Kevin Warsh’s AI Comments: “Interest Rates Get in the Cracks”

Key Takeaways

• Kevin Warsh, the 17th Chair of the US Federal Reserve since 22 May 2026, made two sets of AI remarks in October 2026 that pulled markets in opposite directions: optimistic in Bangkok on 16 October, and cautionary at the FOMC press conference on 28 October.
• At the 28 October press conference, Warsh said, verbatim from the Fed’s official preliminary transcript: “Interest rates get in the cracks, interest rates affect directly or indirectly, almost everything.”
• He did NOT call AI a bubble. The official transcript contains no “AI bubble” warning, and any headline suggesting otherwise is interpretation, not quotation.
• After the Bangkok remarks, traders cut bets on an October rate hike to about 15 percent, from more than 25 percent the day before, and Asian markets rallied — although stocks were already rising before he spoke, so the rally is not attributable to Warsh alone.
• After the 28 October press conference, commentary framed his words as a warning that debt-financed AI data-centre spending is sensitive to interest rates. The Motley Fool’s description of the remarks as driving “a dagger through Wall Street’s AI-driven rally” is the author’s interpretation, not Warsh’s words.
• Search interest in “kevin warsh ai stock market comments” rose sharply over seven days, according to the trend data behind this story — a rise in search interest, not a measure of trading volume.
• Warsh has ended forward guidance in FOMC statements and press conferences since taking over, which makes his remarks harder to read and gives individual sentences more market weight than they used to carry.

Who is Kevin Warsh, and why his words move markets

Kevin Warsh is the 17th Chair of the US Federal Reserve. He took office on 22 May 2026, succeeding Jerome Powell, for a four-year term running to 21 May 2030. He is not a newcomer to the central bank. He previously served on the Fed’s Board of Governors from 2006 to 2011, a period that included the 2008 financial crisis, and after leaving the Fed he became a fellow at the Hoover Institution and a partner at Duquesne Family Office.

That biography matters for understanding his October remarks, because Warsh speaks as someone who has sat inside the Fed during a crisis and as someone who has spent the years since watching markets from the outside. When a Fed Chair talks about artificial intelligence, traders listen for two things at once: what he thinks AI will do to the economy, and what he thinks interest rates will do to AI. In October 2026, Warsh gave them material on both questions, on two different days, in two very different tones — and the market heard both.

The first set of remarks came on 16 October 2026 in Bangkok, at the IMF–World Bank Annual Meetings, in a fireside chat with IMF Managing Director Kristalina Georgieva. The second came on 28 October 2026, at the press conference following the Federal Open Market Committee (FOMC) meeting. Between the two dates, the Fed’s own policy picture also changed in the background: on 16 September 2026, the FOMC had raised rates by 25 basis points, and at the 28 October meeting it voted 10–2 to hold the federal funds rate at 3.75–4.00 percent, with Lisa Cook and Beth Hammack dissenting in favour of a hike. Inflation, meanwhile, had run above the Fed’s 2 percent target for 67 consecutive months on an annualised basis. That is the context in which Warsh’s words landed — a central bank that had recently tightened, was now pausing, and was publicly divided about what to do next.

Bangkok, 16 October: AI as “the most consequential change”

In Bangkok, Warsh was expansive about AI. He called it “the most consequential change” and “the most profound economic transformation of a lifetime”, language that places AI alongside the largest economic shifts of the modern era rather than treating it as another technology cycle. He said AI would push prices up in the short term but bring a “huge benefit” over time, while warning that the path would be “very bumpy” and that its short-term employment impact was unknown.

That combination — long-term optimism, short-term disruption — is worth holding in mind, because it is the same combination that runs through his later remarks. Warsh was not presenting AI as a smooth productivity story. He said AI is “hitting the demand side of the economy before it’s going to hit the supply side”, a distinction with real policy consequences. If AI first raises demand — for chips, for power, for data centres, for the capital to build them — before it raises the economy’s capacity to supply goods and services more cheaply, then in the near term it can add to price pressures rather than relieving them. That is a reason for a central banker to be watchful even while being enthusiastic about the technology.

