Sam Altman wrote yesterday about a future of abundant intelligence, imagining a world where we add a gigawatt of new AI infrastructure every week. This week, we already saw their partnerships with Nvidia and Oracle. He teased about more partnerships and details coming soon:

“If AI stays on the trajectory that we think it will, then amazing things will be possible. Maybe with 10 gigawatts of compute, AI can figure out how to cure cancer. Or with 10 gigawatts of compute, AI can figure out how to provide customized tutoring to every student on earth. If we are limited by compute, we’ll have to choose which one to prioritize; no one wants to make that choice, so let’s go build.”

It’s a striking way to frame the scale of what’s coming: not in terms of chips or dollars, but in raw power capacity. But what does a gigawatt actually mean in everyday terms—and what does it unlock for AI?

Breaking Down a Gigawatt

A gigawatt is one billion watts of continuous power. Numbers that large can feel abstract, so let’s put it in context:

  • Homes: One gigawatt running nonstop for a month can power over 750,000 U.S. homes.

  • Light bulbs: It could keep roughly 100 million LED bulbs shining.

  • Laptops: Enough to charge around 10 million laptops at once.

Now scale that up. Ten gigawatts, operating year-round, would supply 8 million American homes for a full year.

Why Power Matters for AI

Today’s largest AI training runs already consume tens to hundreds of megawatts. Moving to gigawatt-scale AI means building infrastructure that rivals the energy footprint of whole cities. If Sam Altman’s vision of “a gigawatt a week” became reality, we would be adding the equivalent of a new metropolitan power system every seven days, all dedicated to intelligence.

This reframing matters because it shifts the conversation. The bottleneck for AI progress is no longer just better algorithms or more efficient chips. It’s also about energy, grid capacity, and the geopolitics of resource allocation.

What Becomes Possible

At gigawatt scale, AI systems could:

  • Accelerate medical breakthroughs, from drug discovery to personalized cancer treatments.

  • Deliver universal, personalized tutoring at negligible cost.

  • Simulate entire economies or ecosystems in real time to improve policy and planning.

The real leap is not any single application, but the sheer abundance of compute that enables exploration across thousands of domains simultaneously.

For product managers and technologists, the key insight is that the next breakthroughs may hinge less on clever model tweaks and more on infrastructure scaling. Questions worth asking now:

  • Who controls the supply of energy and compute at this scale?

  • Where are the bottlenecks in data center construction and energy delivery?

  • What new products become feasible when compute is no longer scarce but abundant?

Closing Thought

One gigawatt may sound like an engineering abstraction. But in human terms, it’s millions of homes powered, or tens of millions of devices running. Thinking of AI in those terms forces us to recognize the scale of what’s coming: intelligence as an energy industry, not just a software industry.