Limited supply of Nvidia's H20 AI GPUs - new B30 model being developed exclusively for China to meet export regulations.
Nvidia is set to release a new GPU, the B30, based on its Blackwell architecture, in the fourth quarter of 2025. The B30 is designed to cater to the Chinese market, with a focus on US export compliance while maintaining competitiveness in China's AI market.
The B30 GPU is expected to be around 10-20% slower than the H20 GPU but will be offered at a 30-40% lower cost. This price difference could help Nvidia maintain or increase its gross margins.
Currently, Nvidia has halted production of the H20 silicon due to US government restrictions. The company is focusing on the B30 GPU, with testing already underway in China. However, the B30 is not expected to be sold there before September 2025.
There is significant demand projected for the B30 in China, with expectations that H20 inventories will mostly clear before the B30 fully launches. The B30 GPU is being prepared by Nvidia for AI applications.
It is unclear whether the H20's successor will be based on the original Blackwell or the Blackwell Ultra architecture. Nvidia plans to fulfill only existing orders for H20 GPUs using current inventory, not produce any more.
The B30 GPU needs to offer competitive performance compared to AI accelerators from companies like Biren or Huawei. The B30 could also be smaller than the H20 GPU while still meeting US export control requirements for performance density.
Nvidia's leadership plans to discuss the China-focused Blackwell chip in upcoming keynotes and meetings, signaling its strategic importance. However, it is important to note that Nvidia has not confirmed anything about these reported Blackwell-based GPUs for the Chinese market.
References:
- Nvidia’s B30 GPU: A China-focused Blackwell chip targeting AI applications
- Nvidia’s B30 GPU: Expected to launch in Q4 2025
- Nvidia’s B30 GPU: A cheaper, slower alternative to the H20
- Nvidia to discuss China-focused Blackwell chip in upcoming keynotes
The B30 GPU, a China-focused Blackwell chip, is prepared by Nvidia for AI applications in the technology industry. This product, set to launch in Q4 2025, is expected to compete with AI accelerators from companies like Biren or Huawei in the finance sector, offering competitive performance while being smaller and more cost-effective than the H20 GPU.
