Key Takeaways:
XPeng's push to run robotaxis without safety drivers by 2027 rests on a second-generation autonomous driving model that the company says cuts reliance on high-definition maps, a bet that pits the Guangzhou-based automaker against Tesla, Pony.ai and WeRide in China's fast-consolidating driverless ride-hailing market.
"To accelerate the commercialization of physical AI, XPeng recently formed a group-level business development team and is in active talks with domestic and overseas partners to bring the Turing AI chip, the second-generation VLA model, Robotaxi and humanoid robot technology to global markets faster," Chairman and CEO He Xiaopeng said, according to Securities Times.
The pre-production Robotaxi equipped with VLA 2.0 has completed more than 2,000 internal test orders in Guangzhou, running the full passenger-carrying demonstration process, He said. The company has finished developing its cloud-based takeover platform and targets safety-driverless passenger operations next year. XPeng formally established a dedicated Robotaxi Business Unit on March 23, 2026, tasking the tier-one division with product definition, project integration, R&D testing and daily operations.
VLA 2.0 processes environmental data from seven cameras at 5,000 inputs per second, with inference speed twelve times that of the prior-generation system, according to the company. XPeng announced three robotaxi models at its AI Day event in November 2025 — a five-seater, six-seater and seven-seater, all priced under 200,000 CNY (about $29,100) — each carrying four in-house Turing AI chips delivering up to 3,000 TOPS of compute. The vehicles use a dual-redundancy hardware architecture across perception, steering, braking, communications, energy and compute, with no LiDAR and no HD-map dependence.
XPeng's approach differs from pure-play autonomous driving companies. Rather than building a separate hardware platform, it is leveraging existing vehicle production infrastructure and sharing VLA 2.0 across its consumer lineup and robotaxi fleet. The GX, a six-seat flagship SUV sharing the same VLA 2.0 hardware, began road testing in Guangzhou in February 2026 after regulators granted L4 intelligent connected vehicle test approval, and is slated for commercial launch in the second quarter of 2026 as the first consumer model built to support L4-level hardware and software.
Volkswagen has been named the first external customer for both the Turing system-on-chip and VLA 2.0, with XPeng targeting roughly one million chip shipments in 2026. Amap, with 873 million monthly active users, is XPeng's first global ecosystem partner for booking and dispatch, and XPeng plans to open its robotaxi SDK to additional fleet operators over time.
The competitive stakes are rising. WeRide has deepened its partnership with Geely and plans to deliver 2,000 robotaxi units in 2026, with a fleet that already exceeds 1,000 vehicles operating fully driverless in Guangzhou, Beijing and Abu Dhabi. Pony.ai, partnered with Toyota and GAC Toyota, reported positive unit economics in Guangzhou and Shenzhen as of March 2026. Tesla's CyberCab rolled off the production line in February 2026 and is ramping toward several hundred units per week, though its unsupervised fleet has logged just 380,000 miles across six cities versus Waymo's more than 220 million fully driverless miles.
The robotaxi push sits inside a broader physical AI strategy. XPeng's robotics business raised more than $900 million in its first external round at a post-money valuation above $6.3 billion, led by IDG Capital with participation from Gaorong Ventures and strategic support from Tencent and Alibaba. The unit's IRON humanoid robot carries 76 degrees of freedom across its body and 21 in each hand, runs on three Turing AI chips delivering up to 2,250 TOPS of on-device compute, and is expected to enter mass production by the end of 2026 with commercial deliveries in China and overseas in 2027.
XPeng has committed 7 billion CNY to Physical AI R&D in 2026, up from 4.5 billion CNY allocated to AI in 2025. The spending comes after the company reported its first-ever quarterly net profit — 380 million CNY (about $55.2 million) in Q4 2025 — alongside full-year 2025 revenue of 76.72 billion CNY, up 87.7 percent year-over-year. XPeng targets 550,000 to 600,000 vehicle deliveries in 2026, though Q1 guidance of 61,000 to 66,000 units reflects a 30 to 35 percent year-over-year decline attributed to reduced government subsidies and seasonal factors.
Bernstein cautioned late last year that while mass production is planned for 2026, accumulating the safety mileage required for fully driverless permits could push L4 commercial operations into 2027 or later — a timeline that broadly aligns with XPeng's own stated targets. Whether the new business unit can translate VLA 2.0's technical gains into a commercially viable, scalable fleet operation remains the central question heading into the second half of the year. XPeng shares trade on the NYSE and Hong Kong exchange, and the robotaxi and robotics roadmap gives investors a second growth vector beyond the core EV business, though the near-term delivery slowdown and heavy AI spending will test margins before any driverless revenue materializes.
This article is for informational purposes only and does not constitute investment advice.
