TSMC deploys NVIDIA AI across chip manufacturing

NVIDIA and TSMC Bring AI Into Fabs at GTC Taipei

NVIDIA announced on Sunday that TSMC is deploying its accelerated computing and AI technologies across the semiconductor design and manufacturing lifecycle, marking a deepening of the nearly three-decade partnership between the chipmaker and the world’s leading foundry.

AI Across the Fab Floor

The announcement, made at GTC Taipei on June 1, details how TSMC is using NVIDIA’s CUDA-X libraries and AI models to tackle workloads spanning lithography, transistor simulation, process control, and fab operations optimization.globenewswire

TSMC is applying NVIDIA cuLitho, a GPU-accelerated computational lithography library, to achieve a 20–50% improvement in cost effectiveness or cycle time compared with CPU-based approaches at the same cost of ownership. For transistor and process simulation, the foundry is using NVIDIA cuEST, which delivers 50x faster chemistry simulations on average for semiconductor material design. TSMC is also leveraging NVIDIA’s cuML machine learning library to accelerate large-scale analytics, distilling hundreds of thousands of process parameters across thousands of steps to reduce process variation.yahoo

GPU-accelerated scheduling computation using CUDA on NVIDIA H200 GPUs has led to improvements in fab productivity by helping TSMC manage complex constraints and streamline production paths.globenewswire

Defect Inspection and Digital Twin

Beyond compute acceleration, TSMC is using the NVIDIA Metropolis platform and NVIDIA TAO Toolkit to advance automated defect classification with vision AI, improving detection of nanometer-scale defects while reducing the need for repeated labeling and retraining as process conditions change.electronicsforyou

TSMC is also exploring NVIDIA Omniverse to build “FabTwin,” a virtual fab environment for evaluating process tool layouts and simulation workflows. By testing design scenarios digitally before physical implementation, TSMC aims to identify potential constraints earlier and accelerate decision-making before capital commitments are made.globenewswire

A Partnership in Context

“TSMC is bringing NVIDIA AI and accelerated computing into the fab itself, tackling some of the world’s most complex design and manufacturing challenges,” NVIDIA CEO Jensen Huang said in the announcement. TSMC Chairman and CEO C.C. Wei said the collaboration strengthens TSMC’s “technology leadership and manufacturing excellence to support our customers’ future products and success”.globenewswire

The announcement came as part of a broader week of activity in Taipei, where Huang delivered a keynote covering topics from the Vera Rubin AI platform entering full production to a reinvented PC lineup and advances in physical AI.nvidia