The National Compute Grid launched on 7 October with one argument. The AI capacity crunch is partly a scheduling problem, because a lot of the hardware is bought and barely switched on.
- About 750 megawatts of capacity is connected or in sight, against a target of 2 gigawatts by 2030.
- Independent single tenant data centers average under 15 percent net compute utilization, on the group’s own estimate.
- Average GPU utilization runs near 5 percent across tens of thousands of Kubernetes clusters, and 2 percent on AKS.
- CPU and memory requests sit 69 percent and 79 percent above what workloads actually consume.
- Members contribute idle machines, reserve larger clusters for planned runs, and pay only for the power drawn.
Read the full analysis of why idle hardware is the cheapest capacity left
Get the next one before it is old news
Independent analysis of cloud-native infrastructure, Kubernetes and data center economics. No vendor spin.