US President Donald Trump recently stunned attendees at an AI summit in Washington, DC with his remarks about the huge amounts of electricity that will be needed to support his AI expansion plan. Speaking at the “Winning the AI Race” event, Trump openly questioned Nvidia CEO Jensen Huang about the great power demands of building AI data centers, saying, “You’re going to need more electricity than any human beings ever in the history of the world.”
Trump’s comments came as he unveiled his vision for expanding AI capabilities nationwide, but even he seemed taken aback by the extent of the power this would require. He joked about needing an explanation from Jensen someday on why so much energy is necessary, adding, “Couldn’t you do it with a little bit less? My father always used to say, ‘turn off the lights son.’ But you guys are turning up the lights.”
These remarks aren’t just empty words. Official US figures from the Energy Information Administration show that in 2023, the US produced 4,178 billion kWh of electricity, with a majority coming from fossil fuels and a growing share from renewables and nuclear. Doubling that output just for AI would be an enormous challenge. It’s unclear if Trump’s estimate refers solely to AI data centers or includes the entire US power demand, but either way, the scale is staggering.
Data centers running AI workloads utilize GPUs that can consume significant power. Nvidia’s upcoming Blackwell AI GPUs can consume up to 1200 watts each, while the popular H100 models draw around 700 watts. Multiply that by thousands or even millions of GPUs being deployed by companies like Meta, xAI, and OpenAI, and you get a picture of why electricity demand might skyrocket.
Trump also praised US Secretary of the Interior Doug Burgum for efforts to lower energy costs, mentioning a goal to reduce electricity prices below $64 per megawatt-hour. Still, expanding power generation to meet AI’s appetite will require more than just a few extra electricians or “sparkies.”
US energy demand was already expected to grow by about 2% annually between 2025 and 2027, but this new AI surge could push that far higher. While renewable energy sources, such as wind and solar, have recently overtaken coal in electricity production, the closure of coal plants and a reliance on methane gas turbines raise questions about how the gap will be filled.
Interestingly, nuclear power might make a comeback as part of the solution. New York announced plans for the first major new nuclear power plant in 15 years, potentially signaling a shift toward more stable, high-output energy sources to support AI infrastructure.
With AI data centers cropping up rapidly and NVIDIA chips selling out, big announcements about power infrastructure upgrades seem likely in the near future. Until then, if you live in the US, maybe consider turning off some lights early; every bit helps when the grid might be under serious strain.
What do you think? Could the US power grid handle such a massive increase in electricity demand? Will renewable energy keep pace, or will older, dirtier methods make a comeback? Speak for yourself in the comments below!