There’s a particular kind of irony that’s hard to script better than reality just did. This week, Amazon confirmed layoffs inside its own Artificial General Intelligence organization, the very division building the models meant to define the company’s future, at the same time it’s on track to spend roughly $200 billion on AI infrastructure this year, more than 50% higher than 2025. The company that is spending more than almost anyone alive on artificial intelligence just told a chunk of the people building that intelligence they’re no longer needed. If you’ve been telling yourself that AI job disruption is coming for someone else’s role, not the people actually working on AI itself, this is the moment to reconsider that assumption.
The layoffs aren’t isolated. They’re the latest chapter in a restructuring that has now eliminated roughly 30,000 corporate roles at Amazon since October 2025, nearly 10% of its corporate workforce, and the largest workforce reduction in the company’s three-decade history, larger even than the 27,000 cuts it made during the 2022-2023 downturn. Notably, warehouse and fulfillment staff, the roles most people assume are first in line for AI-driven automation, have been largely spared. It’s the office workers, the credentialed, white-collar, “safe” jobs, taking the hit.
Also Read: How the Internet Turned Sexual Assault Into Content
What makes the AGI cuts specifically worth sitting with is what they reveal about the company’s real strategy. Amazon isn’t stepping back from AI. Senior Vice President Peter DeSantis, who now oversees the unit, has openly acknowledged that Amazon’s models haven’t yet reached “frontier tier” performance for the most demanding workloads, meaning the company is still chasing rivals like OpenAI, Anthropic and Google. But rather than throw more people at that problem, Amazon appears to be doing something more calculated: externalizing the labor. Instead of large internal teams doing model customization and post-training work by hand, the company is pushing that work onto customers themselves, who now pay for compute on Bedrock and Trainium chips to do it. The layoff and Amazon’s aggressive AI product expansion are, as one industry analysis put it, the same strategy viewed from two different angles. The humans doing bespoke work get replaced by a self-serve platform, and the company collects a fee either way.
CEO Andy Jassy has been unusually candid about where this is heading. He told employees back in June that AI adoption would reduce the company’s overall corporate headcount over time, framing it not as a risk but as a deliberate outcome of automating “routine tasks.” That’s a remarkable thing for a CEO to say plainly, out loud, to a workforce of hundreds of thousands of people. Most executives dress this kind of announcement in euphemism, “operating more nimbly,” “reducing bureaucracy,” “removing layers.” Jassy did use that language too, but he also said the quiet part: some jobs are going away because a machine can now do them, and that’s the point.
Also Read: The Rise of “Alpha Males” or Trash Males?
There’s a version of this story that reads as a straightforward business case: a company optimizing its cost structure, redirecting capital toward its biggest bets, business as usual in a competitive industry. That framing isn’t wrong, exactly, and Amazon’s AWS backlog of $364 billion and $20 billion-plus annual run rate from its Trainium chip business suggest the bets are paying off in dollar terms. But it’s worth being honest about who bears the cost of that optimization, and it isn’t shareholders. Affected employees get 90 days of pay, outplacement support, and severance, which is more generous than plenty of companies offer, but it’s still a layer of financial cushioning wrapped around the fact that a job that existed a year ago no longer does, and won’t be coming back in the same form.
The deeper unease here isn’t really about Amazon specifically. It’s about what this pattern signals for the tens of millions of people whose careers were built on the assumption that “knowledge work,” the kind requiring a degree, requiring specialized skill, requiring the ability to reason through ambiguous problems, would be the last thing automation came for. AGI researchers and model trainers were, until very recently, considered among the most protected roles in the entire tech industry, the people at the top of the value chain building the tools rather than being displaced by them. If that role isn’t safe from restructuring in the name of AI efficiency, it’s worth asking seriously what role actually is.
None of this means mass unemployment is imminent, or that AI is about to hollow out the labor market overnight. Amazon itself insists it continues hiring in “strategic areas,” and companies routinely reshuffle headcount for reasons that have nothing to do with any single technology. But the specific shape of these cuts, hitting the AI team itself while capital expenditure on AI infrastructure surges, is difficult to read as anything other than a preview. The technology isn’t just changing what jobs exist. It’s starting to change what “essential” even means inside the companies building it, and the people who thought they were standing closest to the machine are discovering they were standing in its path the whole time.
