Peter Smith
⬤ A recent Financial Times analysis shows the AI boom is being powered primarily by fossil fuels. In the US, natural gas dominates data-center electricity supply, while coal remains China's main energy source. The International Energy Agency estimates global data-center electricity consumption will jump from 460TWh in 2024 to over 1,000TWh by 2030 and around 1,300TWh by 2035. This makes energy cost and availability critical factors for AI infrastructure expansion worldwide.
⬤ The IEA data shows data centers will drive nearly 50% of US electricity demand growth through 2030. Even by 2035, fossil fuels are expected to supply over 40% of America's total power. US electricity prices have already climbed 38% since 2020, and the heavy reliance on gas-powered generation means prices will likely stay higher and more unpredictable compared to wind and solar, which keep getting cheaper. This creates real cost risks for companies since electricity is a direct expense for training and running AI systems.
⬤ China faces similar challenges but with a different energy mix. Coal powers most Chinese data centers today, with output projected to grow until around 2030 before leveling off as renewables take on more of the load. Chinese policymakers are moving some data-center capacity to coastal areas with better renewable resources to reduce long-term fossil fuel dependence. There's also a water problem: fossil-fuel plants and cooling systems use massive amounts of water, and about two-thirds of new US data centers built since 2022 sit in areas already dealing with water stress, which could worsen during droughts and conflict with agricultural needs.
⬤ The AI race is turning into an energy race. Long-term competitiveness won't just depend on computing power but on which regions can deliver cheap, clean electricity near major data hubs. As AI demand accelerates, the tension between fossil fuel reliance, renewable expansion, price stability, and water scarcity will increasingly shape where and how fast AI infrastructure can scale globally.
Peter Smith
