Humanoids crossed 20,000 units in 2025 — only 10% actually shipped to work

Humanoid robot production crossed 20,000 units in 2025 — roughly a tenfold jump from fewer than 2,000 in 2024. Only around 10% of those units reached real-world deployment. The remaining 90% stayed in research labs, data-collection exercises, and showcase demos.

That gap is the number worth watching.

Interact Analysis' Humanoid Robots 2026 report frames the surge as a supply-side event, not a demand-side one. Most 2025 deployments were small proof-of-concept pilots, underwritten by government subsidies, OEM strategic capital, and supply-chain partnerships — not by a buyer paying standalone unit economics for the robot.

Solar panel capacity (2010-2012), Chinese EV manufacturing (2015-2018), and lithium-ion gigafactories (2017-2020) all ran the same script: balance-sheet and industrial-policy funding pulled hardware production ahead of commercial viability. Production prints an impressive curve. Earned demand lags by years. The supply curve eventually gets written down, consolidated, or repriced once the subsidy window narrows.

The next 12 months will test which side of that analogy humanoid robots land on. Three signals matter: OEM earnings disclosing pilot-to-paid-conversion rates, write-downs on humanoid inventory at Chinese integrators, and the renewal posture of provincial subsidy programs in Guangdong and Shanghai.

Interact Analysis separately projects humanoid revenue to reach $15bn by 2035. That case only holds if the 10% deployment share climbs toward 30% or higher. Until it does, the framing isn't "humanoids are taking off." It's "humanoids are being built."