Detailed narrative
Rebrand to a physical AI company
In Q1 XPeng formally changed its official Chinese name from XPeng Motor to XPeng Group (ASR renders this 'Xplan Motor to Xen Group'; HSBC's analyst confirms the change to 'XPeng Inc.'), marking its transformation from a smart-EV company into a 'physical AI' company built on three applications: VLA 2.0 intelligent driving, robotaxis, and humanoid robots.
The CEO argued the scaling law holds in both autonomous driving and robotics, making accelerated AI R&D investment the path to breaking technical ceilings, and called physical AI one of the most significant global strategic opportunities of the next decade.
The roadmap sees B2B markets taking off first and international markets generating greater commercial returns than domestic, leveraging the Volkswagen partnership experience. On business-model evolution, he described three compounding effects — hardware scale economies, a software platform with network effects, and emergent multi-agent ('ant colony') network effects — while near term the priority remains scale, brand equity, and gross profit to fund R&D.
Note: all figures on the call are RMB despite USD metadata; the CEO's remarks were delivered through an interpreter.
Q1 financials: seasonal trough with widened losses
Total revenue was RMB 13.03B (-17.6% YoY, -41.4% QoQ), with vehicle sales of RMB 11.0B (-23.5% YoY, -42.3% QoQ) on lower deliveries (the transcript's '$11 billion' and similar '$' figures are ASR errors — the CFO stated he references RMB only).
Services and others revenue of RMB 2.03B grew 41.2% YoY on technical R&D services and parts/accessories, but fell 36.1% QoQ due to a significant technical-R&D-services milestone catch-up📎 recognized in Q4 2025 and no carbon-credit trading revenue this quarter. Loss from operations widened to RMB 1.87B (vs RMB 1.04B YoY and RMB 0.04B in Q4 2025) and net loss was RMB 1.78B (vs RMB 0.66B net loss YoY and RMB 0.38B net profit QoQ), driven by R&D expansion.
Management emphasized supplier payment terms remain at an industry-leading level and that gross margins show strong resilience against cost pressures as EV scale economies strengthen.
GX launch and the 2026 product cycle
The GX flagship SUV launched May 20 as a model 'built for the L4 era' — China's first pre-installed, mass-produced robotaxi-capable model with full hardware redundancy — and is performing above expectations: the Ultra flagship trim above RMB 350,000 took over 80% of initial firm orders with lead times past 30 weeks, while the MAX version is under 5% of mix (below expectation) and the extended-range (EREV) version is approaching BEV-version popularity in western and northern China.
Management is working with supply-chain partners to ramp GX capacity and stressed that from this year all new vehicles are priced and configured with commercial value as a key priority, targeting long-term stable sales rather than a big launch followed by decline.
April's 2026 Mona M03 refresh (MAX with Turing AI SoC, Ultra supporting VLA) completed the Turing SoC upgrade across the lineup; three more models launch in H2, all with Turing SoC and VLA 2.0 and all built for global markets, plus a plan to launch and begin delivering all-new SUV models within the next six months starting with the GX.
International expansion and localized production
The April overseas launch pushed monthly international deliveries above 6,000 units for the first time, and international reached close to 20% of group volume in the latest month versus roughly 10% of volumes in 2025. International vehicle sales generate 'significantly better' gross and net profit contribution than domestic despite tariff issues and cost increases this year.
XPeng has established three localized production bases since last year — Indonesia and Malaysia (serving local Southeast Asian demand) and the Magna partnership in Austria (serving Europe) — all three adding capacity and new models this year, with the majority of European sales expected to be locally manufactured; Brian Gu spoke from the Austrian partner site while securing capacity for expected European and global growth.
The Munich R&D center is the company's fastest-growing research hub, and management is evaluating further localization in other large markets to satisfy local-content rules.
VLA 2.0 and the autonomous-driving roadmap
In April, ADAS mileage penetration on VLA 2.0-equipped XPeng vehicles surpassed 50% for the first time, which management reads as advanced intelligent driving becoming a must-have purchase driver. VLA's second version arrives in Q3 (August) — smarter, calmer, better generalization with lower human-takeover rates — versus VLA 1.0's focus on basic safety and engineering; combined with the VLM model it should approach a 'butler-like' multilingual conversational experience, delivered via three OTA releases in August and by year-end.
The stated end-goal is L4 software capability on L2 hardware, which management expects to transform the business model (details withheld). VLA operates without HD maps, enabling rapid overseas deployment; current cloud VLA models see about 200 million uses per hour (verbatim; possibly ASR-garbled units), and full cloud processing would consume ~100 GB of data per hour — a key reason the humanoid robot is designed for local/edge deployment.
Goal: undisputed #1 in domestic ADAS and a critical step toward global L4 leadership.
Robotaxi strategy: partner-operated, post-2028 commercialization
Robotaxi exploration is currently limited to Guangzhou, where XPeng holds a license; after validating technology, product, and business model there, it will expand through partnerships across China and internationally, with many inquiries already received. XPeng will focus on the product and take a commission working with operating partners rather than running operations itself.
In 2027 it plans to launch an economy car model to demonstrate and validate the robotaxi business model, with the major commercial opportunity expected after 2028. Recent tightening of China's autonomous-vehicle regulation 'hasn't placed any adversity to our rhythm of development.' Management sees B2B robotaxi work benefiting the B2C business — speed modes and low-intervention driving strategies borrowed from robotaxi development — and believes its offering beats incumbent robotaxi companies on generalization, cost efficiency, and scalability, since the GX's full-redundancy hardware/software is decoupled from the vehicle platform and deployable across the lineup including Mona.
Humanoid robot program: mass production by year-end
The mass-production humanoid robot (name garbled in transcript as 'Ron'/'RM'/'Aram'/'iron rain') is entering the software-hardware integration stage, built to automotive-grade safety and reliability standards with many existing auto supply-chain partners becoming component suppliers. XPeng recently completed a proprietary next-generation dexterous hand that is significantly more agile at substantially lower cost, and claims to be the only robotics company in China with full-stack in-house capability from SoCs and physical-AI foundation models to data generation, training, hands, and motion control — everything self-developed except batteries.
Today the robot's cost structure is similar to a car's, but its retail price will naturally exceed a car's, giving superior hardware margins plus software/licensing revenue on top; management believes payback periods for business buyers will be much shorter overseas, making international the more attractive market.
A next-generation robot will be showcased in Q3 with multilingual communication and human-like motion; mass production is targeted by year-end with initial trial deployment in XPeng stores (as guides/assisted shoppers, later cashier-type roles with ecosystem partners) and commercial customer deliveries in China and overseas in 2027.