Analysts speculate that Huawei Technologies [RIC: HWT.UL] may be able to capitalize on a gap in the $7 billion Chinese market created by the United States’ decision to restrict the shipment of sophisticated artificial intelligence (AI) chips to China.
Chinese companies like Huawei have been building their copies of Nvidia’s best-selling processors, such as the A100 and the H100 graphics processing units (GPU), despite Nvidia (NVDA.O.) traditionally being the dominant provider of AI chips in China with a market share of over 90%.
Analysts and certain AI companies, including China’s iFlytek (002230. SZ), have noted that Huawei’s Ascend AI processors are competitive with Nvidia’s regarding raw computational power. However, they still fall short in terms of performance.
Brokerage Guotai Junan Securities’ top market analyst Jiang Yifan said that the U.S. limitations might help Chinese companies overcome a major barrier: their reliance on Nvidia’s processors and software ecosystem.
Jiang remarked on his Weibo account, “This U.S. move, in my opinion, is actually giving Huawei’s Ascend chips a huge gift.”
There are risks associated with this possibility, though.
Nvidia’s CUDA programming architecture is used in various cutting-edge artificial intelligence (AI) applications. This architecture has spawned a vast worldwide ecosystem that can train complex AI models like OpenAI’s GPT-4.
Huawei’s CANN is its implementation; nevertheless, analysts have noted that it has far fewer AI models it can train than Nvidia’s, making Huawei’s chips a poor drop-in replacement.
According to Woz Ahmed, a former executive in chip design turned consultant, Huawei has to help customers migrate their data and models to Huawei’s platform to win over Nvidia’s Chinese clientele.
According to Ahmed, intellectual property rights are another issue because several American companies now own crucial GPU patents.
“To get something that’s in the ballpark, it takes 5 or 10 years,” he said. Reuters’ attempts to reach Huawei and Nvidia for comment were silent.
POWER OF COMPUTATION
Since 2019, the United States has targeted Huawei with export limits; if Huawei successfully takes Nvidia’s market share, it will be another triumph for Huawei. That same year, Huawei released its first Ascend graphics processing units (GPUs), one of several products (including the Harmony OS) the firm claims were developed in-house.
The telecommunications giant has shown a cutting-edge smartphone chip and made claims of improvements in chip design tools in the previous year, indicating a pushback against restrictions imposed by the United States.
Chief Financial Officer Meng Wanzhou stated last month that Huawei aimed to create a computer base for China and provide the globe with a “second option” to the United States, which is now the main supplier.
One of Huawei’s Chinese partners, iFlyTek, is a pioneer in artificial intelligence software; it’s training its models on the Ascend 910. In 2019, the United States also placed IFlyTek on its blocklist.
During Thursday’s earnings call, iFlytek Senior VP Jiang Tao compared the capabilities of the Ascend 910B to those of Nvidia’s A100 and revealed that the company was working with Huawei to build up a general-purpose AI infrastructure in China.
“Our partnership now aims to enable domestically developed LLMs to be built with both homegrown hardware and software technology,” added Jiang.
Tsinghua Tongfang and Digital China are also involved, both of which are under Chinese government control. At a conference in July, Huawei said that its AI processors are used to power more than 30 large language models (LLM) in China, which has more than 130 LLMs and is experiencing a generative AI boom.
Nvidia’s domination of the ecosystem is not “an insurmountable obstacle if domestic players are given sufficient time and a big customer base,” according to 86 Research analyst Charlie Chai.
President Xi Jinping’s drive for greater self-sufficiency in China will contribute to this. “In short, a small disruption to near-term supplies, but a big boost to the long-term self-sufficiency agenda,” noted Chai.