Huawei may Add More Business Thanks to U.S. Export Restrictions on AI Chips

The United States’ measures to limit the export of artificial intelligence (AI) chips to China may pave the way for Huawei Technologies to strengthen its presence in its $7 billion domestic market, Reuters news report said.
Huawei @ MWC Shanghai 2023As restrictions impact chip leader Nvidia and others like AMD and Intel, Huawei sees an opportunity to compete more effectively, even though it faces significant challenges, the report said.

Nvidia has maintained a dominant position in the Chinese AI chip market with a market share exceeding 90 percent. However, Chinese companies, including Huawei, have been developing their versions of Nvidia’s popular chips, such as the A100 and H100 graphics processing units (GPUs).

Huawei’s Ascend AI chips are recognized for their raw computing power, though they are perceived as lagging behind Nvidia’s chips in terms of overall performance, according to analysts and AI firms like China’s iFlyTek.

Jiang Yifan, chief market analyst at Guotai Junan Securities, suggests that a key limiting factor for Chinese companies has been their reliance on Nvidia’s chips and software ecosystem. However, the U.S. restrictions may alter this landscape. He stated on his social media Weibo account, “This U.S. move, in my opinion, is actually giving Huawei’s Ascend chips a huge gift.”

Nonetheless, Huawei faces multiple hurdles. Many advanced AI projects use CUDA, a widely adopted programming architecture pioneered by Nvidia, which has given rise to a global ecosystem capable of training sophisticated AI models like OpenAI’s GPT-4. Huawei’s alternative, CANN, is seen as less capable in terms of AI model training, making it less straightforward to substitute for Nvidia.

Woz Ahmed, a former chip design executive turned consultant, emphasized that to win over Chinese clients from Nvidia, Huawei must replicate Nvidia’s ecosystem, including supporting clients in transitioning their data and models to Huawei’s platform. Intellectual property rights are another concern, with many key GPU patents held by U.S. firms.

Huawei’s success in capturing Nvidia’s market share would mark another victory against the United States, which has imposed export controls on the company since 2019. Over the past year, Huawei has been pushing back by unveiling advanced smartphone chips and claiming breakthroughs in chip design tools. It aims to become a key provider of computing power for AI, emphasizing its commitment to building a computing base for China and providing a “second option” for the world.

Huawei’s partners in China, including iFlyTek, are embracing the Ascend 910 to train their AI models. During an earnings call, iFlyTek’s Senior Vice President Jiang Tao declared that the Ascend 910B’s capabilities are “comparable to Nvidia’s A100” and announced a collaboration with Huawei to develop a general-purpose AI infrastructure in China. Other partners include state-owned software firms Tsinghua Tongfang and Digital China. At a recent conference, Huawei revealed that its AI chips are used to power over 30 large language models in China, reflecting the nation’s growing interest in generative AI.

Charlie Chai, an analyst with 86Research, believes that Nvidia’s ecosystem dominance is not insurmountable if domestic players are given enough time and a substantial customer base. China’s self-sufficiency drive, championed by President Xi Jinping, is expected to support this endeavor. In short, these restrictions may pose a minor disruption to near-term supplies but offer a significant boost to China’s long-term self-sufficiency goals.

TrendForce says the landscape for HPC chips is set for a shake-up, as the current ban expands control over the A800, H800, and L40S series. This development is poised to see China’s tech giants — ByteDance, Baidu, Alibaba, and Tencent — curb their appetite for NVIDIA’s high-end AI servers from 5–6 percent of the global AI server market to a scant 3– 4 percent.