IBM CEO, Ratings Agency Sound Alarms On AI Data Center Spending

IBM CEO Arvind Krishna and Fitch Ratings are separately questioning the wisdom of trillions of dollars in planned data center spending by the world’s largest tech companies, adding a pair of credible voices to a rising chorus of concern about a potential AI bubble.

As Big Tech leaders describe their data center spending spree as demand-driven and already fueling revenue growth, Krishna emerged as a rare voice from within the tech sector’s top ranks to cast doubt on whether this unprecedented investment in artificial intelligence infrastructure will ever deliver appropriate returns.

Krishna, speaking on an episode of the Decoder podcast released this week, said there is “no way” tech giants will see enough new revenue to justify their data center spending spree. 

Krishna's skepticism aligns with a report published last week by Fitch Ratings that identified AI spending as a key bubble risk for the global economy. The report warns that a slowdown in hyperscale spending could have harmful implications.

“There is significant value in AI, but over-investment and over-valuation is a major risk with periods of correction and moderation likely,” Fitch analysts wrote. “Issuers with exposure to the AI investment boom go beyond the tech companies themselves, especially with data centres being a principal focus of investment.”

Amazon, Microsoft, Google and Meta are ramping up spending on new data centers, chips and other computing equipment. Their combined projected capex for 2025 is approaching $400B — tens of billions of dollars above what they predicted just months ago — and is expected to increase meaningfully in 2026.

Adding to this barrage of spending is growing capex from other AI cloud providers such as Oracle and CoreWeave, as well as build-out by AI-native end users like OpenAI.

The tech industry’s spending spree has continued even as rumblings of a looming AI bubble have grown louder. There is growing anxiety among investors, economists and industry observers that spending on data centers and chips could amount to a hype-driven boondoggle that never produces promised returns. 

Leaders of the tech giants driving this spending have been almost universally dismissive of these concerns, arguing that their escalating capex isn't a reckless spending spree but rather cautious capital deployment that is already driving revenue.

“You're going to see us continue to be very aggressive in investing in capacity because we see the demand,” Amazon CEO Andy Jassy said on the company's earnings call last month.

Krishna’s comments make him one of the lone tech leaders lending credence to the idea that Big Tech’s data center spending may never deliver its promised returns. 

By his estimates, it costs around $80B to build and provide the computing power for a one-gigawatt data center campus. With roughly 100 gigawatts of capacity being pursued across the data center landscape, according to Krishna, the total cost of providing that capacity would be close to $8T. 

“It's my view that there's no way you're going to get a return on that because $8T of capex means you need roughly $800B of profit just to pay for the interest,” he said.

Cost estimates also have to taken into account the short lifespans of the GPUs and other IT equipment that account for roughly half of AI capex, according to Krishna. As with traditional data center equipment, processors in AI data centers need to be replaced in regular cycles, typically every four to six years. Because of this, tech giants have only a few years to generate commensurate returns from their GPUs before they are thrown away and new ones are purchased. 

Krishna also responded directly to confidence being expressed by tech leaders — particularly OpenAI CEO Sam Altman — that they are on track to see a return on their data center spending. Such pronouncements, he said, amount to wishful thinking more than an evidence-based conclusion.

“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 didn't delve into the potential economic fallout should he be proven correct, but the stakes are enormous for the U.S. and global economies, according to the Fitch report. 

While the ratings agency stops short of concluding that the AI arms race has created an economic bubble, the firm suggests that the unprecedented pace of Big Tech’s AI infrastructure spending, which accounted for nearly 90% of U.S. GDP growth in the first half of this year, may be all that’s standing in the way of an economic downturn. 

“The robots have come to the rescue," Fitch Chief Economist Brian Coulton said in a statement. “The AI revolution has prompted additional private-sector spending on a scale that is heavily cushioning the adverse impact of tariff hikes on the US economy.”

If AI revenue falls short of expectations, tech giants' reduced capex could shake the financial system.

Valuations of market-propping tech firms would likely collapse. The spending slowdown would also impact sectors with data center exposure, such as REITs, utilities, asset-backed and mortgage-backed securities markets, and project finance lenders, according to Fitch.

The ratings agency conducted a “stress test” that modeled the impact of a 50% pullback in AI capex. The results suggest that declining capex represents a significant threat to tech hardware firms in particular, with companies like Dell and Seagate at the highest risk of credit downgrades. 

Still, the degree to which a slowdown in Big Tech’s data center spending could act as an economic contagion remains unclear. While the Big Tech giants are well positioned to withstand the fallout from an AI slowdown, according to Fitch, there is more uncertainty around newer, AI-specific entrants like OpenAI, Anthropic and xAI that account for a growing share of the data center market. 

These firms, as well as the growing ecosystem of neoclouds and AI startups, offer limited transparency into their balance sheets and counterparty risk. At the same time, the growing use of special purpose vehicles to push data center development off tech firms’ balance sheets also limits visibility into risk exposure for developers, lenders and tech firms alike in the event of a market downturn. 

“The opacity of some AI counterparties could raise credit risk,” Fitch analysts wrote. "Some corporates are turning to joint ventures and off-balance sheet vehicles to manage risk, further reducing transparency.”