NVIDIA and Applied Materials speed up chip simulation

Applied Materials has teamed up with NVIDIA to build an end-to-end digital development model that unifies atomic-scale discovery, process engineering, and factory optimization. By integrating NVIDIA GPU-accelerated platforms—including Ginestra with cuDSS, cuEST, PhysicsNeMo, and Omniverse—the collaboration targets the massive compute demands of modern semiconductor manufacturing. This digital thread allows engineers to virtually explore material properties, optimize chemical recipes, and simulate entire fabrication facilities before physical production begins.

For materials modeling, Applied Materials integrated the NVIDIA cuDSS (CUDA-X Direct Sparse Solver) into its Ginestra simulation platform, achieving up to a 10x speedup over CPU-only setups for sparse linear algebra. To accelerate density functional theory workflows, the collaboration utilizes the NVIDIA cuEST library. Running on NVIDIA B200 systems, quantum chemistry simulations that previously required five days of compute time on 64 CPU cores can now be completed in roughly two hours on a single GPU, representing an approximate 55x speedup.

Further down the pipeline, the Applied Materials ACE+ platform uses NVIDIA PhysicsNeMo to turn multiphysics models of chamber designs into real-time digital twins. This integration accelerates ACE+ topography simulations by up to 35x, allowing engineers to quickly refine process recipes. Once validated, these processes can be deployed to hardware like the Endura metallization system. Finally, the partnership leverages NVIDIA Omniverse to construct physically accurate digital twins of entire factories, optimizing material flows and layouts. This ecosystem-wide approach aligns with Applied's broader EPIC platform initiative.

For semiconductor practitioners, this integration fundamentally shifts the workflow from slow, physical trial-and-error to rapid, interactive virtual testing. Instead of waiting over a week for complex reaction searches or days for chamber simulations, engineers can now evaluate thousands of material combinations and process variations in a single day. This drastically reduces the reliance on expensive physical experiments, lowers development costs, and accelerates the time-to-market for next-generation silicon.