XPeng's IRON humanoid robot walked off a production line under its own power on September 8, marking what the company described as a shift from R&D prototyping to production-line manufacturing3. IRON features a fully enclosed flexible lattice design with 76 degrees of freedom across the body, including 21 in each hand, and runs a physical AI large model deployed at the edge on three of XPeng's self-developed Turing AI chips delivering 2,250 TOPS of effective computing power.
XPeng has begun operating what it calls the world's first automated production line for advanced humanoid robots at its Guangzhou plant, with more than 80% of core processes automated1,2. The company said it applied automotive-grade manufacturing capabilities to robot production. XPeng aims to start mass production by the end of 2026, targeting 1,000 units per month.
On August 8, XPeng CEO He Xiaopeng placed an employee badge reading "IRON" around the 178-centimeter robot's neck, a gesture the company framed as onboarding a new worker. Earlier this year, XPeng had already begun trials in controlled scenarios such as store guidance and campus services to validate reliability, task performance, and operational costs.
Supply-chain overlap and financing
A central element of XPeng's cost argument is parts commonality: over 85% of suppliers for IRON overlap with the company's existing automotive parts suppliers, a figure He Xiaopeng cited at last month's second-quarter earnings briefing. XPeng's robotics division raised $900 million from investors including IDG Capital, Tencent, and Alibaba Group Holding in August, a round described as the largest single-round private fundraising in China's physical-AI sector, at a post-money valuation exceeding $6.3 billion. The funds will be used for R&D in hardware and software, physical AI model training, and the build-out of mass-production capabilities.
XPeng plans to deploy IRON first at its own retail stores for guide and information services, then begin commercial sales inside and outside China in 2027.
Automaker convergence on humanoid production
XPeng is not alone in repurposing EV infrastructure for robots. Tesla earlier in the first half converted Model S and Model X production lines at its Fremont, California plant into Optimus production lines and is reportedly conducting supplier qualification reviews with Chinese EV parts vendors ahead of Optimus's year-end launch. Tesla is also pushing to design and produce actuators and AI semiconductors for Optimus in-house, and its next-generation AI5 chip completed design in April with mass production slated for next year. Tesla CEO Elon Musk unveiled a plan to build a factory called Terafab to produce AI chips directly.
Hyundai Motor Group is building initial annual production capacity of 30,000 units for its ATLAS humanoid, with deployment to manufacturing sites planned starting in 2028. The group is distributing roles across affiliates: Hyundai Motor and Kia provide automotive production infrastructure and process-control technology, Hyundai Mobis supplies actuators, and Hyundai Glovis handles logistics and supply-chain management.
ANALYSIS The common thread across all three programs is the bet that existing automotive supply chains, factory tooling, and volume-manufacturing discipline can compress the timeline and cost curve for humanoid robots in ways that pure-robotics startups cannot replicate. XPeng's 85% supplier overlap figure and Tesla's direct conversion of EV lines both point to the same logic: the hardest part of scaling hardware is the supply chain, and automakers already have one.
XPeng operates three factories in Zhaoqing, Guangzhou, and Wuhan with annual EV capacity of 600,000 units and sold roughly 430,000 vehicles last year, ranking third among EV-specialist companies. The company is also pursuing L4 robotaxis without safety operators starting next year, an effort it advanced on October 8 by branding its robotaxi YOYO and opening public registration in China[1].