AI debt costs rise as Treasury yields rise

As Treasury yields climb this week to their highest levels since 2007, debt-dependent companies are poised to see their borrowing costs rise. This means that building AI infrastructure, which has already reached historic levels, is about to become even more expensive.

JPMorgan Chase estimated in June that $4.1 trillion in AI-related debt would be issued through 2030, as data center companies and others are linked to the artificial intelligence boom to build capacity to meet what many industry experts see as insatiable demand for AI services.

As borrowers return to the market, they now face a 10-year Treasury yield this stands at almost 5.17%, up about 1 percentage point since the start of the year, meaning that companies issuing debt will have to offer more attractive rates of return to attract investors.

The market is not in panic mode, at least not yet. Shares of heavily indebted Neocloud CoreWeave held up well, up nearly 8% this week, while Oraclewhich relied on the debt market to develop its AI, had a more difficult time, falling 7% over the week and around 30% this year.

Meanwhile, Japan SoftBanka leading provider of capital for AI projects, this week raised $11.1 billion in a junk bond sale, with yields as high as 9.75% for the 7-year tranche.

“They are basically price insensitive to this rise, which means they are price takers,” Mark Malek, chief investment officer at Siebert Financial, said in an interview. “In my opinion, a lot of these companies have to be price insensitive. They need to get as much capital as possible to be competitive.”

At the center of the AI ​​craze are leading model developers OpenAI And Anthropicwhich are each valued at nearly $1 trillion in the private market. To provide the infrastructure needed for their advanced models, as well as the models and services of a multitude of other companies, technology hyperscalers — Amazon, Google, Meta And Microsoft – have committed to investing hundreds of billions of dollars this year in capital spending, with an increase expected in 2027.

Although much of this investment is financed through debt raises, these tech giants all benefit from investment-grade credit ratings, providing them with access to capital at lower costs. But for the rest of the pack, bigger challenges lie ahead, according to some market players.

A prominent private investor, who asked to remain anonymous in order to speak candidly on the matter, told CNBC that in the future, NeoCloud deals will be harder to finance because companies have less room to absorb costs.

Riley Thompson, vice president of Mitsubishi HC Capital America, said in an interview that lenders are increasingly demanding about the projects they are willing to finance, even if the borrower agrees to pay a higher rate.

“Instead of a list of 50 neoclouds, there are probably 20 that the market is really interested in,” Thompson said.

CoreWeave, which went public last year, warns of rising rates in its SEC filings. In its latest quarterly report, the company said that starting in June, each 100 basis point (1 percentage point) increase in rates could result in a $30 million increase in its interest expense, based on its variable-rate debt balance.

A warning sign may have emerged this week, when Oracle shares fell following a Bloomberg report that the company had sent a “force majeure” notice related to its data center project in New Mexico to protect itself from higher expenses. The company plans to delay payment at the campus, dubbed Project Jupiter, if it fails to come online as planned in 2028, the report said. Oracle said the project “remains on schedule.”

Rising interest rates aren’t the only problem. Before this week’s yield surge, the CEOs of Anthropic and OpenAI had begun advocating a slowdown in the pace of AI development after industry researchers went public with concerns about the risk of advanced models slipping out of human control.

At the same time, a nationwide backlash against AI data centers has become a major issue in the run-up to the November midterm elections, with 69% of respondents to a recent NBC News Decision Desk poll, conducted by SurveyMonkey, saying they are opposed to building such facilities in their area. On Monday, Texas Republican Gov. Greg Abbott, who is in the midst of a tight reelection race, ordered a temporary halt to all environmental permitting related to data centers following a moratorium on network approvals last month.

Yet demand for AI services is exploding. The latest example is Meta’s personal assistant app Muse, which has grown in popularity since its launch in early September. Muse recorded more than 2.5 million downloads worldwide in its first two weeks, surpassing ChatGPT at the top of Apple App Store and Evercore’s Mark Mahaney told CNBC this week they could reach 100 million users within six to 12 months.

Andrew Giudici, global head of corporate, project and infrastructure finance at credit rating agency KBRA, said that while rising rates may affect future transactions, he doesn’t see a major impact on borrower demand.

“In a normal environment, people might step back and take a break,” Giudici said. “But I don’t think that’s going to happen here. I think you’re going to continue to see relatively large shows.”

Haim Zaltzman, vice chair of Latham & Watkins’ emerging companies and growth practice, said there’s no doubt that as costs rise, “someone is going to have to absorb them.”

“But absorbing it into this type of demand structure, where the demand is so great, is much easier,” said Zaltzman, who works on financing AI infrastructure.

The equation is even simpler for Bernie Margulies, CEO of American Compute, who advises on managing GPU financing risks. He said borrowers are eager to obtain financing even at higher costs, especially if they have commitments with OpenAI and Anthropic, which have signed contracts to guarantee computing capacity in the coming years.

“If you have a deal with Anthropic, will 50 basis points really stop you?” said Margulies.

WATCH: Meta’s Muse will likely reach 100 million users