Home Robots Face a Tunnel Period as Open-Source Strategies Emerge

Humanoid robot shipments are climbing sharply. IDC data shows global shipments in 2025 reached about 18,000 units, up 508% year-over-year; in the first half of 2026, China alone shipped over 40,000 units, accounting for 97% globally. Yet this surge contrasts sharply with household penetration—entertainment, education, and data collection remain the main uses, while few robots actually enter ordinary homes.

01 The 'Tunnel Period': An Unavoidable Stage Qian Dongqi, chairman of Ecovacs, called the path to home entry a 'tunnel period,' estimating 'three years in a optimistic case, five in a pessimistic one.' He cited Ecovacs' floor-cleaning robot, which took nearly a decade from R&D initiation in 2000 to real home entry in 2009. Window-cleaning robots followed a similar cycle: launched in 2011, only recently hitting market explosion with sales exceeding 1 million units in 2025.

Qian argues the core bottleneck is not whether a technical demo works, but that the industry lacks scale effects, keeping key component costs high and preventing reasonable consumer prices. For early floor-cleaning robots, a CMOS chip for a laser motor module cost 300 yuan; now the entire LDS module costs around 30 yuan. This cost reduction relies on 'chain leaders' driving volume so upstream and downstream suppliers are willing to invest.

02 Data Gap: From 89.4% to 12% A more thorny issue than price is 'environmental mismatch.' Stanford HAI's 'AI Index 2026' shows humanoid robots achieve 89.4% task success in simulation, but only about 12% in real homes—a 77-point drop pointing to unstructured household environments far beyond lab settings. Home layouts, furniture placement, wet floors, pet interference, and scattered children's toys all pose unpredictable challenges.