Humanoid Robots 2026: 108 Models, $38B Market & What's Actually Shipping
Humanoid robots have been a fixture of science fiction for a century. What's different in 2026 is that they're becoming a fixture of factory floors. The gap between demo video and deployed product is still wide — but it's closing faster than almost anyone predicted two years ago. This guide covers what humanoid robots actually are, what they can and can't do right now, who's building them, and what the realistic timeline to broader deployment looks like.
What Makes a Robot "Humanoid"?
A humanoid robot is defined by form factor, not capability. It is a robot designed to approximate the human body plan: upright bipedal locomotion, two manipulator arms, and a head-mounted sensor cluster. The defining characteristic is that it can operate in spaces designed for humans — stairs, doorways, vehicle interiors, kitchen counters — without those spaces needing to be modified.
This is the key distinction from other robot categories. A warehouse AMR (autonomous mobile robot) is fast, reliable, and commercially mature — but it needs flat floors and clear lanes. A robotic arm on a factory line is extraordinarily precise — but it's bolted to one spot. A humanoid can, in principle, do either job and move between them. That flexibility is the entire economic rationale for the category.
Not every "humanoid" robot is bipedal. Wheeled humanoids — robots with a human-shaped upper body on a wheeled base — trade walking capability for stability and are often more deployable in the near term. Browse humanoid robots on Geppetto →
Where the Industry Actually Stands in 2026
The honest picture of humanoid robotics in 2026 is more nuanced than most coverage suggests. The category is real, the progress is genuine, and the economics are moving fast — but the gap between announcement and deployment remains significant.
According to humanoidindex.org, of all humanoid robots currently tracked globally:
| Commercial Stage | Number of Platforms |
|---|---|
| Mass production | 2 |
| Commercial deployment | 38 |
| Pilot programme | 15 |
| Prototype | 13 |
| R&D only | 8 |
Source: humanoidindex.org, March 2026
The "38 commercial" figure sounds substantial until you consider that most of those deployments involve single-digit unit counts. The total global stock of commercially deployed humanoid robots is estimated at fewer than 10,000 units — compared to 4.7 million industrial robots and 50–75 million consumer robots. How many robots are in the world? →
The robots closest to genuine commercial scale are Agility Robotics' Digit — the first humanoid with contracted Robot-as-a-Service agreements, currently deployed at Toyota Canada, Amazon, and GXO logistics — and Unitree's G1, H1, and R1 series, which have collectively shipped approximately 5,500 units primarily to research institutions and early industrial adopters.
What Humanoid Robots Can Actually Do
This is where marketing collides with reality. Impressive demo videos have conditioned observers to overestimate current capability. The humanoid.guide skill scoring system, which rates robots on ten real-world tasks, provides a useful calibration.
Most commercially available humanoids score 1.0–1.5 out of 5.0 on real task capability. The most capable currently available platforms — 1X NEO, Unitree H2 — score approximately 2.5 out of 5.0. No currently available humanoid scores above 3.0 on the full task suite.
| Task | Current Best-in-Class Performance | Timeline to Reliable Automation |
|---|---|---|
| Sorting / pick and place | Good — deployed commercially | Now |
| Walking on flat surfaces | Excellent | Now |
| Climbing stairs | Moderate — reliable in controlled conditions | 2025–2026 |
| Recovering from falls | Improving — not yet reliable | 2026–2027 |
| Folding clothes / fine manipulation | Early demos only | 2027–2028 |
| General household tasks | Lab demonstrations | 2028–2030 |
Source: humanoid.guide skill scores; company deployment data — March 2026
The tasks humanoids can do reliably today are concentrated at the simpler end: moving standardised objects between known locations, navigating predictable environments, performing repetitive logistics operations. The tasks that require dexterity, adaptability, and judgement — the ones that would make a humanoid genuinely useful in a home — remain 2–4 years away for the leading platforms.
The Companies Shaping the Category
The humanoid robotics landscape has stratified into three tiers: companies with verified commercial deployments, companies with serious funding and active pilots, and the long tail of startups — particularly Chinese — racing to establish hardware platforms.
Tier 1 — Commercially deployed:
| Company | Robot | Key Deployment | Price / Model |
|---|---|---|---|
| Agility Robotics | Digit | Toyota Canada, Amazon, GXO (RaaS) | ~$10/hr RaaS |
| Unitree Robotics | G1, H1, R1 | Research, education, early industrial | $5,900–$90,000 |
| UBTECH | Walker S series | BYD, Geely, Foxconn EV factories | Enterprise pricing |
Tier 2 — Funded and piloting:
| Company | Robot | Pilot Partner | Funding |
|---|---|---|---|
| Figure AI | Figure 03 | BMW Spartanburg | $1.9B+ ($39B valuation) |
| Boston Dynamics | Atlas (Electric) | Hyundai RMAC | Hyundai-owned |
| 1X Technologies | NEO / NEO Gamma | Consumer pre-orders | $125M+ (OpenAI-backed) |
| Apptronik | Apollo | Mercedes-Benz, Jabil | $400M Series A |
| Tesla | Optimus Gen 3 | Internal Tesla factories | Self-funded |
Source: humanoid.guide, humanoid.press, company announcements — March 2026
Compare humanoid robots side by side →
The China Factor
Of the approximately 180 distinct humanoid robot models tracked globally, 108 are Chinese-origin versus 31 from the United States. China's 3.5× lead in model count reflects a combination of national policy support, lower development costs, and an existing manufacturing ecosystem for the actuators, sensors, and compute hardware that humanoids require.
