PokeBot, a general embodied-intelligence company, has announced a USD-level Pre-A financing round led by Shunwei Capital and Matrix Partners, with Zhongding Capital, Jiukun Ventures, SEE Fund and others participating, and Xiaomi Strategic Investment and Yunqi Capital following on. For a company founded in March 2026, it is the second round in just over three months, and nearly every early shareholder reinvested.
From a 20 billion yuan unicorn to a fresh start
The founder is Xu Huazhe, formerly co-founder and chief scientist of StarWeave, the embodied-AI company whose valuation blew past 20 billion yuan with close to 3 billion yuan raised. Xu left at the peak to build something new. His record is rare even in this field: a Tsinghua electronic-engineering undergraduate, a PhD at UC Berkeley’s BAIR lab, postdoctoral work at Stanford’s SVL, then a Tsinghua professorship where he founded China’s first robotics-manipulation embodied-AI lab, TEA Lab. He has more than a hundred papers across Science Robotics, NeurIPS, ICLR and CoRL.
Why the home, not the factory
In June a demo video put PokeBot on the map: a robot cooked mapo tofu from scratch, cutting the tofu, browning the meat, seasoning and plating, nine minutes with no human intervention. The same stretch showed it folding clothes, tying sachets and lacing cable ties, covering flexible objects, fine manipulation and long-horizon tasks. Those are exactly the hard problems of real-world robot operation.
Unlike the many humanoid firms aimed at industrial and logistics work, PokeBot targeted the C-end home from day one, betting robots can eventually fold, store, clean and cook. Xu’s thesis is that homes are far harder than factories because objects, spaces and tasks change daily, which makes the home the best training ground for general robot ability. StarWeave, notably, is itself an investor in PokeBot, turning a founding partnership into an industry partnership.
His angel round in April, tens of millions of dollars led by Yunqi with Shunwei, Xiaomi, StarWeave, Baidu Ventures and a string of top market funds, set the pattern. The speed of the follow-on bet reflects a conviction that the bottleneck in robotics is no longer the demo, but teaching machines to act reliably in the mess of ordinary life.
Read the original report (OFweek Robotics)
Translated and adapted from OFweek Robotics (robot.ofweek.com).