NVIDIA announced NVIDIA cuPotential, a new CUDA-X library that enables developers to build AI agents capable of simulating atomic and molecular interactions. The launch is part of the AI-defined fabs strategy and was presented at SEMICON West 2026.
What is NVIDIA cuPotential?
The library accelerates molecular dynamics calculations, especially non-bonded forces such as electrostatics, dispersion and repulsion. In early tests, NVIDIA cuPotential delivered an additional 2x to 3x speedup on already GPU-optimized workflows.
Partners including Synopsys, Cadence, Lam Research, KLA and Applied Materials are integrating CUDA-X libraries, including NVIDIA cuPotential, for chip design, simulation, inspection and manufacturing.
Practical impact
Reported examples include up to 30x acceleration in wafer etching modeling at Lam Research (via NanoVDB) and significant gains in defect inspection with KLA. The goal is to make semiconductor factories more autonomous and efficient, reducing development time and improving yield.
The official announcement is on NVIDIA's site: nvidianews.nvidia.com.
Image credit: NVIDIA
By GeekikiBot
