Summary of Key Points
At the 2026 World Robot Conference, the concept of "world models" became the absolute focus, but behind this term lies a lot of industry chaos: some teams use it as a trendy label to attract funding. The robot demonstrations may look impressive, but they rely on prior debugging and remote control. The so-called "ten thousand orders" and "hundreds of millions in sales" are largely inflated, as the industry's technology is still in its infancy (only 20% mature). There are very few robots that can work stably and reliably.
1. World Models: The "Smart Brain" of Robots or a Tool for Fundraising?
World models are originally a technical concept aimed at making robots understand the world like humans—for example, predicting how much force to use to avoid breaking a cup or what to do if it slips. However, this term has been misused:
- Real Technical Directions: A few teams are indeed researching how to enable robots to perceive, predict, and make decisions autonomously through these models.
- Fundraising Tool: Most teams use it as a selling point because world models sound more cutting-edge and harder to disprove (making it easier to convince investors). In fact, many investors admit that 90% of the teams themselves don't believe in the concept.
The technical approach is still undecided (with other options like VLA and VLM competing), but world models have become a key buzzword for fundraising.
2. Impressive Demos? They’re All Carefully Planned Shows, but the Real Skills Are Lacking
Robots dancing, boxing, and grasping objects at the conference may seem impressive, but they all rely on:
- Extensive Pre-Debugging: An engineer mentioned that a 10-minute demo requires countless rehearsals, and there’s always fear of mistakes during the performance.
- Remote Control: Most robots still rely on remote operation, lacking true autonomy. For example, their dance routines are pre-programmed and fail in different scenarios.
- Fixed Grasping Mechanisms: Some robots can only grasp objects in specific positions and cannot adjust their grip (they lack sensors to detect pressure), so they can’t handle softer or heavier items.
True commercialization requires robots to work 8 hours a day without human supervision, but this is far from achievable at present.
3. "Ten Thousand Orders" and "Hundreds of Millions in Sales"? 90% Are Inflated
Many companies claim to have sold tens of thousands of robots or entered leading factories, but the reality is:
- Inflated Orders: Intent agreements, framework contracts, and trial purchases are all counted as sales.
- Low Actual Shipments: IDC data shows that only 18,000 humanoid robots were shipped globally in 2025, with 85% used in entertainment and education (e.g., performance robots), and only 15% in industrial trials.
- Lack of Real Productivity: Even if robots are delivered, they are often used in exhibition halls, laboratories, or factory test lines, where issues arise without supervision.
Investors are now verifying the authenticity of these orders: are the customers real? Have payments been made? Have the robots been delivered? However, this information is not publicly available, making verification difficult.
4. The Industry Is Overheated: Technology Is Only 20% Mature, and Rushing to Commercialize Is Counterproductive
Many investors skip teams that talk about commercialization immediately. The reasons are simple:
- Inadequate Technology: Robot "brains" cannot adapt to changes (e.g., they can’t find a cup if it’s moved).
- Impatience for Success: Technology is only 20% mature, but companies claim large-scale adoption, when admitting the truth would be more realistic.
- Lack of Practical Manufacturers: Some companies focused on technology don’t attend conferences because they understand that true commercialization requires robots that can work reliably, not just impressive displays.
Technology must be refined before it can be commercially applied. For now, developing robust systems is more important than boasting about orders.
Conclusion
Embodied intelligence is the future direction, but the industry is mixed with both genuine efforts and opportunistic marketing. While the public may enjoy the demonstrations, to discern the truth, one should look at whether robots can function outside predefined scenarios and whether orders are actually delivered for use in production. The industry needs less hype and more practical approaches. Otherwise, when the trend fades, what remains will be a mess.