Energy-hungry data centres may have to go nuclear to meet rising demands for artificial intelligence (AI) if the world wants to keep its word and cut greenhouse-gas emissions, experts say.
With AI takeup growing at a clip, big tech - from Microsoft to Google - is looking at nuclear energy as a possible new power source of the vast storage units that power its products.
From China to the United States, tech giants are eyeing the many gains that AI promises - whatever power it may swallow - while campaigners question the potential environmental cost.
Here is all you need to know about nuclear energy and its future in the booming AI sector.
How much power does AI consume?
AI is swallowing ever more energy as global takeup booms.
Over the past year, big technology companies have been racing to build data centres needed to power their generative AI applications, complicating government efforts to decarbonise the power sector and combat climate change.
The needs of Big Data are huge.
Every time an AI tool analyses data or answers questions, it uses a graphic processing unit - an electronic circuit that can process multiple pieces of data simultaneously - that is usually deployed in servers in data centres to boost computing power.
A simple ChatGPT query, for example, uses 10 times more energy than a Google search.
And it takes as much energy to generate one simple AI image as it does to charge a smartphone, according to research by the AI startup, Hugging Face, and the Carnegie Mellon University.
This thirst for power is only set to grow; a new study predicts AI could use as much energy as the Netherlands by 2027.
The information and communication technology (ICT) sector already emits 2 per cent to 4 per cent of global carbon emissions, according to 2020 research from Lancaster University, with that share set to rise as takeup of AI grows.
Is nuclear a sustainable solution?
The carbon footprint of AI depends on where it is used.
AI-use in fossil-fuel reliant countries, such as the United States, leaves a larger carbon footprint than in countries such as France, whose grid runs on cleaner energy.
Nuclear energy produces no greenhouse gas emissions during operation. With the same impact as wind when it comes to its carbon dioxide output, nuclear energy emits a third of the emissions as solar, according to the World Nuclear Association.
The International Energy Agency says using nuclear would not only cut emissions but make the transition to net zero cheaper.
What is the current state of play?
Interest in using small modular reactors is rising among tech companies such as Microsoft, as they can be built at a fraction of the cost of traditional reactors and at speed, too.
So far only China has deployed the technology.
More ambitious plans, backed by OpenAI’s CEO Sam Altman, bank on a breakthrough in fusion energy to provide almost limitless, carbon-free electricity that produces no long-lived radioactive nuclear waste.
The International Atomic Energy Agency (IAEA) says it could generate four times more energy than current fission technology.
Nuclear power plants could also benefit from AI - automating routine tasks, bolstering efficiency, cutting fuel consumption and maximising reactors’ output, according to the IAEA.
What are the concerns though?
Nuclear power fell sharply out of favour after accidents at Ukraine’s Chernobyl and Japan’s Fukushima plants, in 1986 and 2011 respectively, raised global alarm bells over safety.
In Japan, more than a million metric tons of water - enough to fill 500 Olympic-sized pools - was contaminated from contact with fuel rods at the reactor.
The Fukushima disaster prompted Germany to shut six of its nuclear plants and phase out remaining reactors.
Accidents aside, waste generated by nuclear plants, be it through uranium mining or reprocessing spent fuel, stays radioactive for hundreds of thousands of years, creating storage problems for generations, says the environmental NGO Greenpeace.
Reactors also come with a steep price tag and a long build.
Overall costs are in the billions, requiring hefty taxpayer investment, and projects then often rack up significant delays.
The newest US nuclear power reactors, at the Vogtle plant in Georgia, were years behind schedule and billions over budget when they entered commercial operation in 2023 and 2024.
European efforts have met similar delays.
French utility EDF has pushed back the start date of its Hinkley Point C reactor in Britain to at least 2029, although it was originally planned to be powering homes by 2017.
Add to that the cost of decommissioning a nuclear plant at the end of its useful life, between 40 and 60 years. The costly, labour-intensive process can top a billion dollars.
This story was published with permission from Thomson Reuters Foundation, the charitable arm of Thomson Reuters, that covers humanitarian news, climate change, resilience, women’s rights, trafficking and property rights. Visit https://www.context.news/.
