# Nvidia 的新 AI 運算平台搭載台積電製晶片實現突破。

*semiconductor · news · 2026-01-07 · Republic World*

## Key points

- Nvidia 發表搭載六顆由台積電製造的新型並行晶片的 Vera Rubin 平台。
- Vera Rubin 現已全面量產，擺脫 Nvidia 以往一至兩顆晶片的發布週期。
- Rubin 平台採用台積電先進的 3 奈米製程，並包含全新晶片設計。
- 推論成本降至七分之一，MoE 模型所需 GPU 數量較 Blackwell 減少 75%。
- AWS、Meta、Google、Microsoft 與鴻海將成為首批採用與製造者。

Nvidia's new AI computing platform makes a leap with new TSMC-made chips as the company moves into full production of its latest technology. Nvidia Corp. CEO Jensen Huang introduced the supercomputer platform on Monday, which features six concurrent chips manufactured by Taiwan Semiconductor Manufacturing Co. (TSMC). According to a report by Focus Taiwan, the rollout marks a strategic departure from the company's previous development cycle of releasing only "one or two chips" at a time. Speaking during his keynote speech at the 2026 International Consumer Electronics Show (CES) in Las Vegas, Huang confirmed that the platform, named Vera Rubin, reached the production stage. "If Vera Rubin is going to be in time for this year, it must be in production by now. And so today I can tell you that Vera Rubin is in full production," the report quoted Huang. This transition follows a roadmap first teased during Huang's visit to Taiwan in August 2025, where he confirmed the designs were "taped out" at TSMC. He explained the necessity of this multi-chip development by citing the rapid growth of the industry. He said more than one or two new chips needed to be developed in a world where AI models are growing 10-fold, and the tokens generated are increasing fivefold annually. "It is impossible to keep up with those kinds of rates ... unless we deploy aggressive extreme co-design, innovating across all the chips, across the entire stack, all at the same time," the report quoted Huang. Advertisement The Vera Rubin suite includes the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, and Spectrum-X Ethernet Switch. These components succeed the Blackwell architecture and utilise TSMC's advanced 3-nanometer process. The flagship system of this platform is the Nvidia Vera Rubin NVL72, a liquid-cooled supercomputer weighing nearly 2 tons. The platform honours American astronomer Vera Rubin, whose work provided evidence of dark matter. Advertisement Efficiency gains provided by the new system include a reduction in inference costs to one-seventh of the Blackwell platform. It also reduces the GPU count required for training Mixture-of-Experts (MoE) models by 75 per cent. Huang previously described the Rubin as "revolutionary" because all six chips were new designs. "Leading AI labs, cloud service providers, and system builders, including Amazon Web Services (AWS), Meta, Google, and Microsoft, are expected to be among the first to adopt the new platform, the company said in a press release on Monday. Meanwhile, another Taiwanese company, Hon Hai Precision Industry Co. (Foxconn), is said to be the main manufacturer of AI servers using Nvidia's Rubin platform," the report said.

**Companies:** Nvidia Corp., Taiwan Semiconductor Manufacturing Co. (TSMC), Amazon Web Services (AWS), Meta, Google, Microsoft, Hon Hai Precision Industry Co. (Foxconn)
**Countries:** United States, Taiwan

[Read the full story on Republic World](https://www.republicworld.com/tech/nvidia-announces-full-production-of-vera-rubin-ai-platform-with-tsmc-chips)

---

Canonical: https://newsio.io/zh-TW/n/483bc828-ab69-4ff5-ba24-a6221cae11e5/nvidia-ai-vera-rubin
Summarized by Newsio from Republic World. https://newsio.io/how-it-works