At the 18th Xuanyuan Automotive Bluebook Forum in Guangzhou, held from May 15 to 17, 2026, XPeng’s founder and CEO He Xiaopeng delivered a prediction that sent ripples through the automotive and technology industries: fully autonomous vehicles, once considered a distant dream, could hit the roads far sooner than many experts anticipated. In a notable shift from his earlier skepticism, He stated, “The probability of achieving Level 4 autonomous driving software capabilities by 2028 is extremely high,” and went so far as to predict that an initial prototype of Level 5 autonomy could emerge around 2030.
For context, the distinction between Level 4 and Level 5 autonomy is not just technical jargon. According to the framework developed by SAE International and adopted globally—from the US Department of Transportation to automakers and insurers—Level 4 represents “high automation,” where the vehicle can handle all driving tasks within a defined domain, such as a city center or specific highway routes. If the system encounters a situation it cannot handle, it safely brings the vehicle to a stop without any need for the driver to intervene. Waymo’s robotaxis in Phoenix and San Francisco are current real-world examples of Level 4 systems. Level 5, on the other hand, is the holy grail: a car that can drive anywhere, under any conditions, with no steering wheel, no pedals, and absolutely no human intervention. As of today, no commercial vehicle has reached Level 5.
He Xiaopeng’s optimism is rooted in the rapid advances in artificial intelligence (AI) and the “flywheel effect” of data-driven development. “The core competitive factors in the automotive industry are undergoing a fundamental transformation,” He said at the forum, emphasizing that AI and data capabilities are now the decisive competitive advantages. He admitted that, until recently, he believed Level 5 autonomy would not be achievable in his lifetime. However, after restructuring XPeng’s AI research and development framework, he witnessed the evolution speed of the company’s autonomous driving technology increase sixfold. This breakthrough compressed XPeng’s internal timelines for deploying advanced self-driving capabilities.
XPeng’s commitment to AI is not just talk. In 2025, the company invested 9.5 billion yuan ($1.4 billion) in research and development, with nearly half—4.5 billion yuan ($652 million)—allocated specifically to AI. Looking ahead, XPeng plans to ramp up AI-related R&D spending to 7 billion yuan in 2026. “Pushing the boundaries of physical AI is very exciting for my team and me,” He remarked during the company’s fourth-quarter earnings call earlier this year. “We’re more committed than ever to intensifying our R&D investments.”
One of the most tangible results of this investment is XPeng’s second-generation VLA (Vision-Language-Action) autonomous driving system. The company has doubled the on-vehicle model’s complexity, expanding from 10 billion to 20 billion parameters. The target? A five- to tenfold improvement in safety disengagement mileage—a key metric for how far a self-driving car can go before a human must take over. VLA 2.0 is set to receive at least one major over-the-air update each quarter, broadening its reach from highways and city roads to smaller streets, parking lots, and even campus environments.
XPeng’s ambitions don’t end at private vehicles. Last November, the company announced plans to launch three robotaxi models in 2026—a 5-seater, a 6-seater, and a 7-seater—each powered by a vision-only solution and four in-house Turing AI chips delivering up to 3,000 TOPS (trillions of operations per second) of compute power. Manned pilot operations with safety drivers are set to begin in the second half of 2026, with a target for fully driverless operation in early 2027. Volkswagen Group, recognizing the leap in technology, has agreed to adopt XPeng’s VLA 2.0 autonomous driving solution, expanding their collaboration beyond software co-development for VW’s China lineup. XPeng is also preparing for gradual beta testing and deployment of VLA 2.0 in overseas markets by the end of 2026 into early 2027, with initial international tests already showing promise—even without local road data.
But how does XPeng’s technology stack up against the competition? In a widely discussed real-world showdown, two prominent U.S.-based Tesla enthusiasts, auto vlogger Rich Rebuilds and his collaborator Joe, traveled over 10,000 kilometers from the U.S. to China to compare Tesla’s Full Self-Driving (FSD) system with XPeng’s VLA 2.0. The test, documented in a video posted by XPeng’s official account on X, pitted the two systems against each other on identical routes through dense urban traffic, scooter-filled streets, pedestrian crossings, and all the unpredictability of Chinese city driving.
The results? According to XPeng’s video, the XPeng P7 with VLA 2.0 required only two driver takeovers across the tested routes. In contrast, the Tesla Model 3 with FSD needed seven takeovers. The footage shows XPeng’s system smoothly navigating around scooters, bicycles, parked vehicles, and oncoming traffic, while Tesla’s system hesitated, attempted incorrect turns, or required manual intervention at congested intersections and construction zones. XPeng framed the test as a fair, same-road comparison under “real-world conditions” and “the same chaos,” positioning its VLA 2.0 technology as more adept at handling the complex, high-density traffic of Chinese cities.
Reactions online were mixed. Some users praised the demonstration as evidence that XPeng is pulling ahead in the autonomous driving race. Others questioned the fairness of the test, pointing out that Tesla’s FSD rollout in China uses restricted or older versions compared to what’s available in the U.S. Tesla, for its part, is actively seeking approvals to expand FSD into Europe and China. Just last month, the Netherlands became the first European country to approve the technology.
The stock market’s response has been telling. Retail sentiment around Tesla stock remained “extremely bullish” in the past 24 hours, with message volume at high levels. Meanwhile, sentiment around XPeng stock was “bullish,” though not quite as exuberant. Tesla shares have soared about 28% this year, while XPeng’s have fallen by 23%—a reminder that technological leadership doesn’t always translate immediately into investor enthusiasm.
Beyond the technology, XPeng is rethinking its identity. He Xiaopeng explained at the forum the reasoning behind the company’s recent renaming from XPeng Motors to XPeng Group. The move reflects a reassessment of the company’s business boundaries, as the lines between automobiles and robotics blur. “Modern electric vehicles are evolving from traditional physical products into hybrid entities that combine both the physical world and digital world,” He said, explaining that this shift drove changes to XPeng’s internal organizational structure and broader strategy.
As the auto industry moves beyond simple electrification into a new era defined by intelligence and robotics, XPeng’s vision—and the rapid progress of its AI-driven systems—signal a fundamental reshaping of the sector’s competitive landscape. The next few years promise to be a defining chapter in the story of self-driving cars, with XPeng, Tesla, and others racing toward a future that, not so long ago, seemed out of reach.