Xpeng L03: First Chinese EV with Proprietary AI Chips in Mass Production

What Happened?

Last Wednesday, at the Munich event, Xpeng Motors unveiled the L03, an electric coupe-SSUV that marks a milestone in the Chinese automotive industry. The vehicle launches simultaneously in 65 international markets, including Europe, Asia, and the Americas, making it the company's most ambitious global rollout to date. But what truly sets the L03 apart is its brain: the Turing AI chip, designed and manufactured internally by Xpeng. Each L03 unit incorporates at least one Turing AI chip, while the top-of-the-line version, dubbed Ultra, mounts three, achieving a combined computing capacity of 2.250 trillion operations per second (TOPS). For context, the Nvidia Drive Orin chip, used by many competitors, offers 254 TOPS per chip. Even Nvidia's upcoming Thor, announced for 2025, reaches 2,000 TOPS but is not yet in mass production. The L03 is thus the first mass-produced vehicle to surpass the 2,000 TOPS barrier, positioning it as a computing platform on wheels.

The launch was not merely a media event. Xpeng has already begun deliveries in China, and the first export units will roll off production lines in the coming months. The company has confirmed that the L03 is manufactured at its Guangzhou plant, with an initial capacity of 100,000 units per year, expandable based on demand. The price, though not officially disclosed, is estimated at around 40,000 euros for the base version, placing it in direct competition with the Tesla Model Y and Volkswagen ID.4.

Why It Matters

This launch represents a paradigm shift in the Chinese and global automotive industry. Historically, automakers relied on external suppliers like Nvidia, Qualcomm, or Mobileye for the AI chips needed for functions such as autonomous driving, infotainment, and battery management. By developing its own silicon, Xpeng breaks that dependency and gains full control over hardware-software integration. This not only reduces long-term costs but also allows performance optimization for its proprietary autonomous driving algorithms, which the company has been developing for years under the name XPILOT.

Xpeng's move is comparable to Tesla's with its Full Self-Driving (FSD) chips, but with a key difference: while Tesla designs its chips but outsources manufacturing to Samsung, Xpeng has opted for an even more vertical approach, closely collaborating with Chinese manufacturer SMIC to produce the Turing AI chips. This gives it an advantage in supply security, especially amid geopolitical tensions and export restrictions on advanced semiconductors from the United States and its allies.

Moreover, the L03 is not just a car but a vertically integrated AI platform. The Turing AI chip not only handles autonomous driving but also manages the infotainment system, 5G connectivity, and over-the-air (OTA) update functions. This means Xpeng can remotely roll out new software features, improving the vehicle over time—something Tesla and Nio already do, but now with custom-designed hardware.

Market Consequences

The impact of the L03 on the global automotive market will be significant. First, it pressures other Chinese manufacturers like Nio, Li Auto, and BYD to consider similar in-house chip development strategies. Nio has already announced its own chips for autonomous driving but has yet to integrate them into a production vehicle. Li Auto, meanwhile, has opted for an alliance with Horizon Robotics, a Chinese chipmaker, while BYD continues to primarily use Nvidia solutions. If Xpeng proves its vertical approach works, it could trigger a race in the Chinese industry for proprietary silicon, reducing reliance on foreign suppliers.

For chip suppliers like Nvidia, Qualcomm, and Mobileye, Xpeng's move represents a direct threat. China is the world's largest electric vehicle market, and losing a customer like Xpeng (and potentially other manufacturers) could affect their long-term revenues. However, these suppliers still hold advantages in terms of software ecosystems and relationships with other global manufacturers. For example, Nvidia remains the dominant supplier for brands like Mercedes-Benz, BMW, and Volvo, which are unlikely to develop their own chips in the short term.

Globally, the L03 aims to position itself as a benchmark in autonomous driving. With 2,250 TOPS, the Ultra system far surpasses current competitors: the Tesla Model Y Plaid offers around 144 TOPS (with hardware 3.0), while the Nio ET7 reaches 1,016 TOPS with four Nvidia Orin chips. However, computing power is not everything. The actual implementation of autonomous functions depends on local regulations and software maturity. In Europe, for instance, UNECE regulations limit Level 3 and higher autonomous driving capabilities, which could delay the deployment of the L03's most advanced features. In China, where regulations are more permissive, Xpeng already offers highway and city autonomous driving functions, which could be transferred to the L03.

Additionally, the L03 competes on price: the base model is expected to cost around 40,000 euros, similar to the Tesla Model Y (44,000 euros) and Volkswagen ID.4 (41,000 euros). But with an estimated range of 600 km under the WLTP cycle (not yet officially confirmed), it surpasses both: the Model Y offers 533 km and the ID.4, 520 km. If Xpeng can maintain that price and deliver a superior autonomous driving experience, it could capture a significant share of the European electric SUV market.

What Readers Should Know

• The Turing AI chip was designed internally by Xpeng and manufactured by SMIC on a 7 nm process, giving it energy efficiency and performance. Each chip consumes approximately 50 watts, well below the 100 watts of Nvidia's Orin.
• The Ultra version with three chips offers 2,250 TOPS, enough to run next-generation autonomous driving neural networks, including computer vision, sensor fusion, and route planning.
• The L03 launches in 65 markets, including Germany, France, the UK, Norway, the Netherlands, Japan, South Korea, Australia, and several Latin American countries. Deliveries in Europe will begin in the fourth quarter of 2025.
• The estimated range is 600 km under the WLTP cycle, though the exact figure has not been confirmed. The battery is a 100 kWh lithium-ion pack, with LFP chemistry in the base version and NMC in the Ultra.
• The L03's XPILOT 4.0 autonomous driving system includes 12 cameras, 5 millimeter-wave radars, and 12 ultrasonic sensors, plus an optional LiDAR on the Ultra version.

"The L03 is not just a car; it's a statement of intent about the future of intelligent mobility," said He Xiaopeng, CEO of Xpeng, during the Munich presentation. "With our own chip, we can innovate faster and offer customers an experience that no other manufacturer can match."

In summary, the Xpeng L03 represents a turning point in the automotive industry, where vertical integration of hardware and software becomes a key competitive factor. The coming months will be crucial to see if Xpeng can deliver on its promises of range, price, and autonomous features, and whether other manufacturers will follow suit. What is clear is that the race for the smart car has entered a new phase, and the Turing AI chip is the engine driving it.