Silvaco and NVIDIA Announce GPU and AI Partnership to Build Semiconductor Digital Twins

Overview

• The companies issued a joint announcement that included a proof-of-concept 3.2 billion-node, 3D FDTD photonic edge coupler simulation that ran on 32 NVIDIA GPUs in under four hours and did not converge on CPUs.
• Silvaco said it will adopt NVIDIA technologies such as CUDA-X libraries, PhysicsNeMo and Nemotron models, and Omniverse visualization to combine physics-based TCAD/EDA with accelerated computing and AI.
• The partners identified four focus areas: GPU-accelerated physics simulation, AI-driven surrogate models, Omniverse-based visualization and collaboration, and cloud-native, scaled engineering workflows.
• Silvaco projects the work will cut simulation turnaround from multiple weeks to days, which could speed design iterations, shorten time-to-market and support predictive manufacturing when product integration and customer deployments follow.
• Shares of Silvaco rose about 6% in premarket trading after the announcement, and the firms included typical forward-looking caveats as the collaboration moves from demonstration toward technical integration and commercial development.