
The Fed's data center debt box: $680 billion issued, 31 percent of investment-grade borrowing
By Zain·
The Federal Reserve's November 13, 2026 Financial Stability Report puts numbers on AI infrastructure borrowing: more than $680 billion issued by October, plus over $3 trillion of leases and guarantees that sit off the balance sheet.
The Federal Reserve has put a number on how much of the AI buildout is being paid for with borrowed money, and the number is large. In its Financial Stability Report dated November 13, 2026, the central bank says debt issuance tied to advanced computing infrastructure rose to more than $680 billion by October 2026. That is about five times the $130 billion issued across all of 2023.
This is a regulator's document, not a market opinion, and it gives the AI financing debate something it has lacked: a baseline from a primary source.
What the Fed counted
Box 2.2 of the report, titled "Leverage in the Data Center Buildout," tracks debt raised to finance computing capacity. The mix is shifting. It used to be mostly corporate bonds. It now combines public and private debt, structured credit and project finance.
Investment-grade technology companies issued most of it. Advanced computing borrowing was less than 6 percent of nonfinancial investment-grade corporate debt issuance in 2023 and about 31 percent in 2026. Outstanding public debt tied to advanced computing is estimated at more than $1.1 trillion as of October 2026. The Fed adds that private credit deals make the true total harder to measure.
The part that is not debt
The bigger story is what does not show up as debt. The Fed says large technology firms are "rapidly increasing future lease commitments, backstop guarantees provided to other firms, and other complex and contingent obligations."
Fed staff estimate that leases not yet started have undiscounted values above $1 trillion. Take-or-pay contracts and residual value guarantees add undiscounted values above $2 trillion. The Fed warns these figures are not directly comparable with on-balance-sheet debt, which is discounted. Even so, the direction is plain: obligations are growing faster than reported debt.
The mechanism is the one our readers have seen in company filings. A hyperscaler signs a long lease with a data center developer, and the developer uses that lease to borrow. A chip maker guarantees the resale value of GPUs so a smaller AI lab can finance them. Equity stakes often sit on top. The Fed calls this a "circular financing" dynamic. We unpacked how it works in our circular financing explainer.
“these firms' economic exposure to such a downturn may be substantially larger than suggested by their on-balance-sheet liabilities.”

Who holds the paper
The Fed says the on-balance-sheet debt of hyperscalers and GPU designers is held mainly by mutual funds, pension funds and insurers. The riskier subordinated debt and equity tranches of structured financings are held by high-yield bond funds, life insurers, business development companies and other private debt vehicles. The largest U.S. bank holding companies have close to $1 trillion of exposure to large technology firms, data center developers, special purpose entities, private credit vehicles and utilities. Those exposures are evaluated in the annual supervisory stress test.
The report also compares the cycle with the late 1990s broadband telecom buildout. Hyperscaler cumulative capital spending and debt issuance since 2020 have tracked telecom's since 1995 relatively closely, the Fed says, and AI investment has been accelerating, "suggesting that the pace of the current cycle's investment may surpass that of the 1990s telecom cycle relatively soon." The Fed's five hyperscalers in that comparison are Amazon, Alphabet, Meta, Microsoft and Oracle.
What the survey adds
The report's survey box covers 23 market contacts, polled by the New York Fed in September and October. AI was the top-cited risk, with concerns about elevated equity valuations, AI-enabled cyberattacks and complex AI financing. In the spring survey of 20 contacts, AI ranked third behind geopolitical risks and an oil shock. Reuters reported that 87 percent of contacts named AI this time, up from 50 percent in the spring, and listed a risk asset correction at 52 percent. We could not read the chart values in the report itself, so treat those percentages as Reuters' reading.
Two caveats matter. First, a survey of 23 people measures worry, not probability. Second, Reuters notes the survey ran before the Nvidia leadership upheaval that hit AI shares this month, so it does not capture that shock.
The other pressure points
The AI items sit inside a wider report. Reuters notes that leverage among the largest hedge funds is at the highest level since the Fed began collecting the data in 2013, which matters because hedge funds finance positions with repurchase agreements and other borrowing. Respondents in the survey also cited higher long-term Treasury yields, cyberattacks and fiscal debt sustainability. The Fed's survey notes say respondents saw cyberattack and fiscal risks as having the most severe potential consequences.
The AI answers broke into three groups, in the Fed's wording: elevated equity valuations and the chance that lower-than-expected earnings force a correction, cybersecurity risk from state actors or from damage "caused by frontier AI models themselves," and complex forms of AI-related financing. Several respondents linked a fall in AI-related equities to weaker business investment and household consumption, since the buildout is now a large part of corporate capital spending.
This is the first Fed report to treat frontier model behavior as a financial stability input. That is a notable shift for a document whose usual subjects are mortgages, money market funds and bank capital.
Our take
The report does not say a crash is coming, and we would not read it that way. Its own words on the cushion are fair: the biggest firms may absorb losses, which limits spread to the wider system. But the same sentence says their true exposure may be far larger than their balance sheets suggest.
We would watch three things in the next Fed report: the lease and guarantee totals, the bank exposure figure, and whether private credit deals stay opaque. Our earlier coverage of Governor Bowman's debt warning cited a similar scale. The new report is the first time the Fed's own staff have laid the components side by side.
For readers tracking the race, the point is simple. The labs' compute plans are no longer only a technology bet. They are now a credit market, and the Fed is reading the filings.
Frequently asked questions
How much debt has the AI data center buildout raised?
The Federal Reserve's November 13, 2026 report says advanced computing related debt issuance exceeded $680 billion by October 2026, about five times the $130 billion issued in all of 2023. Outstanding public debt tied to the sector is estimated above $1.1 trillion.
What share of corporate borrowing is AI infrastructure?
The Fed says advanced computing borrowing was about 31 percent of nonfinancial investment-grade corporate debt issuance in 2026, up from less than 6 percent in 2023.
What are off-balance-sheet AI obligations?
Fed staff estimate not-yet-started leases above $1 trillion and take-or-pay contracts plus residual value guarantees above $2 trillion, undiscounted. These are not directly comparable with on-balance-sheet debt, the Fed notes.
Do banks have exposure to the AI buildout?
Yes. The Fed says the largest U.S. bank holding companies have close to $1 trillion of exposure to large technology firms, data center developers, special purpose entities, private credit vehicles and utilities, and those exposures are tested in annual stress tests.
Did the Fed say AI is a bubble?
No. The report describes survey concerns and leverage data. It says large firms may absorb losses in a downturn, but that their economic exposure may be substantially larger than their balance sheets suggest.
Where does the 87 percent figure come from?
Reuters reported that 87 percent of surveyed contacts named AI as a risk, up from 50 percent in the spring. The Fed's report ranks AI first among 23 contacts but its chart values were not readable in our copy.
Sources
What each one is, and whose it is.
- 1
Financial Stability Report, November 2026, Federal Reserve Board (November 13, 2026)
FilingIndependent of the vendor - 2
Financial Stability Report landing page, Federal Reserve Board (November 13, 2026)
DocumentationIndependent of the vendor - 3
AI rockets to top of worries about financial system in latest Fed survey, Reuters via The Star (November 14, 2026)
Press reportIndependent of the vendor