Nvidia Is About to Lose Two Major HPC Partners in China, Is Still Optimistic
Nvidia believes that if the US government further restricts sales of A100 and H100 data center grade products to Huawei and Inspur, it will be able to find suitable partners for computing GPUs in Asia Pacific. However, these companies are large and consume hundreds of millions of dollars worth of hardware, so this may not be particularly easy. But Nvidia has no choice.
Colette Cress, Chief Financial Officer of Nvidia, said: Morgan Stanley Technology Media and Telecom Conference(via Seeking Alpha). “As we move forward, we will probably work with other partners to launch computing in Asia Pacific or other parts of the world. Make sure your focus is on the law and that you follow export regulations very closely.So in this case, we will consider helping us from the perspective of our other partners.”
Huawei has been on the U.S. Department of Commerce Entity List since 2019, and virtually all U.S. partners are required to obtain an export license to sell hardware, software, technology, or services to the company. bottom. This has worked for many companies, including his Nvidia, which sold expensive computing GPUs to Huawei.But the Biden administration is looking to further restrict the items that American companies are allowed to ship to Huawei, which is expected to cost Nvidia millions of dollars. Reuters It was recently reported citing reports by government contractors.
Meanwhile, the US government last week blacklisted China-based Inspur, the world’s third-largest server maker, further complicating Nvidia’s attempts to sell its products to Chinese companies.
The US government imposed new export regulations in August, requiring AMD, Nvidia, and Intel to obtain special export licenses to sell or ship high-performance computing hardware to China. The U.S. will allow chip makers to sell products made using American technology that would allow Chinese companies to build supercomputers with over 100 FP64 petaflops or over 200 FP32 petaflops within 41,600 cubic feet (1178 cubic meters). It is virtually prohibited to sell.
Even under existing export restrictions, Nvidia cannot sell its A100, A100X, and H100 series products to Chinese companies. So Nvidia invented his A800 computing GPU. This is an Ampere-based product that cannot be used to build supercomputers with over 100 FP64 PetaFLOPS. performance.
Nvidia’s revenue could run into the hundreds of millions of dollars if the US government stops granting Nvidia export licenses to sell A30, A800 and other GPUs to Huawei and Inspur. Nvidia has little competition in the AI and HPC GPU markets, so given the popularity and ubiquity of the CUDA platform in these segments, it should be very easy to find new partners in the APAC region and deliver the hardware to end customers. (you must obtain a license to ship products to Chinese clients).
But Huawei and Inspur are more than just distributors. They are solution providers with many customers who rely on their hardware and services. Replacing these two customers won’t be easy for Nvidia, so the new restrictions on selling computing hardware to Chinese entities will have a noticeable economic impact on Nvidia.