Concern over the sustainability of the AI investment boom has intensified as rising borrowing costs put heavily indebted companies such as CoreWeave under renewed scrutiny.
The cloud computing firm has seen its shares fall and the cost of insuring its debt rise, prompting fresh questions about whether the billions being spent on data centres and computing capacity will generate sufficient returns.
CoreWeave has become a particular focus because it relies heavily on borrowing to fund its expansion. Dan Greenhaus, strategist at Solus Alternative Asset Management, described the company as “the riskier of the riskier AI buildout play”.
He said the pressure on CoreWeave did not necessarily signal that the wider bull market was about to end, noting that previous warnings had been followed by further gains in technology shares and the broader market.
“How many canaries have we killed in the coal mine over the last three years as everybody has latched on to every new sign of weakness so to speak as evidence that the bull market was about to end only to see the broad market, both tech and otherwise, go on to make new highs?” Mr Greenhaus said.
Higher oil prices and the war in Iran have pushed bond yields to their highest level in more than 20 years, while the Federal Reserve has resumed raising interest rates to tackle persistent inflation. That has increased the cost of financing an AI sector that has become increasingly dependent on the bond market.
Mr Greenhaus said the sector’s more speculative companies could face mounting difficulties if rates continue to climb, although he added that earnings from AI chip suppliers such as Taiwan Semiconductor pointed to continued underlying demand.
“I don’t know whether this is the breaking point or another 25 or 50 basis points higher is the breaking point,” he said. “But as rates move higher, the more speculative tails so to speak of the buildout are going to be called into question, and right now that’s CoreWeave.”
CoreWeave’s chief executive, Mike Intrator, has argued that markets can “misunderstand and act irrationally”. In a memo to staff, he said the company’s backlog of deals was more than sufficient to meet its debt obligations.
Jensen Huang, Nvidia’s chief executive, also defended CoreWeave, in which Nvidia is a shareholder, saying the company did not require a financial backstop.
“CoreWeave’s strength is improving every day,” Mr Huang said. “I’m optimistic. Their cost of capital is actually decreasing.”
Some investors continue to view the recent sell-off as a buying opportunity, pointing to strong demand for AI technology. But concerns are growing that even exceptional revenue growth may not justify the scale of spending planned across the industry.
Gartner has estimated that AI data-centre operators will need to generate cumulative revenue of $8.5 trillion through 2030 to cover capital expenditure, financing costs and a 12 per cent return. It said enterprise and consumer AI revenue reaching data centres was likely to total no more than $3.6 trillion, leaving more than $4 trillion of asset value exposed to potential write-downs through 2032.
Bank of America analysts said the AI investment boom was following a familiar capital cycle, in which high early profit margins encourage overinvestment, excess capacity and increased competition.
That competition has intensified, particularly from Chinese open-weight models. At the same time, business spending on AI has begun to weaken as falling prices for tokens outweigh higher usage volumes.
Bank of America said the resulting pressure on monetisation, combined with intense competition, was “precisely the failure mode seen in past capital cycles”.
