Every few weeks, another company announces a layoff, and the memo says some version of the same thing: AI made this possible. It’s a simple story to believe because it gets repeated so often and because the technology is improving fast. Yet, the data does not align with the statements in the memos.
In 2025, US employers announced 1.17 million job cuts, the most since 2020, according to Challenger, Gray & Christmas. AI was named as the reason for almost 55,000 of them. That’s roughly one in twenty. The rest came from closures, restructuring, market conditions, and in 2025, cuts to federal agencies. These figures cover every sector, not just tech.
2026 complicates the picture, so here’s the update. Through September, AI has been cited in 120,136 announced cuts this year, about 21% of the total and still the leading reason year to date. In September alone, it fell to the fifth most-cited reason, around 9% of the month. Two caveats come with those numbers. They record the reasons employers give, not an independent finding about what ended the job. And an employer who wants a better story for its investors has every incentive to give this reason.
That last point is the thread running through everything below.
What hit tech in 2022: overhiring, rate hikes and a tax rule
Picture a founder in 2021. Interest rates are close to zero, investors are hunting for growth, and the pandemic boom looks like the new normal. You need five developers, so you hire twenty because a growing headcount is a signal that attracts the next round of funding. Two years later, your revenue still supports five salaries, and you’re paying twenty.
Then three things land at once. Rates started climbing in 2022, and the easy money disappeared. A change to Section 174 of the US tax code takes effect that year, which forces companies to spread the deduction for engineers’ salaries over five years instead of writing it off immediately, so employing developers suddenly costs more at tax time. And the growth you promised doesn’t arrive to cover the bill.
Google is the cleanest example I found. It ended 2022 with about 190,000 employees, cut 12,000 in January 2023, and finished 2025 at 190,820.
Same company, same headcount, three years and a layoff in between. Sundar Pichai’s own memo at the time said Google had hired for a different economic reality.
Those layoff waves landed in late 2022 and early 2023. The developer I’m following dates the first model he’d call a dependable coder to Claude 3.5 Sonnet, which shipped on June 20, 2024. You can argue about that cutoff, but it’s hard to credit AI with layoffs that landed about a year and a half before it.
AI washing: how a layoff gets a better story
Back to the founder. For two years you’ve told investors that growth is strong. Laying off fifteen people flips that story, because it says growth stalled and maybe it’s time for investors to leave. So the announcement needs a different frame, and “we’re restructuring around AI” is a frame the market applauds.
Block is the case study. In February, it said it was cutting over 4,000 of its roughly 10,000 employees, down to just under 6,000, and the stock jumped close to 20% in pre-market trading. Block had about 3,800 people in 2019. Responding to critics the next day, Jack Dorsey wrote on X, “Yes, we over-hired during COVID,” while adding that blaming overhiring alone “misses all the complexity.”
I’ll give him credit for saying it out loud, because plenty of CEOs wouldn’t. He argues that AI-assisted teams can be smaller and flatter, and he may be right. But that’s a claim about the future, and the headcount chart is about the past.
The Company Building the Machines Told You the Machines Will Take Your Job. Then It Published a Plan.
I want to be straightforward about what I think is happening right now. We are in the opening months of the largest transfer of productive capacity from human labor to machines in the history of civilization. That is not a prediction. It is a description of events that are already measurable, already documented, already produc…
Amazon ran a version of this at a larger scale. It cut about 30,000 corporate jobs, and Andy Jassy first talked up generative AI to shrink the workforce, then clarified that the cuts were “not really AI-driven, not right now at least.” One detail stuck with me. Since March 2025, New York has let employers tick a box for “technological innovation or automation” in legally required layoff notices. Wired found that none of the 160 companies filing notices, Amazon among them, ticked it. Memos get AI. Legal filings don’t.
