China’s electric car race is becoming all about semiconductors

BEIJING — Chinese electric car companies that are already engaged in an intense price war are turning up the heat on another front: Chip-powered tech features such as the driver-assist function.

Nio and Xpeng have announced that their in-house designed auto chips are ready for production. So far, many of the major Chinese electric car makers have relied on Nvidia chips, with the company’s automotive chips business over the past few years bringing in more than $300 million in revenue a quarter.

“It’s hard to point to your product being superior when your competitors use the exact same silicon to power their infotainment and intelligent driving systems,” said Tu Le, founder of consulting firm Sino Auto Insights, explaining why EV makers are turning to in-house chips.

Le said he expected Tesla and Chinese electric car startups to compete on designing their own chips, while traditional automakers will likely still rely on Nvidia and Qualcomm “for the foreseeable future.”

Nvidia reported a 37% year-on-year increase in automotive segment revenue to $346 million in the latest quarter.

“Automotive was a key growth driver for the quarter as every auto maker developing autonomous vehicle technology is using NVIDIA in their Data Centers,” company management said on an earnings call, according to a FactSet transcript.

“I think the main reason why Chinese [automakers] pay attention [to] self-development system-on-chip is the success of Tesla in full-self driving,” said Alvin Liu, a Shanghai-based senior analyst for Canalys.

In 2019, Tesla reportedly shifted from Nvidia to its own chip for advanced driver-assist functions.

By designing their own chips, Chinese automakers can customize features, as well as reduce supply chain risk from geopolitical tensions, Liu said.

Liu does not expect significant impact to Nvidia in the short-term, however, as Chinese automakers will likely test new tech in small batches in the higher-end of the market.

Nio in late July said it had finished designing an automotive-grade chip, the NX9031, that uses a highly advanced 5 nanometer production technology.

“It is the first time that the five-nanometer process technology has been used in the Chinese automotive industry,” said Florence Zhang, consulting director at China Insights Consultancy, according to a CNBC translation of her Mandarin-language remarks. “It has broken through the bottleneck of domestic intelligent driving chip research and development.”

Nio, which had teased the chip in December, plans to use it in the high-end ET9 sedan, set for delivery in 2025.

The 5 nanometers technology is the most advanced one for autos because the 3 nanometer tech is mostly used for smartphone, personal computer and artificial intelligence-related applications, CLSA analyst Jason Tsang, said following the Nio chip announcement.

Xpeng at its event on Tuesday did not disclose the nanometer technology it was using for its Turing chip. The company’s driver-assist technology is widely considered one of the best currently available in China. 

While Xpeng revealed its chip on Tuesday, Brian Gu, Xpeng president, emphasized in a CNBC interview the day before that his company will primarily partner with Nvidia for chips.

The two companies have a close relationship, and Xpeng’s former head of autonomous driving joined Nvidia last year.

Giants in China’s electric car industry are also recognizing the importance of chips for autos.

If batteries were the foundation for the first phase of electric car development, semiconductors are the basis for the industry’s second phase, as it focuses on smart connected vehicles, BYD‘s founder, Wang Chuanfu, said in April at a press conference held by Chinese driver-assist chip company Horizon Robotics.

Wang said more than 1 million BYD vehicles use Horizon Robotics chips.

BYD on Tuesday announced its Fang Cheng Bao off-road vehicle brand would use Huawei’s driver-assist system.

U.S. restrictions on Nvidia chip sales to China haven’t directly affected automakers since the cars haven’t required the most advanced semiconductor technology so far.

But with increasing focus on driver-assist tech, which relies more on artificial intelligence — a segment at the center of U.S.-China tech competition — Chinese automakers are turning to in-house tech.

Looking ahead to the next decade, Xpeng Founder He Xiaopeng said Tuesday the company plans to become a global artificial intelligence car company.

When asked about the availability of computing power for training driver-assist tech, Xpeng’s Gu told reporters Monday that prior to the U.S. restrictions the company had been working with Alibaba Cloud. He claimed that access now probably gives Xpeng the largest cloud computing capacity among all car manufacturers in China.

Government incentives, from subsidies to support for building out a battery charging network, have helped electric cars take off in China, the world’s largest auto market.

In July, penetration of new energy vehicles, which includes battery-only and hybrid-powered cars, exceeded 50% of new passenger cars sold in China for the first time, according to industry data.

That scale means that companies involved in the country’s electric car development are also contributing to new standards on tech for cars, such as removing the need for a physical key to unlock the door. Instead, drivers can use a smartphone app.

How that app or device securely connects drivers to their cars is part of the forthcoming set of standards that the California-based Car Connectivity Consortium is working on, according to president Alysia Johnson.

A quarter of the organization’s members are based in China, including Nio, BYD, Zeekr and Huawei. Apple, Google and Samsung are also members, Johnson revealed.

She said the organization is looking to enable a driver of a Nio car that uses a Huawei phone to securely send the car “key” to a partner who uses an Apple phone and drives a Zeekr car, for example.

“Digital key tech is becoming a lot more accessible than people would think,” she said.