Issue #003 — The Rest

Published 2026-07-13 · Updated 2026-08-16

The humanoid headline is always the same two names: Unitree and Agibot, shipping thousands of units, raising the largest rounds, getting the press. The story underneath is less reported.

I spent the last few weeks going through every humanoid company's public footprint I could verify — roughly 90 companies, looking at what they actually shipped, who bought it, and what the unit economics look like. Three things stand out, and they don't match the press.

1. The headline unit number is mostly noise

Trade press and analyst estimates put China's 2025 humanoid shipments at roughly 13,000–17,000 units (TechTimes, June 2026; SDV Guru), with Chinese vendors accounting for ~90% of global volume. That number is real, but the label "humanoid" is doing a lot of work. If you restrict to bipedal human-form robots only, the picture compresses considerably — and here I owe you an honest caveat: the tight, bipedal-only total is my estimate from patching vendor-level data, not a figure any research firm publishes cleanly.

What the sourced data does support:

• Unitree and Agibot together dominate volume. TrendForce (April 2026) calls them "clear leaders, supported by strong monetization capabilities and progress in mass production." My working estimate has them combining for roughly 10,000 units in 2025 — call it ~60% of a ~17,000-unit bipedal base, not the ~80% you sometimes see quoted (that 80% figure only works if the denominator shrinks to bipedal-only, and I no longer trust the precision of either number). An earlier draft of this piece made exactly that mistake; the honest answer is "roughly 60%, possibly higher, base definitionally fuzzy."
• The leadership shifted in 2026. Smart Analytics Global (August 10, 2026) reports global humanoid shipments surged 272% YoY to 19.1K, with Agibot overtaking Unitree as the world's largest humanoid vendor. The 50/50 framing from 2025 is already stale.
• Everyone else is thin. Leju (Kuavo + Roban + Aelos): high hundreds to low thousands depending on product line — and the Aelos volume is education robots, not factory-floor humanoids. The second tier (Zhongqing Robotics, Stardust Intelligence, LimX Dynamics, Fourier, Galbot) sits at a few hundred units each by my estimate. The long tail has shipped under 100 units, or is still in prototype, or is a software platform with zero hardware shipped.

The "1,000+ unit" claims you see in press releases are often industrial arms, cleaning robots, or earlier-generation quadrupeds repackaged as humanoids.

2. Most units went somewhere that isn't a factory

This is the part that matters most, and where I'm least willing to fake precision. Industry reporting (e.g., Live in the Future's April 2026 analysis of the "deployment gap") consistently describes a market where education, research, data collection, and demonstration deployments dwarf genuine production-environment deployments. My own read of vendor-level destinations suggests industrial and logistics is a single-digit percentage of 2025 shipments — but I can't source an exact 4%, and a percentage breakdown to the decimal implies data nobody actually has. Treat "under 10%, plausibly under 5%" as the defensible range.

The direction is clear even where the decimals aren't: the humanoid "boom" is currently a research-and-development market, not a deployment market. The customers are paying for future optionality (a robot that might be useful in 18 months) or for data collection (running the robot in a controlled environment to train the next model). Very few customers are buying because the robot is doing useful work today.

3. Unitree and Agibot are winning for different reasons — and both are now investing in hands

Unitree's edge is vertical integration: the company builds motors, reducers, encoders, and dexterous-hand components in-house, and its cost position is widely reported as a large fraction below overseas equivalents. (I've seen the specific in-house BOM percentages and margin figures quoted from Unitree's prospectus, but couldn't independently verify them this cycle — treat those as unconfirmed until the prospectus documents are in hand.)

Agibot's edge is ecosystem and deployment speed: a standardized "A-chain" parts supply chain and a ship-now, debug-in-the-field posture that put units into real consumer-electronics and automotive production lines. Note the relevant August development: procurement activity in the pack (e.g., the Ruixinxing R300 embodied follow-and-carry robot announcement, August 5) shows the broader embodied-robot procurement channel staying active even outside the bipedal headline segment.

The most telling post-July signal, though, is hands. In the last month the catalog logged three new dexterous-hand entries: BrainCo's Revo3 (21 DOF, palm and finger tactile, backdrivable), Unitree's Dex3-1 (three-finger, 7-DOF, 710 g, seven force-control micro joints, positioned for the G1 humanoid), and Figure's 03 hand (in-house tactile sensing down to ~3 grams of force). When the top three contenders all ship new end-effectors within weeks of each other, the bottleneck has officially moved from locomotion — where Unitree and Agibot already won — to manipulation. That's where the next differentiation round gets fought.

What this means for the other ~90 companies

The companies between "shipped a few hundred units" and "still in prototype" are in a genuinely hard place. The tier-2 names have real engineering, real customers in research settings, and a real path to maybe 500 units a year. Even at optimistic ASP assumptions (order of $20K–$50K), that's roughly $10M–$25M in revenue — not enough to cover a 100-person engineering team, sales, and supply-chain overhead in a competitive Chinese labor market. And note the only company credibly reported with >$10M in outside-customer humanoid revenue in 2026 is Figure — a US firm (Value Add VC, June 2026), not any of the Chinese second tier.

My estimate — and it is an estimate, not a sourced figure — is that break-even for a standalone Chinese humanoid company sits somewhere in the 1,000–2,000 units/year range, well above where the tier-2 players are. They are subsidized by venture capital or a parent with strategic reasons to keep them alive.

This is not a permanent state. Two of the second-tier names will likely consolidate into the top two (Agibot's newly claimed #1 spot suggests consolidation pressure is already flowing uphill). Another 10–15 will pivot to niches — medical rehab, agricultural inspection, industrial cases where bipedal is genuinely the right form factor. The rest will quietly fold or pivot to software: training data, simulation environments, model APIs sold to whoever ends up shipping robots.

The headline number is real. The "humanoid revolution" framing is not. The 2025–26 market for bipedal humanoids is, in the truest sense, a research-tools market — and with the 2026 volume tripling on 2025's base while deployments still skew toward labs and demos, the next 18 months are about which research tools turn into products factories actually buy.

That's Issue #003. Next week (Thursday's slot — funding flow): a look at the new companies that raised in the last 30 days and the handful of disclosed valuations that didn't come from AI-generated tables.

→ [Browse the sourced funding table → mazeintelli.com/funding] → [Hit reply — especially if you have unit-shipment data for the second tier I'm missing, or the Unitree prospectus margins.]