Amazon EC2 R9g and R9gd instances powered by AWS Graviton5 processors are now generally available | Amazon Web Services

SEATTLE — In a major development for cloud computing infrastructure, Amazon Web Services (AWS) has announced the general availability of the Amazon EC2 R9g and R9gd virtual server instances. Powered by the groundbreaking AWS Graviton5 processor—heralded by the company as the most energy-efficient silicon architecture AWS has ever engineered—the new memory-optimized instances deliver up to a 25% boost in compute performance compared to their Graviton4-based predecessors, the R8g line.

Designed to address the escalating demands of data-intensive workloads, modern containerized microservices, high-throughput databases, and complex distributed analytics platforms, the R9g family arrives alongside an advanced security paradigm. By integrating the Nitro Isolation Engine (NIE), AWS has introduced what it claims is the industry’s first formally verified cloud hypervisor, establishing a mathematically proven standard for multi-tenant isolation.

As enterprise IT budgets face continuous scrutiny and corporate sustainability mandates demand lower carbon footprints per workload, the rollout of Graviton5 marks a pivotal inflection point in custom silicon design for the public cloud.

Executive Overview

The debut of the Amazon EC2 R9g and R9gd instance families represents the convergence of three distinct technological imperatives in modern enterprise computing: extreme compute density, uncompromising energy efficiency, and mathematically rigorous security.

For nearly a decade, AWS has systematically displaced commodity x86 hardware with its proprietary, ARM-based Graviton processor lines. Each successive generation has widened the performance-per-watt gap against legacy architectures. The Graviton5 processor continues this trajectory, yielding a remarkable 25% uplift in raw compute power over the Graviton4 architecture while consuming fewer overall resources.

The R-series instances are specifically tailored for memory-bound applications—databases, in-memory caches such as Redis, Valkey, and Memcached, real-time analytics engines, and large-scale enterprise applications written in languages ranging from C++ and Rust to Java, Python, Go, and .NET Core. By pairing these capabilities with faster memory subsystems, expanded L3 caches, higher networking limits, and flexible bandwidth allocation, AWS is positioning the R9g lineup as the definitive home for mission-critical enterprise workloads.

Furthermore, the introduction of the R9gd variants—equipped with ultra-fast, low-latency local NVMe-based SSD block-level storage—provides an ideal scratchpad and caching layer for distributed databases and big data pipelines that demand microsecond-level local I/O responses.

Underpinning these performance gains is the architectural foundation of the AWS Nitro System. By migrating virtualization overhead, storage control, and network packet processing to dedicated hardware accelerators, AWS achieves near-bare-metal performance. More importantly, the integration of the Nitro Isolation Engine brings formal verification—a mathematical methodology traditionally reserved for aerospace engineering and cryptographic protocols—directly into the cloud hypervisor layer.

Detailed Chronology and Technological Evolution

The journey toward the R9g instance family is rooted in a multi-year strategy by AWS to vertically integrate its hardware stack. This long-term product evolution highlights the steady advancement of cloud silicon engineering.

[Graviton1 (2018)] ➔ [Graviton2 (2020)] ➔ [Graviton3 (2021)] ➔ [Graviton4 (2023)] ➔ [Graviton5 / R9g & R9gd (2025)]

• 2018 – The Inception (AWS Graviton1): AWS introduced its first custom ARM-based processor, aimed at simple, horizontal, scale-out web tier workloads. While modest in compute power, it proved the viability of ARM architecture in the public cloud.
• 2020 – The Enterprise Breakthrough (Graviton2): Delivering a 7x performance improvement over Graviton1, Graviton2 proved that custom silicon could run demanding general-purpose and memory-intensive enterprise workloads efficiently.
• 2021–2023 – Scaling Performance (Graviton3 & Graviton4): Graviton3 introduced DDR5 memory and advanced security features like pointer authentication. Graviton4 subsequently expanded core counts, memory bandwidth, and architectural density, establishing the R8g instance family as a dominant memory-optimized option.
• Late 2025 – The Graviton5 Milestone (R9g & R9gd Launch): The official general availability of R9g and R9gd instances marks the deployment of Graviton5. Alongside significant microarchitectural improvements, this launch debuts the Nitro Isolation Engine, establishing formal verification as a core pillar of AWS infrastructure security.

Hardware Enhancements in Graviton5

Transitioning from the R8g (Graviton4) to the R9g (Graviton5) generation involved structural updates across the entire compute complex:

1. Higher vCPU Performance: Delivering a 25% performance uplift per vCPU, allowing existing applications to scale without requiring horizontal cluster expansion.
2. Expanded L3 Cache: Significantly larger L3 cache allocations mitigate memory latency bottlenecks, directly benefiting databases and in-memory key-value stores.
3. Faster Memory Subsystems: Higher memory bandwidth speeds up data retrieval cycles, a critical bottleneck for analytical queries and real-time machine learning inference pipelines.
4. Energy Optimization: Advanced power management ensures that performance scaling does not come at the cost of proportional thermal or electrical penalties, aligning with corporate net-zero targets.

Supporting Context, Architecture, and Performance Metrics

Understanding the operational profile of the R9g and R9gd instances requires a granular examination of their system specifications, storage configurations, and networking capabilities.

