Our Take
The AI-will-kill-coding narrative has collided with hiring data; engineering is the job function most resistant to automation layoffs, not least because deployed AI makes engineers more productive, not redundant.
Why it matters
Policymakers, investors, and engineers themselves have spent two years bracing for mass displacement in software roles. This data suggests the displacement thesis was premature, and the productivity case for agentic AI is real enough that it's driving *more* hiring, not less.
Do this week
Engineering leaders: audit your internal productivity metrics against industry benchmarks (ask your VC or board for SignalFire's full State of Talent Report data) before committing headcount reductions based on AI adoption alone.
Engineering hires outpaced the tech contraction in 2025
SignalFire, a venture research firm, analyzed career data across more than 80 million companies to track real hiring trends in 2025. The finding contradicts the dominant narrative in tech layoffs: engineers are not being cut at higher rates than other job functions. In fact, they're being cut at lower rates.
At the 12 largest tech companies (Alphabet, Meta, Apple, Amazon, Microsoft, Netflix, Nvidia, Tesla, Uber, Airbnb, Block, Stripe), total hiring dropped 25% compared to 2019 levels. Engineering roles, by contrast, fell only 11%. Engineers now comprise 55% of all new hires at those companies, up from 46% in 2019 (per SignalFire data). Early-stage startups hired 7% more engineers in 2025 than in 2019.
Asher Bantock, SignalFire's head of research, noted the contradiction: companies cite AI-driven automation as the reason for cutting headcount, yet engineering is the job function that should be most vulnerable to code-generation tools. "If AI were truly substituting for engineering talent, engineering hiring would be the first to fall," Bantock said.
The Jevons paradox is operating faster than expected
The data aligns with recent statements from executives at companies deploying agentic AI. Nvidia CEO Jensen Huang rejected the idea that AI replaces engineers in April 2025, arguing instead that "software engineers are busier than ever" because agents write code instantaneously, forcing engineers to generate "the next idea." Anthropic's head of economics, Peter McCrory, told TechCrunch in March that he had not yet observed material AI-driven workforce displacement, even among roles most exposed to Claude (technical writers, data entry clerks, software engineers).
This pattern describes the Jevons paradox: efficiency gains do not reduce demand for a resource; they expand the work to fill the new capacity. In the current cycle, AI coding tools make engineers more productive, creating more work downstream rather than eliminating the need for engineers themselves.
The hiring data does not erase Dario Amodei's 2024 warning that AI could eventually displace half of entry-level white-collar jobs. But it suggests the displacement timeline is longer than the crisis rhetoric implied, and that the intermediate effect is reinvestment, not contraction.
Why hiring patterns matter more than layoff counts
SignalFire's focus on hiring rather than layoffs is methodologically important. Layoff tracking is notoriously noisy because employees often delay updating their LinkedIn profiles after job cuts. Hiring data, by contrast, is contemporaneous and reflects real-time staffing decisions at scale.
The resilience of engineering hiring cuts through two competing narratives. First, it undermines the claim that AI is already substituting for human engineers in production. Second, it suggests that companies adopting coding agents are not responding by shrinking engineering teams; they are responding by re-deploying engineers toward higher-leverage work.
This does not mean engineering jobs are permanently safe, nor that all engineering roles experience the same demand pressure. It does mean that the near-term risk to engineering employment is lower than the layoff headlines suggested, and that productivity gains from AI tooling are generating work faster than they are eliminating it.