A 'Foxconn of humanoids' is forming in China. Two listed industrial investors are backing the bet.

A new B2B middle layer is forming in humanoid robotics, and the proof is a Chinese contract manufacturer that ships its own 800 N·m planetary joint motors, fielded a marathon-running robot last month, and has now collected two listed industrial investors in twelve months.

BXI Robotics (半醒具身), the company in question, operates on a simple pitch: a brand that wants its own humanoid robot can skip 2–3 years of in-house hardware R&D and outsource the joints, the controller, and the assembly to a contract manufacturer. The model is structurally analogous to the Foxconn layer that built the smartphone industry, though the analogy is the writer's, not BXI's.

The news peg is modest. Shanghai-listed Suochen Technology (索辰科技, 688507) just added a "tens of millions of yuan" position, in the low millions of U.S. dollars at recent exchange rates, to a round that also included an earlier investment from Shenzhen-listed thermal-management supplier Feirongda (飞荣达). Two listed industrial names inside twelve months is the moment a supply layer crystallizes rather than a single bet.

The hardware footprint is what separates BXI from a typical startup profile. The company builds its own planetary hollow-shaft joint motors in three sizes (BXI-50, BXI-70, BXI-85), runs them through a 1,000 Hz controller, and ships 800 N·m peak-torque versions with internal water cooling. A reference humanoid build uses 31 degrees of freedom. Production capacity is "dozens per month" today, with a stated path to 100 or more per month on short notice, according to the company's 36氪 profile.

That capacity gets stress-tested in public. BXI's 精灵2 finished the 2025 Beijing humanoid half marathon in 4 hours 2 minutes 19 seconds, sixth overall, as an unmodified production unit, per Xinhua's 2026 marathon coverage. The newer 精灵3 ran the April 2026 Beijing Yizhuang humanoid half marathon in a 150-minute net time, top of the leaderboard, according to Chinese Wikipedia's event entry. Independent confirmation from the Beijing municipal government's 2026 marathon release and the BXI company site lets a reporter treat the marathon numbers as corroborated rather than company-only claims.

The buy-vs-build trade-off is where BXI sells the deal. A brand that wants a humanoid product line can either staff a 2–3 year in-house program, covering actuators, controllers, supply chain, and integration, or buy from a contract manufacturer that already runs the marathon, ships the joint motors, and can ramp production in Shenzhen and Changzhou via Feirongda's existing facilities. Company guidance for July 2026 to July 2027 puts monthly order value above ¥10 million (~$1.4 million, approximate at ~7 RMB/USD) and full-year revenue above ¥100 million (~$14 million, approximate). Those numbers are BXI's own projection, not audited revenue.

The customer mix is named by category only, and that is the story's main gap. BXI says it serves "top tech and internet firms," "mainstream domestic automakers," robot companies, embodied-AI "brain" developers, and vertical buyers in textiles, photovoltaic cleaning, and unmanned car washes. The one named overseas target is a "top-tier US tech customer," not identified. Suochen board secretary Wu Weizi frames the synergy as closing the loop between Suochen's physics-AI simulation stack and embodied-AI deployment: simulation training, real-machine landing, data return, model retraining.

The thesis holds that bipedal mobility is "solved" at the proof-of-concept level; the next competitive frontier is execution, including dexterous hands, perception, and compute. The CEO's 2027 framing is that humanoids enter daily life in some categories — a CEO opinion that runs ahead of more conservative industry timelines from analysts and from the companies themselves. The CEO also frames the overseas opportunity as "every country wants to avoid single-source dependency in humanoid robots." That is a thesis, not a market study.

The falsifier for the "Foxconn of humanoids" frame is narrow but real. If the ODM customer base turns out to be proof-of-concept demos and pilot units rather than commercial SKUs, or if brand-side customers keep the high-end units in-house and only outsource low-end work, the contract-manufacturer layer compresses back into a small-volume integrator. Industry coverage of competing humanoid ODMs, the actual customer logos behind the "top tech firms" line, and the unit economics on a single shipped joint motor would all be the next reads.

Watch item: whether the Suochen position is followed by a second listed investor in the next two quarters, and whether BXI names the US customer publicly before the end of 2026. Either would turn the supply-layer thesis from a one-fund datapoint into a pattern.