Xu Huazhe Founds 'Poke Robotics,' Targeting Home-Scene Embodied Intelligence

Embodied Intelligence Entrepreneurs Collectively Pivot to Home Scenarios

In 2026, embodied intelligence robot entrepreneurs worldwide have simultaneously turned their attention to a space brimming with imagination yet fraught with challenges — the home.

Fueling this enthusiasm are recent signs of Scaling Law emerging within the industry. Silicon Valley embodied intelligence company Generalist AI validated the certainty of data scaling effects with its GEN-1 model: when they fed massive volumes of data to their robots, success rates on fine manipulation tasks soared from 64% to an astonishing 99%. Shortly after, Silicon Valley's hottest unicorn Sunday Robotics also attempted to crack the data challenge in home scenarios, not only launching the Umi glove data collection solution but also sending its robot Memo directly into real homes to perform chores — clearing dining tables, brewing coffee, and folding laundry — attracting substantial capital investment in the process.

On the other side of the Pacific in China, a familiar figure has officially entered the home robotics arena. He is Xu Huazhe, assistant professor at Tsinghua University's Institute for Interdisciplinary Information Sciences and founder of Poke Robotics.

"Within two years, China will have usable home robots," he said of his outlook for the industry — optimistic and aggressive.

From Xinghaitu to Poke: A Decision to Go Solo

Xu Huazhe's name is well known in China's embodied intelligence community. As one of the "four Berkeley returnees," he has built deep influence in both academia and industry. In 2023, Xu joined Xinghaitu as chief scientist and co-founder, helping build the company into a star enterprise in China's embodied intelligence sector.

However, at the very peak of Xinghaitu's trajectory — with its valuation surpassing 20 billion yuan and nearly 3 billion yuan in funding — Xu made a decision that surprised the outside world: he left Xinghaitu to found an entirely new embodied intelligence company, Poke Robotics.

Behind this choice lies Xu's independent thinking about the industry's development path. If Xinghaitu represents the exploration of embodied intelligence in industrial and commercial scenarios, then Poke Robotics has its strategic focus squarely locked on the home. In Xu's view, home robotics is the true "endgame market" for embodied intelligence, and the current level of technological maturity and data accumulation is gradually approaching that tipping point.

Why Home Scenarios Are the 'Ultimate Testing Ground' for Embodied Intelligence

Compared to industrial and commercial settings, the home environment poses far more complex challenges for robots.

First, there is the highly unstructured nature of the environment. Factory assembly lines operate in relatively fixed and controllable settings, whereas every household differs in layout, object placement, and daily habits, requiring robots to possess exceptionally strong generalization capabilities. Second, there is the diversity and precision of tasks. From folding clothes to cooking, from tidying rooms to caring for the elderly, the range of tasks in home scenarios is vast, and many operations involve soft objects and fine force control, placing extremely high demands on a robot's perception and manipulation abilities.

Finally, and most critically — data. Collecting data in home scenarios is far more difficult than in industrial settings. Factories can use fixed sensors to collect standardized data at scale, but operational data in homes is scattered, private, and lacks unified standards.

For precisely these reasons, whether Scaling Law holds true in embodied intelligence finds its most rigorous litmus test in home scenarios. The recent breakthroughs by Generalist AI and Sunday Robotics have excited the industry precisely because they have preliminarily demonstrated that the logic of "capability emergence driven by data scale" also applies to robotic manipulation tasks.

Opportunities and Challenges for Chinese Players

From a global competitive landscape perspective, China has both advantages and unique challenges in the home robotics track.

On the advantage side, China possesses the world's most comprehensive robotics hardware supply chain — from motors and reducers to sensors — leading globally in both cost and mass production capability. At the same time, China's enormous household market provides a natural foundation for scaled deployment of robots. Additionally, Chinese teams are accelerating their iteration speed in AI large models and multimodal perception technologies.

Challenges are equally significant. The data collection and annotation systems for home scenarios are still in their early stages, and how to efficiently obtain high-quality operational data is a core bottleneck that all players, including Poke Robotics, must overcome. Furthermore, the safety and reliability standards for home robots are far higher than for other scenarios, and taking products from the lab into millions of homes still requires crossing a lengthy engineering chasm.

Outlook: Can the Two-Year Promise Be Fulfilled?

Xu Huazhe's prediction that "usable home robots will appear within two years" has undoubtedly set an aggressive time anchor for the industry. The confidence behind this judgment comes partly from the technological optimism generated by initial validation of Scaling Law in embodied intelligence, and partly from the industrial momentum of global capital and talent accelerating into this track.

Of course, the definition of "usable" is crucial. If it means a robot capable of completing several practical tasks in specific home scenarios at a reliability level acceptable to consumers, then a two-year window is not unimaginable. But if the expectation is for a robot that can flexibly handle every household need like a human, that clearly requires a much longer timeline.

Regardless, from Xinghaitu to Poke Robotics, Xu Huazhe's act of "breaking out of the shell" has written a noteworthy new footnote for China's embodied intelligence industry. As more and more top talent and capital pour into the home robotics track, this field — once regarded as a "science-fiction-level proposition" — is moving toward reality at an unprecedented pace.