AI debt markets face tighter scrutiny as banks, investors pull back

Traders Union
All News
Financial Institutions
AI debt markets scrutiny

Highlights

Banks and investors have tightened scrutiny and reduced willingness to fund AI-related debt deals after a $2.5 trillion Nasdaq 100 selloff and U.S. midterm results triggered sector uncertainty.
Goldman Sachs Research projects global hyperscaler debt issuance will jump to $420 billion in 2027, up over 60% from its $250 billion 2026 estimate, though lenders increasingly question sector exposure.
Spreads on 10-year hyperscaler debt have widened by 50 basis points year to date versus 20 for non-AI industrials, with Meta and SpaceX bonds trading below par and CoreWeave bonds sharply down.

Market selloff reshapes AI lending appetite

As reported by Reuters, banks and credit investors are increasing scrutiny of AI-related debt deals and are showing less willingness to add exposure, even when financings carry support from major technology groups. Several banking and credit sources say some lenders that were actively lining up to fund AI transactions only weeks ago are now reluctant to back additional deals.

The pullback intensifies in the past few days after a stock market decline that wipes out $2.5 trillion in value from the Nasdaq 100 in one session and pushes the index into correction territory. U.S. midterm election results are also adding to investor anxiety about the prospect of greater oversight for data centers.

Sources say investors and lenders are stress-testing cash flows for downturn scenarios and examining contract terms behind AI infrastructure projects more closely, including service obligations, operating deadlines and the conditions under which large technology customers could scale back or end commitments. Some lenders are also hesitating even on deals linked to companies such as Anthropic and Google; both companies decline to comment.

Chris Poggi, a finance partner at Ropes & Gray, says debt providers spent the past year showing strong appetite for neoclouds and data center builders, but many are now assessing whether they have already extended as much credit as they want to this year. He says investor books have filled up as enthusiasm gives way to a more selective approach.

Exposure limits and project risks come into focus

Even with the caution, funding demand from the AI ecosystem remains large. Goldman Sachs Research projects global hyperscaler debt issuance will reach $420 billion in 2027, more than 60% above its estimate of $250 billion for fiscal 2026.

Many of the financings are complex because chipmakers often backstop senior debt used to buy advanced chips and also guarantee payments on long-term data center leases. Those protections had made the loans attractive to banks and private credit firms, but hyperscalers and neoclouds seeking more funding for data centers, chips and computing capacity are increasingly meeting lenders that question how much more sector exposure they want on their books.

At JPMorgan Chase, loans backed by other income-producing properties, a category that includes data centers, stand at $29 billion at the end of the third quarter. That equals 12% of the bank's commercial real estate exposure, making it the second-largest segment after apartments, up from 8% three years earlier; JPMorgan declines to comment.

Investors are also becoming more selective in public debt markets. Bank of America credit analyst Yuri Seliger says spreads on 10-year hyperscaler debt have widened by about 50 basis points year to date, compared with roughly 20 basis points for high-quality industrial issuers not tied to AI, while bonds from Meta sold earlier this year and from SpaceX are trading below par.

In Asia, banks are also tightening risk reviews. OCBC Group CEO Tan Teck Long says during a Friday earnings briefing that the bank structures deals to ensure cash flow support is in place and remains highly focused on risk. The retrenchment is especially evident among riskier AI borrowers such as neocloud companies and standalone data center operators.

According to Baird Equity Research, Oracle and CoreWeave are among neocloud companies that could be affected by possible construction delays or moratoriums. CoreWeave shares are down 35% year to date, and its bonds have also fallen sharply in the secondary market.

Some planned offerings are already being shelved. One source familiar with the matter says Nasdaq-listed neocloud company Nebius explored a high-yield bond sale in September but decides not to proceed last month; the company does not respond to a Reuters request for comment. Nebius enters into a $3.5 billion senior secured loan facility from Citigroup in October for data center development, GPU procurement and general corporate purposes.

Jay Bala, co-founder and CEO of alternative asset manager AIP, says most banks and credit investors are likely to adopt a wait-and-see approach as they gauge how AI growth holds up after Democrats sweep Congress in the midterms.

Our earlier report on U.S. AI data center expansion highlighted how the boom is running into higher financing costs, construction delays, and power-grid constraints that are testing investor confidence. It noted that only a minority of planned projects were viewed as truly viable, while selloffs and cost overruns at infrastructure players underscored the risk that missed deadlines and rising capital needs could undermine long-term economics.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.