The AI infrastructure boom has been financed on an epic scale. Now the bill is getting more expensive. Treasury yields climbed this week to their highest levels since 2007, and companies that rely on debt to fund data centers, chips and power capacity are poised to see their borrowing costs rise — meaning the AI buildout, already at historic levels, is about to get even pricier.
The scale of the dependency is enormous. JPMorgan Chase estimated in June that $4.1 trillion in AI-related debt will be issued through 2030, as data center operators and other companies tied to the artificial intelligence boom race to build capacity for what many industry experts view as insatiable demand for AI services. For more context on this story, see our ongoing artificial intelligence updates.
Rates Are Up, and So Is the Cost of Money
Borrowers returning to market now face a 10-year Treasury yield near 5.17%, up about one percentage point since the start of the year, according to CNBC. That means companies issuing debt must offer more attractive returns to lure investors, raising financing costs across the sector.
The market is not in panic mode — at least not yet. Shares of debt-heavy neocloud CoreWeave have held up fine, rising almost 8% this week. Oracle, which has counted on the debt market for its AI expansion, has had a tougher time: its stock fell 7% for the week and is down about 30% this year.
The week's marquee transaction showed what capital now costs. SoftBank, a principal provider of funding for AI projects, raised $11.1 billion in a junk-bond sale, with yields reaching as high as 9.75% on the seven-year tranche.
"They basically are price insensitive to that raise, which means they're price takers," Mark Malek, chief investment officer at Siebert Financial, told CNBC. "In my view, a lot of these companies need to be price insensitive. They need to get as much capital as possible to compete."
The Credit Divide: Hyperscalers vs. Everyone Else
At the center of the AI spending craze sit the leading model developers OpenAI and Anthropic, each valued at close to $1 trillion in the private market. To provision the infrastructure their advanced models require — along with models and services from a host of other companies — the tech hyperscalers Amazon, Google and Meta have committed historic sums.
A healthy share of that investment is funded through debt raises, but the giants carry investment-grade credit ratings, giving them cheaper access to capital. For the rest of the pack, the challenge is steeper. A senior private credit investor, speaking to CNBC anonymously to discuss the market candidly, said neocloud deals will become more difficult to finance going forward because those companies have less cushion to absorb higher costs.
Lenders are also getting pickier about which projects they will fund, even when borrowers accept higher rates. "Instead of a roster of 50 neoclouds, there's probably 20 that the market's truly interested in," said Riley Thompson, a vice president at Mitsubishi HC Capital America.
The sensitivity is not theoretical. CoreWeave, which went public last year, flags rising rates in its SEC filings. In its latest quarterly filing, the company said that as of June, every one-percentage-point increase in rates could add $30 million to its interest expense, based on its outstanding floating-rate balance.
Warning Signs From Oracle
An early tremor arrived this week, when Oracle's stock slid after a Bloomberg report that the company sent a "force majeure" notice tied to its New Mexico data center project — a move that would let it delay payment on the campus, dubbed Project Jupiter, as it seeks protection from higher expenses.
Rising rates are not the sector's only headwind. Even before this week's yield spike, the chief executives of Anthropic and OpenAI had begun urging a slowdown in the pace of AI development after industry researchers went public with concerns that advanced models risk spinning out of human control.
A nationwide backlash against AI data centers has also emerged as a major issue heading into November's midterm elections. In a recent NBC News Decision Desk Poll powered by SurveyMonkey, 69% of respondents said they oppose construction of such facilities in their local area.
Demand Still Outruns Doubt
Yet the demand picture keeps defying the bears. The latest example is Meta's Muse personal assistant app, which has rocketed in popularity since launching earlier in September, logging more than 2.5 million global downloads in its first two weeks and passing ChatGPT at the top of Apple's App Store.
"In a normal environment, people might take a step back and pause a bit," Andrew Giudici, global head of corporate, project and infrastructure finance at credit rating agency KBRA, told CNBC. "But I don't think that's going to happen here. I think you're going to continue to see relatively large issuance."
Haim Zaltzman, vice chair of Latham & Watkins' emerging companies and growth practice, said there is no doubt that as costs rise, "somebody will have to absorb it." But absorbing it in a demand structure this strong, he argued, is far easier than it would be in a weaker market.
For Bernie Margulies, CEO of American Compute, which advises on risk management for GPU financing, the math is simpler. Borrowers are eager to lock in financing even at higher costs, especially those with commitments to OpenAI and Anthropic, which have been signing contracts to secure compute capacity years in advance.
"If you have a deal with Anthropic, will 50 basis points really stop you?" Margulies said.
For now, the answer across the AI complex appears to be no. But with yields at levels not seen since before the global financial crisis, the era of cheap money that supercharged the buildout is ending — and every future data center, GPU cluster and power plant will carry a higher price tag as a result.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
