How AI is helping fuel a nuclear renaissance

Euan Sinnot

The nascent technology’s insatiable need for power is leading tech giants, keen to keep to their climate goals, to strike deals to reopen and build reactors across America.

The AI boom has taken the world by storm, and left tech companies scrambling to build out the most advanced models to try and capture as much of the industry as possible. Tech giants are all-in on an AI future, with the five largest tech companies spending a staggering $59bn in capex in the last quarter alone, up 63% year-over-year. But this money is not just being spent on engineers and programmers - tech companies now need vast amounts of energy to power their data centres, something they have been increasingly turning to nuclear energy for.

Data centres power much of the internet, and are needed to train and operate AI models, including large language models (LLMs) like OpenAI’s GPT-4. These complex models require both larger amounts and more advanced computers and processors to operate. This means that tech companies need to buy a lot more electricity. In 2022, AI was estimated by the IEA to have consumed around twice as much energy as all of Britain. This will only accelerate, with the amount of energy required by data centres projected to more than double by 2030.

This has led tech giants to look to strike deals with energy producers to ensure a reliable and affordable supply of electricity for their data centres. Complicating matters, all major technology companies have ambitious climate goals – Microsoft and Google promise to be net zero carbon emitters by 2030. Traditionally, these companies have looked to solar and wind to fulfil their energy needs while maintaining their green credentials. But, both are intermittent, and Microsoft cannot afford to have their servers shut down when the wind isn’t blowing – so tech companies have begun to look for alternatives. When it comes to continually-running, reliable energy, there are only two real options – fossil fuels (usually gas), or nuclear fission. However, efforts in climate mitigation mean that tech companies dislike the former, leading the latter as the best placed option. 

The most notable example of AI’s impact on nuclear power is Microsoft’s recent deal to purchase electricity from a reopened Three Mile Island Power Plant, in Pennsylvania. Three Mile Island was previously operational until 2019, but shut down because of high running costs. The power plant is notable for being the site of a partial nuclear meltdown of one of its reactors, which led to it closing for six years. There were no deaths or injuries, but there was still a large, yet misinformed growth in opposition to nuclear power in the US, both nationally, and amongst residents near Three Mile Island. This led to increased regulations driving up costs, and a stagnation in nuclear power. Since the mid-1980s, nuclear has made up about a fifth of America’s electricity mix, but there has been little growth in generation.

But, there is evidence that the tide is turning when it comes to nuclear power. In 2023, Gallup found that 62% of Republicans, 56% of Independents and 46% of Democrats supported nuclear power, the greatest level of public support nuclear has had in the last decade. As climate change has raised in saliency amongst Democrats, but support for big-spending initiatives like the Green New Deal has faltered, promoting nuclear has become an increasingly popular solution to decarbonising the grid amongst policymakers. The Biden Administration has been supportive of nuclear, and recently announced a $900mn investment into small nuclear reactors.

This renaissance in nuclear energy comes at an ideal time for tech companies’ sudden demand for reliable power that aligns with sustainability objectives. As well as deals to take on electricity from full-scale reactors, tech companies are also getting in on the small modular reactors (SMRs) train, with Amazon recently announcing a $500mn investment in a Virginian project. SMRs are much smaller than conventional reactors, and can therefore fit in more locations. They can also be prefabricated, which makes construction much easier and cheaper. While the technology is in its early stages, tech firms are natural investors in them, as they are looking for ways to power their data centres sustainably for decades to come, and have vast amounts of available capital to invest with.

The growth in nuclear power as a result of the AI boom shows how the expansion in Artificial Intelligence is changing the world economy beyond just the tech sector. Nuclear, as the only power source that is zero carbon, all-weather running, and suitable in most climates, will play a crucial role in decarbonisation, and investments from tech firms in nuclear both shows their enormous and growing financial muscle, as well how impactful AI may be on the broader economy. Nuclear becoming cheaper and more flexible, via the development of SMRs, is also undoubtedly a good thing for consumers, as well as society at-large.

Outside of the US, investment in nuclear power by tech companies has been more venture-focused, rather than large-scale agreements to take on electricity from existing or planned plants. This mainly reflects a more negative attitude towards nuclear in many European countries - where most data centres operated by large tech firms outside of the US are located. Germany, for instance has fully phased out nuclear from its grid, and Spain is currently in the process of turning off all its nuclear reactors. In the UK, nuclear has stagnated, with the last nuclear plant opening nearly twenty years ago. The main exception to this is France, which generated 64% of its electricity from nuclear. But, as this is all operated by the state-owned EDF, there is little room for private-sector investment.

A very interesting and well researched article. Would you be able to mention at the end about how this might go beyond the US? Will other countries also be  able to harness nuclear energy. Please reference your stats