Amazon GameLift Servers adds opt-in CPU bursting for container fleets

AWS announced CPU burstability for Amazon GameLift Servers container fleets on October 8, 2026. The opt-in feature lets developers pack containers around typical CPU demand while allowing game-server processes to consume idle instance CPU during spikes, according to AWS’s announcement.

From fixed caps to shared headroom

AWS says container fleets previously used a fixed CPU value that controlled both container packing density and a hard runtime cap. That linked placement decisions to runtime headroom: allocating more CPU to accommodate peaks could mean fitting fewer containers onto an instance.

With burstability, game-server processes can use available idle CPU on the same instance during peak demand. AWS identifies physics calculations, AI processing and collision detection as examples of workloads that can produce spikes. The feature is configured per container group definition, according to the launch announcement.

Availability and pricing

AWS says CPU burstability is available in all AWS Regions where GameLift Servers container fleets are supported, excluding China Regions. There is no additional charge for the feature; standard container fleet pricing still applies, AWS states.

Plan for simultaneous peaks

For capacity planning, the important constraint is “available idle instance CPU.” Treat burstability as access to unused capacity, not guaranteed headroom when an instance is already busy.

The deployment question is therefore how often colocated game servers peak together. Before increasing packing density, teams should test synchronized high-load scenarios and measure game-server latency under contention, rather than relying only on average CPU utilization. If servers peak together, they may compete for the same spare capacity that bursting depends on.