Major Chinese EV makers – Nio, Xpeng, and Li Auto – are all making moves in producing their own chips for vehicles, as the former two focus on AI chips for autonomous driving and the latter works on more basic semiconductor components, according to Chinese media outlet LatePost.
Why it matters: The move reflects a growing trend among Chinese EV automakers to bring some chip production in-house, as an ongoing global semiconductor shortage continues to hinder vehicle production.
- Globally, automakers have been forced to cut more than 3.4 million vehicles from their production schedules so far this year, according to the latest estimate by AutoForecast Solutions.
Nio looks to autonomous driving and lidar chips:
- LatePost reported that Nio has formed a team of 300 people to work on chips. The team focuses on developing chips for autonomous driving and lidar. The direct leader of the team joined from Huawei’s chip arm Hisilicon last year.
- Nio first formed a chip unit in the second half of 2020, attached to the autonomous driving arm of the company led by Bai Jian, vice president of the firm’s hardware department. Bai was previously head of chip development at major Chinese phone vendors Xiaomi and Oppo.
- However, Nio has yet to reveal when the chips will reach the mass production phase or be used in their vehicles.
Xpeng develops NPUs for autonomous driving chips:
- Having formed a chip unit of fewer than 200 employees, Xpeng is developing autonomous chips as alternatives to Tesla FSD chips with a computing capacity of over 100 trillion operations per second (TOPS), people related to the firm told LatePost.
- For comparison, the computing capacity of the Tesla chip Xpeng currently adopts in their vehicles has 144 TOPS and Nvidia’s newly released Thor hits 2,000 TOPS.
- However, according to the LatePost report, Xpeng is focused more on network processing units (NPUs), which can be used to accelerate AI and machine learning.
- For other modules like GPU and CPU on car processors, the firm plans to buy well-designed IP from partners such as Arm. The final design and composition work has been handed to US firm Marvell.
- A source related to Xpeng told LatePost that chip development has been slow due to Marvell’s relative lack of experience.
Li Auto focuses on power semiconductor devices:
- Li Auto is less aggressive on the chip-making front. The company is still in the early stages of the process, researching markets with a unit of a few dozen employees.
- The firm has opened some positions for AI and system-on-chip (SoC) architect engineers and AI/NPU sub-system designers with yearly salaries ranging from RMB 3 million to RMB 5 million ($420,000 – $700,000). The firm has yet to amass a large team, with its chip unit currently having about 10 people
- Li Auto has been working on more fundamental semiconductor components that power EVs and launched a product line for power semiconductor devices in March this year. Collaborating with domestic partner Sanan Semiconductor, the product line is under development and is due to be put into production in 2024.
Context: Multiple Chinese automakers have been looking to move into chip manufacturing, having been hit by the ongoing chip shortage amid growing uncertainty caused by multiple supply challenges, including the US chip export ban to China.
- BYD is by far the country’s biggest maker of automotive microcontroller chips and has recently begun trial operations at a new chip plant in the central Chinese city of Changsha, in addition to operating two existing facilities in the eastern cities of Jinan and Ningbo.
- Geely-backed Ecarx Holdings last May showcased its in-house developed seven-nanometer processors with plans to begin shipment as early as next year with partner Arm China. Potential clients include Lotus, Lynk & Co, and Volvo, all owned or associated with Geely, Reuters reported.
- EV startup Leapmotor in August began delivering its first electric crossover the C11, equipped with two proprietary AI chips dubbed “Lingxin 01,” with each delivering 4.2 TOPS at about four watts of power consumption.