, jumpy market. It brushed $100 a pound in January 2024, collapsed to $64 by March 2025, and sat near $86 this July, still below where it started (Cameco). The long-term contract price, which is what utilities actually sign multi-year fuel deals against, has done something completely different. It has risen almost every single month since early 2024, from $72 to $95.50, and TradeTech’s indicator touched $97 in June, an eighteen-year high. Spot is sentiment. The contract price is utilities quietly locking down supply for the 2030s, and it has been going one way.

The nuclear revival will move only as fast as the fuel behind it. Ignore the ribbon-cuttings.
Watch one number instead: how much enriched uranium the West can make without Russia. The honest answer is not “none”. Urenco and Orano together hold 41% of world capacity. The truth is worse in a quieter way. That capacity is already spoken for, America has exactly one commercial plant covering about a third of its own demand, and nothing announced produces anything before 2029.
The reactors can be bought. The centrifuges have to be built.