Morgan Stanley: US-China Humanoid Gap Is Capability, China Shipment Forecast Raised to 50,000

Morgan Stanley's report divides the US-China humanoid robot comparison into two distinct dimensions. The United States leads in frontier models, vision-language-action (VLA) development, and simulation, while China controls the physical inputs and supply chains needed for scalable production.

The report's key judgment is that shipment share is not the correct scoreboard. The real gap between the two countries lies in humanoid robot capability—the actual performance and functionality of the machines.

This conclusion shifts the discussion from shipment volumes back to technological competence. It suggests that the race is not merely about who can produce more units, but about who can build more capable robots. For industry observers, this means paying closer attention to capability benchmarks, such as the humanoid robot games, rather than just quarterly shipment figures. The two dimensions are complementary: US software prowess and Chinese hardware scale together shape the global humanoid robot ecosystem.

Data from China's second World Humanoid Robot Games shows a dramatic leap in mobility. Tiangong Ultra won the large-class 100-meter final in 8.64 seconds, beating Usain Bolt's human world record of 9.58 seconds. Last year, the same robot won in 21.5 seconds.

Morgan Stanley notes that while humans took about a century to shave roughly 10% off the 100-meter record, robots cut the humanoid record by about 60% in just 12 months. This progress came from reducing torso and waist mass and upgrading motors and control systems.

However, the new dexterous hand challenge reveals a stark contrast. It featured eight tasks—power tool assembly, powder weighing, bricklaying, nailing, and more—directly corresponding to value-added tasks in manufacturing and warehouse automation. These tasks exposed the gap between leading robots and other competitors.

The challenge reflects Moravec's paradox: robots can achieve adult-level performance on the track, but their hand manipulation remains like that of a toddler. Running and jumping rely on body structure, motor power, and balance control, which can iterate quickly. Hand operations involve tactile feedback, force control precision, and multi-finger coordination—each step is far more difficult than running. The contrast between rapid locomotion gains and slow manipulation progress is critical for commercial viability, as most valuable tasks in factories and warehouses require dexterous hands.

Morgan Stanley cites this paradox to highlight the uneven advancement of humanoid robot capabilities. Legs are running fast, but hands are still clumsy.