Wang Xiaogan, the chairman of ACE Robotics, anticipates a breakthrough in embodied AI by the end of 2027. He stated that advancements in robotic control models and a significant increase in datasets related to real-world physical interactions will be key drivers, as reported by Reuters.
"We expect a 'ChatGPT moment' for embodied intelligence by the end of next year," Wang remarked.
However, he estimates that widespread commercial adoption of the technology will take an additional four to five years.
Robots Lack 'Brains'
While current humanoid robots can run, dance, and perform specific programmed tasks, they struggle with a broad range of activities in unfamiliar environments.
ACE Robotics identifies AI models, rather than hardware, as the main limitation. Systems of embodied intelligence must perceive their surroundings in real-time, understand instructions, predict the outcomes of their actions, and select appropriate next steps.
The startup is developing a model called Kairos, which integrates perception, multimodal understanding, physical environment modeling, and behavioral planning.
In July, the company announced that Kairos ranked first in four tests for embodied AI, including tasks involving bimanual control, long-term planning, and physical interaction. According to Reuters, the open version of Kairos-4B outperformed Nvidia's Cosmos 3 and Ant Group's Lingbot in public benchmarks, despite having a relatively small scale of 4 billion parameters.
Data Scarcity as a Major Challenge
Wang also pointed out the limited amount of information regarding real-world interactions as another challenge. He noted that the entire industry has amassed approximately 100,000 hours of such data in recent years, which is "far from enough" to train foundational models of embodied AI.
Typically, companies gather data through teleoperation, where a person in an exoskeleton or using controllers repeatedly performs necessary movements while the robot records the trajectory.
ACE Robotics, however, employs a different strategy. The company outfits real workers on production lines with lightweight sensors to capture their actions. Over the next two years, the startup aims to accumulate tens of millions of hours of data.
ACE Targets 10,000 Stores
The firm is testing its models in the fields of autonomous retail, hospitality, and rapid delivery warehouses. The hardware foundation consists of humanoid robots from Chinese manufacturers such as Unitree, AgiBot, and Fourier.
Within the next year, ACE plans to implement its technology in at least 1,000 stores, aiming to expand that number to 10,000 by the end of the second year.
To train its models, the startup utilizes Nvidia accelerators as well as Chinese chips from Rhino Tech and Digua Robotics. Wang explained this diversification as a means to reduce equipment costs and lessen reliance on specific suppliers.
Founded in July 2025, ACE Robotics counts Ant Group and SenseTime among its investors. In the first half of 2026, the company raised over $100 million and intends to go public as soon as Chinese regulations permit.
In June, Morgan Stanley analysts raised their forecast for Chinese humanoid robot shipments in 2026 from 28,000 to 50,000 units, citing commercialization, government support, and large-scale deployments for data collection as contributing factors.