NVIDIA has launched cuPotential, a new software library for simulating materials at the level of atoms, and says leading chip companies are adopting its CUDA-X libraries to bring AI into chip design and manufacturing. NVIDIA announced the news on 14 October 2026, alongside its presence at the SEMICON West trade show.
CUDA-X is NVIDIA’s collection of software libraries that run on its GPUs. NVIDIA describes “AI-defined fabs” as chip factories “where AI helps engineers optimize processes, solve production issues and improve production efficiency.”
What cuPotential does
According to NVIDIA, developers can use cuPotential to build AI agents that simulate how atoms and molecules interact, which matters for modelling chip materials and manufacturing processes. NVIDIA says that in early-access testing, the library “accelerated already GPU-optimized workflows by an additional 2-3x.” cuPotential is available in early access.
“The semiconductor industry made the AI revolution possible. Now, AI is transforming how chips are built,” said Tim Costa, vice president and general manager of industrial and computational engineering at NVIDIA.
What partners claim
NVIDIA named Applied Materials, Cadence, KLA, Lam Research, Siemens and Synopsys among the companies using its tools. The release includes several speed claims, none of which has been independently tested:
- Synopsys: early results show “a potential 30x speedup” in its QuantumATK software and a 12x speedup in S-Device benchmarks, according to NVIDIA.
- Lam Research: adding NVIDIA’s NanoVDB to its SEMulator3D modelling software “is expected to accelerate select wafer-etch modeling by up to 30x”. The feature is “targeted for release by the end of this year.”
- Applied Materials: “up to 35x faster chamber simulations”, according to NVIDIA.
- KLA: a new image-processing kernel that ran “50x faster than earlier CPU versions”.
What this means
- AI moves into the fab. NVIDIA wants its software inside how chips are made, not just in the chips it sells.
- Speedups are claims for now. The figures come from NVIDIA and its partners.
- Some tools are still early. cuPotential is in early access, and Lam’s feature is due by year end.
Sources
- NVIDIA Newsroom, “NVIDIA and Semiconductor Leaders Advance AI Across Chip Design and Manufacturing for AI-Defined Fabs”, 14 October 2026. https://nvidianews.nvidia.com/news/nvidia-and-semiconductor-leaders-advance-ai-across-chip-design-and-manufacturing-for-ai-defined-labs
Related stories
- NVIDIA Pledges $250 Million in Computing Power for UK Research on Safer AI Agents
- GlobalFoundries Selected to Operate Two New Chip Fabs at Malta, NY, Under a US–Japan Plan of Up to $21 Billion
- ASML Warns Its Planned Supply for 2027 and 2028 “Will Not Fully Meet” Customer Demand as AI Drives Record Sales
