Explainer
The Fed Says AI Debt Is Now the Top Risk to Markets
The Fed's own survey now names AI the top risk to financial stability. Here's what $680 billion of AI debt means for your money — and what to do about it.

In this article
In one week, the two institutions that watch over the world's money sounded the same alarm. The US Federal Reserve's Financial Stability Report, published on 13 November 2026, found that artificial intelligence is now the risk its market contacts mention most. Three days later, the head of the International Monetary Fund warned that the global economy is "standing on one leg" — the AI leg — and that a wobble could cause financial-stability trouble.
This is not a fringe blog warning. It is the central bank and the IMF, in the same week, pointing at the same thing: the AI boom is now built on an enormous pile of borrowed money. That matters even if you have never bought an AI stock — if you hold an index fund, a pension or a retirement account, you already own this story.
What the Fed actually said
Twice a year, the Federal Reserve Bank of New York asks market contacts at banks, funds and broker-dealers what risks worry them most. In the autumn 2026 round, conducted in September and October, 87% named artificial intelligence as a salient risk to US financial stability over the next 12 to 18 months. In the spring survey it was 50%; in autumn 2025 it was 30%. AI has gone from third place to first in six months.
Nothing else came close: risk-asset correction 52%, higher long-term rates 48%, cyberattacks 43%, fiscal debt sustainability 43%. The survey drew just 23 responses, and the Fed stresses they are the respondents' views, not the Board's. Two cautions: 87% is not an 87% probability of a crisis, and the survey closed before Nvidia chief executive Jensen Huang resigned on 12 November.
On 16 November, Fed Governor Michael Barr went further. Speaking in Milan, he described how AI could transmit stress through the financial system: similar trading models reacting the same way at once, shared technology providers becoming single points of failure, and autonomous AI agents moving deposits without human instruction — potentially triggering bank runs in minutes rather than days. The warning is structural, not a prediction of imminent collapse.
The $680 billion borrowing binge
The most concrete part of the report is a detailed box on how the AI build-out is being paid for. Debt issued to fund advanced computing infrastructure passed $680 billion by October 2026, about five times the $130 billion issued in all of 2023. Its share of nonfinancial investment-grade issuance rose from under 6% in 2023 to about 31% this year, and the Fed puts outstanding public debt tied to advanced computing at more than $1.1 trillion.
The seven largest AI-related firms have borrowed sharply during 2026, and the Fed expects their combined debt to exceed $800 billion before the year is out — with the maximum nominal value of their long-term contracts, leases, guarantees and similar off-balance-sheet obligations above $3 trillion.
| Measure | Figure | Source |
|---|---|---|
| AI-related debt issued, 2023 (full year) | $130 billion | Federal Reserve |
| AI-related debt issued, by October 2026 | More than $680 billion | Federal Reserve |
| AI-related debt issued globally in 2026 | Nearly $600 billion | Goldman Sachs, via Howard Marks |
| Combined debt of the seven largest AI firms | To exceed $800 billion by end-2026 | Federal Reserve |
| Their off-balance-sheet obligations | More than $3 trillion | Federal Reserve |
Debt, unlike equity, must be repaid on schedule whether or not the investment behind it pays off. And the borrowing reaches into the corporate bond market, where AI-related issuance now accounts for more than a fifth of all US investment-grade supply. If you own a bond fund, you probably own some of this debt too.
Exposure to AI constitutes an unusually powerful "fault line" that runs through portfolios and has the potential to dominate all aspects of the investment environment. — Howard Marks, Oaktree Capital, 11 November 2026
The circular money problem
The second warning came from the International Monetary Fund, and it concerns a pattern with an unflattering name: circular financing. A large technology company invests in a smaller AI company, which then uses some of the money to buy chips, computing capacity or services — from the very company that invested in it. At each stop, the money can look like new revenue and new growth.
