Robetera CEO Says Humanoid Robots Could Be in Homes Within Five Years as VLA...

　　Chen Jianyu, Founder of Robotera and Assistant Professor at the Institute for Interdisciplinary Information Sciences, Tsinghua University

　　AsianFin — The debate over whether to focus on “world models” or broader vision-language-action models in robotics is shifting, according to Chen Jianyu, founder and CEO of Robetera, who believes the next wave of AI breakthroughs will come from general-purpose humanoid robots.

　　Speaking at the main forum of the 2025 World Robot Conference on Aug. 12, Chen — also an assistant professor at Tsinghua University — said humanoid robots, with their mobility and manipulation skills, are poised to transform productivity and social services.

　　“The fastest path to a general-purpose humanoid robot is to learn directly from humans — the only true general embodied intelligence in the real world,” Chen told attendees. “By integrating a general-purpose brain with a general-purpose body, we can define a new paradigm for robotics.”

　　Chen described VLA as any system that combines visual perception, language understanding, and physical action — with world models being one subcategory. He expects the next generation of VLA will be defined less by a specific architecture and more by its ability to operate end-to-end in the real world.

　　While many in the field are focused on collecting more data, Chen argued that model innovation matters more right now. “The total amount of data will grow, but efficiency in using that data is what counts. If you have to focus on just one thing, focus on the model,” he said.

　　Founded in 2023, Robetera has raised three funding rounds in under two years, including a nearly $69 million Series A in July led by CDH VGC and Haier Capital. The company is developing both the “brains” and “bodies” for general-purpose robots.

　　Its flagship VLA model, ERA-42, unifies vision, comprehension, prediction, and action into one end-to-end system. Current prototypes can perform hundreds of tasks on voice command — from sorting objects and scanning barcodes to using screwdrivers and pipettes.

　　On the hardware side, Robetera’s humanoid robots are designed as modular “universal interfaces” for interacting with the physical world. The Stardust L7 is a full-sized biped optimized for logistics, while the Stardust Q5 targets service roles like retail assistance. The company has also developed the XHand 1, a dexterous robotic hand with 12 degrees of freedom and built-in tactile sensors.

　　Chen sees humanoids as the “ultimate form of embodied intelligence,” thanks to their ability to learn from human behavior and operate effectively in human environments. He predicts humanoid adoption will follow a B2B-to-B2C path — starting with industrial and service applications, before entering households at scale.

　　“In real industrial scenarios, our robots already operate at 70% of human efficiency. By next year, we expect 90%,” he said. “Give it time, and robots will match human capabilities.”

　　Despite recent hype around AI, Chen says robotics hasn’t reached “bubble” territory, noting valuations are still far below those of the autonomous vehicle sector. But once large-scale deployments by major players arrive, he expects a flood of investment.

　　“The killer app for humanoid robots will be in the home,” Chen said. “We’re not there yet — but within five years, the tipping point could come.”