He also used the Bangkok appearance to reiterate that the Fed would “deliver price stability” and noted that it had “removed a dose of accommodation about a month ago”, a reference to the September rate hike. And he added, “I don’t prejudge meetings”, declining to signal what the FOMC would do at its next gathering. He also announced an in-person hearing in the San Francisco Bay Area on 16 November 2026 on AI’s effects on productivity and employment — a sign that the Fed intends to study the technology directly rather than merely reacting to market moves.

The market reaction to the Bangkok remarks was positive, but it needs careful attribution. After his comments, traders cut bets on an October hike to about 15 percent, from more than 25 percent on 15 October. Nasdaq 100 futures rose about 0.18–0.2 percent, the MSCI Asia Pacific index rose 1.4 percent, and Korea’s KOSPI rose about 3.4–4 percent, rebounding from its slump on 15 October, while the dollar gauge fell. However, stocks were already rising before his remarks as the chip selloff eased, so the rally cannot be attributed to Warsh alone. What can fairly be said is that his remarks coincided with, and plausibly reinforced, a day when markets were already recovering and rate-hike bets were falling. The Bangkok Warsh was, in market terms, the reassuring Warsh: AI is transformative, the Fed is watching, and no hike is being pre-announced.

The 28 October press conference: “Interest rates get in the cracks”

Twelve days later, the tone was different, and the sentence that defined the press conference was this one, verbatim from the Fed’s official preliminary transcript: “Interest rates get in the cracks, interest rates affect directly or indirectly, almost everything.”

Warsh was answering a question — from Claire Jones of the Financial Times, according to transcript summaries — about whether AI investment is insensitive to interest rates. The argument he was being asked about runs like this: the expected returns from AI are so large, and so different from ordinary investments, that higher borrowing costs will not slow the spending down. Companies will build data centres and buy chips regardless of what the Fed does, because the prize is too big to wait for cheaper money.

Warsh said he was sceptical of that view. His full answer, as recorded in the transcript, was: “I was skeptical of that view [that rate rises would not affect AI because expected returns are so different]… Since early September… market participants [are] discriminating between the kinds of investments they want to make. You’ve seen credit spreads for certain assets widen out, gap out. For others, you’ve seen some tightening in financial conditions, but not significantly.”

That is a precise and deliberately qualified statement, and each part of it deserves attention. First, the scepticism is directed at a specific claim — that AI investment is immune to rates — not at AI investment itself. Warsh was not saying AI spending should stop, or that it will stop. He was saying that the idea that it cannot be affected by borrowing costs is wrong.

Second, the evidence he cited was market discrimination, not a market collapse. Since early September, he said, market participants had been “discriminating between the kinds of investments they want to make”. Credit spreads — the extra yield investors demand to hold riskier debt over safe government bonds — had widened, or “gapped out”, for certain assets. For others, financial conditions had tightened, “but not significantly”. That is a description of selective repricing: some borrowers are finding money more expensive, others are not much affected, and investors are starting to distinguish between them. It is not a description of a credit freeze.

Third, Warsh added that capital expenditure is “broadening” beyond a narrow tech story. That matters because it cuts against the idea that AI spending is a single, fragile bet concentrated in a handful of companies. If capex is broadening, the spending is spreading across more of the economy — which is consistent with his Bangkok point that AI is a broad economic transformation, not a narrow stock-market theme. He also declined to give forward guidance on a December move, which is consistent with his approach since taking over: he has ended forward guidance in FOMC statements and press conferences, making his remarks harder to read and giving each individual sentence more weight.

