en.Wedoany.com Reported - TSMC has become one of Nvidia's accelerated computing customers, as disclosed by Nvidia at Computex.
Building on its annual investment of $15 billion in Taiwan operations, Nvidia is expanding its local application scope by providing this chip manufacturer with Nvidia CUDA-X libraries and AI models to advance semiconductor design and manufacturing. Specifically, TSMC is using CuLitho (computational lithography), CuEST (computational electronic structure simulation), CuML (computational machine learning), and H200 GPUs for GPU-accelerated scheduling to help manage complex fab constraints and streamline production processes.
Nvidia stated that TSMC is also leveraging its Metropolis and TAO Toolkit to improve automated defect detection through visual AI, including detecting nanoscale defects, while reducing repetitive labeling and retraining. Additionally, TSMC is exploring the use of Nvidia Omniverse libraries to build FabTwin virtual fab environments, enabling layout testing and workflow simulation before actual implementation.
"TSMC is one of our key partners and a global semiconductor leader, and now it has become our customer," said Rev Lebaredian, Vice President of Physical AI Simulation at Nvidia. "CuEst accelerates chemical simulations by 50 times, allowing them to discover new materials in hours instead of days. Preparing an advanced chip design can consume 30 million CPU hours; CuLitho reduces that workload from 40,000 CPU systems to 350 H100 systems, cutting cycle time by nearly 50% and significantly lowering fab costs."
This disclosure marks a full-circle partnership: TSMC has long been Nvidia's primary foundry partner, with the vast majority of Nvidia's GPUs produced by this Taiwan-based company at its factories in Taiwan and in Arizona, USA.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
