On January 13,British Prime Minister Keir Starmer unveiled an ambitious plan to bolster the country's artificial intelligence (AI) sector.The initiative,called the AI Growth Zones,aims to establish multiple regions in the UK focused on AI development and related infrastructure.The first of these zones will be located in Culham,Oxfordshire,home to the UK Atomic Energy Authority and its renowned facilities.

The UK Atomic Energy Authority manages the Joint European Torus (JET),Europe's largest nuclear fusion facility.This facility,previously funded by the European Atomic Energy Community before Britain's exit from the organization,has made headlines in the past,notably in 2024 when it reported generating 69 megajoules of energy using just 0.2 milligrams of nuclear fuel.This remarkable achievement set a record for the highest energy output from fusion experiments.Additionally,the authority oversees the upgraded Mega Ampere Spherical Tokamak,a leading compact fusion device worldwide.

The establishment of the AI Growth Zones appears to be a strategic move to harness nuclear energy for powering AI computing centers.According to official reports,the UK plans to create an AI energy committee composed of government representatives and private enterprises,intending to utilize Small Modular Reactors (SMR) to meet energy demands.The energy consumption of AI computing is notably high; Nvidia's CEO Jensen Huang has warned that if only raw computing power is considered,the AI sector could consume energy equivalent to that of 14 Earths in the foreseeable future.Reports from Omdia analyst Wang Zhen highlighted issues with Elon Musk's xAI company,which has seen design energy requirements soaring at a data center in Memphis,demanding 150 megawatts of power while the onsite transformers can only accommodate 8 megawatts.A significant expansion will take one to two years to materialize,underscoring the urgent need for sustainable energy sources in AI development.

Nuclear energy is gaining attention as a more efficient option compared to current electricity grid solutions,thanks to higher energy output per unit of fuel.Nuclear power generation is less susceptible to external factors such as weather and seasonal fluctuations,making it a consistent energy source for AI centers.Lin Li,Director at CIC Industry Research,affirmed,"Although the initial investment costs for nuclear energy projects are high,the advent of emerging technologies like SMR is enhancing the economic viability of nuclear power.”

A number of technology firms are exploring nuclear energy to support their data centers.Larry Ellison,the founder of Oracle,expressed that a new 1-gigawatt (GW) data center under construction would be powered by three small modular reactors.Additionally,Microsoft has partnered with TerraPower to purchase electricity from the Three Mile Island nuclear plant for its operations over the next two decades.

Despite these advancements,significant obstacles remain in actualizing nuclear energy as a power source for AI computing.One of the primary issues has been the feasibility of connecting nuclear plants to AI centers.Amazon,for instance,had planned to collaborate with Talen Energy for a data center adjacent to a nuclear plant but faced rejection from the Federal Energy Regulatory Commission (FERC).Lin noted,“Established nuclear plants are typically connected to the grid,supplying power through that network.If a data center were to overconsume electricity,it could endanger the grid's reliability,making grid stability a core concern that led to FERC's dismissal of the proposal.”

Moreover,Lin added that safety and public acceptance regarding nuclear energy also present critical challenges.Public apprehension regarding nuclear safety remains a hot topic; a nuclear accident could have catastrophic repercussions.Residents near nuclear facilities often express concerns about environmental impacts,noise pollution,and pressure on existing electrical infrastructure.

To mitigate these potential risks highlighted in Talen Energy's proposal,off-grid power generation is becoming a favored trend for supplying nuclear energy.Wang Zhen pointed out,“In overseas data center planning,nuclear energy often integrates off-grid solutions,indicating this method is becoming a strong candidate for the future AI computing.” Off-grid power generation means generating electricity independently without relying on a public grid.

Nevertheless,the technologies required for off-grid nuclear power remain in preliminary stages for large-scale commercial availability.Wang noted,“The off-grid nuclear energy sector is not yet mature and will require fourth-generation,molten salt-cooled micro modular reactors,paired with equally modular turbine systems.Industry experts suggest that commercialization will not occur before 2030,with 2035 being a more widely accepted expectation.” He also indicated a growing interest in an alternative off-grid solution using natural gas,where large AI computing centers are considering gas-powered generators or turbines.

Different geographical regions may adopt varying approaches to powering AI computing centers.“In the United States,where natural gas resources are abundant,off-grid solutions using gas are quite common.Some nuclear companies are also integrating natural gas into their off-grid power strategies,whereas those in Europe,faced with natural gas scarcity and environmental regulations,tend to explore off-grid solutions less frequently.” Wang Zhen further speculated on the application of nuclear power for Chinese AI computing centers,suggesting geography would play a critical role.“If an AI data center is established in the east,acquiring power credentials may be challenging,while building in the west—where renewable energy is plentiful and electricity usage is low—would likely ensure a dependable power supply.” Reports suggest that Sensetime Technology is already considering utilizing nuclear energy to fuel its AI computation centers.