What Warsh did not say

The most important fact-check in this story is a negative one. Warsh did NOT call AI a bubble. The official transcript of the 28 October press conference contains no “AI bubble” warning. Nobody at the press conference, on the record in that transcript, used those words to describe the AI market.

That matters because the search data behind this story — a sharp rise in search interest in “kevin warsh ai stock market comments” over seven days — shows that many people went looking for his comments after the press conference, and much of what they found framed his remarks as a warning about an AI bubble or as an attack on the AI rally. Some of that framing came from commentary, not from Warsh. The clearest example is the Motley Fool’s 3 November piece, which described Warsh as having “driven a dagger through Wall Street’s AI-driven rally”. That is the author’s interpretation of the market impact of his words. It is vivid, it is memorable, and it is not something Warsh said. Readers who treat it as a quotation will misunderstand both the Fed Chair and the market.

The same care applies to the Fool’s substantive framing: that “interest rates get in the cracks” is a warning that debt-financed AI data-centre spending is rate-sensitive. As an interpretation of Warsh’s argument, that is fair — it follows directly from his scepticism about the claim that AI investment is insensitive to rates. But it is still an interpretation, and it should be labelled as one. Warsh’s own words are narrower and more technical than the headline version. He spoke about credit spreads widening for certain assets and financial conditions tightening for others, “but not significantly”. The headline version — a dagger through the rally — drops the qualification and adds a verdict he did not deliver.

There is a related point about the December meeting. CME FedWatch was cited at a 72 percent chance of a 25 basis point December hike in the Fool’s 3 November piece. Other sources give different figures on different dates, so that number should be treated as a snapshot from that day, not a fixed forecast. Warsh himself declined to give forward guidance on a December move at the press conference. Anyone claiming to know what the Fed will do in December on the basis of his October remarks is going beyond what he said.

How markets reacted — two different days, two different stories

The contrast between the two market reactions is the real story of October. After Bangkok, the read was dovish enough to cut October hike bets and accompany a rally in Asian equities. After the press conference, the read was hawkish enough for commentators to describe it as a warning to AI investors. Both reactions were responses to the same person, twelve days apart, talking about the same technology.

Part of the explanation is that the two appearances answered different questions. In Bangkok, Warsh was asked, in effect, what AI means for the economy in the long run — and his answer was that it means a profound transformation with huge eventual benefits. At the press conference, he was asked what interest rates mean for AI investment right now — and his answer was that rates affect almost everything, including this. Both answers can be true at the same time. A technology can be the most consequential change of a lifetime and still be sensitive to the cost of borrowing, especially when the buildout is heavily debt-financed. The tension is not a contradiction; it is the difference between a ten-year view and a ten-week view.

For readers, the practical lesson is about how to consume Fed commentary in an era without forward guidance. Warsh has ended forward guidance in FOMC statements and press conferences since taking over, which means there is no longer a scripted signal about the next meeting to anchor interpretation. In that environment, a single sentence — “interest rates get in the cracks” — carries more weight than it would have under a Chair who pre-announced the path of rates. Traders, journalists, and commentators fill the gap with interpretation, and interpretation travels faster than the transcript. The defence is to read the transcript itself. The Fed publishes it, and the six words that mattered most are there in plain language, with their qualifications attached.

It is also worth noting what the October story does not establish. It does not establish that AI spending is about to collapse, or that data-centre construction will stop, or that the AI trade is over. Warsh’s own evidence — widening spreads for some assets, modest tightening for others, broadening capex — describes a market becoming more selective, not a market shutting down. Selectivity is uncomfortable for the borrowers whose spreads have widened, and it is uncomfortable for investors who bought indiscriminately. It is not the same as a bust. For broader context on how investors have been repricing the AI boom, see our earlier coverage: AI Bubble Fears Surge as Markets Reprice the AI Boom.

Why the December meeting matters

The next test of Warsh’s words comes at the December FOMC meeting, and the reason it matters is precisely that he has given so little away. With no forward guidance, the market will read the decision, the vote split, and the press conference together, and any sentence about AI, credit conditions, or the broadening of capex will be parsed as closely as the October sentence was.

