The AI boom has sparked an unprecedented spending spree on data centers, with tech giants pouring hundreds of billions into infrastructure. But one prominent voice is raising serious doubts about whether these massive investments will ever pay off.
IBM CEO Arvind Krishna recently shared some sobering calculations on the “Decoder” podcast that paint a challenging picture for companies chasing artificial general intelligence. His back-of-the-envelope math suggests the industry might be heading toward a financial dead end.
Is $8 Trillion in Data Center Spending Sustainable?
Krishna broke down the numbers in straightforward terms. Building and filling a one-gigawatt data center costs roughly $80 billion at today’s prices. When companies commit to 20 or 30 gigawatts, something a single major player might do, that translates to $1.5 trillion in capital expenditure.
The math gets even more daunting when you zoom out. Krishna estimates that global commitments in the race toward AGI total around 100 gigawatts. At $80 billion per gigawatt, the industry is looking at approximately $8 trillion in total spending.
“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,” Krishna explained.
The challenge doesn’t stop at the initial investment. Krishna pointed out another critical factor: the AI chips powering these data centers have a limited lifespan. Companies have roughly five years to extract value before the hardware becomes obsolete and needs replacement.
This concern has caught the attention of investors beyond the tech world. Michael Burry recently took aim at Nvidia over depreciation issues, contributing to volatility in AI stocks.
The spending frenzy shows no signs of slowing down. Meta’s recent earnings calls were peppered with references to “capacity” and AI “infrastructure.” Google has even floated the ambitious idea of eventually building data centers in space.
OpenAI CEO Sam Altman has been particularly aggressive, committing to roughly $1.4 trillion across various deals. In an October letter to the White House’s Office of Science and Technology Policy, Altman recommended the US add 100 gigawatts in energy capacity every year to support AI development.
IBM CEO Krishna Skeptical of AGI Progress and Current LLM Path
When asked about Altman’s confidence in generating returns on these expenditures, Krishna offered a diplomatic but skeptical response. “That’s a belief,” he said. “That’s what some people like to chase. I understand that from their perspective, but that’s different from agreeing with them.”
Krishna’s doubts extend beyond just the financial calculations. He questioned whether today’s technologies can actually deliver AGI, the holy grail of AI development where machines can perform complex tasks better than humans across the board.
He pegged the chances of achieving AGI without a major technological breakthrough at just 0-1%. According to Krishna, reaching that goal will require “more technologies than the current LLM path.” He suggested that fusing hard knowledge with large language models might offer a potential route forward, though even then, he remained cautious: “I’m a ‘maybe.'”
Krishna isn’t alone in questioning the AGI rush. Several high-profile figures have expressed similar doubts. Salesforce CEO Marc Benioff said he was “extremely suspect” of the AGI push, comparing it to hypnosis. Google Brain founder Andrew Ng called AGI “overhyped,” while Mistral CEO Arthur Mensch dismissed it as a “marketing move.”
Why OpenAI’s Sutskever is Pushing Beyond Scaling, and IBM’s Krishna Sees Trillions in Value
Even OpenAI cofounder Ilya Sutskever suggested in November that simply scaling up compute power won’t be enough. He declared that the age of scaling was over, noting that even a 100-fold increase in large language model capacity wouldn’t be completely transformative. “It’s back to the age of research again, just with big computers,” Sutskever said.
Despite his skepticism about AGI timelines and returns on investment, Krishna emphasized that current AI tools still hold tremendous value. “I think it’s going to unlock trillions of dollars of productivity in the enterprise,” he said.
The IBM chief’s perspective reflects a pragmatic middle-ground enthusiasm for AI’s near-term business applications, tempered by realism about the extraordinary costs and uncertain path to achieving true artificial general intelligence. Whether the industry’s biggest players share his caution or continue betting big on AGI remains to be seen.
