AI power demand could stifle industry’s growth | FT Energy Source

News

You can enable subtitles (captions) in the video player

AI’s energy use is set to rocket. That could not only constrain the industry’s growth, but threaten green energy targets and force AI providers to consider alternative sources of power. The International Energy Agency estimates global power demand from data centres, AI, and cryptocurrencies could double by 2026, compared to 2022 levels. Microsoft alone is opening a new data centre globally every three days.

The computational power required to develop machine-learning models has doubled every five to six months since 2010. In March, billionaire Elon Musk said the latest bottleneck to cutting-edge AI technology was electricity supply. Amazon chief Andy Jassy said earlier this year that there was not enough energy right now to run new generative AI services. Also, many customers of these data centres have made net zero CO2 commitments, but renewable sources of energy may not be able to cope with the load and reliability of supply needed.

Natural gas bosses are betting it could be a boon for their industry, which they say is the only way to quickly deliver the reliable power needed. But concerns around emissions have driven interest among data centre developers in options such as onsite power generation and nuclear energy. This year, for example, Microsoft hired its first director of nuclear technologies.

Nuclear energy, shunned in many parts of the world because of safety concerns and cost overruns on traditional reactors, has pitched its next-generation technology as a solution. So-called Small Modular Reactors, or SMRs, and even smaller microreactors, are being designed to be produced quickly, cheaply, and at scale. They don’t emit CO2, but the earliest SMR developments are not set to be commercialised until the end of the decade.

So, for now the industry is focusing on more close-at-hand atomic energy solutions, including striking deals to use existing nuclear plants to power data centres. For example, Amazon’s AWS revealed in March that it had paid $650mn for a data centre campus connected to Talen Energy’s Susquehanna nuclear plant in Pennsylvania. But experts believe that ultimately data centres could prove the perfect growth catalyst for SMRs, given Big Tech companies are among the few with sufficient capital to invest in the first iterations needed to get the reactors off the ground.