The Companies Cutting Jobs for AI Aren't the Ones Using It

When a company announces layoffs and cites artificial intelligence, Wharton professor Peter Cappelli has learned to check the verb tense.

The headline is, ‘It’s because of AI,’ but if you read what they actually say, they say, ‘We expect that AI will cover this work.’ — Peter Cappelli

The headline says AI. The fine print says "we expect AI will cover this work" — future tense, not present. The distinction is not academic. It is the mechanism by which the largest sustained wave of white-collar job cuts in a generation is being justified, and it reveals a gap between what companies say and what the data shows that runs through the entire enterprise. The scale of the cuts is not in dispute. Oracle eliminated more than 25,000 positions while explicitly linking the reductions to an AI infrastructure push ^[1]. Amazon CEO Andy Jassy told shareholders that rolling out generative AI agents would reduce the company's total workforce ^[2]. Meta planned to cut up to 20% of its staff to fund a $600 billion investment in AI researchers and data centers ^[3]. Block, the payments company run by Jack Dorsey, cut 40% of its workforce, saying it needed to refocus on AI ^[4]. These are not distressed firms. Meta posted $26.8 billion in net income while eliminating 8,000 roles ^[5]. The same hyperscalers committing more than $700 billion to AI infrastructure this year — Amazon, Microsoft, Alphabet, Meta, Oracle — are the ones leading the headcount reductions, with Amazon alone eliminating approximately 16,000 corporate positions ^[6]. The financial logic, where it is stated explicitly, is a trade: redirect capital from wages to hardware, and the hardware will eventually do the work. Jassy made the link operational. Meta made it budgetary — 20% of staff for $600 billion. To fund the buildup, Meta, Amazon, and Nvidia each raised $25 billion in bonds this year, pushing Big Tech's AI capital expenditures past $234 billion by mid-2026 ^[7]. The problem is that the hardware is not cheaper than the humans. Nvidia vice president Bryan Catanzaro put it plainly about his own team.

For my team, the cost of compute is far beyond the costs of the employees. — Bryan Catanzaro

Uber exhausted its entire 2026 AI coding-tool budget by April, and Microsoft canceled thousands of internal Claude Code licenses because compute costs exceeded expectations ^[8]. The efficiency rationale is circular: companies are cutting labor to fund tools that currently cost more than the labor they replace. Meanwhile, a different set of firms tells a different story. The European Central Bank studied AI-intensive companies across the euro zone and found they tend, on average, to hire rather than fire ^[9]. A study of 21,559 U.S. firms by Ramp and Revelio Labs found that high-intensity AI adopters grew headcount by approximately 10% over two years, with entry-level positions increasing 12% ^[10]. The companies actually deploying AI at scale are growing, not shrinking, their workforces. The two populations barely overlap. A May 2026 Gartner study found that workforce reductions among organizations deploying AI did not correlate with improved financial returns ^[5]. Oxford Economics determined that AI-related job losses accounted for only 4.5% of total U.S. job losses in the first eleven months of 2025, and concluded that firms are trying to dress up layoffs as a good news story rather than bad news, such as past over-hiring ^[11]. Princeton computer scientist Sayash Kapoor found that in 90% of cases where companies cite AI-driven layoffs, they do not have a functional AI application ready to fill the eliminated roles ^[12]. What remains is a capital reallocation dressed as a technology story. Profitable companies are trading wage obligations for depreciating hardware, financed by debt, on a logic the returns have not validated. Apollo chief economist Torsten Sløk drew the parallel explicitly.

But what if the payoff takes longer than consensus assumes? — Torsten Sløk

The bond markets have funded the trade. The math has not yet returned the favor.