Finance

Why Wall Street contacts now rank AI as the top threat to financial stability

A New York Fed survey shows 87% of market contacts naming AI as a risk, and the reasons have more to do with debt than with technology.

11 min read
Rows of dark server racks with small blue indicator lights lining a grey, empty data center room under a line of ceiling lights.
Image by Schäferle from Pixabay

In the Fed's November 2026 stability report, 87% of 23 market contacts named AI as a risk, up from 50% of 20 in the spring. The concern centers on stock valuations, borrowing for data centers and cyber threats. For savers, exposure runs mostly through index and target-date funds.

In the Federal Reserve's Financial Stability Report released on November 13, 2026, artificial intelligence was the risk mentioned most often by the market professionals the New York Fed surveys twice a year. According to the report as relayed by GV Wire's Reuters-based account, 87% of 23 contacts named AI as a salient risk. In the spring survey, 50% of 20 contacts had done so.

The explanation has less to do with machines than with money. The worry rests on how richly AI-linked companies are valued, how much they are borrowing to build data centers, and how some of that borrowing is structured. For ordinary savers, the connection is indirect but real: the same companies sit inside the index funds and target-date funds that fill many retirement accounts. The survey does not say a crash is coming, and the limits of the data matter as much as the headline number.

What the survey actually measured

The phrase "the Fed ranks AI as the top risk" is a shorthand that needs correcting. The 87% comes from a survey of market contacts run by the New York Fed. Responses were gathered in September and October from people at broker-dealers, banks, investment funds and advisory firms. The report states that the summary reflects those contacts' views and does not represent the views of the Fed Board or the New York Fed.

The survey also counts mentions, not severity. A contact who lists AI is saying it is on their mind, not that a loss is probable or large. The percentages add up to more than 100%, which indicates respondents could name more than one risk, although the exact survey instructions were not available for this article.

According to a chart of Fed data published by Axios, AI rose from 50% in the spring to 87% in the fall. A correction in risk assets rose from 35% to 52%, and higher long-term interest rates from 20% to 48%. Geopolitical risk fell from 75% to 30%. Coverage of the report also puts cyberattacks and fiscal debt sustainability at 43% each in the fall.

Share of Fed survey contacts naming each risk, spring and fall 2026 (%)

Spring 2026 survey (20 contacts), % naming the risk

Fall 2026 survey (23 contacts), % naming the risk

Share of Fed survey contacts naming each risk, spring and fall 2026 (%)
Label Spring 2026 survey (20 contacts), % naming the risk Fall 2026 survey (23 contacts), % naming the risk
AI 50 87
Risk-asset correction 35 52
Higher long-term rates 20 48
Geopolitical risks 75 30

How firm is the jump from 50% to 87%?

The panels are small. By simple arithmetic, 50% of 20 contacts is 10 people, and 87% of 23 is 20 people. A move from about 10 respondents to about 20 is a notable shift in mood, but the groups differ in size and probably in membership, and they are not random samples of the industry. Rough statistical ranges around the two results nearly touch, and because the panel is a convenience group the true uncertainty is larger than those ranges suggest.

Part of AI's climb to first place also reflects other worries receding. In the spring survey, geopolitical risk and an oil shock topped the list, with AI cited by half of respondents. As geopolitical concern dropped to 30%, AI moved up the ranking even apart from any growth in AI anxiety.

Three worries bundled into one answer

"AI" in the survey is an umbrella label. As the Fed summarized the contacts' answers, respondents raised three separate concerns. Each has a different mechanism and would hit different parts of the financial system.

Valuations and a possible correction

The first concern is that equity valuations are elevated and could fall sharply if AI-related earnings disappoint. The Fed Board's own view is consistent with that: it finds stock valuations high, which it says are partly driven by market enthusiasm for "super intelligence." A price decline of that kind would first hurt investors, and it matters for the wider economy only to the extent that losses spread through lending, spending or confidence.

Cybersecurity

The second strand is cyber risk. It includes attacks by state actors and what the report describes as damage caused by frontier AI models themselves. This worry is separate from the market-price question, because a major cyber event could disrupt financial firms regardless of what stocks are doing.

Complex financing

The third strand, risks from complex forms of AI-related financing, is the most novel. Many respondents said circular financing and heavy reliance on debt could magnify losses if a correction came. The Fed gave AI financing its own section in a stability report for the first time, which signals how seriously the topic is now taken inside the institution.

