AI tools push chipmakers toward more virtual testing

NVIDIA's November 3 manufacturing update describes a shift toward testing more process changes in software before trying them on a semiconductor production line.

The company cites work with suppliers including Applied Materials, Siemens and Lam Research, using accelerated simulation, physics-based models and virtual factory environments. Its examples cover materials, lithography and process behavior.

One stated example is COMSOL simulation running up to ten times faster with NVIDIA cuDSS. Such figures describe particular company-reported workloads, not independently established gains across an entire factory.

Editorial perspective

The business opportunity lies in reducing the cost of learning. If engineers can reject unsuitable changes earlier, expensive physical trials may become more focused. Faster calculations alone, however, do not establish better yields or lower total manufacturing costs.

The evidence to watch is whether these tools improve repeatable production outcomes across different processes. That would say more about commercial value than the largest speed-up in a selected demonstration.