Has anyone else noticed a sudden lack of capacity for G6 and G7 instance types in eu-central-1? I've used these instances without trouble before, but this morning a broad range of GPU instance options started returning capacity errors. Other regions, including us-east-1 and us-west-2, still appear to have availability.
3 Answers
I’m seeing the same thing in eu-central-1, while us-east-1 and us-west-2 are working normally. That points to a localized capacity constraint. Providers may be adding capacity, but GPU hardware and data-center expansion can have long lead times.
The increase in AI and rendering workloads is likely contributing. G6 and G7 instances use GPUs that are also popular for machine-learning inference and graphics workloads, so demand can exhaust a region even when capacity is available elsewhere.
This is probably a regional capacity shortage rather than an account-specific issue. GPU demand has grown rapidly because of AI training, inference, rendering, and other workloads, while expanding data-center capacity takes time. It may be worth trying another region or instance type temporarily, then checking eu-central-1 again later.
That makes sense, although it’s hard to tell which specific workloads are driving the shortage from the outside. Regional capacity is probably the main factor regardless.