Process modeling gains: Lam Research reports up to 30x acceleration on wafer-etch modeling with NanoVDB and plasma-prediction times cut from days to milliseconds with NVIDIA PhysicsNeMo.
- Library adoption: Semiconductor companies including Synopsys, Cadence, COMSOL, and Voltai are adopting NVIDIA CUDA-X libraries, including the new cuPotential library, for AI-defined fabs.
- Simulation speedups: Reported gains include a potential 30x speedup with cuEST (Synopsys), 8x faster package optimization with cuDSS (Cadence), and up to 10x faster multiphysics simulations (COMSOL).
- Inspection and testing: KLA reported up to 50x faster performance on an NPP kernel, Teradyne reported approximately 85x faster probe-maintenance inference, and Imec reported a 30% improvement in the lithography process window.
- Fab automation: Applied Materials reports chamber simulation speeds up to 35x faster, while FANUC, Foxconn, OMRON, and Analog Devices apply Omniverse, Cosmos, and Isaac Sim tools to factory digital twins and inspection workflows.
NVIDIA has announced that semiconductor companies are adopting NVIDIA CUDA-X libraries — including the new NVIDIA cuPotential library — to support AI-defined fabs. The adoption spans simulation, inspection, testing, and fab automation, with early results from partners pointing to significant acceleration across engineering workflows.
What AI-defined fabs demand
"AI-defined fabs use AI to help engineers optimize processes, investigate production issues and improve manufacturing efficiency."
That requirement is driving interest in GPU-accelerated tooling. Modeling interactions between atoms and molecules is computationally expensive, even on GPU-optimized software — which is where cuPotential comes in.
"NVIDIA cuPotential provides GPU-accelerated functions that developers can add to existing software."
The library supports atomic-scale simulations, including electrical forces and other interactions across large groups of atoms. It is currently available through an early-access program.
Simulation results across the ecosystem
Synopsys is among the companies using cuPotential for materials simulations, alongside NVIDIA cuEST and NVIDIA cuDSS.
"Early QuantumATK results showed a potential 30x speedup with cuEST, while S-Device benchmarks showed a 12x speedup with cuDSS."
Synopsys' AgentEngineer solutions also interoperate with NVIDIA Nemotron open models and NVIDIA NeMo to support AI agents for chip design and verification.
Cadence is using NVIDIA accelerated computing, NVIDIA Nemotron open models, and CUDA-X libraries for AI-supported design and simulation of electromechanical systems.
"AuraStack AI Super Agent optimized advanced packages for heat and stress 8x faster using cuDSS with Cadence Celsius."
COMSOL is using NVIDIA cuDSS to accelerate semiconductor multiphysics simulations, reporting that the technology can provide simulations up to 10x faster. Voltai, meanwhile, uses NVIDIA cuDSS and CUDA-X libraries in Voltaire, its agent-based multiphysics platform, reducing electro-thermal-mechanical simulation workflows from months to a day.
Manufacturing workflows and inspection
On the fab side, NVIDIA cuLitho accelerates calculations for printing chip patterns, NanoVDB handles 3D data processing as an open-source library, and NVIDIA PhysicsNeMo provides an open framework for training physics-based AI models.
Siemens is using cuLitho to accelerate Calibre software for chip-pattern refinement and mask preparation. Lam Research plans to integrate NanoVDB into SEMulator3D.
"NanoVDB can accelerate selected wafer-etch modeling tasks by up to 30x."
Lam Research has also demonstrated NVIDIA PhysicsNeMo reducing plasma-prediction simulation times to milliseconds instead of days. Luminary is applying PhysicsNeMo to semiconductor package design, training models to predict temperature, pressure drop, and thermal stress — reporting that engineers can evaluate 100x more designs within the same schedule.
For inspection and testing, Imec reports a 30% improvement in the lithography process window using NVIDIA cuLitho, while also using NVIDIA AI to remove noise from scanning electron microscope images. KLA reported performance up to 50x faster than earlier CPU implementations on an NPP NCC kernel. Teradyne reports approximately 85x faster probe-maintenance inference using NVIDIA NPP, and Advantest has integrated NVIDIA Nemotron into ACS RTDI, with NVIDIA NeMo Switchyard AI agents reducing response times from hours to minutes.
Fab operations and automation
Applied Materials is using NVIDIA Omniverse and CUDA-X libraries for materials discovery, process development, and fab optimization, reporting chamber simulation speeds up to 35x faster. FANUC and Foxconn are connecting robot models to the NVIDIA Omniverse Blueprint, and OMRON is using NVIDIA Cosmos and the NVIDIA Metropolis Blueprint for video search and summarization in inspection equipment. Analog Devices is combining Synopsys physics simulation with NVIDIA Isaac Sim and Omniverse libraries for sensing simulations and robotic benchmarks.
NanoVDB is available as open-source software within OpenVDB, and NPP is available through the NVIDIA CUDA Toolkit. Lam Research's NanoVDB-accelerated SEMulator3D capability is scheduled for release by the end of this year.
Is this your company?
This article features your business. Claim it to add your logo, contact details, and a link to your website — or upgrade to reach more buyers.
Did you know 80% of Press Releases trigger AI content warnings? Reach out and the M4S team can assist.