CHINA SYSTEM · OPERATING LOGIC
Read the concept through its operating structure
General, intelligent and supercomputing serve different demand
General compute serves enterprise IT and cloud, intelligent compute AI training/inference, and supercomputing precision scientific workloads.
East-data west-compute is a hub-cluster network
China's eight hubs and ten clusters keep low-latency work near demand while shifting suitable large, non-real-time workloads to resource-advantaged regions.
PUE alone cannot prove green efficiency
PUE reflects facility overhead but not useful GPU work, water, power carbon or hardware life. China's action plan targeted average PUE below 1.5 by end-2025.
Installed and usable compute differ
Construction, rack installation, networking, orchestration, customer testing and commercial operation are distinct. Compute totals require precision, chip generation, utilization and billable-use context.
Korea should design a compute-service market
Public investment should buy usable GPU hours, latency, security, price and SME access, while federating centers and contracting demand, upgrades and residual value.