NVIDIA's AI server platform strategy is expanding from GPUs to CPUs, with its next-generation Rosa CPU becoming a market focal point. According to tech media outlet Wccftech, the semiconductor industry speculates that NVIDIA's Rosa could be manufactured using TSMC's 2nm process, and there's even a possibility of adopting the A16 backside power delivery technology. If this speculation proves true, not only TSMC but also three major Taiwanese manufacturers—Chung-Shan Precision, Sunyen, and Sway Win—could benefit significantly.

NVIDIA recently revealed that its next-generation Rosa CPU will adopt a new 'Rigel' CPU core architecture and continue with the Arm v9.2 architecture design, emphasizing massive single-threaded computing power required for AI agent (Agentic AI) workloads. Rosa is expected to launch alongside the next-generation Feynman GPU platform, entering the data center market in 2029, followed by a PC version platform in 2030.

In response, Wccftech reports that NVIDIA is considering adopting two key TSMC manufacturing technologies: the 2nm process or the A16 backside power delivery technology. Currently, TSMC's advanced manufacturing technologies have already been used to produce various chips, including AMD's EPYC Venice and Intel's Nova Lake.

With these reports circulating, industry insiders speculate that if NVIDIA adopts the A16 backside power delivery technology, it could significantly increase demand for CMP (Chemical Mechanical Polishing) processes and consumables. This initiative would not only benefit TSMC but also Taiwanese firms such as Chung-Shan Precision, which specializes in diamond discs and reclaimed wafers; Sunyen, a Tier-1 reclaimed wafer supplier; and Sway Win, targeting the CMP polishing pad market, all of which stand to gain from the ripple effect.

FACT BOX

  • Source: PR Times
  • Category: New Product
  • Organizations: AMD
  • Products / services: Rosa CPU