Amazon EC2 R9g and R9gd instances on Graviton5 reach general availability

AWS announced general availability of the memory-optimized Amazon EC2 R9g and R9gd instances, powered by its fifth-generation Graviton5 processors. The company claims up to 25% better compute performance than the prior R8g generation, targeting memory-intensive workloads such as relational and NoSQL databases, in-memory caches like Redis and Memcached, and real-time analytics engines. The R9gd variant adds local NVMe storage for workloads that need fast instance-attached disk.

The Graviton5 refresh continues AWS's long-running strategy of using in-house Arm silicon to undercut x86 economics and reduce its dependence on Intel and AMD. Every generation that widens the price-performance gap strengthens AWS's ability to offer cheaper compute for the database and caching tiers that underpin most cloud applications — including the retrieval and vector workloads feeding AI agents.

While not a headline AI announcement, the timing is relevant: agentic and RAG workloads are memory-hungry, and cheaper high-memory instances lower the cost floor for running Bedrock Knowledge Bases, vector stores and real-time inference-adjacent data layers. AWS is quietly building the infrastructure cost advantage beneath its AI services.

What to watch: independent price-performance benchmarks versus comparable x86 and the Graviton4-generation instances, and how quickly the new instances roll out beyond the initial launch regions. For cost-conscious teams already on Graviton, R9g is a straightforward upgrade path for the data tier.