Chinese tech companies reject Nvidia chips and opt for Huawei.

After the US banned the sale of high-performance AI chips to China in October 2023, Nvidia engineers quickly designed a new product line to comply with the new regulations.

However, the American company faces a bigger problem: Chinese cloud businesses – some of Nvidia's biggest customers globally – are not interested in buying these "substandard" AI chips.

Alibaba and Tencent are among China's largest cloud businesses that have been testing Nvidia's new chip prototypes since last November. According to WSJ sources, they announced they will order fewer chips from Nvidia this year than originally planned.

In the short term, Nvidia's downgrading of its processors will narrow the performance gap with domestic alternatives, making Chinese-made chips increasingly attractive to buyers.

Alibaba and Tencent are shifting some of their advanced semiconductor orders to domestic companies like Huawei Technologies and relying more on chips they develop themselves. Baidu and ByteDance are doing the same.

In the long term, Chinese buyers are uncertain about Nvidia's ability to continue supplying them, given that US regulators have committed to regularly reviewing chip export controls and may further tighten performance limits.

Technology companies are revising their business strategies to prepare for a future with less access to Nvidia products and to avoid the costly process of constantly adapting their technology to new chips.

Nvidia CEO Jensen Huang acknowledges Huawei as a formidable competitor in China. (Photo: ZUMA PRESS)

For Nvidia, balancing American regulators and Chinese customers is becoming increasingly difficult. The company has yet to fulfill billions of dollars worth of chip orders, while China is one of its largest markets, accounting for about one-fifth of its revenue.

Demand for Nvidia chips is exceeding supply. However, geopolitical tensions raise the risk of long-term sales losses in the world's second-largest economy, which is pursuing AI development as a strategic priority.

Chinese cloud businesses currently purchase about 80% of their high-end AI chips from Nvidia, and this could drop to 50%-60% in the next five years, according to Frank Kung, an analyst at research firm TrendForce. He added that future tightening of US chip regulations will put pressure on Nvidia's sales in China.

Nvidia says it is looking for ways to provide products that comply with U.S. regulations to customers worldwide . They don't see a short-term financial impact from restrictions on selling AI chips to China because they can find other buyers. However, in 2023, Chief Financial Officer Colette Kress warned that, in the long term, a ban on selling AI chips in China would make it impossible for the U.S. industry to compete and lead in one of the world's largest markets.

Switch to Huawei chips

Over the past two years, the administration of U.S. President Joe Biden has imposed two rounds of export sanctions to restrict China's access to chips and advanced technology that the U.S. believes Beijing could use to enhance its military and surveillance capabilities.

Nvidia CEO Jensen Huang still hopes to supply high-end processors to China and is working with customers in China to obtain export licenses.

Following initial restrictions in October 2022, Nvidia modified the chips sold in China by downsizing their performance to below the threshold requiring US government oversight. They sold approximately $1 billion worth of such chips to Chinese customers in 2023.

As the U.S. continues to restrict Nvidia chip exports to China without a license, the chipmaker has developed a new, weaker line of processors for Chinese buyers, scheduled for release early this year, according to the WSJ.

Last month, Nvidia launched the GeForce RTX 4090 D, a revised version of its flagship gaming chip adapted to the latest US restrictions.

Chinese companies have been testing samples of Nvidia's highest-performance AI chip, the H20. This chip allows for efficient data transfer between multiple processors, making it a better option than "homegrown" alternatives.

However, testers reported that they needed more H2O to emulate the same computing power as previous Nvidia chips, increasing the cost.

Like American chips, China's most advanced chips are capable of handling inference processing—where a trained AI model makes a prediction—and less complex training tasks.

Huawei, acknowledged by Nvidia's CEO as a "formidable competitor" in China, is gaining ground thanks to the new circumstances.

According to the WSJ, in 2023, Huawei received orders for at least 5,000 Ascend 910B chips from major Chinese internet companies. This chip is considered China's closest available alternative to Nvidia's high-performance A100 chip, which is banned from export.

Sources reveal that these chips will be delivered throughout 2024 as Huawei faces production restrictions due to US sanctions.

Chinese procurement officials, such as those in state-owned carriers, have called for the use of domestically produced chips like those from Huawei. China Telecom acquired approximately $390 million worth of AI servers equipped with Huawei chips in October 2023, while China Unicom spent at least $20 million in 2022, according to company procurement documents.

Huawei has been working to expand its software ecosystem and plans to launch a new high-end AI chip in the second half of 2024.

Several government-backed AI computing centers have been using Huawei chips since the US imposed restrictions in 2022.

Nvidia is trying to meet the demands of its Chinese customers while complying with US export regulations. (Photo: Bloomberg)

Alibaba's chip division, T-Head, is also developing a new AI processor under the Hanguang brand, sources said. "If restrictions become even tighter in the next few years, you should start thinking about alternatives now," a senior executive at Alibaba Cloud shared.

The AI ​​craze early last year fueled demand for Nvidia's advanced chips as Chinese companies and startups raced to develop their own large-scale language models. Now, many smaller players are scaling back their efforts and shifting their focus to AI applications.

Kenneth Yang, co-founder of a Shanghai-based AI healthcare startup, said he plans to bypass Nvidia's latest chips and instead lease AI processing power from Baidu or Huawei.

"That's spending money wisely," said Yang, who is developing a nursing app.

Engineers at Chinese technology companies say Nvidia chips will remain the preferred choice for the next 12 months, due to Nvidia's broader product ecosystem and the continued shortage of local alternatives.

In the long run, U.S. restrictions could spur China to develop its own technology, commented Kevin Xu, founder of the hedge fund Interconnected Capital.

"Once the current stockpiling phase is complete, Nvidia's business in China will become the scapegoat," he said.

(According to WSJ)

Source: https://vietnamnet.vn/cong-nghe-trung-quoc-che-chip-nvidia-chon-huawei-2236740.html