The collaboration brings AI deeper into chip production, helping TSMC tackle growing complexity in advanced semiconductor design and fabrication.
NVIDIA has announced that Taiwan Semiconductor Manufacturing Company (TSMC) is expanding its use of NVIDIA’s AI and accelerated computing technologies to enhance semiconductor design and manufacturing processes.
As chip production advances to increasingly sophisticated process nodes, manufacturers face growing computational demands in areas such as lithography, transistor modeling, process optimization, and defect detection. TSMC is deploying NVIDIA’s GPU-powered platforms and AI tools to address these challenges, aiming to improve efficiency, productivity, and manufacturing yields across its operations.
NVIDIA CEO Jensen Huang said the two companies’ long-standing partnership is now extending deeper into semiconductor manufacturing, with AI and accelerated computing being applied directly within fabrication facilities to tackle complex design and production tasks.
TSMC Chairman and CEO C.C. Wei said the collaboration is helping the company strengthen its manufacturing capabilities and technological leadership as it develops next-generation chips for customers worldwide.
Among the technologies being adopted, TSMC is using NVIDIA’s cuLitho platform to accelerate computational lithography, a critical step in chip patterning. The company says the technology can improve cost efficiency and reduce processing times compared with traditional CPU-based approaches.
TSMC is also utilizing NVIDIA’s cuEST simulation software for semiconductor materials research and transistor modeling, enabling significantly faster simulations. In process control, the chipmaker is leveraging the cuML machine-learning framework to analyze vast amounts of manufacturing data and reduce process variability.
To improve factory operations, TSMC is deploying NVIDIA H200 GPUs and CUDA-based scheduling systems to optimize production workflows and increase fab productivity.
The partnership also extends to quality control. TSMC is using NVIDIA’s Metropolis platform and TAO Toolkit to enhance defect detection through AI-powered visual inspection, helping identify microscopic flaws more accurately while reducing the need for repeated model training.
In addition, TSMC is exploring NVIDIA Omniverse technologies to develop “FabTwin,” a digital twin environment that enables virtual testing of fab layouts and production scenarios before physical deployment. The approach is designed to improve planning, identify bottlenecks earlier, and support faster decision-making in advanced semiconductor manufacturing.
