AI Stocks Catch a Third Straight Cold as Oracle Keeps Sliding

​The AI trade is suddenly out of breath
U.S. AI-linked tech names stayed under pressure for a third day, and the premarket tape looked like a slow leak rather than a quick dip. Oracle, Nvidia, Micron, and CoreWeave all slipped again, a sign that investors aren’t just taking profits—they’re stepping back from the whole AI cluster and reassessing the story.

Oracle is the stumble that keeps echoing
Oracle was down roughly 0.9% in premarket trading after a brutal drop the day before, when the stock finished about 11% lower. The trigger was simple and painful: revenue came in at $16.06 billion, missing the $16.21 billion analysts expected. In this market, missing by a little is still missing—and “AI excitement” doesn’t get you a hall pass.

Nvidia, Micron, CoreWeave: sympathy sells are real
Nvidia shed about 0.7%, Micron fell around 0.9%, and CoreWeave slipped roughly 1.3%. That’s the vibe of a crowded trade unwinding: when the headline name wobbles, the whole theme catches the wobble. Even Broadcom, fresh off a strong quarter, was down about 5% in premarket action—because right now investors are treating AI-adjacent tech like one big connected organism.

This isn’t a market crash—it’s a rotation
Here’s the plot twist: the broader market hasn’t been falling apart. The Nasdaq dipped on Thursday, but the Dow and S&P 500 still finished at record highs. That split-screen is the story: money isn’t fleeing stocks, it’s just migrating away from tech and into other corners that feel less crowded, less hyped, and less likely to disappoint on a single line item.

The real worry: “circular deals” and the cost of building the AI future
Oracle’s results also revived a set of nagging investor worries—especially about how AI infrastructure gets funded and how sustainable some of the industry’s dealmaking really is. There’s been rising scrutiny of “circular” GPU and cloud arrangements and of how much debt or long-term commitments companies may take on to build capacity fast.

One analyst argued the market may be getting overly punitive, pointing out Oracle still has strong customer stickiness across databases, applications, and infrastructure—yet the same analyst also cut a fair value estimate for Oracle, citing concerns that delivering planned AI capacity on time is proving harder than expected.

Bottom line: AI isn’t broken, but the market’s patience is
​This pullback doesn’t mean AI demand vanished overnight. It means expectations got so high that “good” stopped being enough—and any uncertainty about spending, timelines, or financing gets punished immediately. The AI trade can recover, but first it has to do what markets love most: deliver clean numbers, on time, with less mystery.