IBM CEO claims there is 'zero chance' of recovering capital investment in AI data centers
AI companies are investing billions of yen in data centers, which are essential infrastructure for AI. However, IBM CEO Arvind Krishna warned that at current costs, it is 'impossible' to make a profit from these investments.
Why IBM's CEO doesn't think current AI tech can get to AGI | The Verge
IBM CEO warns there's 'no way' tech companies turn a profit on their billions in AI investment | Fortune
https://fortune.com/2025/12/03/ibm-ceo-no-way-hyperscalers-google-amazon-turn-profit-data-center-spending/
IBM CEO says there is 'no way' spending trillions on AI data centers will pay off at today's infrastructure costs
https://finance.yahoo.com/news/ibm-ceo-says-no-way-103010877.html
Spending on data centers, the cornerstone of AI infrastructure, is on the rise. Words like 'capacity' and 'infrastructure' were frequently mentioned during Meta's earnings call, and Google announced plans to build data centers in space in the future. The question is whether the revenue generated from data centers can justify all the capital expenditures.
Google proposes 'Project Suncatcher' to equip satellites with AI chips and launch them into space - GIGAZINE
Speaking on The Verge's Decoder podcast, Krishna said there's 'zero chance' that companies will ever recoup their capital investment in data centers.
Krishna estimates that it would cost about $80 billion to build a data center with a 1 gigawatt capacity. 'That's today's numbers. So if a single company wanted to do 20 to 30 gigawatts of computing power, it would require a capital investment of $1.5 trillion,' he said.
Goldman Sachs estimates that the global data center market's total power consumption will be approximately 55 gigawatts in early 2025, with AI accounting for 14% of that. Goldman Sachs predicts that as AI demand grows, the data center market's power needs will soar to 84 gigawatts by 2027.
In addition, CEO Krishna mentioned, 'If you add up the power consumption of companies pursuing artificial general intelligence (AGI) in this field, it seems to be about 100 gigawatts.' If the cost per gigawatt is $80 billion, the cost of 100 gigawatts is simply calculated to be $8 trillion (about 1,200 trillion yen). Therefore, '$8 trillion in capital investment would require a profit of about $800 billion (about 12 trillion yen) just to pay interest. I think that's impossible to make a profit on,' he said, arguing that investment in existing AI data centers is not profitable.
Krishna also pointed out that the depreciation of AI chips in data centers is another factor, saying, 'We'll have to throw away all of our AI chips and replenish them with new ones within five years.'
Krishna also stated that he is not convinced that AGI can be achieved using existing technology alone, estimating the likelihood of achieving it without further technological innovation at 0-1%. In fact, even prominent leaders in the AI industry have expressed skepticism, with Salesforce's Marc Benioff
Major AI companies are investing heavily in AI infrastructure, with the amount expected to reach $380 billion by 2025 alone.
Related Posts:
in AI, Posted by logu_ii