In the race for AGI, AI companies are spending billions on data centers. IBM CEO Arvind Krishna has some thoughts on the math behind those bets.
Data center expenses are increasing. During Meta’s recent earnings call, terms like “capability” and AI “infrastructure” were frequently used. Google has just announced that it eventually wants to build them in space. The question remains: will the revenue generated from data centers ever justify all the capital expenditures?
On the “Decoder” podcast, Krishna concluded that there was probably “no way” that these companies would be able to earn a return on their capital expenditures on data centers.
Saying that his napkin math was based on today’s costs, “because anything in the future is speculative,” Kirshna said it takes about $80 billion to fill a one-gigawatt data center.
“Well, that’s today’s numbers. So, if you’re going to do 20 to 30 gigawatts, that’s one company, that’s $1.5 trillion of capital expenditure,” he said.
Krishna also mentioned the depreciation of AI chips inside data centers as another factor: “You have to use it up in five years because at that time, you have to throw it away and refurbish it,” he said.
Investor Michael Burry recently took aim at Nvidia over depreciation concerns, sending AI shares tumbling.
“If I look at the total commitments in the world in this area pursuing AGI, it seems like 100 gigawatts with these announcements,” Krishna said.
$80 billion for every 100 gigawatts, which puts a value of about $8 trillion for Krishna’s computing commitments.
“I believe there’s no way to get a return on that, because $8 trillion of capital spending means you need about $800 billion of profit just to pay the interest,” he said.
To reach that gigawatt number, AI companies will need to spend massively — and insist on outside help. In a letter to the White House Office of Science and Technology Policy in October, OpenAI CEO Sam Altman recommended that the US add 100 gigawatts of energy capacity each year.
“Decoder” host Nilay Patel reported that Altman believes OpenAI can generate returns on its capital expenditures. OpenAI has committed to spend approximately $1.4 trillion across various deals. Here, Krishna said that he has separated from Altman.
“It is a belief,” Krishna said. “That’s what some people choose to pursue. I understand it from their perspective, but that’s different from agreeing with them.”
Krishna clarified that he was not convinced that the current set of technologies would take us to AGI, a technological breakthrough not yet reached, which is generally considered to be when AI is able to complete complex tasks better than humans. He estimated the probability of achieving this without any technological progress at 0–1%.
Many other high-profile leaders are skeptical about the AGI boom. Marc Benioff said he was “extremely skeptical” of the AGI push, considering it akin to hypnosis. Google Brain founder Andrew Ng said that AGI was “overhyped” and Mistral CEO Arthur Mensch said that AGI was a “marketing move”.
Even if AGI is the goal, scaling calculations may not be sufficient. OpenAI co-founder Ilya Sutskever said in November that the age of scaling is over, and that even 100x scaling of LLM would not be completely transformative. “It’s back to the era of research again, just with bigger computers,” he said.
Krishna, who began his career at IBM in 1990 and eventually became CEO in 2020 and chairman in 2021, praised the current set of AI tools.
“I think it’s going to unlock trillions of dollars of productivity across the enterprise, to be absolutely frank,” he said.
But AGI will require “more technologies than the current LLM path,” Krisha said. He proposed linking hard knowledge with LLM as a possible way forward.
What are the chances of it reaching AGI? “Still, I’m ‘maybe,'” he said.
<a href