Xieyue Intelligence (斜跃智能), a young startup, has announced an Angel+ round worth several hundred million yuan for a project that places the home, not the factory floor, at the center of embodied AI. Sina Finance’s republication of Beijing News reported that the company intends to use the money for model training, compute and data infrastructure, team expansion, robot-body research, and scenario validation.
The round arrives while China’s physical-AI sector is attracting capital across several different layers of the stack. Some companies are building robot bodies, others are gathering data, and others are developing models designed to turn perception into action. Xieyue’s stated focus is an embodied foundation model for general-purpose household robots, a setting that forces the company to deal with varied rooms, objects, users, and interruptions rather than a tightly controlled industrial workflow.
A New Startup Bets on the Home as the Test Case
Xieyue was founded in February 2026. The company was co-founded by Chen Wei, formerly Li Auto’s chief AI scientist and foundation-model lead, and Zhang Xiao (张骁), formerly a Li Auto product-line president. The financing report identifies Linear Capital, Junshan Capital, Hongyi Capital, and Yinshan Capital among the participating investors.
The company’s thesis is unusually specific. It says the household is the first environment in which it wants to deploy an embodied foundation model. That choice makes the data problem far more complex than it is in a single industrial process. A home includes changing layouts, fragile objects, unpredictable requests, family members, pets, and tasks that may be stopped or corrected midway through execution.
Xieyue calls its technical approach “Duplex Reasoning.” That is the company’s technical framing, not an independently verified benchmark. The reports do not provide evaluation scores, task-completion rates, or a commercially released robot model that would allow outside comparison.
The financing announcement is therefore better understood as funding for a development path than evidence that a home robot has solved general-purpose autonomy. The company says it is building data-collection equipment and a data platform to support larger-scale training. Its named spending priorities show how much of the work remains upstream of a commercial deployment.
Data and Compute Are Central to the Household-Robot Claim
For a household robot, a foundation model alone is not enough. The model has to be trained on examples that link visual scenes, human instructions, physical movement, and correction. Xieyue’s emphasis on collection, cleaning, labeling, and training infrastructure reflects the same constraint facing China’s broader physical-AI field.
The data emphasis places Xieyue within a broader physical-AI market that includes world-model development. Xieyue’s own stated focus is a consumer home-robot use case, and its announcement discloses no technical or commercial connection to the other initiative.
The company’s stated plan also includes work on the robot body itself. That is a practical reminder that household robotics cannot be reduced to a software problem. A model may interpret a request correctly yet still fail if the machine cannot grasp an object safely, move through a crowded room, manage power, or recover from an unexpected contact event. Xieyue’s sources do not provide a product specification, sensor package, battery design, or delivery timeline, so those details should not be inferred.
The choice of home robotics makes the investment more demanding than a narrow automation deployment. A warehouse can limit tasks and define routes. A factory can arrange tools and workstations. A home contains a much wider range of conditions, and a useful robot needs to operate around people without assuming a repeated script.
The sector is still searching for workable combinations of data pipelines, model architectures, and robot form factors. Xieyue’s round adds another company to that early-stage contest without establishing a common technical path.
Angel Capital Funds a Long Technical Road
The Angel+ label should not obscure the size of the challenge. Xieyue has raised several hundred million yuan, but the reports do not disclose its valuation, ownership structure, revenue, customer pipeline, or a target date for a consumer robot. Those omissions are significant because they prevent a clear assessment of how close the company is to commercial deployment.
The company’s founders bring experience from Li Auto, where AI systems are developed for a demanding but relatively structured vehicle environment. A household robot faces a different operating problem. It must handle unstructured spaces and work alongside users who may give incomplete, changing, or contradictory instructions. Xieyue’s own “Duplex Reasoning” description suggests that interruption and correction are central issues, but the reports do not show whether the approach performs reliably under real household conditions.
China’s investment market is still willing to fund that uncertainty, as other physical-AI financing makes clear. Yet investment scale is not the same as deployment scale. Investors will eventually ask whether data collection can be sustained, whether compute costs are manageable, and whether the resulting robot can do tasks consumers value enough to pay for.
Xieyue’s financing is notable because it makes that consumer-facing ambition explicit. The round gives the company resources to build its data and model pipeline, but the next evidence will need to come from visible robots, defined tasks, safety performance, and repeatable home tests. Until then, the strongest conclusion is that a new team with automotive-AI experience has secured substantial backing to pursue one of embodied AI’s hardest environments.