Amazon has cut more than 57,000 corporate jobs since 2022 while preparing to spend roughly $200 billion on AI infrastructure this year. That's not a contradiction. It's the business model shifting in public.
You don't need a recession to lose your job at Amazon anymore. Sixteen thousand corporate roles went in late January. Fourteen thousand went three months earlier. Then came customer service cuts in April and third-party seller support roles in May. According to CNBC, Amazon's running total since 2022 now tops 57,000 corporate positions, roughly 16% of that workforce.
This is not a company shrinking. It's choosing.
Andy Jassy isn't hiding the logic. In his shareholder letter, he said Amazon wants to operate like "the world's largest startup," which sounds harmless until you see what it means inside a company with more than a million employees: fewer layers, flatter management, and an internal bureaucracy mailbox that CNBC reported had drawn more than 1,500 complaints and led to about 455 process changes by last September. Jassy has also told employees that AI efficiency gains will mean a smaller corporate workforce over time.
That's the trade. The jobs go, and the capital goes somewhere else.
Amazon Is Buying Compute Instead Of Headcount
Jassy wrote in April that Amazon expects to spend about $200 billion in capital expenditures in 2026, much of it tied to AWS data centers, custom chips, and AI infrastructure. He did not frame it as a bet made in the dark. "We're not investing approximately $200 billion in capex in 2026 on a hunch," he wrote, pointing to customer commitments that should start turning into revenue in 2027 and 2028.
The backlog explains why Amazon can cut people and still look aggressive. MarketWatch reported after Amazon's April earnings call that AWS had a $364 billion backlog, excluding a new Anthropic agreement. Jassy also told investors Amazon had more than $225 billion in revenue commitments for Trainium, its in-house AI chip. AWS revenue rose 28% in the first quarter to $37.6 billion, its fastest growth in 15 quarters, according to Barron's.
Put those figures beside the layoff count and the picture isn't subtle. Headcount is the line item Amazon is cutting. Compute is the line item it's expanding, by orders of magnitude.
The Human Cost Is Not A Footnote
The dollar figures don't capture what CNBC found on July 11, when it spoke with the people living through the cuts. Survivors described burnout from absorbing work once handled by colleagues who are gone. Laid-off employees described a tech job market so crowded that searches now stretch 12 to 18 months, up from roughly four months before this wave began. One former Amazon employee told CNBC that a posting could draw 200 to 300 applicants within hours.
That's a different kind of layoff from the ones that defined 2001 or 2009. Those were driven by collapsing demand. Amazon's current cuts are happening while demand for AWS capacity, AI chips, and cloud contracts is rising. Amazon has said AI wasn't the direct cause of most of these specific reductions, framing them instead as an effort to remove bureaucracy and simplify teams. Fine. But you don't need to call every cut an AI layoff to see the shape of the decision.
Look at where the money is going.
Amazon isn't alone in this pattern. Business Insider, citing Challenger, Gray and Christmas, reported in June that employers had attributed 87,714 job cuts this year to AI, already more than the full-year total for 2025. In May alone, AI was cited in 40% of announced U.S. job cuts tracked by the firm. That doesn't prove every eliminated job was replaced by a model. It does show that executives have found a new language for old cost discipline, and investors are rewarding the companies that can pair cuts with a big AI spending story.
For anyone watching this AI capex season, Amazon is the cleanest example yet. The company isn't retreating. Its AWS backlog sits at $364 billion. Its Trainium commitments top $225 billion. Its corporate layoff count has passed 57,000. That's the scoreboard. It's reallocating from people toward machines it still has to build. It's betting the reallocation shows up first in operating income.
If that bet works, expect the pattern to spread across every company with the balance sheet to try it. If it doesn't, tens of thousands of Amazon workers will have paid for a trade that never closed.
Also read: Oratomic raises 300 million dollars just 100 days after leaving stealth • Anthropic's New Lens Shows What Claude Is Thinking Before It Speaks • Meta Is Quietly Becoming One of the World's Biggest Chipmakers
