On July 14th, the Guizhou Provincial Government Information Office held a press conference on the high-quality development of the province's data industry.

During the Q&A session, a reporter asked about the inclusion of "embodied intelligence" data collection as a core funding priority in the 2026 big data development special fund guidelines, a topic of significant interest to media and enterprises. The question sought clarification on what embodied intelligence data collection entails, Guizhou's strategic focus on this field, and the measures planned to drive its industrial development. Provincial Big Data Bureau Deputy Director Zhang Bo provided the following response.

Embodied intelligence data collection represents a new frontier for the future AI industry and a key breakthrough for the province to capture a leading position in the national data industry landscape. The explanation was structured around three main aspects.

Defining Embodied Intelligence Data Collection

Embodied intelligence data collection is a novel data industry format focused on training robots and intelligent equipment. It involves collecting multimodal interaction data—such as human operational movements, visual footage, and force/tactile sensations—in real-world physical environments using motion capture devices and multiple sensors. This data is then cleaned and annotated to create high-fidelity datasets specifically for robot training, forming the core foundation for the large-scale deployment of robotics.

Rationale for Guizhou's Strategic Focus

Guizhou's development of embodied intelligence data collection is driven by three compelling factors. First, there is immense demand. Currently, there are only about 500,000 hours of compliant data from real physical interaction scenes domestically, whereas the commercialization of robots requires tens of millions of hours of data, leaving a gap exceeding 99% and indicating vast market potential. Second, Guizhou possesses unique scenarios. The province boasts a wealth of real-world settings, including urban and rural households, retail supermarkets, sauce-flavor liquor manufacturing, mining, and cultural tourism. These can be opened for low-cost data collection, compensating for the scarcity of urban scenes and high venue costs in eastern regions. Third, the province has solid supporting infrastructure. Guizhou already has a mature data annotation workforce, a nationally integrated computing power base, and specialized digital intelligence industrial parks, enabling the rapid formation of a complete "collection-processing-storage-trading" industrial chain.

Systematic Measures for Implementation

The province is implementing specific, systematic measures to advance embodied intelligence data collection. First, it is opening diverse collection scenarios across the region. In May, the province, in collaboration with 28 industry regulatory departments, issued the "Implementation Plan for Accelerating the Cultivation and Opening of Artificial Intelligence Application Scenarios in Guizhou Province." This plan coordinates the orderly opening of government public scenes, state-owned enterprise industrial scenes, and commercial/civilian scenes. It is projected that by year-end, over 300 scenario demands will be released, more than 200 application scenarios will be constructed, and over 20 demonstration scenarios will be established. Second, the 2026 development guidelines provide precise,倾斜扶持 (tilted support). In a national first, embodied intelligence data collection has been included as a core supported area. The guidelines support the opening of data collection scenarios and the construction of data collection factories, offering special rewards to data collection enterprises for capacity upgrades, growth, and personnel training. Third, a circulation, trading, and full-process security support system is being established. Leveraging the Guiyang Big Data Exchange, a dedicated trading zone for embodied intelligence data will be set up, improving mechanisms for dataset rights confirmation, compliance review, and trade matching. A unified "five-in-one" registration mechanism will be established, utilizing technologies like data sandboxes, encrypted computing, and blockchain traceability to ensure original data does not leave the domain and is "usable but invisible," thereby solidifying the bottom line for data security. Fourth, professional capacity is being built by relying on provincial digital intelligence industrial parks. Using the Gui'an Optical Valley Artificial Intelligence Data Factory as a carrier, the aim is to create an integrated professional platform for scene collection, multimodal data annotation, dataset trading, and security governance. This will attract leading embodied intelligence enterprises such as Paxini, Zhiyuan Robotics, and Zhishen Technology to cluster, driving the formation of a coordinated development pattern across the province featuring "centralized collection in data parks + distributed scene collection in urban and rural areas." The ultimate goal is to establish Guizhou as a new national highland for the supply of high-quality embodied intelligence data.