Amazon Web Services (AWS) has announced the general availability of its new Amazon EC2 C9g and C9gd instances, which are powered by the latest-generation AWS Graviton5 processors. These compute-optimized instances are designed to support high-demand workloads such as real-time analytics, video encoding, scientific modeling, and CPU-based machine learning inference. According to the company’s official announcement, the C9g instances deliver up to 25% higher performance per vCPU compared to the previous-generation C8g instances, aiming to provide improved throughput for compute-intensive applications.
The rollout of these instances marks a significant shift in how AWS handles compute-bound tasks, particularly as artificial intelligence workloads move toward more complex agentic loops—where software systems perform multi-step reasoning and autonomous tasks. These instances are currently available in the US East (Ohio and N. Virginia), US West (Oregon), and Europe (Frankfurt) regions, with plans to expand to additional regions in the future.
Performance Enhancements and Hardware Specifications
The Graviton5 processors integrated into the C9g and C9gd series feature several hardware upgrades designed to reduce data wait times. These include DDR5 8800MT/s DIMMs, which the company describes as the fastest memory available in any cloud processor instance, alongside a 5x increase in L3 cache compared to the Graviton4-based predecessors. The architecture is specifically optimized for workloads that require high packet-processing speeds, with performance metrics indicating up to 3x higher throughput in that category.
For users requiring integrated storage, the C9gd instances provide high-speed, low-latency local NVMe SSD storage. This configuration is intended for scenarios such as temporary scratch space during high-performance computing (HPC) simulations, local buffers for ad-serving engines, or caching for machine learning inference. AWS notes that these NVMe instance store volumes support high-resolution I/O metrics, allowing developers to track latency histograms with up to 1-second granularity through Amazon CloudWatch or the nvme-cli tool at no extra cost.
Both instance families are available in 11 distinct sizes, ranging from medium to 48xlarge, with a bare-metal option for specialized use cases. Performance data released by the provider highlights a 15% increase in network bandwidth and a 20% increase in Amazon EBS bandwidth on average compared to the previous generation. The largest 48xlarge configuration supports up to 100 Gbps of network bandwidth and 72 Gbps of EBS bandwidth, representing a twofold increase in capacity.
Security and The Nitro Isolation Engine
Security remains a primary focus for the C9g and C9gd deployment. These instances are the first compute-optimized types to utilize the "Nitro Isolation Engine," an enhancement to the existing AWS Nitro System. This component is engineered to enforce stricter isolation between virtual machines by mediating access to CPU register states, memory, and I/O devices via a minimal set of application programming interfaces (APIs).
Technical documentation regarding the formal verification results, including the underlying assumptions and scope of the security model, is available through official AWS technical white papers.
Operational Flexibility and Configuration
Beyond raw compute power, the new instances introduce several operational features for enterprise environments. The Instance Bandwidth Configuration (IBC) allows users to adjust the allocation of bandwidth between Amazon EBS and Amazon VPC networking by up to 25%. This granularity is intended to help database administrators and developers optimize performance for specific caching or database requirements. Furthermore, the instances support ENA Express for enhanced networking, and users can attach up to 128 EBS volumes to a single virtual instance.
Pricing and availability details for the new C9g and C9gd instances can be found on the official Amazon EC2 pricing page. The company has confirmed that these instances support a variety of billing models, including On-Demand, Spot Instances, Savings Plans, and Dedicated Hosts, providing flexibility for organizations balancing cost and compute performance.
Users interested in the specific performance metrics or configuration guides are encouraged to visit the official product page for the new C9g and C9gd instances. Feedback on the deployment can be submitted through the AWS re:Post community forum or via standard AWS support channels. As the company continues its regional expansion, further updates regarding global availability are expected to be shared through official channels.