Instance Bandwidth Configuration (IBC)

A standout networking enhancement in the R9g and R9gd series is Instance Bandwidth Configuration (IBC). Enterprise workloads frequently experience competing demands between backend database storage traffic (Amazon EBS) and frontend application networking traffic (Amazon VPC).

IBC empowers systems architects to dynamically adjust the allocation of bandwidth between Amazon EBS and Amazon VPC networking by up to 25%. This fine-grained control prevents resource contention during peak operational hours, ensuring predictable throughput for distributed transactional systems.

The Nitro Isolation Engine (NIE) and Formal Verification

Security in multi-tenant cloud environments has historically relied on hardware-assisted virtualization and robust software boundaries. While effective, traditional hypervisor validation relies on empirical testing—checking expected behaviors against predefined scenarios.

The Nitro Isolation Engine introduces a paradigm shift through formal verification. Formal verification is a mathematical technique used to prove or disprove the correctness of a system’s underlying algorithms against a formal specification across all possible states, rather than just tested cases.

[Guest VM 1]     [Guest VM 2]
     │                │
     ▼                ▼
┌────────────────────────────┐
│  Nitro Isolation Engine    │ ◄── Formally verified via mathematical proofs
└─────────────┬──────────────┘
              │
     [Dedicated Hardware]

Acting as a purpose-built component within the AWS Nitro System, the Nitro Isolation Engine enforces isolation between virtual machines by:

• Mediating every single access attempt to virtual machine memory, CPU register states, and I/O devices through a heavily minimized, highly audited API surface.
• Providing mathematical certainty that memory leaks, side-channel vulnerabilities, or cross-tenant data exposure paths are structurally precluded by design.

This milestone establishes the AWS Nitro System as the first formally verified cloud hypervisor in the industry, setting a new benchmark for cloud security compliance.

Instance Specifications and Comprehensive Technical Data

To accommodate diverse enterprise architectures, both the R9g (EBS-only) and R9gd (local NVMe SSD storage) families are offered across 11 distinct sizing tiers, scaling from lean single-vCPU development nodes to massive metal instances boasting 192 vCPUs and 1,536 GiB of RAM.

Amazon EC2 R9g Specifications (EBS-Optimized)

Amazon EC2 R9gd Specifications (Local NVMe SSD Storage)

R9gd instances mirror the compute, memory, and networking performance of their R9g counterparts while incorporating high-speed, low-latency local NVMe-based solid-state drives. These are ideally suited for temporary scratch space, high-speed transactional caching, and distributed databases that benefit from localized block storage.

Migration Path and Ecosystem Support

Transitioning legacy applications to modern ARM-based architectures has historically presented friction for enterprise engineering teams. However, AWS has engineered the R9g line to support a friction-free migration path from previous generations, particularly the R8g family.

Code Compatibility and Containerized Workloads

For the vast majority of applications migrating from R8g to R9g, zero code changes are required. Standard multi-architecture container images compiled for arm64 execute seamlessly without modification.

• Operating Systems: Comprehensive support is available out-of-the-box for Amazon Linux 2023, Amazon Linux 2, Ubuntu 22.04 and newer, Red Hat Enterprise Linux (RHEL) 8.4+, SUSE Linux Enterprise Server (SLES) 15 SP3+, and Debian 12+.
• Orchestration Platforms: Native integration is provided for Amazon Elastic Kubernetes Service (Amazon EKS), Amazon Elastic Container Service (Amazon ECS), and standard open-source Kubernetes deployments.
• Developer Tooling: AWS provides robust migration enablement tools, including the AWS Graviton Getting Started Guide, the Graviton Savings Dashboard for real-time financial tracking, and AWS Transform, which automates code refactoring and migration workflows for Java applications moving from x86 architectures to Graviton.

Future Outlook and Strategic Implications

The general availability of Amazon EC2 R9g and R9gd instances signals more than just a hardware refresh; it underscores the maturation of custom silicon as the primary competitive differentiator for hyperscale cloud providers.

As enterprises face intensifying pressure to optimize operational expenditure while scaling digital services, the combination of a 25% compute performance leap, enhanced memory bandwidth, and lower energy consumption per vCPU makes Graviton5 an compelling proposition. Furthermore, by introducing mathematically verified hypervisor security via the Nitro Isolation Engine, AWS has raised the bar for enterprise trust in shared cloud infrastructure.

Regional Availability and Purchasing Options

At launch, Amazon EC2 R9g and R9gd instances are immediately accessible in select AWS Regions, including:

• US East (N. Virginia, Ohio)
• US West (Oregon)
• Europe (Frankfurt)

Organizations can procure the new instances via flexible purchasing models tailored to operational needs, including On-Demand, Savings Plans, Spot Instances, Dedicated Instances, and Dedicated Hosts.

Engineering teams can begin provisioning R9g and R9gd workloads immediately through the Amazon EC2 Management Console, utilizing supported Arm-based AMIs. For deep technical documentation, API integration, and architectural guidance, developers can leverage the AWS MCP Server and plugin ecosystem alongside regional support channels.