AWS brings EC2 R8g to European Sovereign Cloud in Germany

Availability: AWS European Sovereign Cloud (Germany)

AWS announced on October 9, 2026, that Amazon EC2 R8g instances were available starting that day in the AWS European Sovereign Cloud (Germany) region. The announcement expands regional availability for the existing Graviton4-powered family; it is not a new processor or instance-family launch.

AWS positions R8g for memory-intensive workloads, including databases, in-memory caches and real-time big-data analytics. For teams deploying in this region, the change adds another compute option to evaluate against their workload requirements.

Family specifications and performance claims

According to AWS’s announcement, R8g comes in 12 instance sizes, including two bare-metal sizes. The family extends to 48xlarge and up to 1.5 TB of memory, with maximum networking bandwidth of 50 Gbps and Amazon Elastic Block Store (EBS) bandwidth of 40 Gbps. These are family-level ceilings, not specifications that apply to every size.

AWS says the instances use the AWS Nitro System to offload CPU virtualization, storage and networking functions to dedicated hardware and software.

AWS reports up to 30% better performance than Graviton3-based instances and up to 40% faster database performance than Graviton3-based R7g instances. These are vendor-reported comparisons, not guarantees for a particular deployment.

Engineering implications: benchmark before consolidating

The deployment question is whether an available R8g size improves cost at the workload’s required latency and throughput—not simply whether the family offers higher maximum specifications.

Before migrating, confirm regional size availability and account quotas, obtain current regional pricing, and benchmark representative queries, cache behavior and storage traffic. Do not assume that a headline performance uplift translates directly into savings.

If considering consolidation onto fewer, larger instances, include recovery tests and capacity requirements in the evaluation. Assess steady-state performance alongside the amount of work each instance would need to recover after a failure.