Breaking08 / Industry
NVIDIA puts chip-making software speed-ups at up to 35 times
The figures are NVIDIA's own maximums, and neither the yield gains at working fabs nor the price of the software has been disclosed.

NVIDIA on 4 Nov 2026 published maximum speed-ups for chip-making process software from Siemens, Applied Materials and Lam Research, with the largest, 35 times, in an Applied Materials simulation platform. The company said the Lam Research figure is a projection for after a planned software integration.
Specifications. According to NVIDIA, Applied Materials' Ginestra simulation platform runs up to 10 times faster than CPU-only approaches after integrating NVIDIA's cuDSS library, while applying PhysicsNeMo to its ACE+ platform makes surface-topography simulations up to 35 times faster. Of the figures in the announcement, only Ginestra's 10 times names CPU-only processing as its baseline.
Computational lithography is the calculation that determines how circuit patterns are printed onto silicon. NVIDIA said these calculations are too intensive to complete on CPUs at production scale. Siemens is integrating NVIDIA's cuLitho into its Calibre software, with up to 17 times acceleration on key simulation workloads.
NVIDIA said Lam Research plans to integrate NanoVDB into SEMulator3D, its semiconductor process modelling platform, and that select wafer-etch modelling is expected to become up to 30 times faster as a result. For the plasma used in etching, PhysicsNeMo can help reduce prediction simulations from days to as little as milliseconds.
Advantest uses NeMo Switchyard AI agents to automate the choice of AI model on the test floor, cutting response times from hours to minutes, according to NVIDIA. Applied Materials builds digital twins of entire fabs, or chip factories, with Omniverse libraries to test material movement, bottlenecks and equipment layouts in a virtual environment.
NVIDIA calls a factory that combines these capabilities an AI-defined fab. In that model, the company said, engineers direct the work and make the decisions while spending less time waiting.
NVIDIA did not disclose how much yield improves at working fabs, what the software costs, or when Lam Research's integration will take place.