EC2 C9g and C9gd instances powered by Graviton5 now available

AWS made its EC2 C9g and C9gd instances, powered by the new AWS Graviton5 processor, generally available. The instances deliver up to 25% better compute performance than Graviton4-based instances, a 5x larger cache, and what AWS claims is the fastest memory of any cloud processor instances, with the C9gd variant adding local NVMe storage.

Graviton5 is the latest in AWS's line of custom Arm-based server chips, which the company has used to steadily reduce reliance on x86 processors and pass efficiency gains to customers. The 5x cache increase is particularly relevant for memory-bound workloads, and the emphasis on memory bandwidth signals AWS's focus on data-intensive and AI-adjacent compute.

The launch continues AWS's vertical-integration strategy across silicon (Graviton, Trainium, Inferentia), infrastructure, and services — a hedge against Nvidia dependence for general compute and a cost lever for customers facing rising AI bills. It also lands as compute allocation becomes a strategic weapon industry-wide, with Google capping Meta's Gemini access over shortages.

For customers, the practical upside is better price-performance for compute-heavy workloads, though realizing gains requires Arm-compatible software. The Graviton ecosystem is now mature enough that most common workloads run without friction, and the C9gd's local NVMe makes it attractive for I/O-intensive applications like databases and analytics. As with prior Graviton generations, the real test is sustained real-world benchmarks versus AWS's headline claims.