On 8 September 2026, XPENG officially commissioned its IRON humanoid robot production line. The robot completed assembly and walked out of the factory under its own power. That is not a demonstration. That is a product.
The distinction matters. The robotics industry has a long history of impressive prototypes that never make it to manufacturing. IRON is now manufactured.
What Happened on 8 September
XPENG opened its humanoid production facility with a system built for scale from the start. More than 80% of core manufacturing processes are automated, a figure that puts most automotive plants to shame. The line is described as high-precision, highly flexible, and intelligent, drawing directly on XPENG's existing expertise building electric vehicles at volume.
The moment that defined the announcement was IRON completing assembly and walking off the line under its own control. Not being wheeled out. Not carried. Walking. The symbolic weight of that detail was entirely deliberate, and it was entirely earned.
Mass production is targeted for the end of 2026. External customer deliveries in China and overseas are planned for 2027.
The Specifications That Define IRON
XPENG has not positioned IRON as a warehouse workhorse or a logistics mule. The dexterity story is the headline.
Seventy-six degrees of freedom across the full body, with 21 in each hand alone. For context, human hands have roughly 27 degrees of freedom. IRON is not far behind, and it is dramatically ahead of most competing platforms in the industry.
The compute stack matches the ambition. Three in-house Turing AI chips deliver up to 2,250 TOPS of effective computing power. That is enough to run sophisticated on-device inference without depending on cloud processing, which matters enormously for real-world reliability. A robot that needs a network connection to decide how to pick up an object is not an industrial tool. IRON is built to work offline.
The production line itself reflects the specification quality. Automotive-grade manufacturing means the tolerances, consistency, and supply chain management that XPENG has built over years of EV production are directly applied to a humanoid platform. That is a structural advantage most robotics companies simply do not have.
Why This Beats Tesla to a Key Milestone
Tesla Optimus has received more coverage than almost any other humanoid project. It has also been the subject of more timeline adjustments than most. In 2026, Optimus remains deployed only within Tesla's own Fremont and Gigafactory environments. There is no confirmed external delivery date for third-party customers.
XPENG has a commissioned production line and a confirmed 2027 delivery timeline. That is the formal production-line milestone, and it belongs to XPENG.
The comparison is not entirely straightforward. Tesla's ambitions are enormous and Optimus is being iterated rapidly. Elon Musk has described Optimus as eventually being the most valuable product Tesla makes. But ambition and manufactured units are different things, and right now the manufacturing milestone is XPENG's.
The Deployment Strategy: Start Close, Then Scale
XPENG's rollout plan is methodical. Initial deployments are going into the company's own retail stores and campuses, controlled environments where XPENG can gather real-world operational data without the unpredictability of a third-party site. The company sees this phase as a systematic proving ground rather than a marketing exercise.
The official launch for China and overseas markets follows in 2027. Specific pricing has not been announced. Given XPENG's automotive-scale manufacturing, most industry analysts expect the company to push pricing below what pure-play robotics manufacturers can match, though the final figure will depend heavily on specification tier and deployment volume.
This staged approach reflects a broader lesson the industry has learned from early humanoid deployments. Companies that rushed external deployment without sufficient internal testing created operational problems that were slow to resolve and expensive to manage. XPENG is using its own estate as a proving ground deliberately.
How China Is Winning the Humanoid Production Race
The September 2026 humanoid landscape is dominated by Chinese manufacturers in a way that was not fully anticipated even two years ago. Global H1 2026 humanoid shipments surged approximately 272% year-on-year to around 19,000 to 22,000 units, with Chinese makers holding roughly 93 to 97% of that volume.
AgiBot shipped approximately 8,400 units in H1 2026, around 44% of the global market, on a 562% year-on-year jump. Unitree came second with roughly 5,900 units. Both companies sell into research, industrial, and increasingly commercial applications at prices that Western competitors have not matched.
XPENG IRON sits above both in specification and ambition. It is targeting enterprise commercial deployment rather than the volume research market. But its manufacturing infrastructure shares the same advantage as AgiBot and Unitree: automotive-scale production knowledge, domestic supply chain depth, and government support for the sector.
Western humanoid companies are building excellent robots. They are building them more slowly and at higher cost, for now.
What IRON Means for the Used Robot Market
IRON reaching production-line status has implications that extend beyond the XPENG business case.
Manufactured products, unlike prototypes, eventually reach secondary markets. When first-generation IRON units are superseded by a second generation, they will be sold on. Pricing data will form. Condition grading will develop. Resale values will become trackable.
That is when a product like IRON becomes part of the Robot AutoTrader story. We are already building the monitoring infrastructure to track IRON's market entry and will add it to the robot database as specification and pricing data becomes confirmed.
Alternatives Available Today
IRON is not yet purchasable. For buyers looking at the commercial and industrial automation space in the UK today, there are established alternatives with track records, available inventory, and real resale data.
The used cobot and industrial robot buyers guide covers platforms like the Universal Robots UR5e and UR10e, which are proven in production environments, available used at 40 to 60% below new prices, and fully supported in the UK.
The commercial-industrial category on Robot AutoTrader lists current used units available for immediate purchase. The Universal Robots UR5e resale page tracks current pricing on one of the most actively traded used cobots in the UK market.
For buyers who want to monitor where humanoid pricing is heading, the Robot Resale Index tracks depreciation curves across robot categories, including early humanoid data as it becomes available.
IRON represents where the market is going. The listings above represent where you can buy today.