Altman and Nadella need more power for AI, but they’re not sure how much

How much power is enough for AI? No one knows, not even OpenAI CEO Sam Altman or Microsoft CEO Satya Nadella.

That puts software-first businesses like OpenAI and Microsoft in a bind. Much of the tech world has focused on computation as a major barrier to deploying AI. And so have the tech companies running from safe powerThese efforts have delayed GPU purchases to the point where Microsoft has apparently ordered too many chips for its contract volume.

“You can’t really predict the cycle of supply and demand in this particular case,” Nadella said BG2 Podcast. “The biggest problem we have right now is not counting, but it’s a power and the ability to get it. [data center] Builds are done fairly quickly near power.”

“If you can’t do that, you might actually have a bunch of chips sitting in inventory that I can’t plug in. In fact, that’s my problem today. It’s not a supply problem of chips; it’s that I don’t have a warm shell to plug in,” Nadella added, referring to the commercial real estate term for tenants ready for the building.

In some ways, we can see that while companies used to working with silicon and code, two technologies that scale and deploy faster than giant power plants, will have to increase their efforts in the energy world.

US electricity demand has been flat for more than a decade. But demand for data centers has grown over the past five years started to ramp upUtilities are outpacing plans for new generation capacity. This has led data center developers to add power to so-called behind-the-meter systems, where electricity is supplied directly to the data center, bypassing the grid.

Altman, who was on the podcast, thinks the problem could be brewing: “If very cheap energy comes online soon on a mass scale, a lot of people will be extremely burned with the existing contracts they signed.”

“If we can continue this incredible decline in the cost per unit of intelligence — let’s say it’s averaging like 40x per year for a given level — you know, that’s like a very scary indicator from an infrastructure building perspective,” he said.

Altman has invested in nuclear energy, including the fission startup ok and fusion startups HellionWith Exowatt, a solar startup that concentrates the sun’s heat and stores it for later use.

None of these are ready for widespread deployment today, however, and fossil-based technologies like natural gas power plants take years to develop. Also, orders placed today for new gas turbines likely won’t be fulfilled until the end of this decade.

That’s partly because of tech companies Adding solar at a fast clipAttracted to the technology’s low cost, emission-free energy, and ability to deploy quickly.

There may also be subconscious factors at play. Photovoltaic solar is in many ways a parallel technology to semiconductors, and one that is de-risked and commoditized. PV solar and semiconductors are both built on silicon substrates, and both roll off the production line as modular components that can be packaged together and tied into parallel arrays that make the whole part more powerful than any individual module.

Due to Solr’s modularity and speed of deployment, construction speed is much closer to that of a data center.

But both still take time to build, and demand can change much faster than completing a data center or solar project. Altman acknowledged that if AI becomes more efficient or if demand doesn’t grow as he expects, some companies may end up with idle power plants.

But from his other comments, he doesn’t think that’s likely. Instead, he seems convinced Jevons ParadoxWhich states that more efficient use of a resource will lead to greater use, increasing aggregate demand.

“If the cost of computing per unit of intelligence or whatever—however you want to think about it—drops by a factor of 100 tomorrow, you’re going to see usage go up by more than 100, and there will be a lot of things that people would prefer to do with that computation that don’t make any economic sense at current costs,” Altman said.