SignalFire's latest talent data cuts against the easy story about AI replacing engineers. The firms using AI hardest are still hiring technical people, while the real damage is showing up at the entry level.
You have heard the claim often enough by now: AI is coming for software engineers. Developers will be automated away. Startups will run with a tiny team, a stack of large language models, and no patience for headcount. SignalFire's 2026 State of Talent Report, which it says draws on hiring data across more than 80 million companies, points in a different direction.
The report's argument is blunt. Engineering roles have held up far better than tech hiring overall, and the companies pushing hardest into AI are not behaving as if engineers have become disposable. They're hiring them. If you're building a company and treating AI as a reason to starve the engineering team, you should look at the numbers before turning that into strategy.
Across what SignalFire calls the "Tech Majors", including Alphabet, Meta, Nvidia and Stripe, total tech hiring has fallen to 75% of its 2019 level. Engineering hiring is down too, but only by 11%. That gap is the story. It suggests AI is cutting some jobs around technical work faster than it is cutting the technical work itself.
Asher Bantock, SignalFire's head of research, frames the pattern through the Jevons Paradox, the old observation from William Stanley Jevons that more efficient steam engines did not reduce coal use in Britain. They increased it, because cheaper energy made more uses economical. Applied to software, the point is simple enough: make engineers more productive and you don't necessarily need fewer of them. You may find more things worth building.
That isn't a romantic view of engineering. It is what the hiring mix now shows. Engineers account for 55% of new hires at those major tech companies, up from 46% in 2019, according to SignalFire. The technical core is taking a larger share of the room. Designers, marketers and generalist business roles are under more pressure around the edges.
For founders, the early-stage data is more useful than the big-company comparison. You might expect startups to be the first to prove the "AI-first, headcount-last" theory, because young companies feel every payroll decision immediately. SignalFire found the opposite. Early-stage startups hired 7% more engineers in 2025 than they did in 2019.
They are running flatter, though. The average engineering span of control has risen by 34%, from roughly 11 engineers per manager to about 15. That is not a small operating change. It means fewer layers, more direct work, and less room for managers whose main job is managing other managers. But it is not a retreat from engineering headcount.
Here's the thing: AI is not making good engineers irrelevant. It is making weak organizational habits harder to hide. A founder who gives strong engineers better tools can move faster. A founder who assumes the tools can replace the people is betting that judgment, architecture, taste, debugging under pressure and product context are all cheap. They aren't.
There is one painful exception, and it should not be waved away. Entry-level hiring has been hit hard. SignalFire's findings show that new role starts for people with less than one year of post-graduate experience fell 50% between 2019 and 2024. That is where AI's pressure is clearest: on the tasks that used to help young developers learn their way into the profession.
Glue code, boilerplate, simple debugging and narrow tickets once gave new engineers a paid on-ramp. Tools can now absorb a lot of that work. The industry may not be firing experienced engineers in the way the loudest headlines promised, but it is becoming less willing to pay for training. Frankly, that is a real pipeline problem, because senior engineers do not appear fully formed after five years of reading documentation.
TechCrunch's coverage of the SignalFire findings, published today, makes the same correction worth making for startup operators: the most AI-forward firms in the dataset are not acting as if software teams are about to vanish. They are concentrating more of their hiring around engineers and asking those engineers to cover more ground.
That should change the conversation you have inside a startup. The useful question is not whether AI lets you avoid hiring developers. It is which engineers can now take on larger problems because the tools have removed some of the drudgery around them. There is a big difference between reducing waste and pretending the core craft has gone away.
The engineering job market is not healthy for everyone. New graduates are feeling the squeeze, and companies that stop training junior people will eventually pay for that decision. But the broader claim that AI has made engineers a bad bet does not survive SignalFire's data. The firms closest to the tools are still hiring the people who know how to use them.
Also read: Qualcomm bets its future on a 250-core data center chip and a $3.9 billion software acquisition • Micron's $41 billion quarter confirms the AI memory supercycle is nowhere near done • Anthropic's distillation problem reveals that export controls alone cannot hold the line in the US-China AI race
