NVIDIA CUDA-X Adoption Signals a Shift Toward AI-defined Fabs

NVIDIA announced that semiconductor companies are adopting NVIDIA CUDA-X libraries for AI-defined fabs, spanning simulation, process modeling, inspection, testing and fab optimization. The pitch is not abstract: use accelerated computing and AI models to evaluate changes before they reach production.

“AI-defined fabs use AI to help engineers optimize processes, investigate production issues and improve manufacturing efficiency.”

The engineering challenge is breadth. A fab has to connect atomic-scale material behavior, package and chip simulation, lithography, inspection data and equipment decisions without turning every question into a months-long analysis queue.

“AI-defined fabs require tools for modeling chip designs, materials behavior and manufacturing processes before changes are introduced into production.”

Simulation moves earlier in the flow

NVIDIA cuPotential supports atomic-scale simulations of electrical forces and interactions across large groups of atoms. Synopsys uses cuPotential, cuEST, and cuDSS for semiconductor simulations, achieving potential speedups of up to 30x with cuEST.

Cadence uses NVIDIA accelerated computing and CUDA-X libraries for AI-powered design and simulation, reporting an 8x faster optimization using cuDSS. COMSOL uses NVIDIA cuDSS to accelerate semiconductor multiphysics simulations, achieving up to 10x faster simulations. Voltai uses NVIDIA cuDSS and CUDA-X libraries in its Voltaire platform, reducing electro-thermal-mechanical simulation workflows from months to a day.

ChipAgents works with NVIDIA on Build-A-Chip Bench, an AI benchmark for evaluating agents across workflows.

Process modeling and package design get GPU leverage

Siemens uses cuLitho to accelerate Calibre software for chip-pattern refinement and mask preparation. Lam Research integrates NanoVDB into SEMulator3D, accelerating selected wafer-etch modeling tasks by up to 30x.

Luminary is using NVIDIA PhysicsNeMo for semiconductor package design. The company trains physics-based AI models to predict temperature, pressure drop and thermal stress across package configurations, and reports that engineers can evaluate 100x more designs within the same schedule.

Inspection and test data become operational inputs

Imec uses NVIDIA AI and accelerated computing to remove noise from scanning electron microscope images, reporting a 30% improvement in the lithography process window. KLA applied production-kernel techniques to an NPP NCC kernel, achieving performance up to 50x faster than earlier CPU implementations.

Teradyne uses NVIDIA NPP and RAPIDS cuDF for GPU calculations, reporting approximately 85x faster probe-maintenance inference. Advantest integrates NVIDIA Nemotron into ACS RTDI, using NVIDIA Grace Blackwell systems for real-time diagnostics and recommendations.

Applied Materials uses NVIDIA Omniverse and CUDA-X libraries for materials discovery, process development, and fab optimization, reporting chamber simulation speeds up to 35x faster.

The common thread is measurable compression of engineering loops: simulation, inspection and maintenance tasks move closer to the pace of production decisions.