Speaking to CNBC on 16 November, IMF Managing Director Kristalina Georgieva said this circular financing among AI companies has grown to a little over $1 trillion, with about two-thirds of it funded by debt. "We have a world economy that is standing on one leg on the AI leg," she said. "If this leg wobbles, that can lead to financial stability troubles." The IMF's October Global Financial Stability Report examined 14 publicly disclosed transactions since 2023: roughly $300 billion in equity and $650 billion in debt, tied to $1.4 trillion in future revenue commitments. Those are different measures from the $1 trillion headline and should not be added together — but the structure is the same: a small number of very large companies financing each other, mostly with debt.
In an IMF downside scenario, US equity prices could fall 20% in 2026 and global output could end up 1.1% below baseline by 2028 if confidence in AI breaks. Georgieva stressed recent swings have been orderly, and called the $1 trillion figure "not a huge size, but a rapid increase."
The key distinction: nobody in authority is saying AI is worthless. The warnings are about financing — how much debt is being used, how interconnected the borrowers are, and what happens if revenues arrive late. Technology can succeed while the people who financed it do not get paid.
Wall Street is already backing away
Bank of America's November fund manager survey (169 participants, $461 billion) found the sharpest retreat from technology stocks in almost twenty years. The net share of fund managers overweight tech fell from 37% in October to 12% in November — the largest one-month decline since June 2006. Strategist Michael Hartnett wrote bluntly that "the AI boom and euphoria are over."
On the biggest tail risk, "AI bubble" ranked first at 34%, followed by an "AI hyperscaler credit event" at 31% — two-thirds of managers now see AI as the greatest threat to stock and credit markets. A record net 38% said companies are overinvesting in AI, and 52% expect a major AI infrastructure player to announce a cut to capital spending within six months. The market has felt it: the Nasdaq-100 is down more than 4% in November, on track for its worst November since 2008, semiconductor stocks have fallen 27% since early October, and Nvidia has lost more than $1 trillion in market value over the past month. Yet cash fell to 3.6%, the lowest since August, and only 3% expect a recession next year — nervous, not panicked.
Why you already own this risk
Per JPMorgan data cited by Howard Marks, 42 AI-related companies accounted for roughly two-thirds of the S&P 500's price gains since January 2024. If your pension, 401(k) or index fund tracks the US market, a large share of your recent gains came from a very small number of companies — now carrying record debts to fund the AI build-out.
In his 11 November memo "A Layman Thinks About AI," Marks called AI the most consequential and least predictable development for investors in at least a generation. He asked Anthropic's Claude to score each part of the industry's predictability from 1 to 10; it rated AI overall at about 3, the frontier labs at 2 — the same band as early-stage venture capital. Whether AI compute will generate returns to justify trillions in spending, he concluded, "is largely a matter of faith."
That concentration is the real meaning of the "fault line": one sector disappointing should be a setback, but when one sector drove two-thirds of the market's gains, its disappointment can dominate everything else. UK investors hold most AI exposure through global index funds, pensions and SIPPs in US-listed companies, since the FTSE 100 has far less AI concentration.
What to do with your money
The right response is boring: a portfolio check, not a trade.
Do not try to trade this warning. Shorting the market because the Fed published a report is a bet that you know more than the market does about timing. Bubble warnings can arrive years before any peak — the dot-com bubble's most famous warnings came in 1996, and the market then doubled before it fell.
Know your concentration. Check what share of your pension or index fund sits in the largest US technology companies. If the top ten holdings are more than a quarter of the fund, you are more concentrated than the word "diversified" suggests.
Check your timeline. Decades from retirement, a 20% fall — the IMF's downside scenario — is painful but recoverable. If you need the money within a few years, your allocation should already reflect that.
Make sure your safety net is real. The best protection against any market shock is three to six months of expenses in cash you can reach immediately — that is what lets you leave investments alone when headlines get frightening.
Resist leverage. Market routs are driven by borrowed money — margin loans, leveraged ETFs, the kind of debt the Fed is warning about. Leveraged products are designed for days, not years.
Rebalance on a schedule, not on fear. If AI gains pushed your equity allocation above target, trimming back is disciplined investing. Selling everything because a report scared you is a reaction, not a plan.
