NVIDIA and TSMC push AI deeper into semiconductor fabs ... - eeNews Europe

NVIDIA and TSMC have deepened their long-standing partnership by integrating artificial intelligence directly into semiconductor fabrication facilities, aiming to accelerate chip production, improve y... Read original →

TSMC (Taiwan, China) and NVIDIA embedding AI directly into fabs signals a paradigm shift from equipment-centric to algorithm-driven semiconductor manufacturing. Technically, cuLitho and cuEST not only alleviate EUV computational bottlenecks but force EDA vendors like Synopsys and Cadence to rebuild GPU-native simulation stacks—or risk irrelevance below 3nm. On compliance, U.S. export controls prevent Chinese foundries like SMIC from replicating this model, yet spur domestic alternatives such as Huawei’s MindSpore integrated with NAURA tools. Samsung and Intel will respond by accelerating in-fab AI adoption, but without a unified software stack akin to CUDA-X, their efficiency lags. Within 18 months, AI-native fabs—featuring digital twins like FabTwin and real-time ML process control—will become mandatory for inclusion in next-gen GPU supply chains; legacy fabs lacking these capabilities will be excluded from H200/B100 production.