en.Wedoany.com Reported - AWS announced that the Amazon EC2 M9g and M9gd instances, powered by the Graviton5 processor, are now generally available. This processor is specifically designed to meet the demands of agentic AI workloads, including real-time inference, code generation, and multi-step task orchestration.

Previewed at the re:Invent conference in 2025, Graviton5 features a core architecture with 192 cores in a single package, along with a 5x larger L3 cache, DDR5-8800 memory, and PCIe Gen 6 support. Compared to the previous generation, M9g instances deliver up to 25% higher compute performance, with web applications, machine learning inference, and database processing speeds improved by 35%, 35%, and 30%, respectively. The new M9gd instances offer up to 11.4 TB of NVMe SSD storage, with IOPS performance 30% higher than the previous generation.

Since the preview, Meta has adopted Graviton at scale, signing agreements to deploy tens of millions of cores for its agentic AI workloads. Uber and Snowflake are also deploying Graviton for their respective agentic applications. Currently, over 120,000 customers are building applications on Graviton.

Both new instance types are built on the sixth-generation AWS Nitro system, which introduces the Nitro Isolation Engine. This is a formally verified security component that provides mathematically provable isolation between virtual machines. The Graviton5 chip is manufactured using 3nm process technology and is optimized for AWS use cases.

Several customers have provided specific performance data for Graviton5. In internal testing, Airbnb observed that instances based on Graviton5 delivered up to 25% better performance compared to other system architectures of the same generation, and 20% better performance compared to Graviton4 instances, noting impressive P95 latency for critical workloads. After migrating Jira to M9g instances based on Graviton5, Atlassian found a 30% performance improvement and a 20% reduction in latency. In OLTP query tests for SAP HANA Cloud, SAP saw performance improvements ranging from 35% to 60%. Early test results from Synopsys showed runtime improvements of up to 35% for its EDA tools Fusion Compiler and PrimeTime on Graviton5, while its partner Arm observed runtime speedups of up to 40% for Synopsys VCS.

C9g instances for compute-intensive workloads and R9g instances for memory-intensive workloads are planned for launch in 2026.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com