Chinese Computing Sector Faces Structural Mismatch Amid High Idling Rates

People's Daily reported that China's computing sector faces a stark contrast between tight high-end intelligent resources and severe idleness. Top large model developers have even paused new user subscriptions due to nearing capacity limits, while the national intelligent computing utilization rate sits at only 36.8%, leaving more than six tenths of capacity unused. Traditional data centers, small and medium facilities, and local state-owned computing platforms constitute the primary idle operations, often running at significant losses. Many projects launched without clear customer profiles, discovering local demand fell far below capacity, while some struggled until vendor technicians helped lift utilization past 80%. A structural mismatch underlies the shortage and idling. Cutting-edge tasks like embodied artificial intelligence require advanced chip performance, yet available capacity consists heavily of general-purpose, low-to-mid-end computing that fails both high-end workloads and smaller enterprises needing lightweight products. Cross-regional scheduling remains incomplete because network latency prevents western resources from handling eastern real-time inference. Heterogeneous infrastructure interconnectivity suffers computing efficiency drops, while the absence of trusted measurement and settlement mechanisms hinders multi-entity coordination across eastern and western centers.

Certain computing facilities suffered from sustained low utilization due to internal teams lacking familiarity with underlying algorithm libraries, until resident technical specialists from original equipment vendors were deployed on-site to optimize operations and elevate overall utilization beyond 80%.