NVIDIA and chipmakers accelerate AI design and manufacturing

NVIDIA said Applied Materials, Cadence, KLA, Lam Research, Siemens, Synopsys and other semiconductor companies are using its CUDA-X libraries and related AI software across chip design, manufacturing and testing.

The new cuPotential library lets developers build AI agents that simulate atomic and molecular interactions. In early-access testing, it added 2–3x speed to workflows already optimized for GPUs.

AI-defined fabs use digital twins, materials modeling and inspection-data analysis to test process changes before applying them in production.

Simulation and design speedups

Synopsys is using cuPotential, cuEST and cuDSS for semiconductor simulations. Early QuantumATK results showed a potential 30x speedup with cuEST, while S-Device tests showed a 12x speedup with cuDSS.

Cadence is using CUDA-X, Nemotron models and accelerated computing for AI agents in semiconductor, packaging and printed-circuit-board design. COMSOL reported up to 10x faster multiphysics simulations.

Voltai’s Voltaire platform combines AI agents with GPU-accelerated solvers, reducing complex 3D-IC simulation workflows from months to a day. ChipAgents and NVIDIA are developing Build-A-Chip Bench to evaluate AI agents from specification through graphic data system.

Manufacturing models and quality control

Lam Research plans to integrate the open-source NanoVDB library into its SEMulator3D platform. Selected wafer-etch models are expected to run up to 30x faster. Siemens is using cuLitho for chip-pattern refinement and mask preparation.

NVIDIA PhysicsNeMo is being used to predict physical processes. Lam Research demonstrated plasma-prediction simulations reduced to milliseconds instead of days, while Luminary said it could explore 100 times more package designs within the same schedule.

For inspection, NVIDIA offers NPP and RAPIDS cuDF. KLA said a new NPP NCC kernel ran 50x faster than earlier CPU versions. Teradyne uses NPP, cuDF and CuPy to achieve approximately 85x faster probe-maintenance inference.

imec uses AI to remove noise from scanning-electron-microscope images and cuLitho for computational lithography, achieving a 30% improvement in the lithography process window. Advantest integrated Nemotron and Kumo-Anomaly into its ACS RTDI production-analytics system.