is ramping up its efforts to develop chips in-house, working on both an AI inference chip for smart driving scenarios, and a SiC power chip for driving motor controllers, according to local media.

(Image credit: CnEVPost)

Li Auto (NASDAQ: LI) is ramping up its efforts to develop chips in-house, working on both an AI inference chip for smart driving scenarios and a silicon carbide (SiC) power chip for driving motor controllers, according to a local media report.

The company is currently building a team in Singapore to work on SiC power chips, LatePost said in a report today, citing people familiar with the matter.

Li Auto recently posted five Singapore job openings on LinkedIn, four of which were related to SiC chips, according to the report.

Notably, Li Auto said on August 24, 2022, that it had started construction of its power semiconductor R&D and production base in Suzhou, Jiangsu province, which will focus primarily on the development and production of SiC power modules, aiming to build in-house capabilities to design and produce the modules for automotive use.

The production base was built by Suzhou Sike Semiconductor, a joint venture between Li Auto and Hunan Sanan Semiconductor, a leading semiconductor company in China.

Sample trials would begin in the first half of 2023, and production capacity was expected to ramp up to 2.4 million SiC half-bridge power modules per year in 2024, Li Auto said at the time.

In addition to SiC chips, Li Auto is also developing AI inference chips for smart driving scenarios, according to a LatePost report today.

AI inference chips are the current focus of Li Auto's chip development, the report said, adding that the company is working on a SoC (System on a Chip), of which the most critical aspect is the front-end design of the inference model acceleration unit NPU.

Li Auto plans to outsource the back-end design part to Alchip Technologies and let TSMC do the manufacturing, according to the report.

Li Auto's chip division currently employs more than 160 people in Beijing, Shanghai, Silicon Valley and Singapore, according to the report.

On the R&D side, car companies need to focus on how costs can continue to come down, the report said, citing Li Xiang, Li Auto's founder, chairman and CEO, at a media conference earlier this year.

If Li Auto develops its own inference chip, it can realize a cost as low as 's, he said.

Tesla's cost on hardware for smart driving is $1,500, Li Auto's is $4,000, Li said, according to LatePost.

Li Auto's local counterpart (NYSE: NIO) is also developing its own chips and unveiled its first chip, Yangjian, at the Nio IN 2023 innovation day event on September 21.

Yangjian is a LiDAR master chip that Nio said at the time would be in mass production in October.

The chip could allow Nio to save about $100 per vehicle, Li said in an interview with English-language media, including CnEVPost, after the September 21 event.

Nio unveils 1st in-house developed chip, mass production to begin in Oct