Debt-laden data center companies raise risk, sending bond yields soaring

With U.S. Treasury yields rising this week to their highest levels since 2007, companies that rely on debt are poised to face higher borrowing costs. This means that building AI infrastructure, which is already at historic levels, is becoming even more expensive.

JP Morgan Chase Data center companies and others involved in the artificial intelligence boom estimated in June that $4.1 trillion in AI-related debt will be issued by 2030, as many industry experts race to build capacity to meet what they see as an insatiable demand for AI services.

As borrowers return to the market, they are now 10 year government bond yield This is up about 1 percentage point since the beginning of the year and remains at around 5.17%, meaning companies issuing debt will need to offer more attractive rates of return to attract investors.

The market is not panicking, at least not yet. Neocloud stock with a lot of debt coreweave This week it has increased by almost 8% and is trending steadily, oracleThe company, which has relied on the bond market to fuel its AI expansion, has struggled further, falling 7% in the week and about 30% this year.

On the other hand, Japan SoftbankThe company, a major funder of AI projects, raised $11.1 billion this week in a junk bond sale, with the seven-year tranche yielding 9.75%.

“They’re basically price insensitive to that increase. So they’re price gougers,” Mark Marek, chief investment officer at Sievert Financial, said in an interview. “In my opinion, many of these companies need to be price sensitive. They need to get as much capital as possible to compete.”

Leading model developers are at the center of the AI ​​boom OpenAI and humaneach worth nearly $1 trillion on the private market. Because we provide the infrastructure needed for our own advanced models, as well as those of many other companies and high-tech hyperscalers, Amazon, google, meta and microsoft — Pledging hundreds of billions of dollars in capital investment this year, with increases expected in 2027.

Although a healthy amount of their investments are funded through debt financing, these tech giants all have investment-grade credit ratings and can access capital more cheaply. But for the remaining companies, bigger challenges lie ahead, according to some market participants.

A senior investor at a private credit firm, who requested anonymity to speak candidly about the issue, told CNBC that financing neocloud deals will become more difficult going forward because there is less cushion for companies to absorb costs.

Riley Thompson, a vice president at Mitsubishi HC Capital America, said in an interview that lenders are becoming more selective about the projects they fund, even when borrowers agree to pay higher interest rates.

“Instead of a roster of 50 neo-clouds, there are probably 20 that the market is really interested in,” Thompson said.

Coreweave, which went public last year, warned of rising interest rates in an SEC filing. The company said in its latest quarterly report that as of June, each 100 basis point (1 percentage point) increase in interest rates could increase interest expense by $30 million, based on its floating rate debt outstanding.

An early warning sign may have come when Oracle’s stock price fell this week following a Bloomberg report that Oracle sent a “force majeure” notice related to a data center project in New Mexico to protect itself from increased costs. The company is considering delaying payments on the campus, called Project Jupiter, if it does not come online as expected in 2028, the report said. Oracle said the project is “progressing as planned.”

Rising interest rates are not the only issue at hand. Ahead of this week’s spike in yields, CEOs at Anthropic and OpenAI had begun urging a slowdown in the pace of AI development after industry researchers publicly expressed concerns that advanced models risked becoming uncontrollable by humans.

At the same time, a national backlash against AI data centers has emerged as a major issue heading into the November midterm elections, with 69% of respondents in a recent NBC News Decision Desk poll conducted by SurveyMonkey saying they oppose building such facilities in their local area. On Monday, Texas Republican Gov. Greg Abbott, who is running a close re-election bid, ordered a suspension of all data center-related environmental permits, following a suspension of power grid approvals last month.

Still, the demand for AI services is exploding. The latest example is Meta’s personal assistant app Muse. It has skyrocketed in popularity since its release in early September. Muse topped ChatGPT with over 2.5 million downloads worldwide in its first two weeks. apple’s App Store and Evercore’s Mark Mahaney told CNBC this week that it 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 higher interest rates could impact future transactions, but he doesn’t expect them to have a major impact on borrower demand.

“In normal circumstances, people might step back and pause for a moment,” Giudici said. “But I don’t think that’s going to happen here. I think we’ll continue to see relatively large issuance.”

Chaim Salzman, vice chairman of Latham & Watkins’ emerging businesses and growth practice, said there’s no question that as costs rise, “someone’s going to have to absorb it.”

“But it’s much easier to absorb that into this kind of demand structure where the demand is so high,” said Salzman, who works on AI infrastructure financing.

For Bernie Margulies, CEO of American Compute, which advises on risk management for GPU funding, the equation is even simpler. He said borrowers are willing to secure financing at higher costs, especially if they have deals with OpenAI or Anthropic that secure computing power for years to come.

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

WATCH: Meta’s Muse user count could soar to 100 million