At the 2026 Technology Innovators Conference held today, Zhang Zhiqi, Chairman and CEO of Weiyi Zhizao, stated that over 80% of projects in the current industry remain stuck in the proof-of-concept stage, struggling to achieve real-world deployment. Citing an IDC survey of hundreds of industrial enterprises, he noted that what companies care about most is not technological leadership, but the stability and reliability of robots, as well as whether the "numbers can be made to work" financially.

He analyzed three core obstacles to deployment. The first is the cycle time issue—industrial workstations operate on cycle times measured in seconds or even fractions of a second, making laboratory-style slow demonstrations unviable in real factory settings. The second is the high manufacturing cost of humanoid bodies with numerous joints, leading to extended payback periods; while prototype projects may be feasible, enterprises are unwilling to pay when it comes to large-scale replication. The third is that anomalies in real factories far exceed laboratory scenarios, requiring systematic engineering refinement.

Furthermore, Zhang Zhiqi emphasized that industrial sites cannot rely on cloud-based inference from day one. Instead, a cloud-edge-end architecture should be adopted, with models running at the edge. Additionally, a human-machine collaboration strategy is necessary—models execute autonomously within their capability boundaries, and when exceeding those boundaries, they automatically raise alarms for human intervention, similar to a human-supervised model in autonomous driving.