Xpeng Brings the Humanoid IRON to the Production Line
Xpeng has put a new production line for its IRON humanoid robot into operation in Guangzhou. At the opening, one unit walked out of the facility under its own power. According to the company, more than 80% of the core production processes are automated.
This takes Xpeng an important step beyond development prototypes and stage demonstrations. The facility is intended to apply quality processes from automotive manufacturing to humanoid robots. Xpeng describes it as the world’s first automated production line for advanced humanoid systems, although this claim has not yet been independently confirmed.
A functioning production line is an important milestone, but it is not yet proof of economically viable mass production or reliable real-world operation.
What Is Inside the Xpeng IRON
According to Xpeng, IRON has 76 degrees of freedom in its body and 21 in each hand. This is intended to give the robot comparatively precise gripping capabilities and more fluid movement. A fully enclosed, flexible mesh structure serves as its outer skin and is also designed to reduce the risk of injury upon contact.
IRON uses three Turing AI chips developed by Xpeng for its computing needs. Together, they are said to deliver up to 2,250 TOPS of computing power. This figure describes the theoretical number of computing operations possible per second, but on its own says little about how reliably the robot can perform real-world tasks.
The Physical AI foundation model is intended to run directly on the robot. Xpeng says this will enable autonomous operation without continuous remote control by a human. Whether IRON can actually perform complex tasks independently and safely over extended periods remains to be seen in everyday use.
IRON Will Initially Work for Xpeng Itself
Under the current plan, mass production is scheduled to ramp up by the end of 2026. The first units are intended for Xpeng stores and the company’s own sites. This internal deployment will likely serve primarily to test processes, collect error data, and improve the software under controlled conditions.
Commercial sales in China and other markets are planned for 2027 at the earliest. There is no specific launch date for Germany, Austria, or Switzerland. In the German-speaking DACH region—Germany, Austria, and Switzerland—IRON therefore remains a technology project with no purchase option in sight for now.
In the longer term, Xpeng has set significantly larger targets. It initially aims to produce more than 1,000 robots per month, with as many as one million units per year being discussed for 2030. The road to that goal is formidable: not only will production and the supply chain need to scale, but enough economically viable applications will also have to emerge.
Xpeng IRON and Tesla Optimus Compared
Xpeng’s progress is often compared with Tesla Optimus. Tesla had at one point suggested a target of around 10,000 robots for 2026. At the beginning of the year, however, the company acknowledged that its existing systems were not yet performing any practically useful work and that the production ramp-up in Fremont would likely be slow at first.
| Project | Current Production Status | Next Announced Step |
|---|---|---|
| Xpeng IRON | Dedicated line in Guangzhou; according to Xpeng, more than 80% of core processes are automated | Mass production by the end of 2026, commercial deliveries from 2027 |
| Tesla Optimus | Production preparations in Fremont, with a slow ramp-up announced | Establishment of production and presentation of the next generation |
The table shows different stages of progress, but no clear winner yet. Neither Xpeng nor Tesla has delivered large numbers of humanoid robots to paying customers or demonstrated their productivity over an extended period. A single completed unit is no more mass production than an ambitious production target is.
Tesla is continuing to work on Optimus, AI, and its own computing hardware in parallel. Our article on Tesla Optimus, FSD v15, and HW4 provides an overview of its most recently announced plans. Xpeng also closely links robotics with vehicle software, as demonstrated by the autonomous ferry maneuver using Xpeng VLA 2.0.
Why the Production Line Matters More Than the Walking Demonstration
Humanoid robots can be presented quite effectively in short demonstrations. Producing thousands of identical units with reliable hardware, safe software, and reasonable costs is much more difficult. This is precisely where it will be decided whether an impressive machine becomes an economically relevant product.
Through its electric vehicle manufacturing operations, Xpeng already has industrial processes that could help it establish robot production. The next real test, however, is not the robot’s walking motion but its practical value in stores and on company premises. Reliable tasks performed for many hours → valuable automation; mere presence without productive work → expensive demonstration technology.
The announced margins also remain only a target for now. Xpeng expects a significantly higher gross margin per robot than for its electric vehicles, but it has not yet delivered a single unit to a paying customer. The new line makes IRON more tangible, but the decisive proof will only begin with the series-production ramp-up.
