The layoff memos blame AI. The hiring data says otherwise.

Software engineering was supposed to be first in line for the chopping block. It writes code, AI writes code, the math seemed obvious. But the numbers don't back the story everyone keeps telling.

SignalFire's latest talent report, which tracked careers across more than 80 million companies, found engineering was the steadiest job function of 2025. Total hiring at big tech sat 25% below 2019 levels. Engineering roles were down only 11%. At the twelve companies SignalFire calls Tech Majors, engineers made up 55% of all new hires last year, up from 46% in 2019. Early-stage startups actually brought on 7% more engineers than they did before the pandemic.

The reason given and the role that grows

Here's the gap worth chewing on. The reason cited for layoffs is almost always AI, and specifically AI writing code so one engineer can do the work of five. "What we're seeing on the ground is a little inconsistent with that," said SignalFire's Asher Bantock. If AI really substituted for engineers, engineering hiring would fall first. It's growing faster than almost anything else.

The better frame is Jevons paradox. Make a resource cheaper to use and people use more of it, not less. Nvidia's Jensen Huang put it plainly: now that his engineers all run agentic AI, they're "busier than ever," because agents write code fast and then push humans to come up with the next idea. The work expands to fill the new capacity. That's a different thing than replacement, and leaders keep collapsing the two.

Where the cuts are actually landing

Follow the layoffs and they don't cluster around code. Seattle customer-data startup Amperity cut staff this month, framing it as a shift in "where we're investing and the shape of the team we need." Its whole business was unifying scattered customer records into one profile, exactly the kind of data plumbing AI is now good at. The cut followed a co-founder takeover and a pitch that AI is a "major opportunity."

Games tell a harsher version. Sony cut a significant number of jobs at Bungie after ending Destiny 2 development, the third round since the $3.6 billion acquisition in 2022. That one isn't about AI at all. It's a shooter that fell short and a studio shrinking to fit. Two industries, two very different forces, both reported under the same headline word.

When AI is the policy, not the explanation

The word "AI" is doing two jobs in these stories, and they're worth separating. Sometimes it's the actual cause. Sometimes it's the cover for a correction, the way Amazon's Andy Jassy blamed last year's cuts on a culture fix rather than the technology. Workers can feel the difference even when the press release blurs it.

The Magic: The Gathering Arena team unionized partly over Hasbro mandating required AI usage, listed right alongside layoff protections and pay. That's the tell. AI here isn't a market force replacing the team, it's a management policy imposed on people who'd rather have a say in it. When a tool becomes a mandate, the fight stops being about whether AI works and starts being about who decides.

Measuring the spend, not the headcount

The sharper question for a leader isn't whether to cut, it's who's actually getting value from the tools. Parker Conrad's pitch for Rippling Data Cloud leans straight into that anxiety. He showed a dashboard cross-referencing Anthropic usage logs, GitHub pull requests, and performance ratings to flag engineers with high AI spend and high peer rejection on code reviews. "If your peers are telling you to go back and do this over all the time, maybe you're just generating a lot of slop."

One employee was burning $30,000 a year asking Claude to read their calendar. The analysis already pushed Rippling to cut some spending limits. I'd hold this loosely. The same instinct that turns AI spend into a leaderboard is how a useful tool becomes a mandate becomes a union grievance. California now runs an AI-unemployment tracker precisely because nobody trusts the layoff memos to say what's really happening. The honest read so far: the panic is running ahead of the evidence, and engineers are busier than ever.

The retraining bet underneath all of it

Running under the whole thing is a hedge worth noting. Amazon, Microsoft, Anthropic and the OpenAI Foundation just anchored RAISE US, a nonprofit with over $1 billion committed to retrain workers displaced by AI. Gina Raimondo's line: build the best AI in the world and leave millions behind, and "we'll have automated our own decline."

The companies building the tools that reshape the work are also funding the safety net for the people it reshapes. Microsoft's already cross-training entry-level lawyers into AI-adjacent roles. You can read that as conscience or as insurance against the backlash. Either way it concedes the point the hiring data keeps making: the jobs aren't vanishing so much as moving, and the question is who gets carried to the next one.