Humanoid robot makers shipped almost 25,000 humanoid robots in the first half of 2026, according to new data from IDC . That’s up 432.1% from the same period last year. As you’d expect, revenue was also up, topping $740 million, up 322.7% from last year. The key takeaway therefore: shipments are up, and robots are getting less expensive as the industry is starting to scale.

The biggest seller, by a fairly wide margin, is Shanghai-based Agibot.

Agibot shipped more than 8,600 humanoid robots in the first six months of the year, according to IDC, which pencils out to about 35% of the global market. Unitree came second with roughly 5,000 units, after that there’s a steep drop-off to a pack of mostly Chinese robot makers shipping hundreds of units each.

This is the top 10, based on IDC’s chart (IDC charted most vendors as bubbles rather than publishing exact figures, so these are approximate):

  1. Agibot: ~8,600
  2. Unitree: ~5,000
  3. Booster Robotics: ~1,000
  4. UBTech: ~900
  5. Leju Robot: ~460
  6. Galbot: ~400
  7. Astribot: ~340
  8. Fourier Intelligence: ~320
  9. Galaxea AI: ~260
  10. Figure AI: a few dozen

Nine of these 10 are Chinese. Overall, IDC says Chinese vendors built more than 95% of all humanoid robots shipped globally in the first half, and Chinese customers bought about 78% of them: more than 19,000 units, up 426.3% year over year. A caveat here: many of those robot sales are likely going to one “customer” … the Chinese government, which operates “robot schools” as training centers for better physical AI.

As I’ve written before, I’m not super-concerned by the fact that major American and European humanoid robot makers aren’t showing up here yet. Figure shows up here, but I think it’s undercounted (the company has been scaling production for months, and its newest model, Figure 04, is coming at some point pretty soon). Similarly, robot makers like Apptronik, Agility Robotics, 1X, Humanoid AI, Agile Robots, Boston Dynamics and many more are perfecting their solutions before shipping to customers in volume.

I think 2027 is the year of scale for humanoid robots doing actual work, and when these estimates of market shipments will really start to mean something.

This is now the third major estimate of first-half 2026 humanoid shipments. Smart Analytics Global counted 19,100 units , up 272%. Counterpoint put it at 22,000 , up almost 300%. Now IDC says almost 25,000. One thing they all agree on is who’s in first place. SAG had Agibot at roughly 8,400 units and 44% share, while Counterpoint had it at about 9,700 units and 43%. IDC’s 8,600 lands in the middle, but its 35% share is lower because IDC sees a bigger overall market.

That bigger total is significant.

In the SAG and Counterpoint data, Agibot and Unitree control about three quarters of all humanoid robot shipments. In IDC’s data, they have just over half, and companies outside the top 10 account for about 28% of the market.

As you might expect given the massive overall growth rate of 432%, some of the category leaders’ numbers are off the charts. UBTech’s shipments grew roughly 1,900% year over year, Astribot’s about 1,400% and Galaxea AI’s about 1,100%. Agibot grew about tenfold. In comparison, Unitree, which listed on Shanghai’s STAR Market in August, grew about 120%, albeit on much higher base numbers, and Galbot less than 100%.

IDC’s bubbles also represent revenue, and UBTech’s is noticeably bigger than its unit count would suggest. The reason: it sells mostly full-size industrial humanoids, not cheaper and smaller research and education models.

Given the numbers and the revenue, the average humanoid robot shipped in the first half of 2026 cost about $29,600. A year ago, the implied numbers – 4,700 units, $175 million – work out to about $37,000 per robot.

Prices are dropping, and fast.

Figure AI is the only American company in the top 10, and IDC’s chart puts Figure’s first-half shipments at just a few dozen units. That’s hard to square with what Figure has said publicly. In April, the company said it was doubling deliveries every month , with roughly 240 Figure 03 robots delivered that month alone. One possible explanation: IDC counts only units shipped to paying external customers, and much of Figure’s fleet works internally or in its own data-collection projects, like renting 30 homes to train Figure 03 on household chores.

So: where are all these robots going?

They are still mostly shipping to labs, classrooms, stages and government data-collection centers. According to the IDC report, research and education, performances and exhibitions, and government data centers together accounted for 69% of first-half shipments. The good news for those who want to see robots get actual jobs is that this is down from 83.8% for all of 2025, so the mix is shifting toward industrial manufacturing, retail, logistics and tourism. But more than two-thirds of humanoid robots still aren’t doing productive work in the traditional sense.

So there’s progress … but it’s still slow progress.

Agibot is pushing hard on real-world deployments. The company says it hit 20,000 cumulative robots produced in September, up from its 15,000th robot in June . This month it began deploying more than 300 robots at Chimelong Spaceship Park in Hengqin for visitor guidance, entertainment, retail and hospitality, and it has put 100 of its smaller X2 humanoids into ASD retail stores for greeting customers and answering product questions. Its wheeled G2 robots are also working a live production line at Longcheer Technology.

Agibot calls 2026 the start of a “deployment stage” for embodied AI. That’s consistent with what Agibot’s Dr. Yao Maoqing told me in May : the tipping point isn’t about demos or production numbers, but whether robots actually deliver productivity in live environments over time.

Looking further out, IDC raised its long-term forecast by about 50%. It now expects global humanoid robot shipments to exceed 750,000 units in 2030. That’s more bullish than SAG, which forecasts 500,000 by 2030. Getting there from roughly 50,000 to 60,000 this year means the market needs to roughly double every year for four straight years.

That’s a tall order, but this is a very nascent early-stage industry.

IDC China research manager Lily Li says the next phase of competition will come down to who can close the loop between technology, data, engineering and real-world use cases fastest.

“Whoever can more quickly close the loop of ‘technology — data — engineering — scenario’ will have a better chance of seizing the initiative in the next stage of competition,” she says (translated from Chinese).

Agibot looks well-positioned. I think some of the American and European competitors are too, but they’ll have to prove that out over the next six to 12 months.