Is The AI Trade Officially In Trouble

Follow the whole story: every piece in this series is on our AI storyline page.

Short answer: a day like Thursday isn't trouble. Trouble has a definition, and it's simple: the four big buyers of AI chips cutting what they plan to spend in 2027. Until that happens, a 9% day is a price story, not a business story. The question that will still matter next year is a quieter one: how much of your money rides on AI, counting your index fund too? We'll come back to it.

What Does Everyone Already Know?

That it was an ugly day. SPY fell 4.4% and QQQ 6.5%. It was the AI trade unwinding: the chip makers and the companies building the power and data centers behind AI fell hardest. Power is the bottleneck, opposition to new data centers is growing town by town, and the new Congress can lean on both; a Wall Street Journal article argued on Thursday that the backlash could play into the hands of Big Tech. We flagged it on Nov 2, in "Nvidia Stock's Real Risk Isn't China. It's Power."

The AI trade has two sides: the companies that sell the build-out (chips, power gear, cooling) and the four that pay for it. Thursday split them cleanly:

The sellers fell about 12% on average; the buyers about 3%. In dollars, the sellers lost roughly $940 billion of market value and the buyers about $350 billion.

Look at the last column. The hardest falls came where the biggest gains had been: Micron had more than tripled in a year and AMD had nearly doubled. Nvidia, the face of the trade, was flat over the 12 months even before Thursday. And Vertiv now sits about half below its high.

What Changed Since Wednesday?

On Wednesday the market paid the buyers for spending less: they rose while Nvidia fell 3.2%. We wrote up that trade in why Amazon and Google stock rose. On Thursday nobody got paid; the buyers just fell far less. Put the two days together and Nvidia is down about 12% from Tuesday's close, while the four buyers, on average, are up about 1%.

That's investors betting that if spending slows, the suppliers get hurt and the buyers keep the cash. It's a guess about the future, priced in two days. It isn't news about the businesses: the latest spending plans are still the ones the four gave with their results in late October. So let's look at the businesses.

Are The Businesses In Trouble?

Here's the classic Trefis check on all eleven: what you pay, how fast sales are growing, and whether profit margins are holding. P/E is at Thursday's close; growth and margins are over the last 12 reported months.

All eleven are growing. Of the ten with five years of history, nine earn margins at or above their five-year average; Meta is the exception, and only by two points.

That isn't what a business in trouble looks like. A few things stand out. Nvidia traded at about 25 times earnings before Thursday and about 23 after, less than Microsoft, while growing sales several times faster. Micron, at about 9 times, is priced as if its margins won't last: they're running at more than three times its five-year average. AMD, at nearly 90 times, is the one priced for a lot to go right. The risk in these stocks is the price you pay for growth that has to keep coming, not growth that has stopped.

Can The Buyers Keep Paying?

This is the real question under the AI trade, because the buyers' spending is the sellers' revenue. Here's their capital spending story in one table:

The four plan to spend more this year than their businesses brought in over the past year. In that year they kept about $61 billion of free cash between them. Nvidia alone kept $127 billion.

That last line is the whole story. The money hasn't vanished; it has moved down the chain, from the buyers' cash flow to the sellers' profits. For scale, the seven sellers in our table took in about $660 billion of revenue over the past year, less than the four buyers plan to spend in 2026. Not every capex dollar lands with these seven, but it shows who depends on whom.

And the buyers say they aren't done. Alphabet has said its spending will rise "significantly" in 2027 and Amazon "meaningfully"; Microsoft says it expects to stay free-cash-flow positive while it builds. Amazon is the one already spending more cash than it makes. There's a twist, too. If towns and states slow new data centers, the capacity that's already built gets scarcer, and the four buyers own a great deal of it. A pause can hurt the builders and help the incumbents at the same time.

What Would Actually Count As Trouble?

One thing: the buyers spending less on AI next year than they said. If two or more of the four hold 2027 spending flat or cut it, the build-out is slowing, and that's the moment to revisit your plan. Two dates bring real data: Nvidia reports its quarter on Nov 17, and the four buyers report in late January and early February, with their 2027 plans. A second, earlier tell sits in the tables above: if the buyers' revenue growth slows while their spending keeps climbing, their free cash flow shrinks further, and pressure to cut rises. A red Thursday isn't either of those.

How Low Can AI Stocks Go?

Lower than this, even with the business fine. Here's how far each fell in the 2022 bear market, how many months it took to get back to its old high from the bottom, and the 2008 fall for those old enough:

In 2022 the buyers kept spending, and Nvidia still lost two-thirds of its value. Thursday left it about a quarter below its high.

That's the honest frame for a day like Thursday. A price can fall much further than this without anything breaking in the business, and it can take a year or more to come back. Which is why the plan below is about size, not prediction.

"A falling price tells you how other people feel. The business tells you what it is worth." - Anonymous

What Should You Do Now?

1. Count your AI, all of it. Your direct shares, plus what's inside your index funds, plus the builders and power names that ride the same spending. Nvidia, Broadcom and the four buyers alone are about a quarter of an S&P 500 fund. Most people who count this way find more than they thought; we show how in how much AI you really own.

2. Set a band, not a target. Pick the share you want in AI, then a range around it. For example, a 25% target with a band of 20% to 30%. Above 30%, trim. Below 20%, add. Inside the band, do nothing.

3. Let the band decide, not the day. The move this replaces is selling after a 9% chip day, or buying everything on the dip. If Thursday pushed your count below the band, the rule says add, at prices you already decided were fair. If you're still above it after Thursday, you were too heavy before Thursday.

The risk: if this turns into a 2022-style fall, a band still loses money. You'll buy on the way down and the price can keep falling, as the table above shows. What the band does is cap how much of your money is exposed, so the loss is one you planned for.

What's The Bigger Idea?

Two ideas, really. Exposure hides in plain sight: your index fund owned every side of Thursday's trade, the sellers and the buyers. And price isn't the business: a day like Thursday for AMD tells you how other investors felt that day; the buyers' 2027 plans and their cash flow will tell you what the business is worth. Watch the plans, size by the band, and let the price do what prices do.

The AI trade isn't officially in trouble; it's officially volatile. The sellers fell about four times as much as the buyers on Thursday, yet all eleven are growing revenue and nine of ten are running at or above their five-year margins. The buyers plan about $730 billion of 2026 capex, and Alphabet and Amazon have said 2027 goes higher. Trouble is two or more of the four holding 2027 spending flat or lower, which you'll see in late January. Until then, count your total AI exposure, set a band, and rebalance only when the count crosses it.

What Should You Ask Next?

Follow this story with us: we'll be back after Nvidia's report and again when the buyers set their 2027 plans, with what changed and what didn't.

Is Your AI Exposure Inside A Band You Chose?

Thursday tests the plan, not the trade. If you have $1 million or more invested, get a second opinion on your AI exposure from a Trefis expert partner advisor in a free 15-minute call, and ask about Invest With Trefis.

For questions about this article or underlying research, email [email protected].