Silvaco Group, Inc. (Nasdaq: SVCO) said it has partnered with NVIDIA Corporation (Nasdaq: NVDA) to develop AI-powered digital twin technologies for semiconductor design and manufacturing, as the industry increasingly turns to accelerated computing to manage the growing complexity of advanced chip development.
The collaboration will combine Silvaco’s physics-based semiconductor simulation software with NVIDIA’s accelerated computing platform, AI frameworks and digital twin technologies to help chipmakers design, validate and optimise semiconductor devices and manufacturing processes more efficiently.
The companies said the initiative will integrate Silvaco’s technology computer-aided design (TCAD), electronic design automation (EDA) and semiconductor IP portfolio with NVIDIA’s CUDA-X libraries, PhysicsNeMo, Omniverse libraries, Cosmos and Nemotron open models. The combined platform is intended to enable customers to build high-fidelity digital twins that replicate semiconductor devices and manufacturing environments with greater speed and accuracy.
The announcement comes as semiconductor companies face rising computational demands driven by advanced process nodes, heterogeneous integration and increasingly complex packaging technologies. Digital twins are being adopted across the industry to simulate manufacturing processes and device behaviour before physical production, reducing development costs and shortening design cycles.
As part of the collaboration, Silvaco plans to accelerate its semiconductor device, process, photonics and multiphysics simulation software using NVIDIA’s GPU computing platform and CUDA-X libraries.
The company said it recently completed a fully scaled three-dimensional finite-difference time-domain (3D FDTD) simulation of a photonic edge coupler comprising 3.2 billion mesh nodes on 32 NVIDIA GPUs connected through NVLink in less than four hours. According to Silvaco, the same workload did not converge using CPU-based systems, while the GPU-based simulation produced results within 0.15 dB of measured performance.
Silvaco also intends to develop AI surrogate models using NVIDIA PhysicsNeMo to complement traditional physics-based simulations. The models are expected to enable engineers to evaluate design alternatives more rapidly by reducing the number of computationally intensive simulations required during development.
The partnership will also extend to collaborative digital twin environments. Silvaco said it plans to integrate its simulation platform with NVIDIA Omniverse libraries and NVIDIA Cosmos to enable real-time visualisation of semiconductor fabrication facilities, manufacturing systems and robotics infrastructure. The companies said the cloud-native environment is designed to support distributed engineering teams working across multiple computing environments.
The companies expect the combined technologies to reduce semiconductor simulation cycles from weeks to days through GPU acceleration and AI-assisted modelling while improving validation accuracy and engineering productivity.
“The convergence of physics-based simulation, accelerated computing, and artificial intelligence is transforming design and manufacturing,” said Walden C. Rhines, president and chief executive officer of Silvaco.
He said combining Silvaco’s semiconductor modelling capabilities with NVIDIA’s AI and accelerated computing platforms would allow customers to model increasingly complex semiconductor systems with greater speed and confidence.
Da Yang, senior director of product for semiconductor and EDA at NVIDIA, said digital twins were becoming an essential engineering tool as semiconductor development grows more complex.
He said the collaboration would combine AI, accelerated computing and high-fidelity simulation to help customers move more quickly from modelling to engineering insights across semiconductor design and manufacturing.
The companies said the collaboration is expected to support applications including semiconductor process and device simulation, advanced packaging, photonics, AI-assisted chip development, factory optimisation and predictive manufacturing.