Frequently asked questions
Does the Fed think an AI crash is coming?
No. The report surveys market contacts about risks; it does not forecast a crash. The 87% figure is the share of the 23 contacts who named AI a salient risk, and the Fed notes the largest tech firms still have substantial financial capacity.
What is circular financing?
Money invested in an AI company flows back to the investor through purchase orders — a chipmaker invests in a cloud company, which then buys chips from that chipmaker. The IMF estimates such financing at just over $1 trillion, about two-thirds debt-funded.
I only own index funds. Am I exposed?
Yes — probably more than you realise. Forty-two AI-related companies drove roughly two-thirds of the S&P 500's gains since January 2024 (JPMorgan data, via Howard Marks), and a global index fund holds those same companies.
Should I sell my tech stocks?
No universal answer — and this is not personal financial advice. The professionals surveyed are cutting tech exposure, but hold very little cash and do not expect a recession. A scheduled rebalance to your target allocation beats selling everything on a warning.
How is this different from the dot-com bubble?
In 2000, the borrowing was mostly done by small, unprofitable companies; today the borrowers are the world's largest and most profitable firms. What is similar is the concentration of gains in one theme and the growing reliance on debt. Howard Marks refuses to call it a bubble — he argues the right question is whether returns justify the investment, which is unknowable in advance.
Key takeaways
- The Fed's November Financial Stability Report found 87% of market contacts name AI as the top risk to financial stability, up from 50% in the spring.
- AI-related debt issuance passed $680 billion by October 2026 — five times the 2023 total — and the seven largest AI firms carry over $3 trillion in off-balance-sheet obligations.
- The IMF estimates circular AI financing at just over $1 trillion, two-thirds debt-funded, and warns the world economy is "standing on one leg."
- Fund managers cut tech exposure by the most since 2006, yet hold little cash — nervous, not panicked.
- If you own an index fund or pension, you already own this risk: 42 AI companies drove two-thirds of the S&P 500's gains since January 2024.
- The right response is a portfolio check — know your concentration, keep a real cash safety net, avoid leverage, and rebalance on a schedule, not on fear.
The story is no longer a debate between AI believers and sceptics. It is a question about plumbing: how much debt, how interconnected, and how much has to go right. The technology may well change the world; whether it pays for itself on the borrowing's schedule is, as Howard Marks put it, largely a matter of faith. Your job is not to resolve that question — it is to make sure your finances can live with either answer.
Sources
- DailyMech — Fed Financial Stability Report: AI Is Top Risk
- The Parameter — AI is now the top financial stability worry in the Fed's survey
- TFTC — Fed's Barr: AI agents could trigger bank runs in minutes
- Oaktree Capital — Howard Marks memo: A Layman Thinks About AI (PDF)
- CNBC Africa — IMF chief warns AI circular financing hits $1 trillion
- Moneycontrol — BofA survey: fund managers cut tech by most since 2006
- Stocktwits — Howard Marks calls AI exposure a 'fault line'
The MoneyHabit editorial desk researches, writes and reviews every article against our editorial policy.
The Sunday Habit
One useful money idea, every week.
A short, practical email every Sunday — no spam, no selling your address, unsubscribe anytime.
Advertisement
More in Money News & Explainers

Money News & Explainers · Explainer
Fed to Hold This Week — December Is the Real Risk
94 of 100 economists expect a Fed hold on 27-28 October, but all 100 expect a December hike. Here is what that split means for savers and borrowers.
25 October 2026 · 10 min read

Money News & Explainers · News
Pakistan National Savings Rates Rise: What Savers Get Now
National Savings has revised its profit rates for November 2026. We break down every scheme, what tax takes, and whether your money still beats 9.6% inflation.
19 November 2026 · 10 min read

Money News & Explainers · News
Your MEXC Card Just Stopped Working: What to Do Now
MEXC has switched off every Global Card until 18 November. Your money is still yours — but the only way out runs through your MEXC Fiat Account.
15 November 2026 · 10 min read