[News] Chinese Firms Reportedly Buy Over 1M NVIDIA Gaming, Workstation Chips as Domestic AI Supply Falls Short

China’s surging AI computing demand is pushing technology companies to seek alternative sources of processing power as domestic chip supply struggles to keep pace. According to Economic Daily News, citing Financial Times, Chinese technology companies have reportedly purchased more than 1 million NVIDIA gaming and workstation chips this year, turning to processors not originally designed for advanced AI systems.

According to the Financial Times, citing industry sources, China’s domestic AI chip production has more than doubled this year, driven by companies including Huawei, Alibaba, Baidu, and Cambricon. The country’s AI chip supply is expected to reach 3.5 million units in 2026, including approximately 1 million manufactured overseas, while total demand is estimated to far exceed 5 million units.

The NVIDIA chips purchased reportedly include the RTX Pro 6000D, RTX Pro 5000, and RTX 5090. The report notes that while these processors generally offer lower performance than advanced Chinese AI chips such as Huawei’s flagship Ascend 910C across most computing tasks, they remain capable of handling simpler AI workloads. These include the prefill, or prompt-processing, stage of inference, when an AI model processes a user’s input before beginning to decode and generate a response.

The supply constraints are also evident at Huawei. According to Reuters, Huawei Rotating Chairman Eric Xu said in September that the company could not produce enough AI computing hardware to meet domestic demand, leading it to prioritize Chinese customers over overseas expansion.

CUDA Ecosystem Supports NVIDIA Demand as New RTX Pro 5500 Draws Interest

As Financial Times indicates, Chinese companies’ preference for NVIDIA workstation chips is partly driven by its CUDA software platform, which many had adopted before turning to domestic alternatives such as Huawei’s CANN, generally considered more difficult to use. The report adds that DeepSeek is working closely with Huawei to improve hardware and software performance, while Huawei is using AI tools to accelerate code conversion from CUDA to CANN.

Meanwhile, NVIDIA is reportedly preparing to introduce the RTX Pro 5500, a new workstation chip expected to outperform its existing offerings in China and better support basic AI computing workloads. According to Seoul Economic Daily, citing The Information, China’s Ministry of Industry and Information Technology has signaled that it may approve purchases of the chip by domestic companies, including ByteDance and Alibaba. The ministry has requested their intended purchase volumes and usage plans, while ByteDance is reportedly considering an order of around 1 million units.

NVIDIA reportedly plans to ship approximately 500,000 RTX Pro 5500 chips per quarter to China starting in late December, with customers required to place orders by September 30 to secure supply. Designed primarily for engineering and graphics workloads, the chip uses GDDR7 memory rather than the high-bandwidth memory (HBM) found in server-class AI accelerators, as noted by The Information. The report adds that it is priced at around RMB 85,000–90,000 in China, comparable to Huawei’s Ascend 950PR.

Read more

• [News] Huawei Says Ascend Market Share Tops NVIDIA in China, Backs NPO Over CPO for Optical Interconnects
• [News] DeepSeek Reportedly Expands Huawei Chips to AI Training; New Supply Could Arrive in 4Q26

(Photo credit: NVIDIA)