As NVIDIA is gradually ramping up AI servers based on the B200 and GB200 platforms, paving the way for its next-gen GB300 AI server product line, the U.S. AI giant is reportedly making another move discreetly. According to a report from Commercial Times, NVIDIA is planning to set up an ASIC (Application-Specific Integrated Circuit) division to develop customized chips.
In addition, the report notes that as NVIDIA has selected Taiwan as the base for its ASIC R&D center, it is aggressively recruiting talents from major IC design houses locally, raising concerns about a potential brain drain from Taiwan’s semiconductor sector.
According to Commercial Times, citing sources from Taiwanese IC firms, there was a wave of talent poaching in mid-2024. As NVIDIA is reportedly joining the game, another round of fierce competition for talent may be around the corner, prompting major IC design firms like MediaTek, Alchip Technologies, and TSMC’s affiliate GUC to brace themselves for the upcoming battle, the sources added.
Notably, tech giants are eagerly seeking alternatives to NVIDIA’s GPUs. According to a previous CNBC report, Apple revealed in July, 2024, that its AI models supporting Apple Intelligence were pretrained on Google’s Tensor Processing Units (TPUs).
Later in December, 2024, Apple executive Benoit Dupin also mentioned that the company is utilizing Amazon’s Trainium2 chips, indicating that non-NVIDIA training solutions are also effective, as noted by Commercial Times.
Therefore, NVIDIA is actively responding by reportedly establishing its own ASIC department and expanding its custom service capabilities, according to Commercial Times. At the same time, it is planning to recruit over a thousand professionals in fields such as chip design, software development, and AI research and development in Taiwan, the report notes.
It is worth noting that Taiwanese ASIC companies have been playing a critical role in the development of Microsoft’s Cobalt and Maia, Google’s TPU, and AWS’s custom chips, which makes them perfect targets for NVIDIA’s ASIC business, as per Commercial Times.
Read more
(Photo credit: NVIDIA)