Amazon Web Services (AWS) today announced the general availability of Amazon EC2 M9g and M9gd instances, powered by the fifth-generation AWS Graviton5 processors. This milestone marks a significant advancement in cloud computing, delivering unprecedented levels of compute performance, price-performance, and energy efficiency across a broad spectrum of workloads, with a particular focus on the burgeoning demands of agentic artificial intelligence (AI). The M9g instances are designed for general-purpose workloads, while the M9gd instances cater to customers requiring high-speed, low-latency local NVMe SSD storage. Both instance types are built on AWS’s custom-designed Graviton5 silicon, representing the most powerful and energy-efficient processor ever developed by AWS.
A Decade of Innovation: The Graviton Journey to Graviton5
The journey to Graviton5 is a testament to AWS’s sustained investment in custom silicon, a strategic pivot initiated nearly a decade ago. The first AWS Graviton processor was introduced at re:Invent 2018, marking AWS’s entry into the custom ARM-based chip market for cloud infrastructure. This initial offering, while foundational, laid the groundwork for a rapid succession of enhancements. Graviton2, launched in 2019, brought significant performance and price-performance improvements, quickly gaining traction among early adopters. Subsequent iterations, Graviton3 in 2021 and Graviton4 in 2023, continued this upward trajectory, consistently delivering double-digit percentage gains in performance and efficiency with each generation. This continuous evolution has positioned Graviton as a formidable alternative to traditional x86 processors in the cloud, offering compelling advantages for a diverse range of applications.
The strategic rationale behind AWS’s custom silicon program extends beyond mere performance gains. By designing its own processors, AWS gains greater control over the hardware-software stack, enabling deeper integration with its cloud services and optimizing for specific cloud workloads. This vertical integration allows AWS to tailor instances precisely to customer needs, offering better security, performance, and cost efficiency than off-the-shelf solutions. The Nitro System, AWS’s underlying platform for EC2, has been instrumental in this strategy, offloading virtualization functions to dedicated hardware and further enhancing security and performance. The Graviton processors, deeply integrated with the Nitro System, represent the pinnacle of this approach.
Unpacking Graviton5: A Powerhouse for Modern Workloads
The Graviton5 processor, at the heart of the new M9g and M9gd instances, represents a monumental leap forward. Announced in preview at re:Invent 2025, it has since undergone rigorous testing by customers across a wide array of demanding workloads, consistently demonstrating superior performance. Early adopters have reported impressive results: ClickHouse, a popular open-source column-oriented database management system, observed a 36% performance boost compared to previous-generation M8g instances, achieved with zero code changes. Observability platform provider Honeycomb, through a six-month A/B test of production workloads, recorded 36% better throughput per core compared to Graviton4. HubSpot, a leading CRM platform, deployed M9g for its MySQL databases, witnessing query durations drop by a remarkable 60%. These real-world testimonials underscore the tangible benefits Graviton5 brings to critical enterprise applications.
Graviton5 instances offer up to 25% better compute performance compared to Graviton4-based instances. This translates into specific workload accelerations, including up to 35% faster performance for web applications, a significant advantage for interactive and high-traffic services. For machine learning inference tasks, Graviton5 delivers up to 35% faster performance, enabling quicker insights and more responsive AI-powered applications. Database operations also see substantial improvements, with up to 30% faster performance, crucial for data-intensive applications and analytical workloads.
Architecturally, Graviton5 is designed for peak efficiency and power. It features up to 192 cores, a substantial increase that enables parallel processing on an unprecedented scale within the AWS ecosystem. The processor boasts a 5x larger L3 cache compared to its predecessor, minimizing memory access latency and accelerating data retrieval for compute-intensive tasks. Inter-core latency has been reduced by up to 33%, fostering more efficient communication between cores and improving the performance of multi-threaded applications. Furthermore, Graviton5 is the first CPU in the AWS fleet to support the latest generation of PCIe Gen6 and DDR5-8800 memory, delivering the fastest memory of any processor instances in the cloud. These enhancements collectively contribute to Graviton5’s superior performance and energy efficiency, allowing customers to meet ambitious sustainability targets without compromising capability.
Enhanced Connectivity and Storage for Data-Intensive Applications
Recognizing that modern workloads are not solely CPU-bound, AWS has also significantly expanded networking and storage bandwidth in the M9g and M9gd instances to keep pace with the Graviton5’s compute growth. These instances offer up to 15% higher network bandwidth and 20% higher Amazon Elastic Block Store (Amazon EBS) bandwidth on average across various sizes. For the largest instance sizes, network bandwidth can reach up to twice that of previous generations, enabling faster data movement and improved throughput for applications with high I/O demands.
A notable feature introduced with M9g and M9gd is Instance Bandwidth Configuration (IBC). This innovative capability allows customers to adjust the allocation of bandwidth between Amazon EBS and Amazon Virtual Private Cloud (Amazon VPC) networking by up to 25%. IBC provides granular control, enabling optimization for workloads with specific bandwidth requirements, such as database read and write performance, complex query processing, and extensive logging operations. This flexibility ensures that resources are precisely matched to workload needs, enhancing overall efficiency and performance.
