AI system automates chip design verification, cutting a 70% time bottleneck

Researchers built an AI agent that automates functional verification—the process of checking whether a semiconductor chip design actually works as intended—by using large language models to generate test cases and verification logic in Python instead of hand-written code. This matters because verification currently consumes about 70% of chip development time, and this automation could compress that timeline significantly while catching design flaws that human testers miss.

If this system scales beyond the lab to real chip teams, it removes one of the largest time sinks in hardware development, which could accelerate the pace at which new chip designs move from concept to production—and directly reduce the engineering cost per chip.