For two decades the tech industry sold a distinctive employment story: this is not a job, it is a belonging. The campus, the mission, the family language, the equity that made everyone an owner. Then, beginning in 2022, the industry conducted a live experiment in what the story was worth under stress, and everyone watched the results in real time on a tracking site.
The aggregate numbers, per the widely cited tracker Layoffs.fyi, run roughly so: about 165,000 tech workers laid off in 2022, a peak of 262,735 in 2023 (59 percent above 2022), and more than 150,000 across 549 companies in 2024 (Layoffs.fyi, via TechCrunch and Crunchbase). Three years, well over half a million seats. But the size is the least informative part. Recessions happen; workforces shrink. What revealed the contract was the pattern.
What the pattern showed
First, synchronization. The cuts clustered tightly in time across companies with very different balance sheets, which is difficult to explain by firm-level economics alone and easy to explain by cover: when everyone cuts at once, no single cut is news, and the reputational price of breaking the belonging story approaches zero. The industry that prides itself on contrarian thinking executed the most conformist personnel event in its history.
Second, the method. The badge that stops working at dawn, the calendar invite from a stranger, the entire teams discovering their status by locked laptop: procedures optimized for legal and security risk, applied to people the same companies had called family. The method mattered because it disclosed the true model of the employee held by the employer at the moment of separation: a risk vector to be isolated, not a member to be dealt with honestly. Employees updated accordingly, and the update, unlike the severance, is permanent.
Third, the composition shifted as the wave matured. What began as pandemic-overhang correction increasingly targeted structure itself, middle management above all. Deloitte's human capital research recorded U.S. employers posting 42 percent fewer middle-management openings at the end of 2024 than in spring 2022, and cites Gartner's prediction that by end-2026, 20 percent of organizations will use AI to flatten their structures, eliminating more than half their current middle-management positions (Deloitte, 2025). Meanwhile Stanford's payroll-data research found AI's early employment effects concentrating on the youngest workers in exposed occupations, with 22-to-25-year-olds in those roles seeing a relative decline of roughly 16 percent while incumbents held steady (Brynjolfsson, Chandar and Chen, 2025). The wave stopped being weather and became architecture: a re-tiering of who is inside the walls at all.
The contract, as revealed
Put the observations together and the implied contract reads differently than the recruiting version. Belonging was real but conditional, and the conditions were never mutual: employees were asked for loyalty measurable in years and evenings, while the institution's reciprocal obligation dissolved on a spreadsheet timeline. The feudal comparison here is almost unfair to feudalism, whose bond, whatever its brutality, ran both directions: protection was owed downward. The 2022-2026 pattern showed protection was a market condition, not an obligation, honored while labor was scarce and repriced the quarter it was not.
It did not have to be designed this way, and the record contains at least one documented alternative. As Ricardo Semler has told it, when Brazil's 1990 crisis hit Semco, the company put the numbers on the wall and let the workers decide, and they chose a general salary cut in exchange for control over spending and a larger profit share in the recovery, which came. The elements that made it possible (open books, practiced participation, a belief that adults can handle real trade-offs) were built years before the crisis. A company cannot improvise that in a bad quarter, which is precisely the point: layoff style is the visible output of governance choices made long in advance.
What survives the disclosure
The revelation has a curious upside: contracts read honestly can be negotiated honestly. A workforce that no longer believes in unconditional belonging can ask for terms instead: severance defined in advance rather than at the employer's mercy, equity treatment on termination written into offers, information flows that let employees see trouble coming, and the personal disciplines (liquidity, networks, maintained alternatives) that make exit a plan rather than a catastrophe. Employers, for their part, can stop renting a loyalty they do not intend to repay; the belonging story, half-believed, now costs more in betrayal than it buys in retention.
What 2022-2026 revealed was not that tech companies are cruel. Most executed their cuts with generous severance by historical standards. It revealed something more useful: where the decision sits, who is told what and when, and which tier of the town the drawbridge protects. That knowledge, unlike the jobs, is not coming back inside the walls. The workforce keeps it now, and every offer letter is read in its light.
Sources
- Layoffs.fyi via TechCrunch & Crunchbase News (2023-2025). Tech layoff tracking. ~165,000 tech workers laid off in 2022; 262,735 in 2023 (59% above 2022); 150,000+ across 549 companies in 2024. https://techcrunch.com/2024/12/31/a-comprehensive-archive-of-2024-tech-layoffs/
- Deloitte (2025). 2025 Global Human Capital Trends. U.S. employers posted 42% fewer middle-management openings at end-2024 vs spring 2022; cites Gartner's prediction that 20% of organizations will use AI to flatten structures by end-2026. https://www.deloitte.com/us/en/insights/topics/talent/human-capital-trends/2025.html
- Stanford Digital Economy Lab — Brynjolfsson, Chandar & Chen (2025). Canaries in the Coal Mine? Workers 22-25 in the most AI-exposed occupations saw ~16% relative employment decline while experienced workers held steady. https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/