Bloomberg / Getty Images
Nvidia $NVDA +4.06% and TSMC $TSM +3.05% announced that the world's largest contract chipmaker is deploying Nvidia accelerated computing and AI tools across its fabrication facilities to improve chip yields, reduce defects, and accelerate production timelines.
Taiwan Semiconductor Manufacturing Co. is using several Nvidia software libraries on Nvidia GPUs to address some of the most computationally intensive stages of chip production. Benchmarks cited by both companies put cuLitho's performance advantage at 20% to 50% over CPU-based workflows, measured in either cost savings or cycle time reduction. Chemistry simulations that once relied on conventional methods can now be completed up to 50 times faster through cuEST, accelerating TSMC's evaluation of new semiconductor materials.
On the process control side, TSMC is using Nvidia's cuML machine learning library to analyze hundreds of thousands of manufacturing parameters across thousands of production steps, allowing engineers to reduce process variation. Production scheduling across TSMC's fabs is handled in part by Nvidia H200 GPUs, whose computing capacity gives engineers greater flexibility in navigating manufacturing constraints and keeping output moving efficiently.
