The humanoid robot market in 2026: what actually shipped, and why the forecasts disagree by 14x
The entire global humanoid robot market shipped roughly 13,000 to 20,000 units in 2025, and the five research houses forecasting 2035 range from 700,000 to 10 million annual units, a 14x spread. Every trillion-dollar headline is a 2050 number.
Somewhere between 13,000 and 20,000 humanoid robots shipped worldwide in 2025. That is the entire humanoid robot market: every buyer, every country, every price tier. By Interact Analysis's count, only about 10% of the humanoids produced up to 2025 had been deployed in real-world operations. The rest went to academic research, training-data collection, and entertainment.
The trillion-dollar figure in the headlines is a 2050 forecast, and the five 2035 forecasts below differ from each other by a factor of fourteen. Every number here is pinned to a primary document and dated.
What actually shipped in 2025
Four organizations published a figure for a year that has already finished. They do not agree.
| Source | Published | 2025 figure | What it counts |
|---|---|---|---|
| Omdia | Jan 8, 2026 | about 13,000 units | Global humanoid shipments |
| Yano Research | Apr 30, 2026 | 16,580 units | Shipments; wheeled and bipedal only |
| BofA Global Research | Mar 12, 2026 | 20,000 units | Annual global shipments |
| Interact Analysis | Jun 2026 | over 20,000 units | Production, not deliveries |
A 54% spread on a year that already happened is the first thing worth knowing. No clean census exists, definitions differ on whether a wheeled torso counts, and most manufacturers are private and publish nothing.
The disagreement runs down to the company level. On January 8, 2026, AgiBot announced that Omdia had ranked it number one worldwide with 5,168 humanoids shipped, a 39% share of a roughly 13,000-unit market. Three weeks later Unitree issued a release headlined "Unitree Ranks No.1 Globally in Humanoid Robot Shipments", claiming more than 5,500 shipped and 6,500 built, pure humanoids only, and noting that "misinformation regarding our company's 2025 shipment volume has been circulating online." Two companies, one year, both number one.
The rest of the Omdia share table, per Robotics & Automation News on January 21, 2026: UBTech 7%, Leju 4%, EngineAI 3%, Fourier 2%, Figure AI 2%, Agility 1%, Tesla 1%, others 10%. On a 13,000-unit base that is about 260 machines for Figure and 130 for Tesla, a full year from the two American names that dominate the coverage. I track what Tesla has actually delivered on the Optimus page.
Interact Analysis agrees from another direction: Chinese vendors made more than 90% of 2025 production, and about 75% of the robots were delivered inside China.
The only audited numbers in the industry
Almost every humanoid maker is private, which is why there is no clean humanoid stock to buy and why nobody has to open their books. Unitree's Shanghai STAR Market listing changed that for one company.
Its prospectus figures, reported by Caixin on July 3, 2026 and Global Times in August: revenue of 159 million yuan in 2023, 392 million in 2024, and 1.699 billion yuan in 2025, about $252 million at the August 6, 2026 rate. Humanoids brought in 868 million yuan of that, 51.78%, the first year they out-earned the company's quadrupeds. Gross margin on core businesses: 60.13%.
That is the biggest humanoid shipper on earth by its own count, and a quarter of a billion dollars is the revenue base under every "market already worth billions" claim.
We've just closed an oversubscribed $403 million Series A, and are thrilled to announce additional investors including @MercedesBenz, Japan Post Capital, RyderVentures (the corporate venture capital arm of @RyderSystemInc), @ARKInvest and more. (@Apptronik, 2025-03-18)
Five forecasts for 2035, side by side
| House | Published | 2035 annual shipments | Value figure |
|---|---|---|---|
| Interact Analysis | May 19, 2026 | over 700,000 | about $15B revenue |
| Goldman Sachs | Feb 27, 2024 | 1.4 million | $38B total addressable market |
| Omdia | Jan 8, 2026 | 2.6 million | not published |
| Yano Research | Apr 30, 2026 | 7.18 million | not published |
| BofA Global Research | Mar 12, 2026 | 10 million | not published |
Three of those five landed within five months of each other in 2026, off overlapping public information, and the top forecast is more than fourteen times the bottom. Goldman's famous $38 billion needs the same care: a total addressable market, the size of the prize rather than money collected.
