Three robotics disclosures landed inside five business days this week, and read side by side they say something the humanoid pitch deck does not want said out loud. Unitree cleared final CSRC review on July 14 for a Shanghai STAR listing at an implied $6.18 billion valuation. 1X Technologies shipped the first ten Neo home units on July 16 with a launch-blog admission that most household tasks route to a human teleoperator sitting in Oslo. Boston Dynamics said on July 17 that its Stretch warehouse robot passed 1,000 deployed units in Q2, roughly twenty times 1X, Figure, and Sanctuary combined. Three companies. Three capital-market venues. Three form factors that are supposedly all racing toward the same “general-purpose humanoid” endpoint. The revenue underneath the three announcements is not going where the pitch deck is going.
Start with the shape of the money. Stretch is a tracked base with one arm and a suction cup. It is not a humanoid. It unloads trailers. Boston Dynamics has been quietly shipping it since 2023, and the customer list runs through DHL, Maersk, GXO, and a rotating group of large US retailers that the company will not name in briefings. One thousand units is not a marketing round-up. It is the largest deployed base of any purpose-built commercial robot outside of AMR fleets and legacy industrial arms, both of which are decades-old categories with their own supply chains. The category leader in commercial robot revenue, in 2026, is a non-humanoid, single-task machine sold to logistics operators against a spreadsheet ROI. Nobody in the humanoid category is close.
Unitree is the second data point, and the second lesson. The company got approval for its $619 million STAR listing in 73 days, which is not a normal Chinese capital-markets timeline for a hardware company. It is what the timeline looks like when a listing is being run as industrial policy, not as an investor process. Reported 2025 revenue was 1.699 billion yuan and net income was 278 million yuan, which is real revenue on real margin, but the multiple at $6.18 billion implies a category read that is not falling out of those two numbers alone. The read is that the Chinese state has decided humanoid manufacturing is a strategic capability worth capitalizing at a policy-adjusted valuation, and that a listing on STAR is a form of subsidized underwriting rather than a market clearing price. Fine. That is what STAR is for. But the American humanoid category is going to have to price against a Chinese peer whose cost of capital is a policy question rather than a market one, and Agility’s SPAC print last week was the opening bell on that comparison.
Then there is 1X, which is the most instructive of the three, because 1X is the closest thing the Western humanoid category has to a shipping consumer product, and 1X is the one that told the truth. The Neo launch blog on July 16 disclosed that the robot autonomously handles a fixed list of tasks (fetch a labeled item, load and unload supported dishwashers, fold supported towels) and routes everything else to a human teleoperator on shift in Oslo. This is not a scandal. It is exactly how you would run a consumer robotics product in 2026 if you were being honest about the capability gap between where the demo robot is and where the household robot needs to be. The interesting question is not whether teleoperation is happening. It is whether the unit economics work when a nontrivial share of the labor is being paid for in Norwegian salaries billed against a $20,000 hardware SKU. It probably does not, at ten units. It might, at 500. It has to, at 5,000, or the category math does not close.
The three disclosures point at the same underlying pattern from three angles. Boston Dynamics is running the pattern where the humanoid is the R&D loss leader and the boring purpose-built machine is the revenue business. Unitree is running the pattern where the humanoid IS the pitch, and the pitch is being underwritten by a state that has decided the strategic value of the form factor exceeds the near-term commercial value. 1X is running the pattern where the humanoid is the front end of a services business, with the manipulation, the vision, and the household reasoning provided by remote humans until the model catches up. All three patterns work as businesses. None of them are the “one general-purpose humanoid runs every task at $10 per hour on foundation-model policies” pitch that the Series C decks have been telling investors for two years.
The Clank read on this week is that the humanoid category quietly bifurcated in July. There is a demo-and-teleop lane, where 1X and Figure and Agility are shipping units into pilots and services contracts with a lot of human labor on the back end, funded by rounds that have to close before the teleop share drops fast enough to make the margin work. There is a purpose-built lane, where Boston Dynamics and the AMR incumbents are booking revenue against bounded, spreadsheet-legible tasks, and using the cash to fund an eventual humanoid pivot from a position of installed-base strength. And there is a policy lane, where Unitree and its Chinese peers are being capitalized at multiples that only make sense if you assume the underwriter is not a market. Every humanoid company on earth is currently in one of these three lanes. The pitch decks pretend otherwise. The disclosures do not.
The follow-on question the next round of Series C boards is going to have to answer is which lane the company is actually in, and to price the round against that lane rather than against the composite humanoid narrative that no shipping company is currently running. The people running the pilots know which lane they are in. The people writing the pitch decks are going to have to catch up before the next quarter of disclosures triangulates the same argument even harder.