I was scrolling through humanoid robot coverage last week and ended up in a spot I didn't expect. The press coverage is going one way, the actual shipment numbers are going a different way, and I can't stop thinking about the gap.
At the start of 2025, there were fewer than a hundred humanoid robots in warehouses anywhere on the planet. By the end of 2026 that number is expected to land in the low hundreds, and by 2027 in the low thousands. So: real, growing, interesting. Not the default.
The three names that get the headlines
Figure AI has the cinematic story. Thirty-nine billion dollar valuation, more than 1.9 billion in funding, and a Figure 02 robot that recently completed a twenty hour continuous shift at BMW's Spartanburg plant. I watched the video. I was impressed.
Agility Robotics is the one I find more interesting because they're the one actually in commercial operation. Their Digit robot, which honestly looks a bit like a polite ostrich in a work vest, has been running at GXO Logistics and crossed more than 100,000 tote cycles. In February 2026 it started at Toyota Motor Manufacturing Canada. Less drama, more boxes moved.
Tesla Optimus is in an internal deployment phase with a target of about 1,000 units in 2025. The planned price sits at 20,000 to 30,000 dollars. If they hit that, the economics change. I'll believe the price when I see the invoice.
The part most coverage skips
In 2025, China shipped roughly 90 percent of all humanoid robot units globally. Unitree shipped around 5,500 units. AgiBot shipped 5,168. If I asked ten people in the industry which company is leading the humanoid race, I suspect none of them would say Unitree, and yet Unitree is the one with the shipping volume.
The loudest robot in the room and the one actually stocking the shelves are often different robots.
What to do with any of this
Bain's framing is the one I come back to: first real commercial applications, case picking, palletising, line feed, are about three years from broad deployment. Battery life is the short-term constraint: around two hours per charge today, maybe six hours by 2030. IDTechEx projects the humanoid robotics market reaches roughly 30 billion dollars by 2035.
If I were designing a warehouse solution today, I wouldn't plan for humanoids. Too early. If I were designing for 2028 or 2029, I'd add a slot in the labor model called "flexible biped resource" and move on. That's the realistic framing. Agility is the safer bet if I had to pick one, and I would keep an eye on the Unitree shipments nobody is talking about.