Just ten years ago, a humanoid robot cost around $3 million and could do little more than stand and wave. Today, a full-size humanoid from Unitree costs $13,500, and a dual-arm manipulator goes for $4,290. A 200x price drop in a single decade. What happened?
Hardware got cheap
Actuators — the joint motors — account for roughly half the cost of a humanoid. Over the past five years, their prices have fallen dramatically, driven by the scale of Chinese manufacturing. Sensors, cameras, and electric motors followed the same trajectory. The humanoid robot went from a one-off lab prototype to an industrial product.
AI gave them brains
Foundation models — large multimodal neural networks — have broken through in robotics. Previously, every robot action had to be manually programmed. Now, models like RT-2, π0, and VLA architectures allow robots to generalize skills: learn to pick up a cup, and it can pick up a bottle it has never seen. According to Capgemini, this is a fundamental shift — from robots that execute programs to robots that adapt to situations.
Money poured in
In 2025, robotics startups raised over $10 billion in venture capital — an all-time record (Crunchbase/Landbase data). Figure raised $1 billion at a $39 billion valuation. Apptronik secured $935 million. Skild AI closed $1.4 billion. These are numbers previously reserved for software companies.
Demographics are pushing
74% of executives surveyed by Capgemini cite labor shortages as the primary catalyst for robot adoption. 69% point to rising labor costs. Populations are aging, workers are scarce, and robots are getting cheaper. Economics is making the decision for us.
Everything converged at once
Cheap hardware + smart AI + record investment + demographic pressure = tipping point. 79% of organizations, according to Capgemini, are already deploying Physical AI or planning to. 67% of executives consider this technology game-changing for their industry.
The last time this kind of convergence happened was with smartphones in 2007–2010. We know how that turned out.
Sources: Capgemini Research Institute “Physical AI” (May 2026), Crunchbase, Landbase