The Federal Reserve said on 13 November that measures of equity valuations “have remained high, partly driven by market enthusiasm for super intelligence (SI)”, in its November 2026 Financial Stability Report. The same report says the largest US bank holding companies “have close to $1 trillion of exposure to large technology firms, data center developers, advanced computing–related SPEs and private credit vehicles, and utility companies”.
In the New York Fed’s survey of 23 market contacts, “AI was the top-cited risk”, named by 87% of them, up from 50% in the spring survey. That is 20 of 23 contacts, against 10 of 20 in March and April, by our arithmetic. Fed staff put advanced computing debt issuance at more than $680 billion by October 2026, about five times the $130 billion of 2023.
Business Pill · Vendor Guarantees
An evergreen Business Engineer explainer on vendor guarantees: when a supplier backs its customers’ financing, the mechanism behind the residual value guarantees the Fed describes. It teaches the general idea only and does not discuss the Fed report.
The key insight: As we read the report, the Fed has moved AI from an equity-valuation story to a credit story: the same buildout it says lifts stock valuations runs through leases, guarantees and SPEs, and the largest banks hold close to $1 trillion of exposure to it.
The report’s summary rates asset valuation pressures as elevated. It says the ratio of equity prices to earnings for S&P 500 companies “stayed well above its historical median”, and an estimate of the equity premium, the extra return investors demand for holding stocks, “remained low”.
In its section on asset prices, the report says broad equity valuation metrics were elevated, “partly driven by market enthusiasm for SI, though moderated by concerns over firms’ capital expenditures and debt issuance plans”. The report generally reflects information available as of 28 October 2026.
The equity-market reading sits next to a debt one. “Robust debt issuance by large technology companies may drive aggregate business debt higher over time”, the report says, adding that “complex financing arrangements and contingent guarantees also were growing rapidly at those firms”.
Share of New York Fed survey contacts citing each of the five most-cited risks, spring 2026 (20 contacts) and fall 2026 (23 contacts), from the Fed’s November 2026 Financial Stability Report. The survey reflects contacts’ views, not the Fed’s; point changes are our arithmetic.
The $680 Billion Debt Number
A box titled “Leverage in the Data Center Buildout” sets out the figures. Advanced computing–related debt issuance “rose to more than $680 billion by October 2026, about a fivefold increase from $130 billion in all of 2023”. Our arithmetic puts the multiple at 5.2.
That borrowing grew from less than 6% of nonfinancial investment-grade corporate debt issuance in 2023 to about 31% in 2026, the box says. It estimates outstanding public debt related to advanced computing at more than $1.1 trillion as of October 2026. Separately, the report says bond issuance by the largest investment-grade data center operators was greater than $150 billion year to date.
The staff count is narrower than some public figures. It includes public deals and Rule 144A securities, but private placements issued under section 4(a)(2) are excluded. Fed Vice Chair for Supervision Michelle Bowman said on 10 November that AI-related debt “is on track to exceed $800 billion this year”, as we reported on her speech.
The box compares five hyperscalers, defined as Amazon, Alphabet, Meta Platforms, Microsoft and Oracle, with the 1990s telecom buildout. It says advanced computing investment has been accelerating, “suggesting that the pace of the current cycle’s investment may surpass that of the 1990s telecom cycle relatively soon”.
Figures from the Federal Reserve Board’s November 2026 Financial Stability Report (Box 2.2 and Section 1). They are different measures and do not add up.
Where the Exposure Sits: Leases, Guarantees and SPEs
The Fed describes how strong-credit firms back weaker ones without an immediate hit to their own balance sheets. Hyperscalers sign long-term leases with third-party developers, and “These leases are then used by the third-party developers to secure financing for the data center’s construction”.
GPU makers, it says, “often provide a guaranteed residual value of the GPU collateral” to support purchases by advanced computing and large language model firms. Combined with equity stakes from hyperscalers and GPU designers, this creates what the report calls a “circular financing” dynamic. Its data on debt and other obligations cover the five hyperscalers plus Broadcom and Nvidia.
The Fed reads this two ways. The largest technology firms may absorb losses in a downturn, it says, but their “economic exposure to such a downturn may be substantially larger than suggested by their on-balance-sheet liabilities”. The riskier subordinated debt and equity tranches of the SPE financings sit with high-yield bond funds, life insurers, BDCs and other private debt vehicles.
The survey box lists three AI concerns: elevated equity valuations, cyber risk “stemming from either a malicious attack by sophisticated state actors or damage caused by frontier AI models themselves”, and complex AI-related financing. Respondents “widely cited the risk of a sharp correction in AI-related equities, driven by uncertain returns on large capital expenditures”.
Many respondents said “stretched valuations, circular financing arrangements, and heavy reliance on debt to fund the AI buildout could amplify losses”. Behind AI, 52% cited a risk asset correction, 48% higher long-term rates, 43% cyberattacks and 43% fiscal debt sustainability. In the May report’s chart data, AI stood at 50% in spring 2026 and at 30% in fall 2025.
The cyber point also appears in the Fed’s own text, outside the survey. It says “Reports that agentic systems have been involved in cyber events have increased in frequency since the May report”.
The Structural Read
As we read the survey, AI went from 10 of 20 contacts in the spring to 20 of 23 in the fall, by our arithmetic, the largest rise among the five risks the report charts.
As we read Box 2.2, the $680 billion issuance figure is a floor, since the Fed says private placements under section 4(a)(2) are excluded and private credit deals make the total harder to measure.
Federal Reserve Board, Financial Stability Report, November 2026, Box 2.2
“The rapid growth of these other obligations represents deepening connections between the largest technology firms and smaller, riskier firms with less access to credit.”
Three Implications
FOR BANKS The Fed says the close to $1 trillion of exposure to large technology firms, data center developers, advanced computing–related SPEs and private credit vehicles, and utility companies is evaluated during the annual supervisory stress test.
FOR AI BORROWERS The report says the debt being issued is shifting from mostly corporate bonds to a combination of public and private debt, structured credit and project finance.
FOR OUR COVERAGE We will add any Fed breakdown of the bank exposure and the next stress-test results when they are published.
The analysis argues: “It is a sovereign-scale financing problem dressed in tech-company clothing — and the financial architecture is starting to invert.”
As we read it, the Fed’s box describes that shift in supervisory language: leases, residual value guarantees and SPEs move risk from the strongest balance sheets toward funds, insurers and the largest banks. The analysis predates this report and does not discuss it.
What Is Not Established
The survey reflects 23 contacts, and the Fed says it “should not be interpreted as representing the views of the Federal Reserve Board or the Federal Reserve Bank of New York”. The report does not name the banks behind the $1 trillion or break it down by borrower type, and the Fed says private credit deals make the total debt “more difficult to measure”. Its data run to 28 October, before this week’s fall in AI shares.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
The Fed now names super intelligence as part of what holds US stock valuations up, and puts large-bank exposure to the AI buildout at close to $1 trillion. AI went from 50% to 87% of its survey contacts in six months. The next number to watch is the stress-test treatment of those exposures, which the Fed says it already evaluates.
A note on sourcing. We read the Federal Reserve Board’s November 2026 Financial Stability Report in full on federalreserve.gov; the fall survey percentages come from the overview graphic on its page 3. The spring 2026 and fall 2025 survey figures come from the May 2026 report and its published chart data. The Bowman quote is from our earlier piece on her 10 November speech. Contact counts and multiples are our arithmetic. We did not contact the Federal Reserve.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
