Technology was the worst S&P sector, down 1.81%, on the same day Zscaler jumped 16.45%, which tells you the AI trade is no longer moving as a single block.
That split mattered more than the index loss. The S&P 500 fell 0.48%, eight sectors finished lower, the VIX rose 7.95%, and WTI crude gained 1.34%. Those are the ingredients of a market getting pickier under pressure: it didn’t abandon growth altogether, it sold the parts of growth that need heavy spending and kept paying for the parts that could benefit if AI makes companies more exposed to security threats.
AI exposure is being sorted by business model
Stockmarketwatch tied the hardware hit to a weekend essay from Anthropic CEO Dario Amodei, with support from OpenAI’s Sam Altman and Elon Musk, urging a slowdown in AI development. The same digest said semiconductor and printed-circuit-board groups fell hard while software services rose. That’s the mechanism to care about: if the AI story shifts from “build everything faster” to “control the risks,” the winners and losers inside your tech allocation can flip even while everyone still calls it an AI market.
The sector map backed that up. Communication services led, health care and staples also gained, and technology lagged. That is an odd combination if you only look at yesterday as a generic tech selloff. It makes more sense as a re-rating of where future AI dollars go. Platforms, security software, and defensive earnings got a pass. Hardware supply chains, optical names, and equipment stocks had to answer for whether the spending boom can keep running at the same speed.
The outliers showed the rotation clearly
Zscaler had the cleanest catalyst: earnings and guidance beat expectations, so the stock’s move was about a company raising the bar and clearing it. CrowdStrike gained 13.88% as AI risk warnings appeared to push money toward cybersecurity demand. Corning was on the other side after announcing a $2 billion at-the-market share offering, adding dilution concern to existing weakness in AI optical-connectivity stocks.
Those names are useful because they separate company-specific news from the broader theme. Zscaler’s rally was earned through results. CrowdStrike’s move leaned on the idea that AI threat growth could raise security spending. Corning’s drop showed how quickly the market punished any stock where the AI infrastructure story met new share supply or questions about demand. If you own a broad tech fund, all of those exposures can sit next to each other in the same wrapper, pulling against each other.
That is why yesterday’s damage felt sharper than the index move. A small headline decline can hide a violent reshuffle underneath. The market wasn’t simply taking down expensive stocks. It was changing the price it will pay for different versions of the same AI theme. Spending beneficiaries with visible software demand held up better. Capital-intensive beneficiaries with supply, dilution, or buildout-speed questions were treated like the crowded side of the trade.
This morning’s bounce still has a catch
The setup before the open is only partly reassuring. S&P 500 futures were up 0.44%, but the average premarket gap was still -1.01%. That means the headline index can look healthier than the stock list you actually own. Eli Lilly, Exxon, Chevron, Arm, Taiwan Semiconductor, and Zoetis were higher early, while Coinbase, MicroStrategy, Fortinet, Rubrik, ServiceNow, and CrowdStrike were lower.
The important part is that several of yesterday’s software and security winners were giving back some ground before the bell. That doesn’t erase the rotation, but it does make the open a test of whether Monday’s move was durable buying in cybersecurity or a fast scramble away from hardware that already needs a breather. If the futures gain is carried by a few large names while yesterday’s winners fade, your portfolio may feel weaker than the index quote on your screen.
Watch whether cybersecurity holds its gains while chip-linked and optical names stop falling. If those two things don’t happen together, yesterday was a warning that AI exposure now needs to be read line by line, not owned blindly through the label.