AI Is Changing How Semiconductor Fabs Work — and Engineers Are Moving Faster Because of It
Reported by NVIDIA and The Tech Buzz2 sources ↓
NVIDIA describes collaborations and AI tools for speeding semiconductor simulations, predicting process behavior and testing fab changes in digital twins. The company reports partner demonstrations of faster calculations and quicker task handling.
Key points
- NVIDIA says COMSOL and Applied Materials simulations run up to 10x faster with its computing tools
- NVIDIA says Siemens’ Calibre workloads run up to 17x faster with cuLitho
- Lam Research demonstrated plasma predictions in milliseconds, compared with days for a full simulation
Sources
Details
NVIDIA describes a set of tools and collaborations intended to help semiconductor engineers evaluate process changes before physical trials. The company says accelerated computing can shorten simulations, physics-based AI can estimate process behavior without rerunning full simulations, and digital twins can test factory changes virtually.
NVIDIA reports that COMSOL’s multiphysics simulations and Applied Materials’ Ginestra platform run up to 10 times faster using cuDSS.
NVIDIA says Lam Research demonstrated PhysicsNeMo predictions of plasma behavior in milliseconds, compared with days for a full simulation. Applied Materials uses NVIDIA Omniverse libraries for fab digital twins, and Advantest uses NVIDIA NeMo Switchyard agents to select AI models for tasks; NVIDIA says this reduced response times from hours to minutes. These performance figures and examples are reported by NVIDIA; the article does not provide independent measurements.
Written from coverage as of
Images
- NVIDIA
- NVIDIA
- NVIDIA
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh