China's 2026 national Outstanding Smart Factory results have been rolling out since October 28, according to reports from Huashang Wang, Autohome and China Household Electric Appliance Research Institute media: NIO's advanced manufacturing plant won the Excellent-level designation, Midea's compressor unit MeiZhi was also named, and Xiaomi's AI-driven super factory entered the MIIT Lighthouse tier.
The Outstanding tier sits near the top of MIIT's tiered smart-factory program — Basic, Advanced, Excellent and Lighthouse — which scores plants across design, production, management and supply chain. This year's list is strikingly broader: EV manufacturing, home-appliance components and consumer electronics all appear in the same batch, a sign that AI adoption is spreading beyond a few showcase lines.
Xiaomi's entry offers the clearest window into what the graders reward. Technical details disclosed the same day show its robotics team built the program around the flexible-manufacturing scenario of its car plant, collecting and validating data across 1,827 standard operating procedures. Simulation ablations found that a dexterous hand with 17-21 degrees of freedom lifted average stable grip success by 15%-38% over 6-15 DOF designs, while gains flattened above 23 DOF. Xiaomi settled on a 21-DOF hand with hybrid actuation and tendon-tension control.
NIO and MeiZhi underline the list's industry spread — one an EV manufacturer, the other a Midea-group appliance-components supplier. For both, the test of intelligence has moved past single-machine automation to whether data actually flows between R&D, production and supply chain.
Chen Duan of Central University of Finance and Economics, commenting on Xiaomi's hand, put it as “scenarios define products”: tactile, force-position and assembly-timing data captured on the shop floor is scarce, scenario-specific capital that trains not just the device but the wider AI system. The next round of competition among smart factories will be decided less by equipment counts than by how reusable a plant's data turns out to be.