Will AI kickstart a new age of nuclear power?

This article is published in association with United Nations.

The rapidly expanding use of artificial intelligence worldwide is putting electrical grids under huge pressure and many believe that, to meet that need without contributing to the climate crisis, a full-scale expansion of nuclear energy is essential.

The global demand for electricity is growing at a vertiginous rate. By 2035, it is expected that it will increase by more than 10,000 terawatt-hours, equivalent to the total consumption of all advanced economies today.

The rise of artificial intelligence has a large part to play: AI technology is powered by data centres and the electricity consumption a medium-sized data centre is equivalent to that of 100,000 households. According to the International Energy Agency, data centres demand increased by more than three quarters between 2023 and 2024 and is expected to account for over 20  per cent of electricity‑demand growth in advanced economies by 2030.

In the United States, where many of the leading AI businesses are based, the power consumption of AI-driven data processing is predicted to exceed the combined electricity consumption of aluminium, steel, cement and chemical production put together by the end of the decade.

In December of last year, policymakers, technology companies and nuclear industry leaders from around the world gathered at the International Atomic Energy Agency (IAEA) headquarters in Vienna to explore the opportunities for nuclear power to enable the expansion of AI and, conversely, how AI could drive innovation in the nuclear industry.

Training cutting-edge AI models requires tens of thousands of central processing units (CPUs) to run continuously for weeks or even months. At the same time, the daily application of artificial intelligence is expanding to almost all sectors such as hospitals, public administration, transportation, agriculture, logistics and education.

Every query, every simulation, every recommendation consumes power. “We need clean, stable zero-carbon electricity that is available around the clock,” says Manuel Greisinger, a senior manager at Google, focusing on AI. “This is undoubtedly an extremely high threshold, and it is not achievable with wind and solar power alone. AI is the engine of the future, but an engine without fuel is almost useless. Nuclear energy is not only an option, but also an indispensable core component of the future energy structure.”

© Unsplash/Geoffrey Moffett

A data centre in Ireland.

Bullish nuclear industry

Mr. Greisinger’s view is shared by IAEA Director General Manuel Grossi, who believes that the nuclear industry is destined to be the energy partner of the AI revolution. “Only nuclear energy can meet the five needs of low-carbon power generation, round-the-clock reliability, ultra-high power density, grid stability and true scalability,” he declared.

The nuclear industry appears to be in bullish mood. Seventy-one new reactors are under construction, adding to the 441 that are currently operating globally. Ten are scheduled to be built in the US, which is already home to 94 plants, the largest amount of any country.

The tech giants that are using the data centres have pledged to support the goal of at least tripling global nuclear power capacity by 2050. Microsoft, for example, has signed a 20-year power purchase agreement that allowed Unit One of the Three Mile Island nuclear power plant in Pennsylvania, USA, to be restarted. 

NOAA/OAR/Great Lakes Environmental Research Laboratory

The Enrico Fermi nuclear generating Station near Monroe, Michigan, US.

The rest of the world is also actively investing in nuclear energy, driven by the growth of AI. “Europe has the world’s densest digital corridors, with Frankfurt, Amsterdam and London as hubs,” explained Mr. Grossi. 

“Traditional nuclear energy powers such as France and the United Kingdom are doubling down on nuclear energy construction, and emerging countries such as Poland are also accelerating their participation.” 

Russia, with a research base proficient in mathematics and computer science, remains the world’s largest exporter in the field of nuclear energy, and is a leading operator and developer of advanced reactor technology, whilst China is making major achievements in both AI and nuclear energy.

“AI technology and the construction of artificial intelligence data centres are advancing simultaneously, and the number of new nuclear reactors in the world also ranks first in the world during the same period,” said the UN nuclear agency chief.

Japan is investing heavily in building and upgrading data centres to meet growing demand whilst, in the Middle East, the United Arab Emirates has established a nuclear energy programme and has emerged as a regional AI hub. 

IAEA/Dean Calma

The IAEA supports training to ensure the safety of nuclear power plants like this one in the Czech Republic.

Are small reactors the answer?

The need for much more energy, and soon, is also driving the construction of small modular reactors, which are very different from the traditional large power plants that require huge investment, and a lead time of around 10 years. 

“These kinds of reactors have a small footprint and upgraded safety systems, and can be deployed in nearby industrial areas, including data centre campuses,” Mr. Grossi said. 

Tech companies that use them don’t have to worry about regional grid supply constraints or transmission losses. This will be a decisive advantage in areas where grid upgrades are slow, and interconnection queues are long.” 

Although this form of reactor still needs to move beyond the R&D phase, the IAEA is working closely with regulators and the industry to make them a viable proposition and we could soon be seeing large numbers of small reactors being deployed to meet the demand. 

Google for example, has signed an agreement with an energy company to buy nuclear energy from multiple small modular reactors, a global first. If all goes well, they could be operational by 2030.

Google is also turning its attention to space, exploring space-based solar networks to enable large-scale machine learning in orbit, taking full advantage of unfiltered solar energy. Two prototype satellites are due to be launched in early 2027 to test radiation tolerance and data processing capabilities in the space environment.

Whether it’s harnessing solar energy in space, restarting old reactors, investing in a new generation of small modular reactors, or building large reactors, all actions point in the same direction – building an energy system based largely on nuclear energy that can support the needs of future civilizations.