The Intelligence from The Economist
The Intelligence from The Economist

Bots on the ground: China leads humanoid race

August 28, 2026

AI Summary

5 min read

In Beijing this week, a humanoid robot named Tiangong Ultra finished a race, then crashed into a wall. Other competitors fell over. One caught fire. The event was the World Humanoid Robot Games, and the scene captures the state of an industry that China is betting on heavily. Companies in the country will probably sell around 50,000 humanoid robots this year, triple last year’s figure. But as Don Weinland, The Economist’s China Business and Finance Editor, explains, these are very expensive machines that cannot yet do very much.

Why robots are so hard to train

The core problem is data. Teaching a chatbot to communicate is relatively straightforward because it can scan vast amounts of digital text. Teaching a robot to interact with the physical world is far harder. Picking up a glass of water, for example, requires understanding the shape of the glass, the fragility of the material, the amount of force needed, and the presence of any obstacles. Weinland reports that experts estimate a hundred million hours of real-world training data may be needed to make humanoid robots truly useful. At current collection rates, that would take more than eleven thousand years.

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

Listen to Audio Summary Open in App

What you'll learn

  • 1 (01:22) **Humanoid Robot Games in Beijing** - The episode opens with a scene from the "World Humanoid Robot Games," framing China's lead in the industry.
  • 2 (02:51) **The Core Challenge: Training Humanoids** - Don explains why making useful humanoid robots is so difficult compared to AI chatbots.
  • 3 (04:00) **Two Paths to Bridge the Data Gap** - Don outlines the two main methods for collecting the training data needed to make robots smarter.
  • 4 (04:55) **Inside a Chinese Robot Training Center** - Don describes his rare visit to the Hubei Humanoid Innovation Center in Wuhan, a key facility in China's data strategy.
  • 5 (06:10) **China's Scaling Advantage and Its Costs** - Don assesses China's progress, highlighting its ability to build infrastructure at scale, but also the immense financial burden.
  • 6 (07:33) **Alternative Data: Egocentric Recording at Scale** - The investigation moves to another, potentially cheaper method of data collection being pursued in China.
  • 7 (08:26) **Who Is Buying These Useless Robots?** - Don addresses the paradox of a booming market for robots that can't do much, revealing the circular nature of the current demand.

+ Full timestamped outline available in the app

Show Notes

Our correspondent goes inside a Chinese robot facility to see thousands of droids being taught to complete useful tasks. Is it worth the time and money? How Meta’s $17bn settlement at its blockbuster trial could change teen social-media use. And 75 years of the Paris bookshop Shakespeare and Company.


Watch extended clips from Insider here, plus the full Elon Musk interview


Guests and host:

  • Don Weinland, China business editor
  • Alice Fulwood, “Money Talks” host and US business editor
  • Alexandra Suich-Bass, culture editor
  • Rosie Blau, co-host of “The Intelligence”


Topics covered: 

  • Robots, humanoids, China, AI
  • Meta, Instagram, Facebook, social media, addiction, 
  • Shakespeare and Company, Paris, bookshops


Listen to what matters most, from global politics and business to science and technology—subscribe to The Economist.



Hosted on Acast. See acast.com/privacy for more information.

The Intelligence from The Economist

More from this podcast

The Intelligence from The Economist →