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

SignalFire says engineering was the most resilient job function in 2025, with engineering hiring down just 11% versus 2019 compared with a 25% drop in total hiring across large tech companies. Engineers made up 55% of new hires across 12 tech majors in 2025, up from 46% in 2019, and early-stage startups hired 7% more engineers than in 2019. The piece argues AI is boosting engineering productivity rather than replacing software engineers, consistent with commentary from Anthropic and Nvidia executives.

Analysis

The key market implication is not that AI is failing to displace labor, but that the first-order beneficiaries are the firms selling the picks and shovels, while the second-order effect is a re-acceleration of software demand. If engineers are getting materially more productive, management will not run a static headcount plan; it will push for more features, more internal tooling, and faster product cycles, which supports sustained spend on compute, developer tools, and cloud infrastructure. That is structurally bullish for NVDA and, by extension, for the large-cap platforms that monetize higher AI usage through inference, cloud, and ads rather than through labor savings.

The more interesting trade is that this narrative is negative for the “AI destroys software” consensus short. If the market had been pricing a sharp deceleration in engineering payrolls and enterprise software demand, the data argues that the timing is off by at least 12-24 months. In the near term, there is even a margin-paradox risk for software companies: they may hire fewer low-cost junior developers but retain or add expensive senior engineers and AI infra spend, compressing operating leverage before true labor substitution appears. That creates a better setup for capex beneficiaries than for labor-disruption beneficiaries.

The biggest contrarian read is that the labor market signal could be masking a later-stage break: companies are using AI to avoid backfilling open roles and to extend productivity, which delays the employment hit rather than eliminates it. If that is right, the downturn would show up first in entry-level hiring, contractor demand, and startup formation, not headline engineering employment. Over the next 6-18 months, watch for a plateau in cloud consumption growth or a slowdown in developer-tool budgets; that would be the first evidence that productivity gains are finally translating into less incremental demand.

Key Decisions for Investors

• Long NVDA vs short a basket of software names most exposed to the 'AI kills coding jobs' narrative for 3-6 months; risk/reward favors NVDA because productivity gains should lift inference and training demand before they pressure labor costs.
• Add to GOOGL, MSFT, and AMZN on any post-earnings weakness over the next 1-2 quarters; the setup is for higher AI utilization to translate into cloud spend and platform monetization, with downside limited if labor displacement remains delayed.
• Initiate a tactical long in META into the next 4-8 weeks: if engineering remains resilient, Meta can keep compounding product velocity and ad-targeting improvements without a near-term labor headwind, supporting both growth and margins.
• Avoid shorting AAPL/TSLA purely on AI-job-loss fears; the article argues the market is early, so direct labor-disruption shorts have poor timing. Better expression is via call spreads on NVDA or cloud leaders rather than outright shorts.
• For venture/private-market exposure, favor late-stage infra and developer-tool names over seed-stage automation plays; if startup engineering hiring is still positive, the next 12 months should reward tools that increase output rather than eliminate headcount.