The Humanoid Robot Report Card Is In: 25,000 Shipped, and China Is Winning the Count

The humanoid robot industry has spent three years selling the future. This week it got a report card instead. Market data released by International Data Corporation (IDC) on October 2 puts global humanoid robot shipments at nearly 25,000 units for the first half of 2026, a 432.1 percent increase over the same period last year. One company, China's Agibot, shipped more than 8,600 of those units, good for roughly 35 percent of the global total. For the first time, the conversation about humanoid robots has actual numbers attached, and the numbers say the field left the lab sometime last year.

For the first time, the humanoid robot industry has a real scoreboard, and the score says this stopped being a demo reel sometime last year.

From demos to shipments

Until now, the public conversation about humanoids ran almost entirely on fundraising announcements, prototype videos, and carefully staged factory clips. IDC's figures give that conversation a unit of account: shipments. The year-on-year growth, from an implied base of roughly 4,700 units in the first half of 2025 to nearly 25,000 this year, is the kind of curve that makes an industry stop being theoretical. Even the skeptics' version of this story has to acknowledge that five-fold growth is not a press release.

IDC's read of the market matters as much as the numbers. The firm says the industry is moving past technology validation toward broader commercial deployment, with humanoids expanding into industrial manufacturing, retail, logistics, tourism, and consumer applications. Competition, IDC says, is shifting away from individual hardware specifications toward broader embodied AI capabilities: the software that lets a robot perceive, reason, and act in an unstructured world. That framing is a quiet rebuke to the last two years of spec-sheet wars over degrees of freedom and battery life.

Why China is ahead on the count

Agibot shipping 8,600 units in six months is not an accident of accounting. Chinese humanoid makers have structural advantages that their Western counterparts keep rediscovering the hard way: dense actuator and sensor supply chains within a few hundred kilometers of Shenzhen, customers (especially automakers and logistics firms) willing to run pilots at real scale, and government policy that treats humanoids as strategic industrial infrastructure rather than a curiosity.

The gap between the two ecosystems is visible in who is actually deploying. Chinese automakers are putting humanoids on production floors; Western deployments remain smaller and more carefully ringed. None of this means Western programs are behind on capability, and IDC's own framing points to embodied AI as the coming battleground, where data pipelines and foundation models may matter more than manufacturing throughput. But shipments are the thing IDC counted, and on that metric the leaderboard is not close.

The gap between shipped and working

A shipped robot is not a working robot, and the industry's own moves this week underline that distinction. On October 6, Agility Robotics announced a deepened partnership with FORT Robotics to build out the safety architecture for Digit 5, the company's next humanoid, as it moves toward commercial deployment. The centerpiece is a jointly developed Offboard Safety Bridge meant to connect Digit 5 with the external safety systems already running across industrial facilities.

The agreement expands a multi-year relationship into a three-part safety setup: a safety pendant with emergency-stop and enabling-device functions, onboard communications, and off-robot interfaces. Digit 5 is being engineered as Agility's first humanoid designed for cooperatively safe work, combining human detection, safety cues, and motion control with FORT's offboard infrastructure. FORT will also handle joint hardware development, regulatory compliance, and deployment support.

The numbers behind the announcement are worth noting. Earlier Digit generations have logged more than 65,000 operating hours, including deployments with Schaeffler, GXO, and Toyota Motor Manufacturing Canada. That is the real metric of the industry right now: not units sold, but hours worked without incident. The Agility-FORT deal is a bet that the binding constraint on humanoid deployment is no longer the robot's hands or legs, but the paperwork and engineering that prove it can work next to people without hurting them. Safety infrastructure, not actuators, is the moat.

Where the money thinks the risk is

Capital is also flowing to the part of the problem where the payoff is starkest: jobs that can kill you. Also on October 6, Minerva Humanoids emerged from stealth with roughly $10 million in pre-seed funding led by General Catalyst, with Long Journey Ventures and Credo Ventures co-leading and participation from, among others, Hugging Face and Codecademy. The company is building Roger, a semi-autonomous humanoid designed for onshore and offshore oil and gas work, explosive ordnance disposal, and hazardous material response, and it is launching its first paid pilots this fall.

Minerva's approach is less ambitious and more honest than most. Roger is not fully autonomous: a trained specialist operates it through a VR headset, seeing through the robot's eyes while the company's Minerva Intelligence software handles balance, fall recovery, and navigation. The operator's judgment stays in the loop; the operator's body stays out of the blast zone. Founders Sandor Felber and Maurice Rahme bring pedigrees that fit the brief: Felber worked on Tesla's Optimus program and MIT CSAIL's humanoid research line, while Rahme spent six years at Boston Dynamics as a technical lead on Stretch, the warehouse robot that actually shipped. Minerva says it went from idea to walking humanoid prototype in five months, with prototypes built in Germany.

The company's headline estimate, that energy work kills roughly 7,000 workers a year, is its own calculation derived from International Labour Organization and International Energy Agency workforce figures, and should be read as a pitch deck number, not a census. But the direction of travel is hard to argue with: the first humanoids to earn their keep at scale will likely do it in the jobs humans least want to do.

What the scoreboard actually proves

The contrarian read on the IDC numbers is that 25,000 units is simultaneously a breakthrough and a rounding error. Breakthrough, because a market growing five-fold while moving from demos to paid deployments is real commercial momentum. Rounding error, because the global workforce numbers in the billions, and 25,000 robots is a rounding error in any labor market on Earth.

That is exactly the point, though. The industry has crossed the line where the relevant question stopped being "does this work" and became "how fast can you build and deploy." The numbers to watch from here are repeat orders, fleet operating hours, and safety certification: the unglamorous metrics that decide whether humanoids become infrastructure or expensive demos. Agibot won the first scoreboard. The second one is measured in hours worked, and Agility, Minerva, and everyone else know it.