Jensen Huang has a talent for making terrifying numbers sound like exciting business opportunities. During a lecture at Stanford’s CS153 Frontier Systems class, the Nvidia CEO dropped a figure that should keep energy planners up at night: the amount of energy needed for computing is roughly 1,000 times more than what’s currently available.
Not 10 times. Not 100 times. One thousand times. And Huang thinks even that estimate might be conservative.
Why 1,000X isn’t hyperbole
The key driver behind Huang’s jaw-dropping projection is a fundamental shift in how AI systems operate. Today’s computing model is largely on-demand. You ask a question, a server wakes up, fetches your answer, and goes back to sleep.
The future Huang envisions is different. He’s describing a world of continuous, always-on agentic AI systems, ones that don’t wait for prompts but instead operate autonomously around the clock.