The cost differential is stark. Morgan Stanley's 2025 Robot Almanac puts the humanoid bill of materials in China at approximately $50,000 versus approximately $130,000 for a comparable US-built platform. That gap is structural, not temporary — it reflects China's dominance of the rare earth magnet supply chain, its actuator manufacturing base, and lower engineering labour costs.
The consumer sentiment gap is equally striking. AlphaWise survey data from the MS Almanac shows 61% of Chinese consumers report a positive view of humanoid robots entering their homes, versus approximately 5% of US consumers. China is not just building more humanoids — it has a domestic consumer market that is already prepared to buy them. The global robotics market in 2026 →
The Price Collapse That Changes Everything
The most important development in humanoid robotics over the past 18 months is not a capability breakthrough — it is a price collapse.
| Robot | Launch Year | Price | Significance |
|---|---|---|---|
| Unitree H1 | 2023 | $90,000 | First "affordable" full-size humanoid |
| Unitree G1 | 2024 | $16,000 | Psychological sub-$20K barrier broken |
| Unitree R1 | Late 2025 | $5,900 | Full-size humanoid under $6K |
| Noetix Bumi | Oct 2025 | $1,400 | Child-size humanoid at consumer electronics pricing |
Source: Company announcements; humanoid.press
Noetix's Bumi sold 100 units in its first hour and 500 units within two days of launch on JD.com — demonstrating demand exists at consumer-accessible price points even for early, limited-capability platforms.
The trajectory implies full-size capable humanoids reaching sub-$5,000 pricing by the late 2020s. At that price point, the industrial ROI case becomes straightforward for a wide range of applications. Which jobs are robots replacing? →
What to Realistically Expect by 2030
2026: First contracted commercial humanoid deployments scaling beyond single digits. Consumer pre-orders shipping (1X NEO). Price floor for capable humanoids at ~$5,000–$6,000.
2027–2028: First humanoid platforms reach hundreds-of-units deployment scale in logistics and automotive. Hyundai begins Atlas deployment across manufacturing network. Tesla Optimus exits internal testing for external customers.
2029–2030: First humanoid platforms in four-figure deployment volumes. Early consumer home deployments in Asia (primarily China and South Korea). Goldman Sachs' $38B market projection for 2035 starts to look achievable from current trajectory.
The Morgan Stanley projection of 1 billion humanoids by 2050 remains an upside scenario requiring costs to fall to appliance-level pricing and reliability to reach consumer-grade standards. It is plausible given current trajectories — but the 2026 reality is that we are in the earliest innings of a technology whose impact will compound over decades, not quarters.
Frequently Asked Questions
What is a humanoid robot?
A humanoid robot is a machine designed to approximate the human body plan — upright bipedal locomotion, two arms, and a head-mounted sensor array — so it can operate in environments built for humans. The defining practical advantage is that humanoids can use the same spaces, tools, and workflows as human workers without requiring those environments to be redesigned for automation.
Can you buy a humanoid robot in 2026?
Yes, within limits. Unitree's R1 is available for purchase at $5,900 and the G1 at $16,000, primarily targeting research institutions and early industrial adopters. 1X Technologies' NEO is accepting consumer pre-orders with deliveries beginning in 2026 at a $20,000 price point or $499 per month subscription. Industrial platforms from Agility, Figure, and Boston Dynamics are available through enterprise agreements rather than retail purchase.
How much does a humanoid robot cost?
Prices range from $1,400 for Noetix's education-focused Bumi to over $250,000 for enterprise industrial platforms. The most commercially relevant range in 2026 is $5,900 (Unitree R1) to $90,000 (Unitree H1) for platforms with real task capability. Consumer-grade humanoids in the $1,000–$5,000 range are expected to emerge by 2028–2030 as manufacturing costs continue to fall.
Which humanoid robot is the most advanced?
By commercial deployment, Agility Robotics' Digit is the most proven — it has contracted commercial service agreements and is the only humanoid running paid shifts at scale. By raw capability scores, the 1X NEO and Unitree H2 lead currently available platforms. By funding and valuation, Figure AI ($39B valuation, BMW deployment) is the most prominent US startup. By volume shipped, Unitree leads all competitors with approximately 5,500 units across its product line.
Why are there so many Chinese humanoid robots?
China accounts for 108 of approximately 180 humanoid robot models globally — a 3.5× lead over the US — for three structural reasons: national policy support through the 15th Five-Year Plan prioritises humanoid robotics development; manufacturing costs are significantly lower (approximately $50,000 BOM vs $130,000 in the US per Morgan Stanley data); and China's existing industrial robot and actuator supply chains provide a ready component base that US startups must build or source from scratch.
Will humanoid robots replace human workers?
Humanoid robots will automate specific tasks — particularly repetitive, physically demanding work in controlled environments — rather than replacing entire job categories immediately. The tasks being automated first are warehouse pick-and-place, factory line feeding, and parts sequencing. Tasks requiring adaptability, fine motor dexterity, and judgement in unstructured environments remain 3–5 years from reliable automation. See which jobs are most at risk →
Data sources: humanoid.guide robot catalog (March 2026); humanoidindex.org commercial stage data (March 2026); humanoid.press deployment tracker (March 2026); Morgan Stanley Global Embodied AI Research (Robot Almanac Vol. 3, December 2025); IFR World Robotics 2025; company announcements. Last updated: March 2026.