Sota Wujie (索塔无界), a four-month-old startup, has announced a strategic partnership with Europe's lar...
Sota Wujie (索塔无界), a four-month-old startup, has announced a strategic partnership with Europe's largest supermarket group to deploy over 1,000 embodied AI robots in real retail environments over the next three years. The company, founded by Hu Debo (胡德波), a former Huawei Europe executive, is positioning itself as the first to commercialize native physical world models, bypassing hardware manufacturing to provide an "original physical brain" for robots. The partnership was unveiled at the WAIC conference where Sota Wujie was listed as a key Shanghai AI project.
Sota Wujie's approach centers on a 4D world model that integrates spatial representation, physical constraints, spatiotemporal reasoning, safety evaluation, and action output, enabling end-to-end understanding and operation of the physical world. Led by Chief Scientist Liu Zhe (刘哲), a PhD from Huazhong University of Science and Technology and postdoc at Hong Kong University, the team has developed a World-Action technology stack based on prior collaborations with Huawei, Changan, Xiaomi, Li Auto, and Daxiao Robotics. The company aims to overcome the scarcity of high-quality physical manipulation data by deploying custom collection systems in partner supermarkets, accumulating 100,000 hours of cold-start data and targeting an annual increase of one million hours.
The choice of supermarkets is strategic: they offer high complexity similar to factories but dynamic and diverse interactions akin to home environments, making them an ideal intermediate for training models that can later generalize to broader scenarios. This partnership not only validates the commercial viability of physical-world AI but also highlights the regulatory and compliance hurdles—data anonymization, local storage, and privacy laws—that Sota Wujie navigates using Hu Debo's experience working with seven of the top ten global retail groups. This move could set a precedent for other embodied AI companies, potentially accelerating the transition from simulated training to real-world deployment.
For builders and investors, this signals a shift from purely virtual or simulation-based approaches to grounded, domain-specific deployments. The success of this model could encourage more targeted investments in vertical AI applications, particularly in retail and logistics, where the near-term ROI is clearer than in broad general-purpose robotics. It also underscores the importance of local expertise and compliance in international AI deployments, a barrier that many Chinese AI firms face when expanding overseas.