China Wants to Put an End to Its Reliance on the U.S. It’s Encouraging Local Companies to Stop Buying Nvidia Chips

  • The government advises Chinese companies to prioritize purchasing Chinese AI chips instead.

  • However, Nvidia expects to deliver more than one million H20 chips to its Chinese customers by the end of this year.

Chips
No comments Twitter Flipboard E-mail

The Chinese government wants the country’s tech companies to stop buying integrated circuits from Nvidia. According to the South China Morning Post, this is just an advisory step at the moment and targets companies purchasing GPUs for artificial intelligence. Specifically, China recommends that they buy chips produced locally rather than from the American Nvidia.

To put this into context, the U.S. government’s sanctions prevent Nvidia from selling its most advanced AI GPUs, such as the A100 or H100 chips, to Chinese customers. This has led Nvidia to develop revised versions of its GPUs that meet the requirements imposed by the Department of Commerce. In fact, the company plans to deliver more than one million H20 chips to its Chinese customers by the end of the year, given that the American authorities allow sales of this GPU unit in the Asian country.

However, Nvidia is losing its competitive edge in China due to the U.S. restrictions. The H20 GPU is currently less capable than the company’s more advanced AI chips. Although China’s Ministry of Industry and Information Technology hasn’t made any official announcements, according to SCMP, the Chinese government has restricted its companies from purchasing Nvidia’s H20 GPUs since last August.

Chinese Alternatives Are Ready to Go

In recent years, dozens of Chinese companies have developed their own AI hardware. Some of the most important ones include MetaX, Alibaba, Biren Technology, Moore Threads, Innosilicon, Zhaoxin, Iluvatar CoreX, DenglinAI, and Vast AI Tech. However, Huawei stands out as the most prominent, with a strong presence in the Chinese market and an annual turnover of around $7 billion in China alone. Additionally, Huawei is benefiting from Nvidia’s decline in the country.

For Huawei, the current situation presents an excellent opportunity to strengthen its presence in its home country.

Over the past five years, Huawei has developed its own AI GPUs, called the Ascend AI chips. The company has been refining and enhancing the capabilities of these chips in order to match or even surpass the performance of Nvidia’s GPUs. Some analysts, such as those at the Chinese company iFly Tek, believe that the power of Huawei’s GPUs matches that of Nvidia’s chips, but they still lag behind in real-world performance.

This presents an excellent opportunity for Huawei to strengthen its presence in China. However, Chinese companies, including Huawei, face a significant challenge in trying to compete with Nvidia, which has long dominated the market. Most AI projects, both in and outside of China, are developed using CUDA (Compute Unified Device Architecture), the compiler and development tool used by programmers for Nvidia’s GPUs.

Huawei has developed CANN (Compute Architecture for Neural Networks) as its alternative to Nvidia’s CUDA. However, analysts believe that CANN is less powerful than Nvidia’s AI model training technology. To strengthen its position in the Chinese AI chip market, Huawei needs to refine CANN and make it easy for Nvidia customers to transition to its ecosystem.

In fact, that’s exactly what Huawei seems to be doing. Its Ascend 910C AI GPU could be the key to success. Many Chinese tech companies are seeking alternatives to Nvidia’s chips, and Huawei is seizing the opportunity by providing samples of that GPU model for testing. This new chip is an improved version of the Ascend 910B, which Huawei claims is on par with Nvidia’s A100, and is expected to be even more powerful than its predecessor.

Image | Richyart (via Midjourney)

Related | Nvidia Already Rules the AI Hardware Market. With the Launch of Its Own Open-Source LLM, It’s Now Going After GPT-4 and Llama

Home o Index