NEW YORK — Banks and credit investors are pulling back from the debt that has financed the artificial intelligence boom, tightening their scrutiny of new loans and growing reluctant to increase their exposure to the sector, just days after a sharp stock market selloff erased $2.5 trillion from the Nasdaq 100 and sent the index into correction. The shift, reported by Reuters on Monday, 9 November 2026, is the first broad sign that the lenders who bankrolled the AI build-out are asking harder questions about repayment.
Illustration: Bankers are taking a harder look at AI loans after a sharp market selloff changed the mood in credit markets.
What Happened
Some banks that only a few weeks ago were lining up to fund AI deals are now telling market participants they would be reluctant to finance additional deals, even with guarantees from marquee technology companies, according to several banking and credit sources cited by Reuters. The change, the sources said, has come in just the past few days.
The lending pullback has been worsened by the recent drop in the stock market, which erased $2.5 trillion from the Nasdaq 100 on Thursday and sent the index into correction. "Over the past year there was unbridled enthusiasm to provide debt to neoclouds, anyone who was building data centers or some sort of AI asset," Chris Poggi, a finance partner at the law firm Ropes & Gray, told Reuters. "Debt investors' books have become full and now they are assessing if they have extended themselves as much as they're going to extend themselves this year."
Lenders are now stress-testing cash flows for economic downturns and scrutinising the contractual details behind AI infrastructure projects, including the conditions under which large technology customers could reduce or terminate their commitments, according to two of the sources cited by Reuters. Many of these deals are complex: chipmakers typically backstop the senior debt that finances the purchase of advanced chips and also guarantee payments on the long-term lease of the data centers. Banks had been attracted to such loans precisely because of those guarantees — but they are now reading the fine print.
The pricing of AI debt already reflects the new mood. The spread on 10-year hyperscaler debt has widened by about 50 basis points so far this year, compared with roughly 20 basis points for high-quality industrial companies not tied to AI, according to Bank of America credit analyst Yuri Seliger. Bond prices of Meta, issued earlier this year, and of SpaceX are trading below par. At JPMorgan Chase, loans backed by properties including data centers stood at $29 billion at the end of the third quarter, 12 percent of its commercial real estate book, up from 8 percent three years ago.
The pullback has been most pronounced among the riskier AI borrowers, such as neoclouds. Oracle and CoreWeave are among the neocloud companies that could be affected by possible construction delays or moratoriums, according to Baird Equity Research. CoreWeave's stock has fallen 35 percent so far this year. Some planned bond offerings are being shelved: Nebius had been exploring a high-yield bond issue in September but decided not to go ahead with it last month. Nebius did, however, enter into a $3.5 billion senior secured loan facility provided by Citigroup in October. "Most banks and credit investors will take a wait and see approach" on the outlook after the midterm elections, Jay Bala, of asset manager AIP, told Reuters.
Background
This story is the sequel to the market rout, not a repeat of it. This blog reported on 6 November (post #841) on the fall itself: the Nasdaq sliding into correction and Wall Street's fear gauge jumping to an eight-month high as AI shares sold off. What Reuters documented on 9 November is what happened next: the lenders who funded the boom stepping back, and borrowing costs rising.
Borrowing for artificial intelligence has more than doubled in 2026, according to research firm Reeve Hayward. The biggest technology companies, the so-called hyperscalers, now account for almost 10 percent of global investment-grade corporate bond issuance, up from 0.5 percent in 2024. The five largest hyperscalers — Alphabet, Amazon, Meta, Microsoft and Oracle — invested $416 billion in 2025, and Goldman Sachs Research projects global hyperscaler debt issuance of $420 billion in 2027, up from $250 billion for fiscal 2026.
At the centre of the lending boom sits the neocloud: a company that rents out powerful AI computing capacity, often built around advanced chips, to businesses developing AI. Neoclouds typically lack the cash to buy the chips themselves, so they borrow heavily and use long-term contracts with big technology customers — and guarantees from chipmakers — as the security that makes the loans possible. For more than a year the structure worked. Then, in October, the warning signs began to pile up.
Illustration: Data center construction depends on borrowed money, and lenders are now asking tougher questions about who will repay it.
On 14 October, the research firm Gartner forecast that more than 80 percent of neoclouds would stop operating independently by 2030, and warned that more than $3.5 trillion of cumulative AI infrastructure spending through 2030 could face substantial asset impairments. The next day, the price of insuring CoreWeave's debt against default — its five-year credit default swap — passed 1,000 basis points, a level generally read as distress, and touched 1,140 basis points, CNBC reported on 15 October. Reuters put the implied chance of default within five years at about 63 percent, up from about 30 percent a year ago. By 21 October, CNBC reported that CoreWeave's unsecured debt was yielding about 15 percent and its five-year default-swap spread stood at nearly 1,067 basis points, while about $20.7 billion of its debt falls due through 2030, with another $14.9 billion after that.
The banks themselves were already sounding the alarm. JPMorgan Private Bank strategist Justin Biemann wrote on 23 October that "companies must show that they can finance investment, service their debt and eventually turn spending into cash flow." JPMorgan's data showed AI-linked bond issuance had already hit $266 billion in 2026. Oracle carries $125 billion of borrowings and $288 billion of data center leases that have not yet started, while CoreWeave pays interest equal to 23 percent of its revenue. The International Monetary Fund now lists an AI-driven market correction as one of three salient risks to the financial system, the report noted.
Why It Matters
It matters because the AI build-out cannot continue at its current pace without borrowed money. If financing slows, construction slows — and a slowdown would curb loan growth and dent the investment banking fees from the loans, bond sales and stock offerings that financed the sector's expansion, Reuters reported.
