Issue #002 — The Players

Ken Z.

— 3 min read
Share
Published 2026-07-10 · Updated 2026-08-18

Last week I admitted I didn't have a defensible funding number. This week I went and got one — not from an AI-generated valuation table, but from primary sources: a STAR Market prospectus, a Hong Kong exchange filing, and a Caixin report. Five companies, every figure sourced. Here's what the actual filings say about who's winning the Chinese humanoid race right now.

The one that's already profitable

Unitree (Unitree) filed its STAR Market prospectus on March 20, 2026 — confirmed across SSE filing records and subsequent CSRC coverage (per TMB and Embodied Global, Apr–Jul 2026). The numbers in it are unusual for this sector:

Unitree is profitable. In a field where every comparable company is burning cash to scale, that's the single most surprising number I found this week. The prospectus also discloses a Hong Kong subsidiary set up in September 2025 and flags U.S. export risk explicitly — overseas revenue is real, and so is its exposure to the geopolitical window.

The one that scaled revenue but not profit

UBTECH (UBTECH, HKEx 09880) is the first humanoid listed anywhere. Its 2025 annual report (via Dongwu Securities) tells a different story:

UBTECH's humanoid line grew 22x in a year and became its largest revenue source — but the company still lost $100M. The growth figure lacks a disclosed base in the summary; treat the multiple as directional, not precise. This is the "scale-first, profit-later" bet. Unitree and UBTECH are the two listed comparables, and they've made opposite bets: Unitree on disciplined margin overseas; UBTECH on volume at home.

The one that's the most valuable unlisted robot company

Galbot (Galbot) raised ¥2.5B (~$350M) in March 2026, per Caixin reporting carried by RobotToday and TechNode (Mar 2, 2026). Three-year cumulative funding: ¥7B (~$980M). Valuation: ¥20B (~$2.8B) — confirmed by Embodied Global's June 2026 coverage citing "cumulative funding of over RMB 6.9 billion at a valuation above RMB 20 billion." Caixin's phrasing: "the highest-valued unlisted robotics company." Backers include CATL (the world's largest battery maker), China Mobile, and the National Integrated Circuit Fund Phase III. That investor list is the tell — this isn't venture capital betting on a robot, it's industrial capital securing a position in a supply chain.

The dexterous-hand pure-play

Linkerbot (Linkerbot) closed a ~¥1.5B (~$210M) Series B in February 2026 (Gasgoo, Feb 14, 2026), and Reuters (May 4, 2026) reported the company is seeking a next round at a $6B valuation — a target, not a closed mark. Linkerbot makes one thing — robotic hands — and claims 80%+ global share in high-DoF dexterous hands (Dealroom, citing Reuters). Whether that share figure holds up or not, the capital signal is clear: a single-component company can now command a multi-billion valuation. The bottleneck for every humanoid in this list may be the hand, not the legs.

A mid-August 36Kr piece — "Robot Hand Costs Hundreds of Thousands Yuan: Are Robot Component Players More Profitable Than Unitree?" — sharpens the thesis further: the question isn't just whether hands are a bottleneck, but whether component specialists can outperform the integrators on margin. That piece didn't exist when I drafted Issue #002; it would have changed the framing here, and it's exactly the kind of debate I'll be tracking in Issue #003.

What I'm not claiming

  • Agibot (Agibot) appeared in a brokerage industry report, but I couldn't find a clean disclosed valuation in a primary source this week. I'm leaving it out of the table rather than importing a number I can't trace. It goes in next issue.
  • Dobot (Dobot) listed in Hong Kong on Dec 23, 2024 (stock code 2432.HK) per its own press release; its prospectus is 8MB and I haven't finished extracting the comparable metrics. Next issue.
  • The "$6B target valuation" for Linkerbot is a target for a round that hasn't closed — marked partial, not verified. Reuters' framing ("will seek") makes the distinction clear.
  • These are five companies out of a field of 50+ — and a wider competitive set than that, once you count Xiaomi's CyberOne demos (Aug 17–18, Lei Jun posting working videos) and the broader component catalog tracked on the funding page. The point of Issue #002 isn't "here's the full map." It's "here's what the primary sources actually say, versus what the secondary tables claim." The gap is large.

The takeaway

One of the five is profitable (Unitree). One is scaling revenue while losing money (UBTECH). One is the most valuable private company in the field, backed by industrial capital rather than VC (Galbot). One is a single-component specialist commanding a multi-billion valuation target (Linkerbot). The fifth — Agibot — is held over to next issue pending a clean primary-source number. The "humanoid bubble" narrative doesn't survive contact with these filings — there's real revenue, real profit in one case, and real industrial buyers on the cap table. What there isn't, yet, is a single company that's both profitable and at scale. That's the watch for 2026 H2.

Issue #003 will add Agibot and Dobot with extracted prospectus numbers, fold in the 36Kr components-vs-integrators profitability debate, and start tracking the deployment side — which factories, which lines, which volumes. The funding table is maintained at mazeintelli.com/funding and updated as primary sources surface. → [Browse the sourced funding table → mazeintelli.com/funding] → [Hit reply — especially if you have Agibot or Dobot numbers I couldn't find.]

Read more