Li Auto Officially Establishes R & D Center in the US with Increased Investment in Intelligence | Exclusive from 36Kr

The ideal R & D system expands its territory in Silicon Valley.

36Kr has learned from multiple sources that Li Auto has officially established an AI R & D center in Silicon Valley, which will be responsible for the R & D of intelligent technologies and started recruiting personnel several months ago.

People familiar with the matter told 36Kr that Li Auto previously had a small R & D team in North America to support chip R & D and other intelligent - related work. "This time, the Silicon Valley team will be upgraded to a real R & D center."

According to 36Kr, the expansion of Li Auto's Silicon Valley R & D center is mainly focused on the field of assisted driving, aiming to recruit high - end algorithm talents with a background in cutting - edge intelligent technologies.

Since October this year, Li Auto has successively completed the channel establishment and product launch in four important international markets such as Egypt and Kazakhstan. In Li Auto's overseas expansion strategy announced to the public, the global R & D layout is also crucial, with R & D centers already established in Germany and the United States respectively.

So far, Li Auto will have a total of four R & D centers globally.

In addition to the Silicon Valley AI R & D center, Li Auto's first overseas R & D center is located in Munich, Germany, and opened in January this year. This R & D center is responsible for forward - looking styling design, power semiconductors, intelligent chassis, and the pre - research of next - generation technologies for electric drive. The two domestic R & D centers are located in Beijing and Shanghai respectively, mainly responsible for core technology breakthroughs and vehicle R & D.

Silicon Valley is a global talent hub for AI technologies. Leading AI companies such as Tesla, Waymo, OpenAI, and NVIDIA are all based here. Here, Chinese companies have the opportunity to directly access cutting - edge technologies and high - end talents.

Among Chinese new - energy vehicle startups, NIO and XPeng made earlier layouts. They established R & D centers in Silicon Valley, the United States in 2014 and 2018 respectively. Li Auto's move to set up a base in Silicon Valley and focus on autonomous driving R & D means that the technological competition among Chinese intelligent electric vehicle companies has further escalated.

In 2025, the competition in automotive intelligent driving is becoming increasingly intense.

Huawei released the ADS 4.0 version this year, launching a commercial solution for L3 - level autonomous driving in high - speed scenarios. Currently, models such as the Zunjie S800 and Wenjie M8 on sale are all equipped with hardware supporting L3 - level autonomous driving functions.

Li Auto released the VLA driver large - model in March this year. Its technical principle is to jointly model vision, language, and action, introduce language models and thought chains, enabling vehicles to understand scenarios, conduct causal reasoning and common - sense judgment, thereby reducing the dependence on rule - based systems and improving the generalization ability of the system in unknown scenarios.

On September 10th, Li Auto started a full - scale push to users. However, Li Xiang, the CEO, said on a social platform that the VLA large - model pushed this time was a "crippled version" with some core functions not yet open.

The technical difficulty of VLA lies in the integration of large - models and autonomous driving. Although it has obvious advantages, it still requires coordinated breakthroughs in algorithms, computing power, and data, and continuous verification through large - scale real - world road tests before it can be mature and useful.

Domestic algorithm technology talents, especially high - end talents with a cross - background in large - models and autonomous driving, are still scarce. Undoubtedly, Silicon Valley is the key to solving these problems.

XPeng Motors, which takes assisted driving as its core competitiveness, has been deploying R & D centers in North America since 2018. Its North American R & D team is distributed in Silicon Valley and San Diego, with a total of about 100 people. Even during the period when XPeng was implementing cost - reduction and efficiency - improvement measures across the company and the recruitment quota for the North American team was restricted, this team was retained.

Now, the North American R & D team has become an important force in XPeng's assisted driving technology R & D, once forming a R & D model where the North American team is responsible for core algorithms and the domestic team is responsible for putting forward requirements and conducting tests.

According to 36Kr, in September this year, He Xiaopeng, the CEO of XPeng Motors, launched an internal mobilization, tilting the company's computing power and other resources towards the North American R & D team to support the installation and delivery of its Foundation Model in vehicles within this year.

This Foundation Model was initiated and promoted by Liu Xianming, a technical executive recruited by XPeng from Cruise. According to the information previously disclosed by XPeng, the parameter volume of its cloud - based base model reaches 72 billion, adopting the strategy of "cloud training and vehicle - end distillation".

That is, the ultra - large - scale "teacher" model trained in the cloud can be compressed and migrated to a smaller "student" model running on the vehicle - end chip through knowledge distillation technology, and finally deployed in the vehicle.

"It is expected that the parameter volume of the model deployed on the vehicle - end will be around 7 billion - 8 billion, and the internal expectations are very high." A source told 36Kr. In contrast, the parameter volume of the VLA model recently delivered by Li Auto is in the order of 4 billion.

All this is inseparable from XPeng Motors' long - term R & D investment in North America. Li Auto has also gradually realized the importance of getting close to the cutting - edge of intelligent algorithm sources and making long - term investments.

In addition, XPeng Motors also has a robot algorithm team of about 20 people in North America. After XPeng's first humanoid robot, IRON, was released in November this year, the stock price soared, indicating that the market recognized XPeng Motors' intelligent capabilities.

At the beginning of 2025, Li Xiang readjusted the company's strategic direction. In the next decade, Li Auto hopes to grow into an "AI company". Subsequently, Li Auto established an artificial intelligence technology committee, and all business units of the company will re - think their business operation logic and methods around artificial intelligence. The VLA driver large - model is Li Auto's first attempt. Establishing a R & D center in Silicon Valley is a further move for Li Auto to get closer to cutting - edge technologies.

He Xiaopeng once publicly stated that currently, core AGI capabilities such as multi - modality to world models, long - term insight, and planning are not yet mature and may require several years of iteration, or await further breakthroughs in some underlying capabilities.

The intelligent vehicle industry is changing rapidly. Whether a car company can capture these changes faster and apply them to its own business will directly affect its intelligent development. That's why Chinese car companies have all chosen to deploy in Silicon Valley to stay close to the forefront of intelligent technologies.

As the "talent war" between Li Auto and XPeng in Silicon Valley quietly escalates, the competition faced by Chinese car companies has shifted from the sales battlefield to the source of intelligent innovation.