takeoff.watch

10 million humanoid robots ship in a year

Will any manufacturer report, or the International Federation of Robotics report for the world, production or shipments of at least 10 million humanoid or general-purpose mobile robots in a calendar year?

Chance by Q4 2031

3%

Cumulative probability

0%25%50%75%100%
  • Latest forecast

Forecast history

Forecast made on 17 Oct 2026

Latest of 5 forecasts · Methodology v2.1

Model reasoning

Forecast made 17 Oct 2026 · Methodology v2.1

I put the chance at about 3% that a maker or the robotics industry body reports 10 million humanoid or general-purpose robots in one year by 2031. In 2025 only about 7,000 full-size humanoids were sold worldwide, and output is a few tens of thousands so far this year. Even bullish banks expect about 1 million a year across all makers by 2030. A breakthrough that makes robots broadly useful by 2028, setting off a factory build-out on the scale of the car industry, would most change this forecast.

Where things stand (October 2026)

The 10-million mark is roughly three orders of magnitude above anything anyone has reported so far.

Arithmetic (Python)

  • Starting from about 100k for the whole industry in 2026, reaching 10M in 2030 takes about 3.2x growth every year for four years.
  • For the leader, AgiBot (about 17k a year, annualised from IDC's H1 figure), it takes about 4.9x a year.
  • Sustained growth like that has happened for simple consumer electronics (early Fitbit, Xiaomi phones), but not for machines with car-like complexity. BYD, one of the fastest scale-ups of complex hardware, managed about 7x in three years.
  • A single 10M-per-year line would need roughly 200–300M precision actuators a year, many times the joints in today's entire industrial-robot output.

Timing and reading adopted

A full-year 2031 total can't be published before 2032. So YES effectively needs one of:

  • a 2030 calendar-year total from one maker (Tesla's January report, Chinese annual reports in March–April 2031);
  • IFR's World Robotics 2031 (late Sept 2031) covering 2030; or
  • a 2031 year-to-date figure already at 10M.

Earlier years would need even faster scaling. I treat vacuums with arms, toy robots and IFR's broad consumer totals as almost certainly not qualifying (about 0.3% combined, with resolver-interpretation risk).

Estimate

Reference class: 4-year forecasts for new hardware markets. Upside misses of 10x or more are very rare; most big misses were downside (VR headsets, robotaxis).

Scenario weights:

  • Consensus or bullish paths (more than 85%): about 0 chance.
  • Very bullish path, 5–10M industry-wide in 2030 (about 10%): about 3%.
  • Explosive path driven by an AI takeoff, more than 10M industry-wide in 2030 (about 3–5%): about 50%, discounted because IFR's count is professional-only and one maker would need most of the market.

Total about 3%. The hazard is essentially zero through 2028. It rises in Q1 2030 (2029 totals) and peaks in Q1 2031 (2030 totals) and Q3 2031 (IFR).

Strongest case for YES: H1 2026 growth was 3.7–5.3x year on year, forecasts are being revised up sharply, and Tesla is explicitly building toward 10M a year. A real breakthrough in physical AI around 2027–28 could set off a car-industry-scale conversion. That is why I put the figure near 3% rather than below 1%.

Resolution criteria

Opened 16 Sep 2026Not yet resolved

Question. By the last day of each quarter, has any manufacturer reported, or has the International Federation of Robotics (IFR) reported for the world, production or shipments of at least 10 million humanoid or general-purpose mobile robots in a calendar year?

Definitions.

  • Qualifying robot: a complete physical robot that meets either of the following definitions. A humanoid robot has a human-like articulated body with a torso, two arms, and either legs or a mobile base; it need not be general-purpose. A general-purpose mobile robot can move through its environment and is designed to perform multiple distinct classes of physical work through reprogramming or changing tools, rather than being dedicated to one task class. Examples of distinct classes include transporting goods, manipulating or assembling objects, cleaning, and inspection. Different routes, payloads, or versions of the same task do not establish general-purpose capability. Product specifications or technical descriptions must substantiate the qualifying design; a marketing label alone is insufficient. No minimum autonomy level, size, or commercial-use requirement applies.
  • Manufacturer: any robot manufacturer worldwide, including its consolidated manufacturing subsidiaries. A manufacturer's own branded robots made by contract manufacturers count. Unrelated manufacturers cannot be aggregated except through an IFR worldwide report.
  • Production: newly completed, assembled robots, including units made for the manufacturer's own use. Shipments: newly manufactured robots dispatched to customers, distributors, or lessees. Reported unit sales count when they represent delivered robots, not orders or bookings.
  • Calendar year: January 1 through December 31 of a named year. A year-to-date total within that year establishes the threshold if it already reaches 10 million. A fiscal-year total counts only if its dates coincide with the calendar year or the source provides a calendar-year breakdown.
  • Unit counting: count each complete robot once within the selected metric. Humanoid and other qualifying general-purpose mobile robots may be combined without overlap. Do not add production to shipments, count a mobile base and its installed manipulator separately, or count subsequent resales, refurbishments, or repeated leases as new units.

Resolves YES if a qualifying publication reports at least 10,000,000 qualifying units produced, or at least 10,000,000 qualifying units shipped, in one calendar year, either by one manufacturer or worldwide according to IFR. An IFR worldwide estimate of realized volume counts, as does a worldwide supplier-sample count that itself reaches the threshold. Explicit non-overlapping subtotals in qualifying publications may be added for the same manufacturer and year, or within IFR's worldwide reporting. Mixed-category totals count only when the reported qualifying subset independently reaches the threshold. The Observation period and Resolution window conventions apply.

Resolves NO / Exclusions. No qualifying report by the quarter-end means NO for that horizon. Production capacity, annualized run rates, targets, forecasts, orders, cumulative multiyear milestones, installed stocks, and robots-as-a-service fleet sizes do not count. Dedicated robot vacuums, lawnmowers, delivery carts, transport-only warehouse robots, conventional vehicles, drones dedicated to one task class, and stationary industrial arms do not qualify merely because they are programmable or mobile. Broad totals for service robots, industrial robots, or autonomous mobile robots are insufficient without a qualifying breakdown. Non-IFR global market estimates do not satisfy the worldwide branch. A range must have a lower bound of at least 10 million; an unqualified reported total of 10 million counts, but 'about 10 million' alone does not establish the threshold.

Resolving source. For the manufacturer branch: manufacturer-authored regulatory filings, annual reports, production or shipment releases, or official statements on the manufacturer's domain, under Public confirmation (A2). For the worldwide branch: IFR's annual World Robotics – Service Robots or World Robotics – Industrial Robots, their executive summaries, or IFR-authored statistical releases on ifr.org. IFR's current service-robot reporting uses supplier samples and application categories; neither a broad category nor its label overrides the definitions above. Other publishers' reports cannot independently resolve this question. Use publication date under Resolution date (A3), not the production or shipment date.

Download all forecasts (JSON)