The CEO of one-time tech giant IBM has says that he doubts that data center mega-projects will prove profitable.

Speaking on the Decoder podcast, Arvind Krishna admitted that he was using napkin math and that "anything in the future is speculative."

With that margin of error, he said that a 1GW data center would cost around $80 billion to build and fill up with IT equipment.

"Okay, that's today's number," he said. "So, if you are going to commit 20 to 30GW, that's one company, that's $1.5 trillion of capex."

He added: "If I look at the total commits in the world in this space, in chasing AGI, it seems to be like 100GW with these announcements," which would cost around $8 trillion.

"It's my view that there's no way you're going to get a return on that, because $8 trillion of capex means you need roughly $800 billion of profit just to pay for the interest."

Some companies are turning to bond markets to fund their capex investments, but they are also selling equity, or funding it with cash from other operations - so it is not accurate to assume it is all debt with interest.

Krishna added that the chips inside the data centers also grow out of date quickly. "You've got to use it all in five years because at that point, you've got to throw it away and refill it," he said.

This is a matter of debate, with Nvidia and Google reporting that their hardware is still being used after six to seven years. Whether this will continue with the shift to an annual chip cadence, and the increase in supply, is unknown.

Generative AI leader OpenAI has pledged to spend as much as $1.4 trillion on the cloud and Stargate, despite currently not turning a profit. The company hopes to fund the buildout with a mixture of equity, debt, and - eventually - profits.

Longer term, OpenAI CEO Sam Altman said that the investment would handsomely repay its costs due to the benefits of AI.

"That's a belief," Krishna said. "That's what some people like to chase. I understand that from their perspective, but that's different from agreeing with them."

He put the chance of current technology achieving AGI at 0-1 percent, but added that AI is still "going to unlock trillions of dollars of productivity in the enterprise."

IBM, which has tens of billions of dollars in debt to repay, operates its own cloud service, but the platform is significantly smaller than the hyperscalers or long-time rival Oracle.