The categories also overlap. Axios notes that the second-ranked risk, a broad sell-off in risk assets, is at least partly tied to the possibility of an AI-driven market downturn. Some of the 52% may therefore be an AI worry under another name, which suggests the 87% could understate AI-linked concern, or that the categories double-count it.

The money behind the worry: debt, leases and circular deals

According to Axios's reading of the Fed report, companies borrowed more than $680 billion to finance data centers and other advanced computing infrastructure in the first 10 months of 2026. That is roughly five times the borrowing for all of 2023. The Fed compares the spending and borrowing by the largest cloud companies since 2020 with the telecom investment cycle that began in 1995, a period of real technology and real investor losses.

Obligations that do not show up as debt

Fortune's account of the report focuses on commitments that sit outside company balance sheets. They cover seven firms: Amazon, Alphabet, Broadcom, Meta, Microsoft, Nvidia and Oracle. Through the second quarter of 2026, the tally was:

  • Leases that have not yet started: more than $1 trillion, undiscounted
  • Take-or-pay contracts and residual value guarantees: more than $2 trillion
  • Combined, in Governor Michael Barr's phrase, a "potential maximum nominal value" above $3 trillion

The Fed itself cautions that these off-balance-sheet figures are undiscounted and not directly comparable with debt, so the $3 trillion should not be read as money the companies owe today. Barr separately said the seven largest AI firms' on-balance-sheet debt is expected to exceed $800 billion before the end of the year.

How "circular financing" works

The mechanics, as the Fed describes them, run roughly like this. Large cloud companies sign long-term leases with third-party developers, and those developers use the leases to secure construction financing. Debt deals linking cloud companies and chip designers with smaller firms are often supplemented by equity investments from the same large firms. The money, in other words, can flow in a loop, with the lender's customer and the lender's investment target being the same company.

The Fed sees a double edge. These arrangements could let the largest firms absorb losses in a downturn and limit the damage to the broader system. But the same arrangements mean their economic exposure "may be substantially larger than suggested by their on-balance-sheet liabilities." One illustration comes from Fortune's summary of Nvidia's latest quarterly filing: $3.5 billion in guarantees on AI cloud partners' data center leases as of July 26, against which those partners had placed $712 million in escrow.

What the Fed itself says about the system

The survey reflects contacts' opinions, so it helps to separate it from the Fed's own assessment. Coverage of the report indicates that outside AI, the Fed found the system broadly stable, with banks and households looking solid. Leverage appears to have shifted toward hedge funds, life insurers and the vehicles that banks lend to, which is a commentators' reading rather than a Fed conclusion. The report also says bank credit commitments to other financial institutions surpassed $2.8 trillion in the second quarter of 2026.

Banks are not untouched by the buildout. Axios reports that they have nearly $1 trillion in exposure to technology companies, data-center developers and other firms and lenders involved in AI infrastructure. The report also says Treasury yields rose markedly over the six months since the previous edition and that volatility increased on net. Newly issued private credit still carried spreads that widened only modestly and remained relatively low.

Bowman's scenario

In a November 10 speech, Vice Chair for Supervision Michelle Bowman said she is focused on risks to the economy and financial stability if revenue from AI customers disappoints. She said investment is increasingly financed with outside funds, including off-balance-sheet liabilities, and that revenues have been rising quickly but are not yet enough to cover the cost of maintaining existing models and data centers and building new ones. Her description of what could follow:

we could see a sharp correction in equity prices for AI firms, higher financing costs, and reduced access to credit.

Michelle Bowman, Fed Vice Chair for Supervision, November 10, 2026

Bowman also said the risks seem to be concentrated in only a few large firms that provide AI services. That concentration cuts both ways. It limits how many companies could be the source of trouble, but it means a problem at one of them could carry outsized weight in markets. She added that lower equity prices could undermine the resilience of consumer spending.

How ordinary savers are exposed

This section offers context, not personal advice. For most households the main channel is indirect: AI-linked companies are large parts of broad stock funds. According to a Bloomberg graphic, technology now accounts for nearly half the value of the S&P 500. As of November 3, nearly $32 trillion in market value had been added to the index since ChatGPT's release, and more than half of that came from just 10 AI-linked companies.

