The AI stock shudder is markets asking for proof, not a verdict

Is the market falling out of love with artificial intelligence? We don't think so. It is falling out of love with taking AI's word for it, and those are very different things.

The latest shudder has been global, and it follows a rough stretch. The LA Times called last Friday the cap on the market's worst week since late June. This Thursday, Reuters reports, stock indexes fell after Meta's earnings report reignited worries about the AI boom's heavy capital spending.

The trigger is the bill, not the technology

Look at what has been spooking people. The Guardian tied the mid-October selloff to fears over rising inflation and future interest rate hikes, and CNN's coverage pointed to rising rates and AI debt. None of that says "chatbots don't work." It says money is getting more expensive just as AI spending is getting bigger, and the talk of AI debt suggests borrowing is part of the picture.

That is a repricing of promises. When capital is cheap, a company can say "we will spend heavily now and the revenue will come." When rates and inflation worry people, the market wants a date and a number. In our view, that's healthy, and arguably overdue.

Good news no longer buys a rally

The sharpest tell isn't from this month. Last November, Nvidia posted strong results and stocks rose briefly, then reversed within a day. As the BBC reported, a StoneX analyst said "what should have happened didn't happen." One portfolio manager quoted by the Guardian made the key distinction: the company selling the chips is benefiting, but the customers buying them may be spending too much.

That is the fault line we would watch. The chip seller has revenue today. The big spender has a capex line and a hope. The market seems to be slowly separating the two, and this week's drop after Meta's report suggests it is increasingly willing to punish the spender, even when the story is exciting.

What the winners will look like

We expect the next phase to reward a boring-sounding trait: revenue that repeats. Customers who renew, usage that pays for itself, margins that survive the depreciation of all those data centers. Capex announcements alone will probably stop working as a catalyst, and might become a liability.

A caveat, and an honest one. Selloffs have bounced fast: Reuters reported AI-linked stocks rebounding on Monday as bond yields and oil eased. So macro conditions, not just AI fundamentals, are clearly driving the swings. That makes any single day a poor guide, and we would not read a bounce as proof either.

Another view worth noting: a Yahoo Finance piece argues worriers should focus more on supply risk in the infrastructure buildout than on valuations. We haven't seen the full argument, so we won't lean on it, but it hints that the debate may be moving from "is it a bubble?" to "can it be delivered and paid for?"

Our verdict

If you're an investor, the lesson is not to flee AI or to buy every dip. Ask each company one question: where does the money come from, and when? Firms with an answer will look better with every shudder. Firms with only a spending plan will look worse. The technology isn't on trial. The business models are, and for once the jury wants evidence.