Kashkari Pushes Back on the 'Only AI' Story of US Growth

Minneapolis Federal Reserve President Neel Kashkari said on Sept. 30, 2026 that he is skeptical of the argument that the US economy would be contracting were it not for a boom in artificial-intelligence-related construction.

Neel Kashkari, president of the Federal Reserve Bank of Minneapolis, has pushed back on one of the most repeated claims in the current macro debate: that strip out the artificial-intelligence building boom and the United States economy would be shrinking. Kashkari said he is skeptical of that argument, according to Bloomberg Technology.

The remark is short, but it lands on a fault line that runs through nearly every forecast being published right now. If AI capital spending — data centers, substations, transmission lines, chip fabs, cooling plant — is the single load-bearing column under US growth, then the economy is one capex pause away from recession, and the Federal Reserve should treat the rest of the expansion as already fragile. If it is one driver among several, as Kashkari's skepticism implies, the policy calculus is very different.

Why the 'only AI' claim took hold

The argument became popular because the construction component of the AI build-out is unusually visible. Data-center projects are large, lumpy and physically obvious. They show up in nonresidential structures investment, in equipment orders, in regional electricity demand, and in the capital expenditure lines of a small number of very large technology and infrastructure companies. That concentration makes the contribution easy to isolate rhetorically — and easy to overstate.

The leap from "AI construction is a meaningful contributor to growth" to "AI construction is the only thing keeping growth positive" is a big one. It requires assuming that consumption, services output, government spending and the rest of business investment are collectively flat or negative. Kashkari's stated skepticism is a direct challenge to that assumption rather than a denial that AI spending matters.

It is worth being precise about what he did and did not say. The reported position is doubt about a specific causal claim — not a forecast, not a vote, not a statement about the next policy meeting. Fed officials routinely separate their read of the data from their read of the appropriate policy response, and nothing in the remark commits him to either side of the rate debate.

What it would change for rate policy

The "only AI" framing is, functionally, a dovish argument. It says the underlying economy is weaker than the headline numbers suggest, that the strength is narrow and reversible, and that policy should ease pre-emptively before the capex cycle turns. A policymaker who rejects that framing is implicitly saying the breadth of growth is better than advertised — which reduces the urgency behind cutting.

That is why a single line of skepticism from a regional Fed president is worth more attention than it might appear. Rate expectations are built from exactly this kind of signal: not from forecasts, but from how individual officials characterize the quality of growth. An official who sees a broad expansion will tolerate a slower pace of easing than one who sees a single sector holding up the aggregate.

There is a second, harder question lurking behind it — one the Fed has no settled answer to. AI investment arguably pushes in both directions at once: it is demand now, in the form of concrete, copper, turbines and labor, and it is potentially supply later, in the form of productivity. Demand-side effects argue for caution on inflation. Supply-side effects, if they materialize, would argue for the opposite. Officials have been publicly wrestling with how to weigh the two, and Kashkari's comment is part of that wider conversation.

How markets closed on the day of the remarks

Equity markets ended Sept. 30, 2026 with a split that tracks the debate itself. The SPDR S&P 500 ETF Trust (NYSEARCA: SPY) finished at $762.63, down 0.21% from the prior close of $764.20, having traded between $762.18 and $769.41 — closing effectively at the low end of its range.

The Invesco QQQ Trust (NASDAQ: QQQ), the Nasdaq 100 proxy where the AI capital-spending complex is most heavily represented, closed higher at $739.77, up 0.25% from $737.93. The SPDR Dow Jones Industrial Average ETF Trust (NYSEARCA: DIA) was the weakest of the three, ending at $508.55, down 0.84% from $512.88.

That leaves roughly a 1.09-percentage-point gap between the Nasdaq 100 proxy and the Dow on the day — an illustrative spread derived from the two closing moves, not a reported statistic. It is a single session and should not be read as a verdict on anything. But it is the pattern that makes the "only AI" argument intuitive to investors: the tech-concentrated index holds up while the broader industrial benchmark gives ground. The market-level version of the question Kashkari was asked is whether that divergence reflects a narrow economy or merely a narrow market.

