Artificial Intelligence

    ACE Robotics Chief Says Humanoid Robots Nearing a Breakthrough Like ChatGPT

    Sunday, August 23, 2026 at 02:31 PM UTCOriginally reported by Decrypt
    ACE Robotics Chief Says Humanoid Robots Nearing a Breakthrough Like ChatGPT cryptocurrency guide featured image

    The chairman of Chinese startup ACE Robotics, Wang Xiaogang, has said that robot intelligence could reach a turning point comparable to the arrival of ChatGPT by the end of next year. He made the remarks to Reuters, pointing to two drivers: world models and the collection of data from real environments.

    What the company said

    ACE Robotics was founded in July 2025 and builds AI models meant to run humanoid robots. Ant Group and SenseTime back the company. According to Reuters, it raised more than $100 million in the first half of 2026 and intends to seek an IPO as early as it is permitted to.

    Wang argued that the shortage of training material is a central problem. He said the whole industry has gathered roughly 100,000 hours of data, which in his view is far too little for training embodied foundation models.

    Key concepts and industry context

    • Embodied AI: systems that let a robot or other physical agent sense its surroundings, reason about them and turn decisions into movement using sensors and actuators.
    • World models: AI that learns how objects and environments behave, so a robot can anticipate the outcome of moving, picking something up or touching its surroundings before it acts.

    Others are working on the same problem:

    • In October, researchers revealed HumanoidExo, a wearable exoskeleton that records human motion to help train humanoid robots.
    • In January, Boston Dynamics showed the production version of its Atlas humanoid, saying AI progress had moved it nearer to commercial use.
    • In June, Alibaba launched its Qwen-Robot Suite, a group of AI models intended to help robots navigate, carry out physical tasks and simulate real-world settings.

    Humanoid robots can already walk, dance and box. Doing useful work reliably in unpredictable physical settings is still among the hardest challenges in AI. Language models like ChatGPT and DeepSeek spread quickly, but robots still have trouble handling many different tasks in unfamiliar places.

    Why it matters

    If you are learning about AI and the technology sector, this story shows why robots have lagged behind chatbots: text is plentiful online, while data about physical movement and environments must be gathered deliberately. The timeline is Wang's own expectation, not an established outcome. It also shows that investor money, large tech groups and research labs are all pursuing the same data and modeling problem, and that startups in this field are already talking about public listings.