www.industry-asia-pacific.com
04
'26
Written on Modified on
TSMC Deploys NVIDIA Accelerated Computing and AI in Fabs
TSMC is integrating NVIDIA AI and accelerated computing platforms to optimize semiconductor design, manufacturing, and defect inspection workflows.
www.nvidia.com
NVIDIA has announced that TSMC is deploying its accelerated computing and AI technologies across its semiconductor design and manufacturing lifecycles. The integration addresses the computational challenges of transitioning to advanced process nodes by applying software libraries and AI systems to massive-scale simulations and real-time operational optimization.
Acceleration via CUDA-X Libraries
TSMC is utilizing NVIDIA CUDA-X libraries and AI models running on NVIDIA graphics processing units (GPUs) to accelerate key production workloads:
- Computational Lithography: The company is implementing NVIDIA cuLitho, a GPU-accelerated library for chip mask design. The technology provides a 20% to 50% improvement in cost-effectiveness or cycle time compared to CPU-based alternatives, without changing the total cost of ownership.
- Material and Transistor Simulation: TSMC uses the NVIDIA cuEST library for electronic structure simulation, achieving an average 50x speed increase for chemistry simulations in semiconductor material design.
- Advanced Process Control: By employing the NVIDIA cuML machine learning library, TSMC executes large-scale analytics on GPUs. This enables the acceleration of algorithms to process hundreds of thousands of parameters across thousands of manufacturing steps, minimizing process variation.
- Fab Operations Optimization: TSMC is executing scheduling computations on NVIDIA H200 GPUs using CUDA-powered processing. This integration enhances the management of complex operational constraints to streamline factory production paths.
Vision AI and Defect Inspection
To identify microscopic manufacturing anomalies, TSMC is deploying the NVIDIA Metropolis platform alongside the NVIDIA TAO Toolkit. This vision AI system improves defect classification at the nanometer scale. The integration optimizes quality inspection routines and limits the need for continuous data re-labeling or model retraining when process conditions, tools, and defect types shift.
Virtual Prototyping with FabTwin
TSMC is exploring NVIDIA Omniverse libraries to construct a virtual fab environment named FabTwin. This digital twin platform enables engineers to model process tool layouts and simulate operational workflows. By evaluating architectural and logistical configurations digitally before physical implementation, TSMC can identify structural constraints and optimize factory planning without premature capital commitments.
"TSMC and NVIDIA have built a long-standing partnership rooted in advancing the technologies that make the next generation of computing possible. By using NVIDIA accelerated computing and AI across fab operations optimization, lithography, process control and inspection, TSMC is strengthening our technology leadership and manufacturing excellence to support our customers’ future products and success." — C.C. Wei, Chairman and CEO of TSMC.
Edited by Romila DSilva, Induportals Editor, with AI assistance.
To identify microscopic manufacturing anomalies, TSMC is deploying the NVIDIA Metropolis platform alongside the NVIDIA TAO Toolkit. This vision AI system improves defect classification at the nanometer scale. The integration optimizes quality inspection routines and limits the need for continuous data re-labeling or model retraining when process conditions, tools, and defect types shift.
Virtual Prototyping with FabTwin
TSMC is exploring NVIDIA Omniverse libraries to construct a virtual fab environment named FabTwin. This digital twin platform enables engineers to model process tool layouts and simulate operational workflows. By evaluating architectural and logistical configurations digitally before physical implementation, TSMC can identify structural constraints and optimize factory planning without premature capital commitments.
"TSMC and NVIDIA have built a long-standing partnership rooted in advancing the technologies that make the next generation of computing possible. By using NVIDIA accelerated computing and AI across fab operations optimization, lithography, process control and inspection, TSMC is strengthening our technology leadership and manufacturing excellence to support our customers’ future products and success." — C.C. Wei, Chairman and CEO of TSMC.
Edited by Romila DSilva, Induportals Editor, with AI assistance.
