Roland Berger’s new report, Humanoid Robots 2026: The Convergence Moment of a New Market, reframes the debate. The question is no longer whether humanoids are a bubble. It is who claims a position first if the industry actually takes off.
The trigger was not a flashy demo. It was three variables colliding: vision-language models, generative AI and edge computing letting robots move from scripted motion to acting on what they see. Core subsystems, compute, sensing, actuators, power and structure, are in pilot validation but not yet at scale. Battery life is mostly 2 to 8 hours against a 16-hour target, and key subsystems still need 50 to 90 per cent cost cuts.
The structural case is the labour math. Roland Berger projects China’s working-age population will fall 24 per cent by 2050 versus 2025, the EU 18 per cent and Japan 25 per cent. Humanoids are not about replacing everyone. They fill the gaps fixed automation cannot: picking, handling, loading, simple assembly, flexible parts.
The firm puts the humanoid OEM market at about USD 300 billion in its base case by 2035, up to USD 750 billion optimistically, and potentially USD 4 trillion by 2050. The number that matters more: operating cost of advanced humanoids could fall to roughly USD 2 per hour. If that holds, the logic of global manufacturing location changes.
The real bottleneck is not hardware. It is data and software.
Translated and adapted from OFweek Robotics (robot.ofweek.com).