Banks Turn Cautious on AI Debt After a $2.5 Trillion Nasdaq Rout, Even When Google and Anthropic Back the Loans

For most of the past year, lending to artificial intelligence looked like one of the safest trades in credit. Chipmakers and cash-rich tech giants guaranteed the debt, data center tenants signed long leases, and banks lined up to write the checks. That mood has changed in a matter of days. According to a Reuters report published Monday, some banks are now telling market participants they would hesitate to fund additional AI deals, even ones backstopped by marquee names such as Anthropic and Google.

The pullback follows a brutal stretch for tech stocks and a midterm election that has raised the odds of tougher oversight of the data centers powering the AI boom. For business readers, it is an early sign that the cheapest, easiest money in the AI build-out may already have been raised.

What Changed, and How Fast

Reuters, citing several banking and credit sources, said the shift has happened only in the past few days. The trigger was a sharp drop in equities: the Nasdaq 100 lost about $2.5 trillion in market value on Thursday and slipped into a correction. Bloomberg, in a market wrap carried by American Banker, described the Nov. 5 slide as the index’s biggest since March 2020, driven by worries about growing political opposition to AI data centers.

Chris Poggi, a finance partner at Ropes & Gray, told Reuters that the past year brought “unbridled enthusiasm” to lend to neoclouds and anyone building data centers, and that debt investors’ books are now full. Lenders are reportedly stress-testing borrowers’ cash flows against a downturn and combing through contract details, including service obligations, construction deadlines, and the conditions under which big tech customers could cut back or cancel their commitments.

The Numbers Behind the Caution

The data shows why lenders are nervous about concentration. Reuters noted that at JPMorgan Chase, loans backed by “other income-producing properties,” a bucket that includes data centers, stood at $29 billion at the end of the third quarter. That is now the bank’s second-largest commercial real estate exposure at 12% of the book, behind only apartment buildings, up from 8% three years ago when the category ranked fourth. JPMorgan declined to comment.

Bond markets are sending the same message. Bank of America credit analyst Yuri Seliger told Reuters that spreads on 10-year hyperscaler debt have widened by about 50 basis points this year, compared with roughly 20 basis points for high-quality industrial companies with no AI ties. Meta bonds issued earlier this year and SpaceX bonds are trading below par, according to the report.

Yet the need for money keeps growing. Goldman Sachs Research projects global hyperscaler debt issuance of $420 billion in 2027, more than 60% above its $250 billion estimate for fiscal 2026, Reuters reported. In other words, supply is set to climb just as appetite cools.

Riskier Borrowers Feel It First

The squeeze is sharpest for neoclouds, the companies that rent out GPU computing capacity, and for standalone data center operators. Reuters cited Baird Equity Research naming Oracle and CoreWeave among the companies that could be hurt by construction delays or local moratoriums. CoreWeave shares are down 35% this year, and its bonds have dropped sharply in secondary trading.

Some planned deals are being shelved. Nebius explored a high-yield bond in September but decided last month not to proceed, a source told Reuters, though it did sign a $3.5 billion senior secured loan facility with Citigroup in October for data center development and GPU purchases. In Asia, OCBC Group CEO Tan Teck Long said on Friday that the bank structures deals to make sure the cash flow is there.

A Bond Market That Is Still Open, Just Pickier

This is not a credit freeze. Bloomberg reported that a heavy slate of new corporate bonds on Monday showed credit markets functioning smoothly, and stocks extended their rebound as oil fell more than 5% to two-month lows. The broader rate backdrop is also tense. CNBC reported the 2-year Treasury yield rose to about 4.56% and the 10-year to about 4.88% ahead of Tuesday’s October CPI report, after payrolls jumped by 165,000 jobs against 59,000 expected.

Monday’s $58 billion 3-year Treasury auction drew soft demand, stopping at 4.627% with a bid-to-cover of 2.53, the lowest since August 2025, Investing.com reported. The same report put the CME FedWatch odds of a quarter-point Fed hike at the Dec. 8-9 meeting at roughly 77%. Higher policy rates would raise the cost of every floating-rate loan funding the AI build-out.

Why It Matters for Business

For years, the AI trade was financed on the assumption that guarantees from the biggest names made the risk almost disappear. Lenders now appear to be pricing a different scenario: projects that slip, customers that renegotiate, and politicians who slow approvals. When banks question even guaranteed loans, the message is that terms, not just headlines, will decide who keeps building.

Several practical effects follow. Expect wider spreads and tighter covenants for neoclouds and independent data center developers, which favors balance-sheet-rich hyperscalers over leveraged challengers. Investment banks could see a dent in the loan, bond, and equity fees that the sector has generated. And companies that resell or depend on cheap AI compute, from software startups to enterprise buyers, may face higher prices if their suppliers’ funding costs rise.

Jay Bala, CEO of alternative asset manager AIP, told Reuters that most banks and credit investors will take a wait-and-see approach after the Democratic sweep of Congress. That pause may prove healthy if it forces better underwriting, but it also means that 2027’s projected record issuance will have to clear a far more skeptical market.

The Bottom Line

The AI debt boom is not over, but its easy phase may be. With hyperscaler spreads widening, data center loans taking a bigger share of bank books, and rate-hike odds climbing, the cost of capital is becoming a real constraint on who wins the infrastructure race. Watch Tuesday’s CPI print, this week’s 10- and 30-year Treasury auctions, and the next big data center financing for clues on how deep the caution runs.

Edited by Coron @QUE.COM
Website: https://QUE.COM Intelligence
Sponsored by: https://MAJ.COM AI Autonomous