A September CNBC report citing Morningstar said five tech giants with AI exposure make up 30% of the S&P 500. It added that the average 401(k) investor may own AI stocks through an S&P 500 index fund or a target-date fund. Such figures vary by definition, since five firms versus ten, or "AI-related" versus the whole tech sector, give different answers, so they should not be compared directly.

Three channels beyond stock prices

  1. Fund values: a sharp repricing of a handful of large AI-linked stocks would flow into index and target-date funds.
  2. Credit conditions: Bowman's scenario includes higher financing costs and reduced access to credit, which would affect borrowers more broadly.
  3. Prices and spending: Bowman said prices of memory chips, semiconductors, computers and other electronics have risen sharply this year, in a few categories by double digits, as data-center demand outpaced supply. She also said lower equity prices could weaken consumer spending.

Recent market moves show the sensitivity. Bloomberg ties an early-November selloff to the possibility that newly elected Democrats would slow data-center development. That episode came after the Fed survey closed in October, so it is not part of the 87%.

A signal of sentiment, not a forecast

The sharpest reading of the survey is that Wall Street's attention has moved. Fewer contacts are focused on oil and geopolitics, and more are looking at how a very large, debt-assisted buildout could unwind if AI revenue lags its costs. That is a meaningful shift, but it is a shift in what professionals mention, not a measurement of what will happen. The Fed does not forecast a crash, and officials stress that the largest firms may be able to absorb losses.

What to watch is concrete. The spring 2027 stability report and survey will show whether the 87% was a peak or a new baseline. Beyond that, the data that would settle the question are AI-company revenue relative to spending, and credit spreads on data-center borrowers. Reports of Bowman's remarks also point to wider corporate bond spreads and higher credit default swap prices for some companies in the buildout. How the circular guarantees behave under stress is the test that no survey can run in advance.

Frequently asked questions

Does the Fed say AI is the biggest risk to the financial system?

Not exactly. The 87% figure comes from a New York Fed survey of 23 market contacts, and the report says those are the contacts' views rather than those of the Fed Board or the New York Fed. The Board did separately flag high stock valuations and examine AI financing in its own section for the first time.

Why did the share naming AI jump from 50% to 87%?

The survey does not explain it, but several factors are visible. AI financing drew heavy attention, with more than $680 billion of data-center borrowing in the first 10 months of 2026 according to Axios. Geopolitical concern also fell from 75% to 30%, which lifted AI in the ranking. The panels are small, at 20 and 23 contacts, so the change is suggestive rather than statistically firm.

Is my 401(k) exposed to AI stocks?

Possibly. CNBC, citing Morningstar, reported in September that five AI-exposed tech giants make up 30% of the S&P 500, and that investors may own them through an S&P 500 index fund or a target-date fund. Exact exposure depends on the funds held, and no authoritative household-level data was available for this article.

Are banks in danger because of AI lending?

The evidence available does not say so. Summaries of the Fed report describe banks and households as solid outside AI, and Fed officials note that the largest AI firms may be able to absorb losses. Axios does report that banks have nearly $1 trillion in exposure to AI infrastructure firms and lenders, which is why the Fed is watching the linkages.

Is the $3 trillion in AI commitments debt?

No. The figure is the Fed's tally of potential maximum nominal value for leases, take-or-pay contracts and residual value guarantees at seven large firms. It is undiscounted and, in the Fed's words, not directly comparable with debt. The seven firms' on-balance-sheet debt is expected to exceed $800 billion by year-end, according to Governor Barr.

Sources (8)

  1. Financial Stability Report, November 2026 , Board of Governors of the Federal Reserve System (2026-11-13)
  2. Financial Stability Report (publication page) , Board of Governors of the Federal Reserve System (2026-11-13)
  3. Speech by Vice Chair for Supervision Bowman on the U.S. economy and considerations for monetary policy , Board of Governors of the Federal Reserve System (2026-11-10)
  4. Federal Reserve flags financial risks tied to AI financing boom , Axios (2026-11-16)
  5. Fed's first close look at AI buildout finds $3 trillion of commitments off Big Tech's balance sheets , Fortune (2026-11-17)
  6. AI Bubble Risk: Stock Selloff Shows Markets' Dependence on Tech , Bloomberg
  7. AI stock swings could affect your 401(k). Here's what retirement savers need to know , CNBC (2026-09-18)
  8. AI Rockets to Top of Worries About Financial System in Latest Fed Survey , GV Wire (2026-11-13)

Share this article

Back to list