An analyst notes that Huawei’s new chip isn’t likely to hold much appeal outside of China

Reports of a new Huawei chip weighed on Nvidia’s stock Monday.

Chinese tech giant Huawei is reportedly ready to test a new artificial-intelligence chip to compete with Nvidia — but the U.S. chipmaker shouldn’t be too concerned, according to one analyst.

Huawei is looking to test its Ascend 910D chip with fellow homegrown tech companies and is expected to get its first samples of the chips next month, the Wall Street Journal reported, citing unnamed people described as familiar with the matter. The Chinese company, which was added to the U.S. Entity List in 2019, reportedly is aiming to take on Nvidia Corp.’s H100 chips, which are used to train some of the world’s most powerful AI models. Nvidia cannot sell its H100 chips to Chinese companies under U.S. restrictions.

News of the Huawei chip, reportedly still in early development, negatively impacted some U.S.-listed chip stocks on Monday, including Nvidia’s. Shares of Nvidia fell 2%, while Micron Technology Inc. saw its shares drop 1.5%, TSMC’s U.S.-listed shares slipped 1% and AMD shares were down 0.3%, though all recovered from intrasession lows. Shares of Intel Corp., meanwhile, gained 2.3%, while Broadcom Inc. rose fractionally.

But while a new AI chip would demonstrate Huawei’s ability to innovate quickly, it likely wouldn’t have the performance benefits of Nvidia’s offerings, which would limit its appeal outside of China.

The Ascend 910D reportedly uses packaging technology that can improve performance by integrating more silicon dies. However, it uses more power and is less efficient than the H100, according to the Wall Street Journal report.

Huawei’s rollout of chips to compete with Nvidia “is driven by the deepening rift between the West and China and China’s need to get rid of its dependence on Western technology (such as AI chips),” Richard Windsor, founder of research firm Radio Free Mobile, said in a note on Monday.

‘I see a substantial disadvantage for Chinese AI when trying to win hearts and minds outside of China, as it will be both more expensive and come with the usual strings attached.’

— Richard Windsor, Radio Free Mobile

This wouldn’t be the first chip in the Ascend line. The Ascend 910 was made using TSMC’s 7-nanometer process and came out in 2019. That was followed by the Ascend 910B in 2023, which was made using China’s Semiconductor Manufacturing International Corp.’s 7-nanometer process. The Ascend 910C, which was also made using SMIC’s 7-nanometer process, fell short of its comparison to the H100 according to some engineers, the Journal reported. The 910C is expected to start shipping next month, Reuters reported.

Meanwhile, Nvidia’s H20 chip — which it developed specifically for the Chinese market to comply with U.S. export controls — was restricted for sale in the country by the Trump administration earlier this month. Some Chinese companies have reportedly sought to order more Ascend 910C chips amid the ban.

But Chinese AI startup DeepSeek, which sent shockwaves through the market after it demonstrated models on par with those from OpenAI and Meta using Nvidia’s less-powerful H800 chips, showed “that while limiting China’s access to silicon is not going to prevent it from developing leading-edge models, it will damage its ability to do so economically,” Windsor said.

Nvidia’s Blackwell chips were built in anticipation for trillion-parameter models, Windsor said, meaning that AI developers without access to the chips are set to face longer training times and higher costs for training and inferencing as models get larger.

“When developing models for government and military purposes, cost is not a major consideration, but when it comes to the civilian market (which will determine global dominance in AI), cost is everything,” Windsor said.

And Huawei’s chipmaking efforts are likely to stay limited at 7 nanometers, “as it is already at the very limits of what multipatterning can produce as evidenced by TSMC’s and Intel’s failure to progress beyond it,” Windsor said.

While more expensive training and inferencing are “not going to hurt the prospects for Chinese AI in China, it is the rest of the world that will determine whether U.S. attempts to isolate China are successful,” Windsor said. “Here, I see a substantial disadvantage for Chinese AI when trying to win hearts and minds outside of China, as it will be both more expensive and come with the usual strings attached.”

However, Nvidia does face the risk of China’s government forcing private-sector companies to switch from the company’s CUDA computing platform in favor of Huawei’s, according to Windsor, which would have an impact on its China revenues.

“The net result is that I don’t think that Huawei is going to be competing directly with Nvidia, and so concerns over substantial market-share loss are overblown, in my opinion,” Windsor said.