Everything About AI
NextFin News - Federal Reserve Chairman Kevin Warsh called artificial intelligence the most consequential change to the US and global economy in his lifetime, forecasting a large long-term payoff for productivity and wealth while warning the transition will be "very bumpy" — and he deliberately refused to signal whether interest rates move at the Fed's October 27-28 meeting.
Speaking Friday, October 16, 2026, in a fireside chat with International Monetary Fund Managing Director Kristalina Georgieva at the IMF-World Bank Annual Meetings in Bangkok, Warsh delivered a message that split cleanly in two. AI is a positive supply shock on the horizon, he argued, but a demand-side inflation driver today. And on the policy question markets care about most — what happens next month — he offered nothing to trade on.
"I'd say the advent of this new technology is probably the most consequential change to the US and the global economy in my lifetime. There really is, I think, a huge benefit to the global economy, to the US economy over the horizon."
But the path there is uncertain. Warsh said the transition "is likely to be a very bumpy one, bumpy in our understanding of the technology, bumpy because we really don't know what the market structure is going to be of these new technologies." He added that AI's short-term effect on employment remains unclear, and noted the Fed would use the technology to analyze data more effectively in conducting monetary policy — itself an example of the positive effects he sees.
On interest rates, Warsh declined to preview the October 27-28 Federal Open Market Committee meeting. "I don't prejudge meetings," he said. "I don't believe in saying now what I might do at the meetings that we have in the balance of this year."
The silence is itself the signal. In September the FOMC raised the federal funds target by 25 basis points to 3.75%-4.00%, the first increase since July 2023. The committee's latest projections showed 16 of 18 officials expecting at least one more hike this year, which would lift the range to 4.00%-4.25% by year-end. Two meetings remain on the 2026 calendar: October 27-28 and December 8-9.
And the inflation backdrop is not cooperating. September's consumer-price index came in at 3.7% year over year, above the 3.6% forecast, with a 0.6% monthly gain — the largest in five months — driven by a 9.1% jump in gasoline. Core inflation, which strips out food and energy, rose 0.3% for the month to 2.6% annually, interrupting a cooling trend that had run since May.
The Two-Sided AI Shock the Fed Must Price
Warsh's framing captures the classic central-bank dilemma in new form: the AI boom is a positive demand shock first and a positive supply shock later. Today it means data centers, semiconductors, power infrastructure and a surge in capital spending that competes for capital and pushes prices up. Tomorrow it means more output per worker and cheaper goods and services.
The timing mismatch is the problem, and it is acute. Monetary policy works with long and variable lags. If the Fed tightens into a demand surge that will later be followed by a supply boom, it risks over-tightening — choking off the very investment that creates the productivity gain. If it waits for the supply benefit to appear in the data, inflation may already be entrenched. Either mistake is costly, and they point in opposite directions.
Warsh was not the only central-bank figure wrestling with this split personality. Georgieva, hosting the session, has described the global economy as being pulled in opposite directions by a negative energy supply shock and a positive AI demand shock, with oil near $100 a barrel and global public debt on track to exceed 100% of world GDP. That triad — AI, energy, debt — is the environment in which the Fed must now operate: a productivity promise wrapped in an inflation problem and financed by record borrowing.
This is not an abstract debate inside the Fed. The committee's own estimate of the longer-run neutral rate — the level that neither stimulates nor restrains the economy — rose to 3.25% in the September projections. The median federal-funds-rate dot sits at 4.125% for both 2026 and 2027. In plain terms, the FOMC is already pricing a world in which the neutral rate has moved structurally higher, which gives the current 3.75%-4.00% target less room to maneuver than a casual glance at the level would suggest.
Warsh's governing philosophy, laid out at Jackson Hole in August, is no forward guidance and no pre-commitment. "I stand here today committed to a discipline, not to a decision," he said then. Friday's refusal to prejudge October is that philosophy applied: the reaction function is data-dependent, and the data are currently sending mixed signals.
Here is the tension in one line: the Fed hiked in September because inflation is too high; it is hesitating now because it cannot yet tell whether AI is the cause of the inflation or the cure for it.
The Bumpy Path Is Already Showing Up in Credit
The "bumpy" transition Warsh described is not a hypothetical. It is already visible in the financing of the AI buildout, where the cost of capital is being repriced in real time.
CoreWeave, the AI-focused cloud provider at the center of the infrastructure boom, has taken on more than $51 billion of debt and obligations to fund its expansion. Investors are demanding to be paid for the risk: its five-year credit-default swaps recently stood around 1,088 basis points, up 33% week over week and 177% from a year earlier. The yield to maturity on its 9.75% notes due in 2031 reached 15.78%. Its stock fell 15.46% on October 14 to close at $65.78.
One stressed borrower does not make a crisis. But CoreWeave is the marginal borrower in the hottest capital trade of the decade, and when the marginal borrower starts trading at distressed credit levels, the second-order question is not whether that single company survives. It is whether the cost of capital for the entire AI infrastructure complex has permanently re-rated higher.
That matters for Warsh's supply-shock thesis in a way most commentary misses. The productivity payoff from AI depends on capital continuing to flow into deployment — chips, data centers, grid upgrades, software integration. If financing costs stay elevated, the diffusion of the technology slows, and the supply-side benefit arrives later and smaller than the optimists expect. In that scenario, the demand shock (spending) persists while the supply shock (output) disappoints — the worst combination for inflation.
