On June 30, 2026, AWS announced the general availability of Amazon EC2 C9g and C9gd instances, powered by AWS Graviton5, its fifth generation of custom-designed processors. For a company already running compute-intensive workloads on AWS —real-time analytics, batch processing, machine learning inference— this launch matters because it changes the performance-to-cost relationship without requiring an application rewrite.

What is AWS Graviton5

AWS Graviton is the family of Arm-based processors AWS designs for its own Amazon EC2 instances. Each generation aims for the same thing: more performance per vCPU and better energy efficiency than the previous one. Graviton5 is the fifth generation of that family, and the C9g and C9gd instances are the first to feature it.

What C9g and C9gd instances bring

According to AWS’s official announcement, C9g instances deliver:

  • Up to 25% higher performance per vCPU compared to previous-generation C8g instances.
  • 5x more L3 cache, reducing the time workloads spend waiting on data.
  • The fastest memory of any AWS compute instance, with DDR5 at 8800 MT/s.
  • Up to 3x higher network packet-processing performance compared to Graviton4-based instances.
  • Available in 11 sizes, from medium to 48xlarge, plus a bare metal option.

C9gd adds high-speed, low-latency local NVMe storage —useful for temporary workspace during HPC simulations, ML inference caches, or local buffers for ad-serving engines— with up to 30% higher storage performance than the previous generation of local-storage instances.

A new security layer: Nitro Isolation Engine

C9g and C9gd are the first compute-optimized instances to feature the AWS Nitro Isolation Engine, a component of the Nitro system built in Rust that isolates virtual machines from one another and mediates all access to memory, CPU registers and I/O devices through a minimal set of APIs. It is an additional security layer, with formal verification documented by AWS, added with no extra configuration required.

Which workloads benefit

AWS’s announcement points to concrete use cases where this performance jump shows up:

  • Real-time analytics and distributed processing, where faster memory and larger cache reduce time spent waiting on data.
  • Batch processing and video encoding, sustained compute workloads that benefit directly from higher performance per vCPU.
  • CPU-based machine learning inference, an option worth considering when a dedicated GPU is not required.
  • Agentic AI workloads, where reasoning steps and concurrent task execution rely on CPU compute and benefit from Graviton5’s higher core count and larger cache.

Where it is available today

At launch, C9g and C9gd instances are available in US East (Ohio and N. Virginia), US West (Oregon) and Europe (Frankfurt), with more regions being added progressively. For companies in Peru, that is concrete good news: AWS Virginia, the region we recommend for its more direct fiber connectivity for Peru and Chile workloads, is already among the first with access to Graviton5.

What it means for a company already running on AWS

If your workload already runs on Arm-based instances (the Graviton family), moving to C9g or C9gd is usually an infrastructure decision —changing the instance type— rather than a code-rewrite project. For workloads still tied to x86 architecture, the prior step is evaluating compatibility before migrating.

Either way, the underlying question does not change: does this workload benefit from more performance per vCPU, and does that benefit justify the effort of migrating the instance? That is an architecture and cost question, not just a matter of technical specs.

How Caleidos helps

At Caleidos we work with clients to evaluate which workloads genuinely benefit from instances like Graviton5, within a broader cloud strategy. Our Cloud Strategy practice reviews the current architecture and the adoption path, and our FinOps practice helps quantify the real cost-benefit of migrating a workload to a new instance generation before committing to the change.

Frequently asked questions

What did AWS announce? The general availability of EC2 C9g and C9gd instances, powered by AWS Graviton5, on June 30, 2026.

What does it improve over the previous generation? Up to 25% higher performance per vCPU, 5x more L3 cache, and the fastest memory of any AWS compute instance.

Is it available for workloads in Peru? AWS Virginia, the recommended region for Peru and Chile, is already among the first with access to Graviton5.

Want to evaluate if your workload benefits from Graviton5?

Let’s talk about your AWS architecture and we’ll help you evaluate the cost-benefit of migrating to the new instance generation.