The 2030 waypoints are just as far apart. Goldman projected "more than 250,000 humanoid robot shipments in 2030." BofA's March 2026 exhibit puts 2030 at 1.2 million, on a path of 20,000 in 2025 and 90,000 in 2026, an 86% compound annual growth rate. Interact Analysis, writing two months later, expects shipments to stay under 100,000 a year near term and calls 2032 the inflection point.
Where the trillion-dollar headline comes from
Morgan Stanley Research's April 29, 2025 report sits behind nearly every "$5 trillion robot market" post. Its own summary says the humanoids market "is likely to reach $5 trillion by 2050, plus related supply chains as well as repair, maintenance and support." A 2050 figure, and not confined to selling robots.
The rest of that page is more useful than the headline. More than 1 billion humanoids in use by 2050, about 90% of them, roughly 930 million units, doing "repetitive, simple, and structured work" in industry and commerce. Only 80 million in homes. Analyst Adam Jonas, quoted there: "We are not going to see a robot in every home overnight."
Nobody's forecast puts one in your kitchen this decade
Every house agrees on the shape while disagreeing on the size, and the shape is industrial first. Goldman's 2030 base case of 250,000-plus shipments is "almost all of which would be for industrial use." Morgan Stanley expects adoption to accelerate in the late 2030s and puts a general-purpose home humanoid about another decade of hardware and AI progress away. Interact Analysis calls household use "a longer-term opportunity, constrained by safety and environmental complexity." BofA, the most bullish on homes, does not have them at 62% of the fleet until 2060.
The consumer slice that exists today is small and sits to one side. TrendForce, in a July 15, 2026 release, puts the companion robot market at US$1.1 billion by 2030, and the products it names are desk and therapy companions. The one general-purpose humanoid sold to households, the 1X NEO, is $20,000 or $499 a month, and 1x.tech still says "US Deliveries start 2026" as of August 7, 2026, with an Expert Mode letting a remote human pilot it for tasks it has not learned. What exists against what gets demoed is its own running comparison.
What one costs to build
The manufacturing numbers explain the timeline better than the shipment numbers do.
- BofA, March 2026: pilot-stage bill of materials runs $90,000 to $100,000 per unit. A China-built humanoid's BOM was about $35,000 in 2025 and is forecast below $17,000 by 2030. Actuators alone are more than half the BOM in its 2030 breakdown.
- Goldman, February 2024: manufacturing cost of $30,000 to $150,000 per unit, down from $50,000 to $250,000, a 40% drop against expectations of 15% to 20% a year.
- Morgan Stanley, April 2025: the cost of one humanoid was "around $200,000 in 2024 in high-income countries," falling to about $150,000 by 2028 and $50,000 by 2050, and as low as $15,000 in lower-income countries.
That last set circulates online stripped of its qualifiers and relabelled as a manufacturing cost curve. It is a price forecast for high-income markets, a different quantity from what a Chinese factory spends on parts, and reading it as a build cost puts Morgan Stanley's 2024 number at nearly six times BofA's China-built bill of materials for 2025.
The retail evidence sides with BofA. A Unitree R1 checks out at $4,900 today, a small research biped rather than anything these forecasts price, and even so it shows Chinese component supply taking the entry price somewhere no 2024 model anticipated. The current ladder is in what a humanoid robot actually costs.
Where it actually stands
Twenty thousand machines a year at the top of the estimates, nine in ten of them in a lab or a demo. That is the humanoid market in August 2026, and everything else on this page is somebody's model. Before reposting the next big figure, pin its year and check whether it counts industry or homes.
The category also carries a long record of demand that looks inevitable in a slide deck and thin at checkout, which is the whole point of the consumer robot graveyard.
This is a ledger, not a prediction. When a house revises, the row changes and the date on the row changes with it.
One email when something actually changes: a delivery verified, a price cut, a robot discontinued. No weekly filler, unsubscribe whenever.