XPeng's IRON walks off the world's first automated humanoid line — you still can't buy one until 2027.

On September 8, XPeng (NYSE: XPEV) said it had commissioned its humanoid robot production lines in Guangzhou and that its next-generation IRON robot completed assembly and autonomously walked off the line, marking what the company called a leap from R&D prototyping to line manufacturing. XPeng describes the setup as the world's first automated production line for advanced humanoid robots, built to automotive-grade quality standards with more than 80% of core processes automated, borrowing the quality systems it uses to build electric cars. 'The robot production lines were created from scratch with no precedent to follow,' chairman and CEO He Xiaopeng said. 'Today's step is small, but XPENG is building the production lines for an entirely new product category.' He then clipped a staff badge onto IRON, a bit of theater meant to signal the robot is now on the payroll. The reason this matters more than another walking demo: manufacturing — building a machine with thousands of parts repeatably, consistently, and cheaply — is the wall the entire humanoid industry keeps running into.

The specification sheet is genuinely strong. The next-generation IRON, first shown at XPeng's AI Day last November, has 76 degrees of freedom across its body plus 21 in each hand, wrapped in a fully enclosed flexible lattice structure that doubles as skin and safety feature. Its brains are three of XPeng's in-house Turing AI chips, which the company says deliver up to 2,250 TOPS — enough to run its Physical AI foundation model directly on the robot so that, XPeng claims, IRON performs complex tasks autonomously without a remote operator in the loop. Those are the same specs the company detailed on August 24, when its robotics arm, Dogotix, signed agreements to raise more than $900 million at a post-money valuation above $6.3 billion — the largest single private round in China's embodied-AI sector, led by IDG Capital with Tencent and Alibaba joining. XPeng's stated ambition is to build more than 1,000 robots a month on the way to a million a year by 2030.

Set against that ambition, the near-term plan is deliberately modest — and the milestone is narrower than the language around it. Commissioning the line does not mean IRON has entered mass production: XPeng reiterated that mass production begins by the end of 2026, with the first units going into its own stores and campuses to handle reception, vehicle introductions, visitor guidance, inspections and small-item handling, before any official launch and deliveries in China and overseas in 2027. The company disclosed neither the line's capacity nor the robot's price, and its promise that IRON's per-unit gross margin will run 'significantly higher' than its EVs is exactly that — a promise. As Electrek put it, one robot walking off a line is not mass production; XPeng has not shipped a robot to a paying customer or proven usefulness at scale, and every humanoid timeline in this industry, XPeng's included, has slipped.

For a buyer, IRON is the clearest evidence yet that a carmaker can point automotive-grade manufacturing discipline at humanoids — but through 2026 it remains a captive-first, look-but-don't-buy proposition. The first units are XPeng's own showroom and campus workers, not stock you can order; there is no price, no published output rate, and no external order book, and China-built humanoids are shut out of new US sale under the FCC's July import ban regardless of how good the line is. The questions that will actually tell you something by December are concrete: how many IRON units the line produces per month, whether they do useful work in those stores rather than stand around looking human, and what a unit costs. Until those are answered, the humanoids you can put to work today are still the ones with dated deployments and real prices — Agility's Digit, Figure at BMW, Apptronik's Apollo — not IRON.