On September 7, 2026, Chinese electric-vehicle maker XPeng commissioned what it describes as the world’s first automated production line built specifically for advanced humanoid robots, and its IRON humanoid robot walked off that line under its own power, marking the project’s move from prototype to actual manufacturing.

What IRON actually is

IRON is XPeng’s general-purpose humanoid robot, first unveiled in November 2025 and significantly upgraded since. The current version has 76 degrees of freedom across its body, plus 21 more in each hand, giving it a range of motion well beyond the rigid, limited-joint humanoids most people picture when they hear “robot.” It’s powered by three of XPeng’s own Turing AI chips, delivering a combined 2,250 TOPS (trillion operations per second) of computing power, enough to run XPeng’s physical AI foundation model directly onboard the robot rather than relying entirely on a cloud connection.

That onboard compute matters for a practical reason: a robot operating in a factory, warehouse or home needs to react to its physical environment in real time, and a round-trip to a remote server for every decision introduces exactly the kind of latency that makes a robot clumsy or unsafe around people and machinery.

Why the production line itself is the actual news

Plenty of companies have shown humanoid robot prototypes walking, waving, or performing choreographed demos. What XPeng announced on September 7 is a different kind of milestone: a production line built to manufacture these robots at scale, with more than 80% of core processes automated. XPeng describes the line as combining automotive-grade quality control systems, the kind of manufacturing discipline it already applies to building electric vehicles, with the precision assembly a humanoid robot requires.

That combination is the real reason this story is bigger than a typical robotics demo. XPeng is not a robotics startup bolting together a single show unit. It’s an automaker applying a manufacturing discipline it already runs at real production volume to a fundamentally different, more mechanically complex product. The company has reaffirmed plans to begin mass production of IRON by the end of 2026, with an official launch and deliveries in China and overseas markets targeted for 2027.

Why this comparison to Tesla’s Optimus keeps coming up

Several outlets covering this story have framed it explicitly against Tesla’s Optimus humanoid robot program, describing XPeng’s production milestone as arriving while Tesla’s own humanoid effort has reportedly stalled on its production timeline. Both companies are approaching the same basic bet from the same starting point, an automaker’s manufacturing expertise applied to humanoid robotics, which makes the comparison a genuinely useful one rather than just a headline hook.

The practical difference the comparison highlights is timeline and execution: a working, automated production line that a robot has physically walked off is a concrete, verifiable milestone in a way that a stated production target or a demo video is not. Whether XPeng actually hits its year-end mass production target remains to be seen, but the milestone it announced this week is real and independently reported, not a projection.

Where a robot like IRON would actually be used first

Humanoid robots at this stage of development are not being aimed primarily at consumer households. The near-term target markets that companies like XPeng, Tesla and others in this space consistently describe are structured industrial environments: factory floors, warehouses and logistics centers, where the tasks are more repetitive and the environment more controlled than a home, and where the economic case for replacing or supplementing manual labor is more immediate and easier to quantify.

XPeng’s own factory floor is a plausible first deployment environment, both as a practical use case and as a public proof point: a company willing to use its own robots in its own production line is making a stronger claim about reliability than one that only shows robots working in controlled demo settings.

Common myths about humanoid robot production

Myth: a robot walking off a production line means mass production has already started. Commissioning a production line and beginning mass production are different milestones. XPeng’s own stated timeline targets year-end 2026 for mass production and 2027 for actual deliveries, meaning IRON walking off the line demonstrates the manufacturing process works, not that robots are yet being built and sold at volume.

Myth: humanoid robots this advanced can already assemble other robots without human oversight. While automated processes handle the large majority of assembly steps, “more than 80% of core processes automated” explicitly means a meaningful share of the process still involves human oversight or intervention, not a fully lights-out, human-free factory.

Myth: this is XPeng’s first robotics product. XPeng has been developing robotics and autonomous driving technology for years as part of its core EV business; IRON represents an expansion into humanoid form factors specifically, building on infrastructure and AI chip development the company had already invested in for its cars.

Frequently asked questions

When will XPeng’s IRON robot actually go on sale?
XPeng has targeted mass production by the end of 2026, with an official launch and deliveries in China and overseas markets planned for 2027. The current milestone is the commissioning of the production line itself, not the start of consumer or commercial sales.

How is IRON different from Tesla’s Optimus robot?
Both are humanoid robots built by automakers applying vehicle manufacturing discipline to robotics, but as of September 2026, XPeng has a working, automated production line with a robot that has physically walked off it, a concrete milestone reports describe Tesla’s Optimus program as not yet having reached on the same timeline.

What can IRON actually do with its 76 degrees of freedom?
That figure describes the number of independent points of movement across IRON’s body, excluding its hands, which have 21 more each. In practical terms, it gives the robot a much closer approximation of human range of motion than earlier, more rigid humanoid robot designs, which is a prerequisite for tasks requiring fine manipulation or complex physical movement.

For more on the humanoid robotics and automation space, see our companion piece The Robots Are Finally Leaving the Factory Floor, or our living AI Technology News Today briefing.