He Xiaopeng challenges Musk

Author | Wang Xiaojun

Editor | Chai Xuchen

This year has not yet passed, and the automotive industry has already begun to prepare for next year's "battle" in the field of intelligent assisted driving.

On December 11, XPeng Chairman He Xiaopeng made a "bet" on social media: if by August 30 next year, XPeng's VLA technology reaches the overall level of Tesla's FSD V14.2 in Silicon Valley, he will set up a Chinese-style cafeteria in Silicon Valley; if he fails, the head of autonomous driving, Head Xianming (Liu Xianming), will "run naked" on the Golden Gate Bridge.

He Xiaopeng made this bet to compete with Tesla, also due to his long-term experience and observation of Tesla's FSD system.

After recently test-driving the latest FSD V14.2 version and Robotaxi in Silicon Valley, He Xiaopeng stated that FSD has progressed from a "smooth experience" of L2 advanced assisted driving a year ago to the current "reassuring stage of quasi-L4."

During the approximately 5-hour test drive, he found that standard vehicles and Robotaxi using the same model performed consistently, and although there were still flaws, it had "significantly exceeded last year's level." He acknowledged the progress of Tesla's autonomous driving technology.

He Xiaopeng believes that the future will be an era of the same autonomous driving system and hardware system, where people can directly enjoy cars with L4 capabilities. This judgment is closely related to XPeng's own technological route choice.

While publicly praising Tesla's FSD, He Xiaopeng did not hide the current gap between XPeng's VLA and FSD. He admitted, "The first version of XPeng VLA cannot fully realize all the capabilities of FSD V14.2." However, he also revealed that XPeng VLA 2.0 will be released in the next quarter.

Regarding the timeline for catching up, He Xiaopeng stated in August this year that the current first-tier players in intelligent assisted driving are "neck and neck," with some relying on computing power, while those with insufficient computing power are combining computing power with lidar. However, he predicts that by 2026, when XPeng's Robotaxi begins trial operations, the model equipped in XPeng's Ultra vehicles, which shares the same source as L4 Robotaxi, will "significantly surpass the capabilities of other intelligent assisted driving products on the market by dozens of times or more."

At XPeng's Technology Day in 2025, XPeng's intelligent driving made new progress. At that time, XPeng launched the second-generation VLA model, which innovatively removed the "language translation" step, achieving end-to-end direct generation from visual signals to action commands for the first time.

With the support of this technology, XPeng's automatic assisted driving can handle various complex small roads and park scenarios. At the same time, XPeng also launched the "navigation-free automatic assisted driving" Super LCC+ human-machine co-driving This means that the feature can be activated globally without relying on navigation. He Xiaopeng introduced that in the first quarter of 2026, the second-generation VLA will be fully pushed to the XPeng Ultra models.

He Xiaopeng emphasized that XPeng is heavily investing in the new generation of VLA large models, aiming to ensure that cars can safely navigate various complex scenarios, including major roads, small roads, highways, outdoors, and even the alleys of Beijing. He Xiaopeng believes that only by solving the problems of these scenarios can true autonomous driving be achieved.

XPeng's heavy investment in autonomous driving is based on its judgment of industry trends. He Xiaopeng believes that in the next decade, we will see larger-scale applications of L4-level autonomous driving. He pointed out that future cars will be "embodied intelligent cars," and the automotive industry and humanoid robotics industry will complement and promote each other.

In terms of hardware chips, in August this year, He Xiaopeng announced that starting from the third quarter of 2025, the Ultra versions of its models, including the XPeng G7 and P7, will fully upgrade their intelligent hardware, equipped with three self-developed Turing AI chips, with effective computing power reaching 2250 TOPS.

Meanwhile, at this year's Technology Day, XPeng's Robotaxi also made new progress. He Xiaopeng revealed that XPeng will launch three Robotaxi models in 2026 and start trial operations. Unlike Tesla, XPeng will focus on the research and development of Robotaxi products and will not be involved in the operational aspects. Currently, Gaode has become the first global ecological partner for XPeng's Robotaxi.

In terms of R&D investment, He Xiaopeng expects that the annual R&D investment of automotive companies will reach 50 billion yuan in the future, with 30 billion yuan allocated to the AI field. For XPeng, investing several billion yuan each year to build computing power centers has become the norm.

Behind this gamble is a microcosm of Chinese smart car companies challenging global leaders in the core technology of autonomous driving. With the release of the second-generation VLA and new progress in the Robotaxi project, XPeng is accelerating its pursuit in the autonomous driving race.

From the enhancement of AEB capabilities to the innovation of the VLA model, XPeng is attempting to build barriers through full-stack self-research. However, Tesla's iteration speed indicates that the competition in intelligent driving is not only a battle of algorithms but also a comprehensive contest of data, computing power, and engineering capabilities.

The end of the gamble will determine whether a Chinese-style cafeteria lands in Silicon Valley across the Golden Gate Bridge or whether the head of autonomous driving runs "naked" on the Golden Gate Bridge, depending on whether XPeng's intelligent driving system can reach Tesla's level and become a true "Chinese Tesla." At that time, XPeng may usher in a new round of revaluation.

Risk Warning and Disclaimer

The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investing based on this is at one's own risk