Summary: US software stocks sold off on February 3, 2026 as investors weighed AI-driven disintermediation of SaaS products, and the article describes the move as a sector-wide re-rating. Palantir rose almost 7% after its revenue and adjusted EPS beat estimates.
U.S. equities finished the day steeply lower on Tuesday, February 3, 2026, with investors continuing to reassess what AI means for software vendors. The S&P 500 lost 0.84% and settled at 6,917.81, while the Nasdaq Composite gave up 1.43% to end at 23,255.19; the Dow briefly set an all-time intraday high of 49,653.13 early in the session, then slid 166.67 points (-0.34%) to close at 49,240.99. Those index moves were only the backdrop — software took the worst of it, and what happened there was not a one-day shock but a sector-wide re-rating still underway.
What the Tape Showed
CNBC’s intraday coverage laid out just how broad the selling ran:
- ServiceNow and Salesforce both finished roughly 7% lower; Intuit was hit hardest, losing nearly 11%
- The iShares Software ETF (IGV) shed about 5% in the session, leaving it roughly 17% down on the year and about 28% under its recent peak
- The WisdomTree Cloud Computing Fund (WCLD) slipped 3.2%, a sixth consecutive losing day
- HubSpot, ServiceTitan, ServiceNow, Atlassian, Klaviyo, DocuSign, Asana, PagerDuty, Salesforce, Workday, and Adobe each marked fresh 52-week lows
- Microsoft and Meta dropped more than 2% apiece and Nvidia slid close to 3%; the standout exception was Palantir, which jumped almost 7% after Q4 revenue of $1.41 billion and adjusted EPS of 25 cents came in ahead of estimates
What Set It Off: Claude Code and Anthropic’s Legal Tools
The story driving the selling is that companies can now assemble software themselves with AI. CNBC’s reporting highlighted Anthropic’s Claude Code and tools like it, which make it possible for enterprises to build in-house the functionality they once bought from SaaS vendors — putting subscription revenue, pricing power, and margins directly in the line of fire. On his show that day, Jim Cramer added a sharper and more specific spark: the new legal tools Anthropic released for its Cowork product. Bill Northey of U.S. Bank distilled the anxiety into a single phrase: the “potential disintermediation that can occur from artificial intelligence.” One nuance worth holding onto: the fear concerns competition that has not happened yet, not today’s numbers. Most of the stocks touching 52-week lows had not reported bad quarters — the variable that moved is the story about who will be building enterprise software three years out.
Valuation Multiples Are Shrinking; Earnings Are Not
Cramer’s reading of the tape deserves to be kept around: what is compressing is the multiple, not the earnings. So far this year ServiceNow has lost about 28%, Salesforce close to 26%, and Intuit north of 34%, yet the earnings estimates attached to these names have not come apart — investors are just no longer willing to pay the same price for those profits. He put the risk plainly: once the multiple is in a sustained squeeze, “you don’t know how low it can go.” His response is to screen hard for individual names; he bought CrowdStrike for his Investing Club portfolio on the argument that cybersecurity is tougher for AI tools to copy than commodity SaaS, and he flagged Procter & Gamble, FedEx, and Union Pacific — software buyers rather than software sellers — as Tuesday’s winners. Put differently, the multiple is falling while the cash flow is not — and that absence of any earnings cut is exactly why this selloff can stretch on: nothing is forcing a bottom.
The Spillover: Private Credit and Rotating Money
The anxiety did not stay inside tech. Private credit lenders carrying software exposure sold off hard the same day: Blue Owl, TPG, Ares, and KKR all posted double-digit declines, Apollo dropped 7%, and BlackRock 5%. Ritholtz’s Josh Brown said the crowded trades had been “absolutely nuked.” Two sideline details round out the day: Citadel’s Ken Griffin cautioned that some companies might use AI as cover while they correct earlier over-hiring, and Bitcoin broke below $74,000, its weakest since November 2024 — proof that the flight from risk ran well past software.
The Read for Builders and Product Teams
Three things to take from it. One, the market’s new baseline is AI agents doing away with per-seat SaaS licenses: should agentic tools of the Claude Code and Cowork variety genuinely let enterprises handle customization in-house, seat-based and usage-based pricing gets repriced before anything else. Two, the gap between the casualties and the one standout carries real information — Palantir climbed on results powered by AI demand, which tells you the market’s objection is not to AI itself but to legacy models with no AI narrative attached. Three, read it as input for your roadmap: the moment investors start assuming companies will assemble their own tools through agents, every SaaS product owes an answer to why an agent cannot simply write it out of existence. Framed as pricing, this selloff promotes software’s pricing power to a fresh risk variable — as the marginal cost of custom software heads toward zero, holding a premium becomes difficult. For the bigger picture of AI moving from a model race to agent deployment at the start of 2026, see our opening outlook.