Humanoid robots: where the field actually stands
AgiBot shipped 5,168 humanoids in 2025 against roughly 150 each for Tesla, Figure and Agility — a 30× gap. Meanwhile Unitree became the first humanoid maker to list publicly and factories are being built for volumes nobody has yet sold. A read of shipments against ambition.
The humanoid field carries three metrics and 19 milestones here. The shipment metric is the one that reorders most people's mental map of the field.
Per Omdia's 2025 ranking, AgiBot shipped 5,168 humanoids for a 39% global share, Unitree 4,200 for 32% (the company self-reports 5,500), and UBTech about 1,000 — the first maker past a thousand full-size humanoids. Figure shipped roughly 150, as did Tesla and Agility. Chinese makers lead Western ones by roughly thirty times on units out the door. Disclosed cumulative private funding tells a different story: Figure sits at $1.9B on a $39B valuation, UBTech at $1B, 1X at $600M, with Apptronik and Agility at $400M each. Research output runs about 2,325 papers a year.
So capital concentrates in the US while units ship from China. Both facts are real, and they measure different things — one is a bet on general-purpose autonomy, the other is a count of machines built, many of them for research, education and exhibition rather than sustained industrial work.
Western progress in 2026 was mostly about proving autonomy and converting pilots to contracts. Figure ran its humanoids 200 hours nonstop in May, sorting 249,558 packages with zero hardware failures and no teleoperation, driven end to end by its Helix network. Agility converted Digit from paid pilots into a commercial robots-as-a-service deal with Toyota at around $30 per robot-hour, and agreed to go public via a $2.5B SPAC. Apptronik raised $520M at a $5B valuation. Germany's NEURA announced a Series C of up to $1.4B at a $7B valuation, though the full amount is milestone-contingent.
The manufacturing buildout is running well ahead of demonstrated demand. 1X opened a 10,000-unit-per-year NEO plant in Hayward in May, targeting 100,000 by 2027. Tesla began a roughly 5.2M sq-ft Optimus factory at Giga Texas aimed at an aspirational 10M robots a year. Hyundai committed to absorbing about 25,000 Boston Dynamics Atlas units and is building toward 30,000 a year by 2028. AgiBot and Minth started Europe's first Chinese-designed humanoid mass-production line in Šabac, Serbia in August, with a stated target of 20,000 units a year — a target, not an output figure, in a converted parts plant.
August also brought the field's first public listing: Unitree began trading on Shanghai's STAR Market on 19 August, closing its debut up about 460% at a roughly $50B valuation, an order of magnitude above its ~$7B private mark. Worth reading carefully — Unitree's revenue still comes mostly from quadrupeds and components rather than humanoids, and the shares gave back much of the debut gain in following days. The milestone is the listing, not the price.
The locked milestone here is mass-produced general-purpose humanoids doing open-ended work, placed in the 2030s. Nothing in the 2026 record contradicts that placement.