Robots already run factory floors and warehouse aisles without much debate about whether they belong there. What's still unsettled is who leads the next phase, the humanoid machines meant to work alongside people in stores, offices, and homes. A photo of Iron standing next to a child in a Wall Street Journal newsprint spread sent me back to XPeng's numbers, and from there to Tesla's, because the same question sits under both: does that lead come from Chinese manufacturing scale or American software and compute. Comparing the two companies' actual disclosures, not their demo reels, is where this piece starts.

Brian Gu put a number on Iron's margin before XPeng has shipped the robot in volume. The bipedal humanoid, which the company describes as built with electronic skin and a bionic spine, could deliver a hardware gross margin above 50%, Gu told the Wall Street Journal, against the 12.1% vehicle margin XPeng posted in the second quarter. The robotics unit, known internally as Dogotix, is now valued above $6.3 billion after its first outside raise. XPeng says mass production starts this year.

The Margin Number Came First

Gu's case rests on reuse, not on Iron proving itself as a new category. Roughly 85% of Iron's supply-chain partners overlap with XPeng's auto network, which cuts the factory spend a robot-only startup would have to absorb from nothing. Software and model updates on units already in the field are meant to be the recurring revenue layer stacked on top of the hardware sale.

That comparison depends on which margin figure you hold constant. XPeng's company-wide gross margin was 20.7% in the second quarter, lifted by services revenue. The vehicle segment alone, the actual comparator to Iron hardware, was 12.1%. If Iron reaches Gu's projected 50%, robotics becomes a materially higher-margin version of the same plant network XPeng already runs. That figure has not been printed yet. It is a forecast, not a result.

The unit is not profitable.

On XPeng's Q2 earnings call, Gu said it was too early to give a profitability date or a volume guide. Training and compute for tasks like staffing a desk or making coffee, he told the Journal separately, will likely cost more than developing a vehicle.

Who Put the Money In

Dogotix took more than $900 million in what XPeng calls the largest single private raise yet in China's embodied AI sector. IDG Capital led the round. Gaorong Ventures participated. Alibaba Group and Tencent Holdings came in as strategic investors. External money accounts for about $600 million of the round. An XPeng subsidiary added roughly $200 million. Entities controlled by He Xiaopeng and Gu together put in $100 million, alongside warrants for additional investment.

Jefferies analysts treated the round as a valuation anchor, a way for investors to price XPeng's robotics business apart from the blended multiple the stock trades on today (Jefferies, 2026).

XPeng still consolidates Dogotix and keeps control of it. The deal documents set an 18-month window for XPeng to work toward a corporate separation of the unit, not a deadline for the robot itself to operate independently. Gu said there is no pressure to spin it off while the company is focused on production and commercialization.

XPeng has priced a projected margin. Tesla has not priced a start date.

Tesla Dropped the Summer Date

Tesla is chasing the same market at a larger scale, and its calendar has gotten looser rather than firmer. The company finished tearing down its Model S and Model X assembly line at Fremont in 46 days this summer, converting the space into a dedicated Optimus production line. Elon Musk told investors in April that limited production would begin in late July or August, and warned output would be slow given that Optimus carries roughly 10,000 unique parts across an entirely new line.

That date did not hold. Tesla's Q2 shareholder update, published July 22, confirmed the Fremont lines were fully decommissioned and that first-generation Optimus lines were being installed, but it replaced the summer month with the broader phrase "anticipated later this year." No start date, unit count, or price has been disclosed, and Tesla's regular production reports have not carried an Optimus line item. Early units are described as going to an internal training program rather than to customers.

Musk has also delayed the reveal of the production-intent Gen 3 robot, a decision he attributed to competitors copying the design from released footage. A second Optimus factory under construction at Giga Texas is targeted for 2027, aimed at a higher-volume generation of the robot.

Same Market, Different Order of Operations

XPeng's first Iron customers are its own stores and campuses. Small commercial and retail deployments follow, with external sales targeted for 2027. Tesla is building a line sized for a million units a year before it has confirmed a robot it will sell to anyone outside the company. XPeng is testing its margin model at a volume its existing plants can already attempt.

Unknowns and Uncertainties

XPeng's $900 million round is a set of committed agreements, not necessarily cash already sitting in the company's accounts, and the timing of when it closes has not been independently confirmed. Separately, a Motley Fool piece published August 27 characterized Optimus as having "just entered production" at Fremont. Tesla has not confirmed that in a production report or an official statement as of this writing, and the claim should be treated as inference until Tesla says otherwise.