There are three things to watch, all of them verifiable rather than speculative. First, the vote itself: the October meeting split 10–2, with Cook and Hammack dissenting in favour of a hike, so the December vote will show whether that division has widened or narrowed. Second, the language on credit conditions: if Warsh repeats or extends the “discriminating” theme — spreads widening for some assets, tightening elsewhere — that would suggest the selectivity he described in October is continuing. Third, the San Francisco Bay Area hearing on 16 November 2026 on AI’s effects on productivity and employment, which Warsh announced in Bangkok. What the Fed learns there may shape how it talks about AI’s supply-side effects, which is the part of the story Warsh said had not yet arrived.

None of those three things will, on its own, settle whether AI investment is rate-sensitive. That question will be settled over quarters, by spending data, earnings, and credit markets, not by a single press conference. But the December meeting will show whether the Fed Chair who said rates “get in the cracks” thinks the cracks are widening.

For the debt side of the story, the scale of AI borrowing is part of the context. Companies have been raising large sums to fund AI infrastructure, and our report on Oracle and Alphabet Borrowing Billions for AI on the Same Day shows how directly debt markets and AI spending are now linked. That is the mechanism Warsh was describing: when the spending is financed with debt, the price of debt is part of the investment decision, whatever the expected returns.

What to watch next, without guessing the outcome

If you want to follow this story honestly, watch a short list of verifiable milestones rather than daily commentary. First, the 16 November hearing in the San Francisco Bay Area — who testifies, and what the Fed says it learned about AI, productivity, and employment. Second, the December FOMC decision and vote split, and whether Warsh’s press conference language on credit spreads and capex changes from October. Third, credit spreads themselves: whether the widening Warsh described for certain assets continues, narrows, or spreads to more borrowers. Fourth, company spending plans in the next earnings season, and whether capital expenditure keeps broadening beyond a narrow set of technology names, as Warsh said it was doing.

The funding environment is also worth watching from the company side. OpenAI’s experience in October — cutting its funding target as AI stocks slid — is a reminder that even the most sought-after AI companies face tougher questions about price when markets turn selective. Our report on OpenAI Cuts Its Funding Target by $10 Billion as AI Stocks Slide covers that episode in detail, and it is a useful companion to Warsh’s remarks: the Fed Chair described selectivity in credit markets, and OpenAI’s round showed selectivity in equity funding at the same time.

The bottom line

Kevin Warsh said two things in October 2026 that are both true and easy to confuse. In Bangkok, he said AI is the most consequential economic change of a lifetime, with a huge benefit over time and a very bumpy path to get there. At the FOMC press conference, he said interest rates “get in the cracks” and affect almost everything, and that he was sceptical of the view that AI investment is immune to borrowing costs. Neither statement is a bubble call. The transcript contains no “AI bubble” warning, and the “dagger through the rally” framing is a commentator’s interpretation, not Warsh’s words.

What he did say is narrower, more technical, and arguably more useful: since early September, investors have been discriminating between investments, credit spreads have widened for certain assets, financial conditions have tightened for others but not significantly, and capital expenditure is broadening. That is a description of a market becoming selective about AI — not a market abandoning it. The search surge behind this story shows how many people wanted to understand that distinction. The transcript, read in full, is the best place to find it.

Sources: Federal Reserve, FOMC Press Conference Transcript, 28 October 2026 (preliminary); Motley Fool, Fed Chair Kevin Warsh Just Drove a Dagger Through Wall Street’s AI-Driven Rally; IMF, Managing Director Fireside Chat with Federal Reserve Chairman Kevin Warsh, 16 October 2026; The Straits Times, Stocks Futures Rise as Fed Bets Ease After Warsh; Bitrue, Warsh AI Stocks Interest Rates.