Experts warn: is the US betting on nuclear technology that could collapse?
While the US government and investors are optimistic and investing about $9 billion in small modular nuclear reactor (SMR) technology to fuel the AI era's energy boom, many in the energy sector warn:
This technology is still far from proven.
Since 2019, more than $6 billion in government funds (mainly from the Department of Energy and Defense) and another $3 billion in private funds have been injected — but progress on the ground? Slow, expensive, and full of challenges.
Big promise – bigger risk. SMR technology promises to provide clean, stable, and cheap electricity to power artificial intelligence models, which consume enormous amounts of energy.
This is one reason for the surge in the value of companies in the field — even those that have not yet supplied a single megawatt.
But Chris Gadomski, head of nuclear research at BloombergNEF, warns: “There is a lot of enthusiasm, but the capital needed to reach the finish line is enormous. It's reminiscent of the dot-com bubble of the early 2000s.” The problem?
Monster costs!
As of today, costs far exceed planning.
NuScale's SMR project was canceled in 2023 after costs jumped by 120%.
Reactors operating in Russia and China exceeded budget by 300%–400%.
According to Wood Mackenzie, the cost of electricity generation from SMR in 2030 will be $182 per megawatt-hour.
Compared to?
- Conventional nuclear – $133.
- Natural gas – $126.
In contrast, companies like Oklo, supported by Sam Altman, promise much lower costs – just $90, and NuScale even talks about just $64.
But is it realistic?
Problematic fuel, limited scale. Most small reactors use a unique fuel called HALEU, which is currently produced mainly in Russia.
In the US, there is only one plant approved for large-scale production – creating a dangerous bottleneck.
According to Bank of America, fuel costs alone could add $20 per megawatt-hour.
Nuclear engineer Nick Touran claims companies “ignore reality”: “Even if they produce like Henry Ford – it won't solve the project's economics.”
And yet – hope is not extinguished. Supporters of the field remind that every new technology starts with losses.
Simon Irish, CEO of Terrestrial Energy, explained: “Every new power plant is not economical at first, but costs decrease over time.” Meanwhile, companies are not waiting – contracts have been signed to supply 32 gigawatts of clean electricity since 2023, including binding agreements with Google and Kairos for 500 megawatts by 2035.
According to David Brown of Wood Mackenzie, this is a “huge vote of confidence in the industry and a significant step towards true commercialization.”
Ultimately, the US is betting on small nuclear to power big artificial intelligence.
But the real question is – will this be the move that sparks a revolution, or a new nuclear bubble?