The bond market is already registering the strain. The 10-year Treasury yield stood at 5.09% on October 16, after a mid-month swing that produced the largest two-day drop of 2026. That volatility is the market's way of saying it does not know whether to price growth, inflation, or a credit event.
What Warsh Is Not Saying — and Why It Matters
The most charitable reading of Warsh's silence is that it buys optionality. With inflation at 3.7% and the dot plot pointing to another hike, the market can reasonably assume the bias is upward. Before his remarks, Fed funds futures were pricing roughly a two-in-three chance of a 25-basis-point hike at the October meeting. But Warsh has refused to lock that in. The asymmetry — hawkish data, neutral language — is deliberate. It lets the Fed react to the October and November inflation prints without having talked itself into a corner.
The counter-thesis is that this is not caution but opacity, and that opacity is itself a tightening force. A central bank that will not say whether its next move is up or sideways is a central bank whose reaction function is hard to anchor. If markets cannot pin down guidance, they price a wider distribution of outcomes, which shows up as a higher term premium and higher long-term yields. The 10-year yield hovering near 5.1% is the live evidence: investors are demanding extra compensation not just for inflation, but for not knowing what the Fed will do.
There is a second, sharper version of the counter-argument, and it comes from Warsh's own intellectual camp. Critics of discretionary policy argue that anchoring a stance on a single forecast of AI productivity is exactly the kind of bet a rules-based central bank should not make. If the productivity data never arrives, the Fed will have kept policy tighter for longer on the strength of a promise rather than a print — and the credibility cost will be paid in the form of higher inflation expectations. The reform-minded answer to that critique is a reaction function that adjusts automatically: if AI delivers, rates come down without anyone having to forecast a product roadmap; if it underdelivers, rates hold.
The falsifying signal is specific. If the FOMC raises rates at the October 27-28 meeting, Warsh's "no prejudgment" stance collapses into an explicit hiking bias, and the market will price a terminal rate above the 4.00%-4.25% year-end dot-plot median. Conversely, if the Fed holds and signals a pause, the supply-shock narrative gains credibility and the disinflation-through-productivity trade comes back into play. Watch the October CPI and the November nonfarm-payrolls report: a string of prints at or above 0.3% monthly core inflation, combined with continued employment strength, makes the hike nearly unavoidable. A pair of soft prints below 0.2% would give the committee cover to hold.
The Verdict: Structural Promise, Cyclical Pain
So is AI a cyclical fluctuation or a structural shift? The answer has to be split, because both forces are present and they operate on different clocks.
The capital-spending boom is cyclical. Investment surges of this kind — railroads in the 19th century, electrification in the 1920s, the internet in the 1990s — follow a familiar pattern: over-enthusiasm, over-building, a financing crunch, and then a shakeout that leaves the survivors with durable advantages. The CoreWeave credit stress is the crunch phase arriving on schedule. Cyclical legs mean-revert; this one will too, and the mean-reversion will be painful for the marginal borrowers.
The technology itself is structural. General-purpose technologies change the production function, not just the level of spending. If AI raises trend productivity growth by even a fraction of a percentage point for a decade, the compounding effect on real output and living standards dwarfs any single rate cycle. That is Warsh's call, and it is defensible — but it is a call about the 2030s, not about the October 28 meeting.
The policy error to guard against is conflating the two: using a structural promise to justify cyclical patience, or using cyclical inflation to dismiss a structural gain. The Fed's job for the next two meetings is to keep them separate.
What to Watch
Short term (the October 27-28 meeting): The base case is a hold with a hawkish tone, keeping the December door open. The upside case is a pair of soft inflation prints that let the Fed signal a pause, which would be constructive for equities and duration. The downside case is the dot-plot's second hike, lifting the funds rate to 4.00%-4.25% before year-end — negative for rate-sensitive assets but consistent with the committee's stated inflation priority.
Medium term (into 2027): If AI-driven investment keeps competing for capital while measured productivity stays flat, the Fed stays stuck, and the neutral-rate debate moves from academic to operational. The 10-year yield's range — roughly 4.9% to 5.3% in recent sessions — is the battleground. A decisive break above 5.3% would signal that the term-premium channel of the counter-thesis is winning; a sustained move below 4.9% would indicate the market is buying the pause narrative.
Long term (structural): Warsh's thesis stands or falls on one number: output per hour. A sustained step-up in productivity growth validates the supply-shock reading and makes today's inflation look cyclical. A persistent absence of productivity gains turns the AI boom into a demand shock with a debt hangover, and the terminal rate stays higher for longer than the dot plot currently implies.
The specific signals that would falsify the structural-supply-shock view: core CPI at or above 0.3% month over month for two consecutive months alongside flat productivity, or a further 200-basis-point widening in AI-infrastructure credit spreads without a corresponding pickup in capital-expenditure efficiency. Either would indicate that the demand effect is dominating and the supply payoff is not arriving.
Warsh's message, distilled: AI will make the economy more productive — but not in time to make the Fed's next decision easy. The central bank can afford to wait for the data; investors financing the boom may not have the same luxury.
Explore more exclusive insights at nextfin.ai.
