-
IBM’s CEO did some math on data centers and said it’s “impossible” to be profitable at current costs.
-
Arvind Krishna told Decoder: “$8 trillion in capital spending means you need about $800 billion in profits to pay interest.”
-
Krishna is skeptical that current technology can achieve AGI, putting the possibility at between 0-1%.
In the race for AGI, artificial intelligence companies are investing billions of dollars in data centers. IBM CEO Arvind Krishna has some thoughts on the math behind these bets.
Data center spending is increasing. In Meta’s recent earnings call, words like “capacity” and artificial intelligence “infrastructure” were used frequently. Google just announced that it wants to eventually build them in space. The question remains: Will the revenue generated by the data center be enough to justify all capital expenditures?
On the “Decoder” podcast, Krishna concluded that it may be “impossible” for these companies to generate returns on data center capital expenditures.
Kirshner, who said his napkin calculations are based on today’s costs “because anything in the future is speculative,” said it would cost about $80 billion to fill a 1-gigawatt data center.
“Well, that’s today’s numbers. So, if you’re committing to 20 to 30 gigawatts, that’s $1.5 trillion in capital expenditures from one company,” he said.
Krishna also mentioned the depreciation of AI chips within data centers as another factor: “You have to use it within five years, because by then, you have to throw it away and refill it,” he said.
Artificial intelligence stocks have tanked recently after investor Michael Burry took aim at Nvidia over concerns about its depreciation.
“If I look at the total commitments in the world chasing AGI in this space, those announcements seem to be in the 100-gigawatt range,” Krishna said.
The cost per 100 gigawatts is $80 billion, which puts Krishna’s computing commitment price at about $8 trillion.
“I don’t think you’re going to get a return because $8 trillion in capital spending means you need about $800 billion in profits to pay interest,” he said.
Reaching this gigawatt number will require significant investment from AI companies and will require outside help. OpenAI CEO Sam Altman recommended in an October letter to the White House Office of Science and Technology Policy that the United States add 100 gigawatts of energy capacity annually.