AWS Graviton5 doubles core count to 192, introduces formally verified Nitro Isolation Engine for enterprise AI workloads
By
Shashi Bellamkonda
Summary
Amazon Web Services has released Graviton5, its fifth-generation custom CPU, now generally available on EC2 M9g instances. The chip doubles core count from 96 to 192, moves to TSMC's 3nm process, and introduces the Nitro Isolation Engine — the first formally verified (mathematically proven) cloud hypervisor. M9g instances deliver up to 25% higher compute performance than Graviton4-based instances. The article discusses how these architectural changes impact enterprise planning for agentic AI workloads and cloud infrastructure strategy.
Source
Key quotes
· 3 pulledGraviton5 doubles core count from Graviton4's 96 to 192, moves to TSMC's 3nm process, and ships alongside the Nitro Isolation Engine, a hypervisor whose security properties are established through mathematical proof rather than audit.
M9g instances deliver up to 25% higher compute performance than Graviton4-based instances.
For enterprises planning agentic AI infrastructure, both developments affect how they should evaluate AWS compute going forward.
You might also wanna read
AWS Launches Graviton5 Processor with 192 Cores and 25% Performance Improvement
AWS has launched its fifth-generation Graviton5 processor, featuring 192 cores per chip and 5x larger cache than previous generations. The n
AWS Lambda Arm64 vs x86_64 Performance Benchmark Analysis Across Node.js, Python, and Rust Runtimes
This article presents a comprehensive performance comparison between AWS Lambda's Arm64 (Graviton) and x86_64 architectures across multiple
chrisebert.net·7mo agoCloud Hardware Trends Analysis 2015-2025: Network Bandwidth Improves 10x While NVMe Storage Stagnates
This research paper analyzes cloud hardware trends from 2015 to 2025, focusing on AWS and comparing with other cloud providers and on-premis
IBM's Granite 4.1: 8B Parameter Open-Source Model Competes with Models Four Times Its Size
IBM released Granite 4.1, a family of open-source enterprise language models (Apache 2.0 licensed) trained on 15 trillion tokens. The stando

Microsoft Launches Maia 200 AI Accelerator Chip to Compete with Amazon and Google
Microsoft announces the Maia 200, its latest in-house AI accelerator chip built on TSMC's 3nm process. The chip features over 100 billion tr


Comments
Sign in to join the conversation.
No comments yet. Be the first.