Siemens has expanded its strategic partnership with Nvidia to bring self-verifying agentic AI workflows to electronic design automation (EDA) — the software engineers use to design the world's most complex chips. The centerpiece is an upgraded Fuse EDA AI Agent that validates its own decisions against deterministic, physics-based EDA engines before they are handed to engineers.
The move, unveiled at the Design Automation Conference (DAC) in early August, is an answer to a central problem of agentic AI in engineering: trust. An AI that proposes a chip layout or verifies a design is only useful if its reasoning is grounded in the same physics that will ultimately decide whether the silicon works. Siemens says its agent checks work against 'ground truth' simulation engines rather than pattern-matching its way to an answer.
The expansion builds on a partnership that has already pushed chip verification to 'trillion-cycle' scale, with Siemens tools running on Nvidia hardware and software stacks. The companies position the work as a step toward autonomous design teams where AI agents and engineers collaborate — with the machine double-checking itself at every step.
For chipmakers wrestling with designs of staggering complexity, self-verifying AI agents could compress verification timelines from months to days. But the approach also raises the stakes of AI quality: an agent that confidently validates a broken design is worse than none at all — which is exactly why Siemens is making physics, not confidence, the referee.
