NVIDIA announced that semiconductor companies are adopting its CUDA-X software libraries to support AI-defined fabs, which are chip factories where AI helps engineers optimize processes and solve production issues. As demand for advanced chips grows, these fabs must expand production while managing increasingly complex manufacturing steps. To support this, NVIDIA introduced cuPotential, an early-access library that aims to help developers build AI agents that simulate how atoms and molecules interact. Other tools, including cuEST, cuDSS, cuLitho, NanoVDB, NPP and the PhysicsNeMo framework, aim to support simulation, image analysis and physics AI. Engineers can also test process changes in digital twins before applying them in production.
- Synopsys, Cadence, COMSOL and Voltai use these tools and AI agents for chip design, packaging and multiphysics simulation. ChipAgents builds chip design AI models and is working with NVIDIA on Build-A-Chip Bench, a benchmark for AI agents.
- Siemens uses cuLitho for chip-pattern refinement, and Lam Research plans to integrate NanoVDB for wafer-etch modeling. Luminary applies PhysicsNeMo to package design, and Applied Materials uses Omniverse and CUDA-X for materials discovery, process development and fab digital twins.
- imec, KLA and Teradyne apply NVIDIA tools to chip inspection and image-sensor testing. Advantest uses AI analytics and NeMo Switchyard agents for test workflows. OMRON uses Cosmos and Metropolis to help investigate circuit board defects.
- Analog Devices, FANUC and Foxconn use Isaac Sim and Omniverse for sensing and robotic simulation.
cuPotential is in early access, NanoVDB is open source, and NPP comes with the CUDA Toolkit. Lam Research’s NanoVDB capability is targeted for release by the end of the year.
Image generated by: ChatGPT