For applications demanding high-speed, low-latency local storage, the M9gd instances are equipped with up to 11.4 TB of NVMe SSD storage. These local SSDs provide 30% higher IOPS and storage performance compared to Graviton4-based M8gd instances. This makes M9gd instances ideally suited for general-purpose workloads requiring a balanced blend of compute and memory alongside rapid local storage access. Use cases include application servers, microservices, gaming servers, midsize key-value data stores, caching fleets, data logging, media processing, batch and log processing, and applications that rely on temporary storage like caches and scratch files.
Pioneering Security: The Nitro Isolation Engine
Security and isolation remain paramount for cloud workloads, and with M9g and M9gd instances, AWS is raising the bar even further through the introduction of the Nitro Isolation Engine. As an enhancement to the foundational AWS Nitro System, the Nitro Isolation Engine rigorously enforces instance isolation and leverages formal verification to provide assurances of isolation with mathematical precision.
The Nitro Isolation Engine is a purpose-built component specifically designed to enforce strict isolation between virtual machines. It mediates all access to virtual machine memory, CPU register state, and I/O devices through a minimal set of highly controlled APIs. The cornerstone of its reliability lies in formal verification, a sophisticated technique that mathematically demonstrates that hardware or software behaves exactly as intended under all possible conditions, rather than just in specific test cases. This intensive verification process establishes Nitro as the first formally verified cloud hypervisor, setting a new industry standard for mathematically proven cloud security. This level of assurance is particularly critical for highly sensitive workloads and regulatory compliance, offering customers unparalleled peace of mind regarding the integrity and confidentiality of their data in the cloud.
Graviton’s Expanding Ecosystem and Impact on Agentic AI
AWS Graviton’s footprint has expanded dramatically since its inception, now powering over 350 instance types and serving more than 120,000 customers, ranging from nascent startups to global enterprises. This widespread adoption is supported by a robust ISV partner ecosystem and a broad suite of managed services, ensuring compatibility and ease of migration for a vast array of applications. Customers can leverage Graviton for diverse workloads including web applications, microservices, analytics, databases, machine learning (ML) inference, electronic design automation (EDA), gaming, and video encoding.
The timing of Graviton5’s general availability aligns perfectly with the explosive growth of artificial intelligence, particularly the emergence of agentic AI. As AI models evolve from simply answering questions to executing actions, running code, utilizing tools, evaluating results, and orchestrating complex multi-step tasks, the demand for high-performance CPU compute is soaring. Graviton5 is purpose-built to address this critical shift. With its 192 cores, massive L3 cache, low inter-core latency, and high-bandwidth DDR5 memory, Graviton5 helps AI agents spend less time waiting on CPU-bound steps. This enables them to process more instructions, handle large numbers of concurrent environments, and keep accelerators like GPUs efficiently supplied with data, thereby accelerating the entire AI pipeline.
A prime example of this impact is Meta, which is deploying Graviton at scale, starting with tens of millions of cores, to support its agentic AI initiatives. This makes Meta one of the largest Graviton customers globally. Agentic AI workloads, encompassing real-time reasoning, sophisticated code generation, and the orchestration of multi-step tasks, are inherently CPU-intensive. They significantly benefit from Graviton5’s superior compute performance, larger caches, higher memory bandwidth, and impressive core density, enabling Meta to innovate faster and more efficiently in the rapidly evolving AI landscape.
Instance Specifications and Availability
The M9g instances offer a balanced configuration of one vCPU for every four GiB of memory, making them suitable for a wide range of general-purpose applications such as application servers, microservices, midsize data stores, gaming servers, caching fleets, containerized applications, large-scale Java applications, code repositories, web applications, and agentic AI.
The M9gd instances, as detailed, add high-speed local NVMe SSD storage, providing up to 11.4 TB of capacity. Both M9g and M9gd instances are available in a comprehensive range of sizes, from medium with 1 vCPU and 4 GiB of memory, up to 48xlarge and metal-48xl with 192 vCPUs and 768 GiB of memory, ensuring scalability for virtually any workload requirement. Network bandwidth ranges from "Up to 15 Gbps" for smaller instances to a formidable 100 Gbps for the largest configurations, with corresponding EBS bandwidth scaling up to 72 Gbps.
M9g and M9gd instances are generally available today in key AWS Regions, including US East (N. Virginia), US East (Ohio), US West (Oregon), and Europe (Frankfurt). Customers can acquire these instances through various purchase options, including Savings Plans, On-Demand, Spot Instances, Dedicated Instances, or Dedicated Hosts, providing flexibility to optimize costs based on workload predictability and usage patterns.
Accelerating Migration and Optimization
AWS continues to invest in tools and resources to facilitate the adoption and optimization of Graviton-based instances. The AWS Graviton Getting Started Guide provides comprehensive technical guidance on building, running, and optimizing workloads on Graviton. The Graviton Savings Dashboard helps customers track and measure the significant cost savings achieved by migrating to Graviton. Furthermore, AWS Transform, an AI-powered service, automates code transformations for migrating Java applications from x86 to Graviton-based Amazon EC2 instances, handling compatibility analysis, automated recompilation, dependency updates, and validation, thereby streamlining the migration process.
The general availability of Amazon EC2 M9g and M9gd instances powered by Graviton5 processors marks a pivotal moment in cloud computing. It not only solidifies AWS’s leadership in custom silicon but also provides enterprises with the advanced, efficient, and secure infrastructure needed to power the next generation of AI-driven applications and complex cloud workloads, while also contributing to more sustainable cloud operations.
