On July 6, TechCrunch published its running tally of major tech layoffs in 2026 where employers cited artificial intelligence. The numbers are bracing: roughly 120,000 roles cut so far this year, according to Layoffs.fyi. Microsoft eliminated 4,800 positions — 2.1% of its global workforce — while posting record revenues. Oracle disclosed it had shed 21,000 employees over twelve months, a 13% reduction. In each case, management pointed to AI as both the engine of future growth and the reason for the cuts.

Three weeks later, the Stanford Institute for Economic Policy Research released a policy brief titled “What Is Really Happening to Jobs? Separating AI Hype from Reality.” Authored by economists Neale Mahoney, Erika McEntarfer, and Karsen Wahal, the brief quickly shot to the top of Hacker News, where it was received as a welcome dose of sobriety. The message, as the commentariat understood it: calm down. The aggregate data does not show AI triggering mass job destruction. The panic is premature.

Both things can be true. That is the problem.

The Brief vs. the Pink Slips

The SIEPR brief is almost certainly correct about the past. Aggregate employment statistics are lagging indicators by design, and they have not yet registered the kind of dislocation that would confirm the most apocalyptic predictions. The U.S. economy continues to add jobs on net. The unemployment rate remains low. If you are an economist looking at a chart of total nonfarm payrolls, the AI revolution looks like a lot of noise around a steady upward trend.

But the 120,000 people laid off this year do not live in the aggregate. They live in specific Slack channels that went quiet, specific all-hands where a CEO read from a script about “AI-enabled efficiencies,” specific Monday mornings when a calendar invite from HR appeared without warning. The brief’s distinction between hype and reality — between what the data shows and what people fear — is analytically useful. It is also, for a growing number of workers, beside the point.

A product manager at a firm that cut 8% of staff last quarter put it this way in a message to a former colleague: “We’re not replacing people with AI. We’re replacing people with the expectation of AI. The code doesn’t work yet, but the headcount reduction does.”

Anticipation Effects Are Real Effects

Economists have a term for this: anticipation effects. When market participants believe something will happen, they act on that belief before it materializes, and their actions can bring about the very outcome they anticipated. If every Fortune 500 CEO becomes convinced that AI will let them run the company with 30% fewer people, they do not need to wait for the AI to work. They can start cutting now and call it “restructuring ahead of the AI transition.” The layoffs show up in the data long before the productivity gains do.

This is not a hypothetical. The TechCrunch tracker is full of companies that reported strong earnings in the same quarter they announced AI-related layoffs. A separate June 2026 paper from the Ramp Economics Lab, cited in the SIEPR brief’s own references, examined firm-level AI spending and workforce adjustment and found patterns that complicate any simple aggregate story. The pattern is not “AI arrived and made workers redundant.” The pattern is “management decided AI will make workers redundant, so workers became redundant to the budget.” The distinction matters because it means the job losses are not being driven by technological capability. They are being driven by a narrative that has captured the C-suite, and narratives do not need to be true to have consequences.

The Narrative Is the Policy

Here is the uncomfortable implication the SIEPR brief does not address: when enough powerful people believe a story about the future, that story becomes a form of policy, even if no regulator writes it down. The Federal Reserve does not need to issue a directive for companies to reduce headcount in anticipation of AI. The consulting firms, the earnings calls, the business press, and the peer pressure of quarterly capitalism do the work on their own.

The brief’s authors are not wrong to want a clear-eyed look at the data. But the data they are looking at describes a world that no longer exists — the world before the narrative took hold. By the time the aggregate statistics catch up to what 120,000 laid-off workers already know, the “hype” will have become the reality, and the economists will publish another brief explaining how it happened.

The people getting the calls will not need to read it.

Sources