We’ve been here before. As with so many other transformational inventions — railroads, electricity, the internet — outsize enthusiasm for artificial intelligence is supporting a level of spending that’s gotten ahead of plausible profit expectations, a defining characteristic of a financial bubble. Now that the stocks of some A.I. builders and suppliers are wavering, we think it’s important to ask the next critical question. What happens to the American economy if this bubble bursts?
The damage could run deep.
It was just over a year ago when we argued that the enormous sums being poured into A.I. hearkens to the go-go internet era of the 1990s and the housing boom of the 2000s, both of which turned out to be financial bubbles whose bursting spilled over into the broader economy, leading to job and income losses that affected most Americans. Now, we’re already starting to see signs that this latest boom may be fading: The tech-heavy Nasdaq index fell by more than 5 percent Thursday and is down nearly 10 percent since Oct. 1. Growing investor nervousness led the prominent A.I. lab OpenAI to dial back its latest funding-round target by $10 billion.
What happens if the A.I. boom goes bust? There’s a consensus building among investors, academics and market analysts that any downturn would be relatively mild, similar to what happened in the early 2000s, after the dot-com bubble burst. The inflation-adjusted value of the overall economy, or real gross domestic product, shrank less than half a percent. And while unemployment rose, it was considerably lower than it was at the peak of the financial crisis.
Unfortunately, this consensus looks too optimistic. First, the A.I. bubble is by all measures substantially larger than the dot-com bubble; the stock market has increased by more than it did between 1996 and 2000 and is currently worth 240 percent of G.D.P., versus 160 percent at the peak of the dot-com bubble.
This difference in magnitude means that if the stock market fell as much as it did over the course of the dot-com bust, the amount of lost wealth could translate into a reduction in real consumer spending that would be twice as large. That reduction would show up as a lower demand for goods and services, which could in turn reduce business investment, hiring and overall growth.
Our current economy is also more fragile than it was back in 2000. Hiring has been sluggish and real wages are falling, the latter due to elevated war-induced inflation. The Federal Reserve has been gradually raising interest rates, leading to higher borrowing costs and potentially dialing back economic activity.
Additionally, today’s fiscal conditions and outlook are much worse. In 2001, the federal government ran a budget surplus of 1.2 percent of G.D.P.; today it is running an historically large deficit of 6.2 percent of G.D.P. That could limit Congress’s willingness to inject even more borrowed money into the economy to offset any downturn.
To get a rough read on the magnitude of the damage triggered by what the A.I. bubble burst might do to the economy, we ran the same economic model that the Federal Reserve uses to simulate the workings of the macroeconomy. The results confirmed our concerns. Mapping a dot-com style sell-off onto today’s markets would lead real G.D.P. to fall by more than three times as much as it did during the 2001 recession: 1.4 percent versus 0.4 percent. The unemployment rate would steadily climb from its current rate of 4.2 percent to above 7 percent.
Back when we first expressed our concerns about the stability of the A.I. investment boom, we noted that the companies building the A.I. models and the infrastructure to support them were mostly spending their own cash, meaning a bust would hurt them, but not their lenders. That’s no longer the case: $445 billion worth of A.I.-related debt, or 1.4 percent of G.D.P., was issued in the first eight months of 2026. If we throw in the opaque off-balance-sheet debt deals companies are relying on to pay for this build-out, we estimate that total A.I.-infrastructure debt is roughly 7 percent of G.D.P., approaching levels of subprime mortgage lending that helped trigger the financial crisis.
Households and the banking system are both less exposed and less leveraged now than they were before the financial crisis. That said, debt-triggered recessions can be particularly intractable, as the credit system can take considerable time to recover, as borrowers regain solvency and lenders regain confidence. Should such debt financing continue to inflate the bubble, this risk factor will grow.
One of our biggest concerns this time is whether our leadership can effectively respond to a crisis. The consistently weak economic stewardship of the current administration is evident in persistently elevated inflation, interest rates and uncertainty. The extent to which officials are bogged down in their war of choice in Iran does not inspire confidence regarding their willingness and ability to respond effectively to a bubble-induced downturn, should it occur. We hope that the newly elected Congress will have a mitigating effect.
We’ve been careful to approach these thorny questions with humility. Trying to identify a bubble is hard enough. Predicting what happens if it pops requires even more speculation. But if current investment jitters deepen, what happens next could hit the economy harder than the last time we were here.
Jared Bernstein is a senior policy fellow at the Stanford Institute for Economic Policy Research, where Ryan Cummings is an economist.
Source photographs by xiaokebetter and Jun Zhang via Getty Images.
The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our email: [email protected].
Follow the New York Times Opinion section on Facebook, Instagram, TikTok, Bluesky, WhatsApp and Threads.
The post Tech Stocks Had a Terrible Day. What’s Coming Could Be Worse. appeared first on New York Times.
