Ola Dimensions raises hundreds of millions of yuan in Angel+ round for home robots
Zhou Shunbo's home-robotics startup closed its fourth round in six months, betting on an action-native model for household tasks.
Ola Dimensions, the home embodied-AI startup founded in March 2026 by former Huawei "genius youth" Zhou Shunbo, has closed an Angel+ round worth several hundred million yuan. The round was joined by Chengyuanxiang Fund, Xinneng Venture, Xizhou International, Topband, and Hony Capital, with existing shareholders over-subscribing and Zhiguan Capital acting as sole financial adviser.
The pace is the story as much as the money. Four rounds in about six months — seed, seed+, angel, and now Angel+ — have stacked up several hundred million yuan in total from backers including Hillhouse, 5Y Capital, China Merchants Venture, Junshan Capital, SAIF, Unity Ventures, and Baidu Ventures. Zhou previously ran Huawei's embodied-intelligence innovation lab as its first head, building the company's embodied AI stack from zero and leading its CloudRobo platform.
The technical bet is what the company calls Real-world RSI — recursive self-improvement: robots execute real tasks, collect feedback, and feed it back through data processing, training, and evaluation. Its OLA-Sem and OLA-Geo models, built on an action-native foundation-model route where action participates in pretraining rather than being bolted on afterward, have topped RoboTwin benchmark settings, the company says.
Commercially, Ola Dimensions is deliberately anti-consensus. The body is a mobile base with dual arms — not necessarily a bipedal humanoid — and the first customers are makers and developers, not ordinary households. Partnerships with home-services firm HaoKang@Home and D-Robotics are meant to pipe real household data into training. A first-generation product goes through small-scale validation in the first quarter of 2027.
Notably, Zhou has walked back an earlier goal of selling a thousand units in 2027, citing immature supply chains and the risk of shipping hardware whose experience doesn't justify its cost. The metrics he names instead: task success rate in real homes, and how often and how long each household actually uses the machine. For a sector addicted to demo videos, that's the right scoreboard — if the company can keep investors patient while it climbs it.