NVIDIA Expands CUDA-X GPU Libraries for AI Semiconductor Manufacturing

NVIDIA expanded its CUDA-X software portfolio at SEMICON West 2026 to accelerate semiconductor design, inspection, and fab automation, introducing the cuPotential library for modeling atomic and molecular interactions with an additional 2-3x speedup over existing GPU-optimized workflows. Chip design and equipment suppliers reported major simulation gains across the CUDA-X stack: Synopsys saw a potential 30x speedup in QuantumATK using cuEST and a 12x speedup in S-Device using cuDSS, while Cadence optimized advanced package stress analysis up to 8x faster. In wafer fabrication and process modeling, Lam Research expects integrating NanoVDB into SEMulator3D to accelerate selected wafer-etch workloads by up to 30x and demonstrated PhysicsNeMo reducing plasma-prediction simulations from days to milliseconds. For inspection and testing, Teradyne reported approximately 85x faster probe-maintenance inference using NPP, RAPIDS cuDF, and CuPy, KLA achieved up to 50x faster performance than earlier CPU implementations, and Applied Materials reported up to 35x faster chamber simulations. "The semiconductor industry made the AI revolution possible. Now, AI is transforming how chips are built." — Tim Costa, VP and General Manager of Industrial and Computational Engineering at NVIDIA.