Nvidia’s newsroom said on 14 October 2026 that the company had introduced cuPotential, a CUDA-X library for simulating atomic and molecular interactions, as semiconductor companies adopt its software across chip design and manufacturing.
Developers can add cuPotential’s GPU-accelerated functions to existing software to calculate forces and other interactions among large numbers of atoms. Nvidia says AI agents could use those calculations to run and refine materials simulations. In early-access tests, the company reports that cuPotential made workflows already optimised for GPUs an additional 2-3x faster.
Inspection uses a different library, NPP, to process images on GPUs. Nvidia says a new NPP kernel informed by KLA’s production software ran 50x faster than earlier CPU versions and performed comparably to KLA’s hand-tuned implementations.
Checking chips for defects could benefit from faster image processing, since the inspection work Nvidia describes depends on analysing images quickly. The reported gain applies to the kernel tested against earlier CPU versions.
For developers building simulation software, Nvidia reports that early results from Synopsys’s QuantumATK show a potential 30x speedup with another library, cuEST, while S-Device benchmarks show a 12x speedup with cuDSS. Nvidia also says imec achieved a 30% improvement in the lithography process window using cuLitho.
