The exponential growth of artificial intelligence — if it continues — could push total AI peak power requirements to 50 gigawatts by 2030, according to a new analysis by the Electric Power Research Institute.
That figure would exceed half of the energy capacity of all U.S. nuclear plants, for comparison. To gain deeper insights into future power demand, the EPRI report focuses on AI training, when computer models are given huge libraries of text and data and instructed to find patterns and relationships that can turn the models into problem-solving machines.
While the report lists things that may limit electricity demand from AI trainers, it focuses on today’s surge in electricity demand from OpenAI, Meta, Amazon, Google and Microsoft.
Grid operators also face the uncertain impacts of President Donald Trump’s AI strategy, which aims to secure American leadership in the technology by paving the way for tech companies to build some of the world’s biggest data centers. That could include projects on federal land.