What would actually settle the argument

The claim is testable, which is more than can be said for most macro slogans. Several strands of data bear on it directly:

The composition of GDP growth — specifically how much of quarterly growth comes from nonresidential structures and equipment versus consumption and services.
Private nonresidential construction spending, where data-center and computing-facility categories are reported separately from other commercial building.
Capital expenditure guidance from the handful of hyperscalers and infrastructure firms financing the build-out, which is where any deceleration would show up first.
Payrolls by sector, since construction and utilities hiring tied to the build-out is separable from health care, leisure and other service employment.
Regional Fed surveys and the Beige Book, which capture whether demand outside the AI corridor is genuinely flat.

None of these resolves the question alone. Together they determine whether the skeptics or the "only AI" camp are right — and whichever way they break, the answer will shape how fast the Fed moves.

The risk on both sides of the argument

For investors, the practical stake is concentration risk. If growth really is as narrow as the strongest version of the AI thesis claims, then a pause in data-center commitments would hit not just the technology complex but industrials, utilities, materials and the regional economies hosting the projects. Diversification away from the AI trade would, in that world, be less protective than it looks, because so much of the cyclical economy has been pulled into its orbit.

If Kashkari is right and growth is broader, the opposite risk applies: portfolios heavily weighted toward the build-out are paying a premium for an exceptionalism that the wider economy does not actually require. The next few prints of construction spending and capex guidance will matter more to that question than any single speech.

Key facts

Who: Neel Kashkari, President, Federal Reserve Bank of Minneapolis
Claim disputed: That the US economy would be shrinking but for AI-related construction
QQQ close (Sept 30, 2026, 20:00 GMT): $739.77, +0.25% vs prior close $737.93
SPY / DIA close (same time): SPY $762.63, -0.21%; DIA $508.55, -0.84%

Frequently asked questions

What exactly did Kashkari say?

Federal Reserve Bank of Minneapolis President Neel Kashkari said he is skeptical of the argument that the US economy is shrinking but for a boom in construction related to artificial intelligence. The remark, reported on Sept. 30, 2026, challenges the idea that AI capital spending is the only thing keeping aggregate US growth positive.

Why does the 'only AI' argument matter for interest rates?

It is functionally a dovish case. If growth depends on a single sector's capital spending, the expansion is fragile and reversible, which argues for faster easing. A policymaker who rejects that framing is implicitly saying growth is broader and more durable, which reduces the urgency behind cutting rates.

How did US equity benchmarks close that day?

As of the last trade on Sept. 30, 2026 at 20:00 GMT, the S&P 500 proxy SPY closed at $762.63, down 0.21%. The Nasdaq 100 proxy QQQ closed at $739.77, up 0.25%. The Dow proxy DIA closed at $508.55, down 0.84%, making it the weakest of the three.

Does AI spending push inflation up or down?

Potentially both. In the near term it is demand: construction, equipment, energy and labor, which is inflationary at the margin. Over a longer horizon, if it raises productivity, it adds supply capacity, which is disinflationary. Fed officials have publicly been weighing how to balance those two opposing effects.

What data would show whether AI construction is really carrying growth?

The composition of GDP by category, private nonresidential construction spending with data-center categories broken out, hyperscaler capital-expenditure guidance, sector-level payrolls separating construction and utilities from services, and regional Fed surveys including the Beige Book on demand outside the AI corridor.

Is Kashkari's comment a signal about his rate vote?

No. What was reported is a judgment about how to read growth data, not a policy commitment or forecast. Fed officials routinely distinguish their assessment of economic conditions from their view of the appropriate policy response, and nothing in the remark binds him on a future decision.

Sources

Fed’s Kashkari Says AI Not Only Driver of US Economic Growth — Bloomberg Technology

Photo: Hans · Pixabay Content License — source