It also matters because the guarantees that made AI loans look safe are only as strong as the contracts behind them. Lenders are now examining the conditions under which large technology customers could reduce or terminate their commitments. If a big customer can walk away from a lease, or delay taking delivery of capacity, the cash flow that repays the loan disappears — leaving the lender holding a claim on fast-depreciating chips.
The strain is not evenly spread. The strongest borrowers can carry their debts: the hyperscalers' net debt stood at 1.3 times their earnings before interest, taxes, depreciation and amortisation in the second quarter, against 2.7 times for the median United States investment-grade company. The pressure sits further out, with the neoclouds and standalone operators that borrowed the most relative to their cash flow. The market is drawing that line sharply.
The midterm results have raised expectations of regulatory scrutiny and delays to building data centers, and credit-desk sources told CNBC in October that the high-yield debt market was "effectively on hold until after the midterm elections". SoftBank, which has spent $64.6 billion on a 13 percent stake in OpenAI, has become a bellwether for the strain: its shares fell 11 percent on Friday and 4.2 percent on Monday, leaving them down 23 percent for the year, and the cost of insuring its debt against default has hit a four-year high, Reuters reported ahead of the company's quarterly earnings on Tuesday. SoftBank's net interest-bearing debt stood at 10.8 trillion yen ($68 billion) at the end of June, against 6.6 trillion yen a year earlier — and its planned $5.4 billion takeover of ABB's robotics unit was blocked by China.
For the wider economy, the stakes are visible in the market itself. Technology remains by far the biggest sector in the S&P 500, holding a greater than 35 percent weight as of Thursday's close, Reuters reported. The swoon culminated in the biggest one-day drop for the S&P 500 technology sector in over six years. Yet the rout has made valuations look less stretched: the technology sector's forward price-to-earnings ratio fell to 17.1, after rising above 32 in October 2025, against 17.9 for the overall S&P 500 — the first time since 2017 that tech has traded below the broader market.
What Happens Next
Illustration: Falling technology shares have pushed borrowing costs higher for the companies building the AI boom.
The coming week will give the market its first hard evidence of whether the lending pullback is a pause or a turning point. CoreWeave reports its quarterly results on Tuesday, and SoftBank reports second-quarter earnings the same day. Nvidia is due to report on 17 November, and Dell reports on 24 November.
Investors will also watch whether new AI debt deals get done — and at what price. The spread on 10-year hyperscaler debt, the Nebius decision to shelve its high-yield bond issue, and the terms of any new neocloud borrowing will show whether lenders are merely being selective or are stepping away altogether. Nebius's $3.5 billion loan facility from Citigroup shows that lending has not stopped; the question is how much more of it there is, and for whom. If construction schedules slip, Baird's warning about Oracle and CoreWeave will look prescient.
A separate test of sentiment is coming in the stock market. Reuters reported on Thursday that potential investors were raising doubts about whether the artificial intelligence company Anthropic could proceed with its initial public offering roadshow next week in time for a November listing. Anthropic is not the subject of this story, but its listing will signal how much appetite remains for AI risk.
Sponsored: explore today's featured sponsored offers
Frequently Asked Questions
What exactly are banks pulling back from?
From new lending to the companies building AI infrastructure — data center operators, neoclouds and chip buyers — and from increasing their existing exposure to the sector. According to Reuters, some banks are now reluctant to finance additional AI deals even when the deals carry guarantees from major technology companies such as Anthropic and Google. Existing loans are not being called in; the pullback is about new money.
How is this different from the Nasdaq selloff reported on 6 November?
Post #841 on this blog covered the stock market fall itself: the Nasdaq sliding into correction and Wall Street's fear gauge reaching an eight-month high. This story covers what happened next in the credit markets — banks tightening their lending and bond spreads widening. The selloff was the trigger; the lending pullback is the consequence.
What is a neocloud, and why is CoreWeave the focus?
A neocloud is a company that rents out powerful AI computing capacity, often built around advanced chips, to businesses developing AI. CoreWeave is the largest of the independent AI cloud providers: it has borrowed heavily to buy chips, pays interest equal to 23 percent of its revenue, and its five-year credit default swap spread passed 1,000 basis points in October. Its stock has fallen 35 percent this year.
Does this mean the AI boom is over?
Not on the evidence so far. The largest technology companies remain lightly indebted relative to their earnings, and technology is still the biggest sector in the S&P 500. What has changed is the price and availability of credit for the riskiest borrowers. Goldman Sachs still projects global hyperscaler debt issuance of $420 billion in 2027.
What should readers watch next?
CoreWeave's results and SoftBank's earnings, both due on Tuesday, 10 November; Nvidia's results on 17 November; whether Anthropic proceeds with its planned IPO roadshow next week; and whether new AI debt deals are completed and at what spreads. Any delay to construction schedules would be the clearest sign the pullback is biting.
Sources
- Reuters, 9 November 2026, "Banks, investors grow wary of AI debt, spooked by market selloff"
- Reuters, 9 November 2026, "Tech set to still sway US stock market's course despite AI-driven swoon"
- Reuters, 9 November 2026, "SoftBank to face scrutiny of AI strategy at second-quarter earnings after selloff"
- CNBC, 15 October 2026, "CoreWeave credit default swaps hit distress levels as investors sour on AI debt"
- Startup Fortune, 27 October 2026, "JPMorgan says lenders now demand proof AI firms can repay debt"
- Reeve Hayward, October 2026, "Who is lending to AI, and can the borrowers repay?"
- FourWeekMBA, 21 October 2026, "CoreWeave Debt Yields 15% as Credit Desks Wait for Midterms" (summary of CNBC reporting, 21 October 2026)
- Gartner, 14 October 2026, press release: "More than 80% of neoclouds to cease independent operations by 2030"
Comments
Post a Comment