AI was supposed to kill engineering jobs, but new data suggests they’re the most resilient

Whether AI is already replacing jobs is the taxable of fierce debate.

Tech layoffs deed their highest azygous month total successful years successful May, and AI was the most-cited reason, according to outplacement steadfast Challenger, Gray & Christmas.

Software engineering, successful theory, is the nonrecreational tract astir susceptible to automation, fixed the accelerated adoption of AI-powered coding tools. However, researchers astatine task steadfast SignalFire accidental the hiring information tells a antithetic story.

“The rationale fixed for tons of layoffs is consistently AI, and specifically they’ll accidental AI with respect to code; they’ll accidental 1 technologist could bash the occupation of nevertheless galore engineers successful the past,” said Asher Bantock, SignalFire’s caput of research. “What we’re seeing connected the crushed is simply a small inconsistent with that.”

SignalFire’s analysis, which tracked the careers of millions of employees crossed much than 80 cardinal companies, suggests that engineering was the astir resilient occupation relation successful 2025. Instead of focusing connected layoffs, which are hard to way due to the fact that radical often hold updating their employment presumption aft occupation cuts, SignalFire examined hiring information arsenic a much close indicator of real-time workforce trends.

While full hiring crossed ample tech companies dropped 25% compared to 2019 levels, engineering roles saw a overmuch smaller diminution of conscionable 11%, according to SignalFire’s latest “State of Talent Report.”

In fact, engineers comprised 55% of each caller hires successful 2025 crossed the 12 companies SignalFire classifies arsenic “Tech Majors” — Alphabet, Meta, Apple, Amazon, Microsoft, Netflix, NVIDIA, Tesla, Uber, Airbnb, Block, and Stripe. This is simply a important leap from 2019, erstwhile engineers represented lone 46% of caller recruits, according to the report.

The continued request for engineers was adjacent much evident astatine early-stage startups, which collectively brought connected 7% much engineers successful 2025 than they did successful 2019, SignalFire’s information shows.

If AI were genuinely substituting for engineering talent, Bantock argued, engineering hiring would beryllium the archetypal to autumn amid the existent tech hiring contraction. Instead, SignalFire’s information shows that engineering headcount is increasing faster than astir different occupation functions successful tech.

While Anthropic CEO Dario Amodei warned past twelvemonth that AI could wipe retired half of each entry-level white-collar jobs and propulsion unemployment arsenic precocious arsenic 20% wrong 5 years, the company’s ain caput of economics, Peter McCrory, told TechCrunch successful March that helium had not yet seen immoderate important AI-driven effects connected the workforce.

Said McCrory astatine the time: “There’s astatine slightest nary larger worldly quality successful unemployment rates” betwixt workers who usage Claude for the “most cardinal task of their occupation successful automated ways” — similar method writers, information introduction clerks, and bundle engineers — and workers successful jobs little exposed to AI that necessitate “physical enactment and dexterity with the existent world.” 

Nvidia CEO Jensen Huang went further still, outright rejecting the mentation that AI volition regenerate engineers. “Somebody said that AI is going to destruct each of the bundle engineering jobs,” Huang said successful an interview astatine the Stanford Graduate School of Business successful April. He past argued the other is true. Now that each engineers astatine Nvidia are utilizing agentic AI, “software engineers are busier than ever,” he’d said.

Huang added that portion agents are penning codification adjacent instantaneously, they are perpetually pushing engineers to make “the adjacent idea.”

For present astatine least, it seems that equipped with AI, engineering has go a classical illustration of the Jevons paradox — the thought that greater ratio doesn’t trim request for a resource; it increases it, due to the fact that the enactment expands to capable the caller capacity. As Bantock said of engineering endowment successful this moment: “They’re abruptly a batch much productive, and there’s endless enactment for them to do.”