OpenAI and Microsoft AI boom is running out of juice

Just a heads up, if you buy something through our links, we may get a small share of the sale. It’s one of the ways we keep the lights on here. Click here for more.

Sam Altman and Satya Nadella have a problem most people can only dream of: too many chips, not enough power outlets. 

On a recent episode of the BG2 podcast, the OpenAI and Microsoft CEOs confessed that their companies are bumping up against a new kind of bottleneck, electricity.

“The biggest issue we are now having is not a compute glut,” Nadella said, “but power.” 

In other words, Microsoft has bought so many GPUs to feed the AI boom that it doesn’t have enough power, or finished data centers, to plug them in. 

“You may actually have a bunch of chips sitting in inventory that I can’t plug in,” Nadella admitted. “That is my problem today.”

For decades, software scaled faster than hardware, and hardware scaled faster than power plants. Now the tables have turned: data centers are demanding more energy than utilities can generate. 

US electricity demand stayed flat for years, but thanks to AI, it’s suddenly rocketing upward, so much so that some companies are skipping the grid entirely and wiring power straight to their data centers.

Altman, ever the futurist, sees storm clouds on the horizon. If cheap new energy suddenly comes online, say, through nuclear or solar breakthroughs, companies could get stuck with pricey long-term power contracts. 

Still, he’s betting big on the next wave: investing in nuclear startups Oklo and Helion, plus solar innovator Exowatt. 

None is ready for prime time yet, though, and natural gas plants take years to build.

That’s why Big Tech is racing to add solar power. It’s clean, fast to deploy, and made of silicon, just like the chips AI runs on. 

The parallels aren’t lost on Altman: both are modular, scalable, and theoretically limitless.

But even solar can’t keep up with the speed of AI’s appetite. Altman predicts demand will only skyrocket. Meaning: the cheaper the intelligence gets, the more we’ll use it. 

And right now, it’s the world’s power grid that’s struggling to keep up with the machines.