Could an AI Spending Crunch Cause a Recession?

Investors are raising concerns about a slowdown in artificial intelligence spending. While the market doesn’t appear to be factoring in significant recession risk yet, it’s worth a closer look given how important AI has been on the demand side of the economy. We think that an AI spending crunch, if it occurs, is unlikely to produce a recession.

The shock to aggregate demand should be small enough that the Federal Reserve can offset it by lowering interest rates. If an economic slowdown did occur, the worst case would be something like the 2001 recession after the dot-com bust, which was very mild.

Recessions are of obvious importance to any participant in the economy, given the accompanying job loss and other economic disruptions. From the investor perspective, they also tend to cause large declines in the prices of risky assets. While we’d be hard-pressed to recommend that investors ever try to time a market selloff, it’s still important to be prepared. Steeling oneself for market volatility can make it easier to avoid panicked decisions, namely selling at the bottom, which leads to permanent loss of capital.

A recession is one scenario that would be very positive for bonds, given the ensuing declines in interest rates. While we don’t expect a recession to occur, adding more bonds is one way to prepare against it. Moreover, we’re optimistic about bonds even if a recession doesn’t occur.

A Recession Isn’t on the Radar, Despite Increasing Investor Pessimism

Investors have started to sour on some of the biggest AI spenders. Amazon.com AMZN and Alphabet’s GOOG/GOOGL third-quarter earnings disappointed the market, and doubts have swelled over the past month about the financial viability of neocloud companies like CoreWeave CRWV.

In turn, that’s led investors to trim expectations for spending on AI infrastructure. The Morningstar US Technology Index has dropped around 9% since a month ago, driven heavily by chipmakers and others who benefit from AI spending.

AI spending worries even show up in the bond market: The 10-year US Treasury yield is down around 0.4 percentage points since its peak in early October, which followed a 1.2 percentage point runup year to date. This supports our prior argument that AI has been the main factor driving up bond yields in 2026. As spending expectations have been trimmed, bond yields have come off their recent peak.

However, the market doesn’t seem to be pricing in much probability for a scenario where AI spending slows sufficiently to cause a US recession. Outside of tech, US stocks have done well over the past month, which isn’t consistent with a recession scenario.

Likewise, consensus rates the likelihood of a recession as low.

The median economic forecaster calls for a 24% probability of a recession in the next 12 months. That may sound high, but keep in mind forecasters have a lot of incentive to inflate this probability. A 24% probability is pretty low within the range of historical forecasts—it’s closer to an environment like the later 2010s when there were few recessionary clouds on the horizon.

The Fed Should Be Able to Offset an AI Demand Shock

But is the market right to be nonchalant about recession risk?

After all, AI is contributing a big part of economic growth from a demand-side perspective. GDP growth averaged 2.3% year over year in the first three quarters of 2026. Split by expenditure, a full 1.1 percentage points of that—nearly half—is coming from tech-related investment. Furthermore, by boosting stock prices and thereby household wealth, AI is also a major prop to consumer spending.

If spending on AI cratered, that demand-side boost to GDP would shrivel up.

But that’s not the end of the story: As we detailed in our earlier piece, the Fed has been offsetting AI’s demand-side shock by keeping interest rates higher than they would otherwise be. Right now, the federal-funds rate is about 3 percentage points above its 2017-19 prepandemic average, and the 10-year Treasury yield is 2.5 points above it. That whole process would run in reverse should AI spending falter.

The Fed would respond to falling AI spending by loosening monetary policy. That would unleash spending in parts of the economy, like housing, that have been suppressed by elevated interest rates.

The main recessionary risk comes if the Fed runs out of room to cut. That is, if the federal-funds rate abuts the lower bound at zero, where it was for nearly all of 2009-15. The Fed has some other tools in that case (namely quantitative easing), but they’re less powerful than federal-funds rate cuts. This helps explain the economy’s meager recovery coming out of the Great Recession.

Still, the Fed has a lot of room to cut today. The federal-funds rate stands at 3.75%-4.00%, and more importantly, bond market expectations are currently factoring in hikes all the way up to 4.25%-4.5% by mid-2027. Any shortfall vis-à-vis that latter figure would constitute monetary policy loosening.

An AI Slowdown Would Likely Be Even Milder Than the Dot-Com Recession

In the event of an AI spending crunch, the best historical analogy is probably the bust following the 1990s dot-com bubble. That suggests that if a recession did occur, it would be mild. It’s quite possible that monetary easing would be sufficient to counteract the shock from AI, meaning the main impact of an AI spending crunch would be lower interest rates, not a recession.

Indeed, the 2001 recession pales compared with the Great Recession. GDP growth was never negative in year-over-year terms, troughing at 0.5% in the third quarter of 2001, versus the negative 4% posting in the second quarter of 2009 during the Great Recession. The unemployment rate increased by about 2 percentage points during the 2001 recession, versus about a 5.5-percentage-point increase in the Great Recession.

Many of the factors which made the 2001 recession mild are relevant this time around. The tech investment boom of the 1990s was heavily equity-financed. US business sector debt as a share of GDP did increase by about 10 percentage points in the second half of the 1990s, but this was far smaller than the increase in household debt/GDP of nearly 30 percentage points in the 2000s housing bubble.

Debt-fueled financial bubbles tend to produce much more violent corrections when the bubble bursts. One reason is the vicious cycle that emerges when the fall in collateral values and availability of credit proves mutually reinforcing. In the housing bubble, for instance, financial panic led to shrinking mortgage credit as well as distressed asset sales. This sent housing prices downward, spurring further financial panic.

Today, we see few signs of financial instability that could trigger a meltdown in the event of slowing AI spending. Sure, there could be isolated instances of financial distress, such as in private credit. But overall financial leverage in the private sectors (household and business) is not elevated by historical standards.

The hyperscalers financing much of the AI buildout have done so mostly using their own operating cash flow so far. That has started to change in 2026, and if the AI boom continues to grow and becomes more debt-reliant, then a spending crunch could have more dire consequences for the overall financial system. As of right now, we don’t appear to have a financial house of cards waiting to collapse.

While the dot-com bust is the best template we have, it probably is more of a bear case than a base case for the ramifications of an AI spending crunch. The S&P 500 P/E ratio, based on forward 12-month consensus earnings, stood at 19 in October, while it was over 24 at the market peak in 2000. That suggests less room for a stock market correction.

We’ve also argued that monetization of AI is in a much better place than that of the internet circa 2000. The AI infrastructure that’s been built so far is being fully utilized and is generating real revenue from end users. That’s an entirely different situation from the speculative buildout of the internet infrastructure in the 1990s. That puts a definite floor on how far AI spending could feasibly fall.

Finally, there were other shocks hitting the economy in the early 2000s at the same time as the dot-com bust. These included the 9/11 terrorist attacks and the displacement of US manufacturing from the rise of China.