Why Tech Layoffs Keep Happening Even as AI Investment Booms

Scroll through tech news on any given week this year and you’ll see two headlines that shouldn’t sit next to each other. One says a company just cut thousands of jobs. The other says the same company just committed billions to AI infrastructure. Read them back to back and it stops making sense, until you realize they’re not contradicting each other. They’re describing the same decision.

As of June this year, tech companies have cut more than 185,000 jobs across roughly 267 separate layoff events, and over half of those layoffs explicitly named AI or automation as a factor. Meanwhile Amazon, Microsoft, Alphabet, and Meta have collectively pledged something close to 700 billion dollars in capital spending this year, nearly double what they spent in 2025, almost all of it going toward AI compute, chips, and data centers. Companies are firing people and buying GPUs with the money they save. That’s not a metaphor, it’s roughly what’s happening on the balance sheet.

The money is moving from payroll to compute

Meta’s internal communications around its May layoffs described the cuts as a way to offset the cost of its AI investments. That’s about as direct as a company gets about the connection. Atlassian’s CEO, Mike Cannon-Brookes, said something similar when the company cut 1,600 jobs in March, framing it plainly: AI changes the mix of skills and the number of roles a company needs in certain areas. Coinbase’s CEO pointed to engineers using AI to ship in days what used to take a team weeks, and used that as the basis for cutting 700 positions.

This is the part that’s easy to miss if you only read the layoff headline. The job cuts aren’t happening despite the AI spending, they’re functioning as a funding source for it. Budgets that used to go toward headcount are being redirected toward infrastructure. PayPal’s CEO described the company’s plan to cut roughly a fifth of its workforce as resting on two things at once: removing layers from the organization, and accelerating AI adoption. Those aren’t two separate initiatives, they’re the same initiative described from two angles.

It’s not just struggling companies doing this

If layoffs were purely about companies in trouble, the pattern would be easier to explain away. But some of the biggest cuts this year have come from companies reporting strong or record financial performance. Oracle disclosed it had reduced its workforce by 21,000 people over the past year, citing AI adoption directly in a regulatory filing. ASML, despite a strong year, is planning to cut around 1,700 management roles. These aren’t distressed companies trying to survive, they’re profitable companies reallocating where their money goes.

That reallocation is the actual story. The tech industry isn’t shrinking, it’s restructuring around a different mix of what it pays for. Less spent on certain categories of human labor, more spent on the infrastructure and the smaller number of specialized people needed to build and run AI systems.

Some of this is AI getting blamed for things AI didn’t do

It would be too simple to say every layoff this year is purely about AI, and some of the most credible voices in the industry have pushed back on that framing. OpenAI’s own CEO has acknowledged that some companies are blaming AI for cuts they would have made anyway. Venture capitalist Marc Andreessen has pointed to pandemic-era overhiring and higher interest rates as the real driver in plenty of cases, calling AI “the silver bullet excuse.” Deutsche Bank analysts used the phrase “AI redundancy washing” to describe this pattern, and outplacement firm Challenger, Gray & Christmas has noted that AI actually ranks behind market conditions, restructuring, closures, and general cost-cutting as a stated reason for job cuts.

So there are really two things happening under one umbrella. Some layoffs are genuinely driven by AI tools doing work that used to require a team. Others are ordinary cost-cutting that gets labeled as AI-related because it’s a more palatable explanation than admitting a company simply overhired or is responding to investor pressure. Both versions are true at the same time, in different companies, sometimes in the same announcement.

Who’s actually losing roles, and who isn’t

The data on this is more specific than the headlines suggest. Stanford’s AI Index found that employment for software developers between 22 and 25 fell nearly 20 percent since 2024, concentrated specifically in boilerplate coding, scripted testing, and routine bug fixes, the kind of tasks AI tools now handle reasonably well. Developers over 30 at the same companies saw their headcount grow rather than shrink. The pattern isn’t junior versus senior exactly, it’s routine versus judgment-heavy. Roles in machine learning infrastructure, model evaluation, and AI safety remain in high demand and are reportedly hard to fill, even as roles built around repetitive execution get absorbed into AI tooling.

This is also why the same companies cutting jobs are hiring at the same time. Meta is cutting close to 8,000 roles while actively hiring for specialized AI positions. It looks contradictory until you stop thinking of it as one workforce and start thinking of it as two: one shrinking, one growing, inside the same company.

What this actually means if you’re trying to make sense of it

The honest takeaway isn’t that AI is coming for every job, and it isn’t that AI has nothing to do with any of this either. It’s that the money funding the AI buildout has to come from somewhere, and payroll is the most direct lever companies have. Some of what gets cut is genuinely being replaced by AI capability. Some of it is decisions that were coming anyway, dressed up in the language of AI because that’s the story investors want to hear right now.

If there’s a practical thread to pull on, it’s the one Stanford’s data points to: the roles holding up best aren’t necessarily the most senior or the most technical in the traditional sense, they’re the ones built around judgment, oversight, and the kind of work that’s hard to hand off entirely to a model. The roles getting absorbed fastest are the ones that were always closer to repetition than decision-making. That’s a more useful way to read this moment than treating every layoff announcement as proof of the same story.