The first global census of humanoid robots, published this week by the International Federation of Robotics, counted 7,000 units sold worldwide last year. A single Saudi framework agreement, signed in Riyadh, provides for up to 10,000 over five years. The Kingdom’s order book is larger than the world market it is buying from, and that inversion is the strategy, not an accident of it.
The census numbers first. The Frankfurt-based federation began collecting humanoid-specific data for 2025, its first attempt at counting a category that until recently existed mainly in demonstration videos. It found 7,000 humanoids sold for industrial and professional service use, against 542,000 conventional industrial robots installed in 2024 and 199,000 service robots. Humanoids remain “a fraction” of the robot population, in the words of the federation’s secretary general, Susanne Bieller, and many of last year’s buyers were research institutions, or companies collecting data to train artificial-intelligence models, rather than operators putting the machines to productive work.
The Saudi position in that small market is outsized by design. The UK-based manufacturer Humanoid and the Saudi firm QSS AI and Robotics operate a framework for up to 10,000 units over five years, a showroom in Riyadh that has been running demonstrations since November, and local assembly at a Riyadh robotics facility that has produced Saudi-built models named Sara, Mohamad and Saud. “Saudi Arabia is moving faster than any region in the world to integrate humanoid robotics into industrial use,” the manufacturer’s chief executive, Artem Sokolov, said this week. Sellers say such things; the census gives the claim a denominator. Ten thousand units against a 7,000-unit world market is not a customer, it is a market maker.
The play follows a sequence the Kingdom has now run several times: commit demand at scale before the market matures, and let the commitment pull the industry onshore. Anchor purchases preceded the cloud regions that go live in Riyadh this autumn. A state-owned carmaker preceded the automotive supply chain now forming at the coastal industrial cities. The humanoid framework applies the same logic one technology earlier in the cycle, buying position in a market that barely exists yet, at prices set before anyone can prove what the machines are worth.
The federation’s caveat is the discipline in the story. Most humanoids sold last year are not doing productive work, and the distance between demonstration and deployment is where the money is lost. Forecasts are steep, Bank of America’s research arm projects 90,000 shipments in 2026 and 1.2 million a year by 2030, but a framework is an option on that curve, not a delivery schedule. The Saudi use cases that would justify exercising the option are specific: logistics inside the giga-projects, inspection work in heat that limits human shifts, and repetitive industrial tasks in the new factories that are themselves short of labor. Each is testable, and none requires the machines to be interesting, only economic.
The number to follow is not the framework total. It is the count of deployed units doing measured work in Saudi facilities, and whether the Riyadh assembly line graduates from final assembly toward component manufacture. The census will repeat annually; the useful question is whether the Kingdom shows up in it next year as a buyer, or as a place where the working population of humanoids actually lives.