$650 billion in AI spending. Software engineer job postings up 11%.
Citadel Securities analyst Frank Flight has been watching this disconnect. Indeed's data shows hiring demand holding even as AI investment soars. The St. Louis Fed's Real Time Population Survey tells a similar story. Daily AI use for work has stayed flat, not accelerating the way you'd expect if mass displacement were around the corner.
The problem is confusing what AI can do with how fast it gets deployed. As Flight puts it, "recursive capability does not imply recursive adoption." AI adoption is tracking a classic S-curve, much like PCs and the internet. Infrastructure bottlenecks in compute and power create natural economic boundaries. When demand for compute rises, so does its marginal cost. At some point it becomes cheaper to just hire people.
But there's a wrinkle.
Corvi Careers data shows 119,717 open tech positions, with the market heavily favoring mid-senior engineers over entry-level roles. Business software applications account for 28,213 openings and general software engineering adds 22,621, while specialized AI engineering sits at just 10,387. The on-ramp for junior developers appears to be shrinking. Tools like Copilot and Cursor handle the boilerplate code that used to teach newcomers. IBM says it plans to triple entry-level hiring to build AI workflows, but the broader data says companies want experienced engineers who can manage system architecture and stop codebases from rotting when AI-generated code piles up.
The macro picture holds. New business formation is climbing, prime working age employment sits near historic highs, and unemployment printed at 4.28%. Every major productivity shock, from steam power to computing, has been disinflationary and growth-enhancing in the medium term.
The real question is whether the workforce can retool fast enough for a market that suddenly wants fewer junior coders and more senior architects.