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 has put software-first companies like OpenAI and Microsoft in trouble. Much of the technology world is focused on computing as a major barrier to AI deployment. And as tech companies race to secure power, those efforts have lagged behind GPU purchases to the point where Microsoft has apparently ordered too many chips for the amount of power it has contracted.

“You can’t really predict the cycles of supply and demand in this particular case,” Nadella said on the website BG2 podcast. “The biggest problem we face now is not a surplus of computers, but it is a force and a kind of ability to [data center] builds close to power quickly enough.”

“If you can’t do that, you may have a bunch of chips in your inventory that I can’t plug in. That’s basically my problem today. It’s not a supply of chips issue; it’s the fact that I don’t have any hot shells to plug into,” Nadella added, referring to the term commercial real estate for buildings that are ready for tenants.

In a sense, we’re seeing what happens when companies accustomed to dealing with silicon and code, two technologies that can scale and deploy quickly compared to massive power plants, have to step up their efforts in the energy world.

For more than a decade, U.S. electricity demand was flat. But over the past five years, demand from data centers has increased, outpacing utilities’ plans for new generation capacity. That has led data center developers to add power in so-called behind-the-meter arrangements, where electricity is delivered directly to the data center and bypasses the grid.

Altman, who was also on the podcast, thinks problems could arise: “If a very cheap form of energy comes online on a large scale soon, a lot of people are going to be extremely burned by the existing contracts they have signed.”

“If we can continue this incredible reduction in the cost per unit of intelligence – let’s say it’s on average 40 times higher for a given level per year – that’s a very scary exponent from an infrastructure buildout perspective,” he said.

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

However, none of these are ready for widespread deployment today, and fossil fuel-based technologies such as natural gas plants take years to build. Furthermore, orders placed today for a new gas turbine will likely not be fulfilled until later this decade.

That’s partly why tech companies have been rapidly adding solar energy, attracted by its cheap cost, emissions-free energy and ability to deploy quickly.

Unconscious factors can also play a role. Solar photovoltaics is in many ways a parallel technology to semiconductors, one that has been de-risked and commoditized. Both PV solar panels and semiconductors are built on silicon substrates, and both roll off production lines as modular components that can be packaged together and linked into parallel arrays, making the finished part more powerful than any individual module.

Due to the modularity and speed of solar deployment, the construction pace is much closer to that of a data center.

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

But from his other comments he doesn’t seem to think that’s likely. Instead, he seems to believe in it wholeheartedly Jevons paradoxstating that more efficient use of a resource will lead to greater use, increasing overall demand.

“If the price of computers per kind of unit of intelligence or whatever – no matter how you think about it – were to fall by a factor of 100 tomorrow, you would see usage increase by much more than 100 and there would be a lot of things that people would love to do with that computing power that just don’t make economic sense at today’s costs,” Altman says.