この日は FinOps 向けの AI エージェントと生成AI活用が中心でした。AWS は FinOps 実務者向けの新しいフロンティアエージェント「AWS FinOps Agent」をプレビュー公開し、コスト質問への回答、最適化機会の提示、コスト異常の自動調査、定期ワークフロー実行を可能にします。前日発表の Claude Fable 5 は日本語ブログと News Blog でも取り上げられ、Bedrock と Claude Platform on AWS での提供が紹介されました。コンピュートでは EC2 P6-B200 (NVIDIA Blackwell GPU) が GovCloud (US-East) に、EC2 M8/R8 系の metal-48xl/96xl サイズが一般提供開始。ストレージでは FSx for OpenZFS の Intelligent-Tiering が 8 リージョン追加、SageMaker Unified Studio Notebooks が EMR Serverless に対応しました。ML ブログでは Strands Agents と AgentCore Browser Tool による保険請求の初期受付自動化、Amazon Quick と New Relic を用いたインシデントトリアージ、NVIDIA Isaac Lab を使ったロボット強化学習が解説されました。
FinOps: AWS FinOps Agent のパブリックプレビュー提供開始
生成AI: Claude Fable 5 の Bedrock/Claude Platform on AWS 提供
コンピュート: EC2 P6-B200 の GovCloud 提供、M8/R8 系 metal サイズの一般提供
分析・ストレージ: SageMaker Unified Studio の EMR Serverless 対応、FSx OpenZFS の Intelligent-Tiering 拡大
エージェント活用: 保険請求受付やインシデントトリアージの自動化事例
AWS is announcing the general availability of metal-48xl and metal-96xl sizes for Amazon Elastic Compute Cloud (Amazon EC2) M8in, M8ib, M8idn, M8idb, R8in, R8ib, R8idn, and R8idb instances. These instances are powered by custom sixth generation Intel Xeon Scalable processors, available only on AWS and feature the latest sixth generation AWS Nitro cards. These instances deliver up to 43% better compute performance per vCPU compared to previous generation M6in, M6idn, R6in, and R6idn instances.
M8in, M8idn, R8in, R8idn instances deliver 600 Gbps network bandwidth, the highest network bandwidth among enhanced networking EC2 instances. M8in and R8in instances are ideal for workloads such as real-time big data analytics, distributed web scale in-memory caches, caching fleets for AI/ML clusters, and Telco applications such as 5G User Plane Function (UPF). M8idn and R8idn instances are ideal for network-intensive general purpose workloads requiring local storage, such as distributed compute, data analytics, and high-performance file systems.
M8ib, M8idb, R8ib, R8idb instances deliver up to 300Gbps EBS bandwidth, the highest among non-accelerated compute EC2 instances. M8ib and R8ib instances are best suited for workloads that benefit from high block storage performance, such as high-performance file systems and NoSQL databases. M8idb and R8idb instances are ideal for storage-intensive general purpose workloads such as large commercial databases, data lakes, and NoSQL databases that benefit from both high EBS throughput and low-latency local NVMe storage.
M8in, M8ib, M8idn, M8idb, R8in, R8ib, R8idn, and R8idb instances support Elastic Fabric Adapter (EFA) networking on 48xlarge, 96xlarge, metal-48xl, and metal-96xl sizes. EFA networking enables lower latency and improved cluster performance for workloads deployed on tightly coupled clusters.
The new metal-48xl and metal-96xl sizes are available in the AWS US East (N. Virginia) region. To get started, see AWS Management Console, AWS Command Line Interface (AWS CLI), and AWS SDKs.
Today, AWS announces the preview of AWS FinOps Agent, a frontier agent for FinOps practitioners and engineering teams that answers cost questions, surfaces optimization opportunities, automatically investigates cost anomalies, and runs recurring FinOps workflows on a schedule you define.
With the AWS FinOps Agent, you can ask questions about your AWS costs and generate cloud cost reports for finance and engineering teams. The agent surfaces rightsizing, idle resource, and Savings Plans recommendations from AWS Cost Optimization Hub and AWS Compute Optimizer, and can open Jira tickets on your behalf. When a cost anomaly is detected, FinOps Agent can automatically investigate the root cause and can post the findings to a Slack channel, so engineering teams are notified without manual triage.
AWS FinOps Agent (preview) is available in the US East (N. Virginia) Region and includes cost and usage data covering all AWS Regions, except AWS GovCloud (US) Regions and AWS China (Beijing and Ningxia) Regions. AWS FinOps Agent is offered at no additional charge during the preview.
Learn more about AWS FinOps Agent in the User Guide, product details page, and the blog. Get started by visiting the AWS FinOps Agent page in the AWS Management Console.
You can now create Amazon S3 Access Grants in the AWS European Sovereign Cloud (Germany) Region.
Amazon S3 Access Grants map identities in directories such as Microsoft Entra ID, or AWS Identity and Access Management (IAM) principals, to datasets in S3. This helps you manage data permissions at scale by automatically granting S3 access to end users based on their corporate identity.
Visit the AWS Region Table for complete regional availability information. To learn more about Amazon S3 Access Grants, visit our product page.
Amazon SageMaker Unified Studio Notebooks now support Amazon EMR Serverless with Apache Spark Connect, giving data engineers and analysts more flexibility in choosing their Spark runtime for interactive analytics and data engineering workloads. In addition to Amazon Athena Spark, users can now leverage Amazon EMR Serverless as their Spark runtime, selecting the optimal engine based on their requirements.
With this launch, you can run PySpark and Spark SQL on an EMR Serverless Spark Application in Notebook cells. Users can select their Spark runtime from the Notebook side panel, and the selected runtime applies to both Python and SQL cells. Additionally, users can leverage SageMaker Data Agent, the built-in AI assistant, to generate code and execution plans from natural language prompts, accelerating Spark development workflows with EMR Serverless. Organizations can leverage pre-initialized capacity to improve session start times, while benefiting from unified Spark UI monitoring across all supported engines for consistent visibility into job execution and performance. Additionally, EMR Serverless provides VPC connectivity support for workloads requiring network isolation.
This feature is available in all AWS Regions where Amazon SageMaker Unified Studio is available, supporting both SageMaker Unified Studio notebooks and JupyterLab IDE environments. To get started, see Amazon SageMaker Unified Studio User Guide.
You can now create Amazon FSx for OpenZFS file systems with the Intelligent-Tiering storage class in 8 additional AWS Regions across the US, Europe, Asia Pacific, and South America.
FSx Intelligent-Tiering is built for general-purpose file workloads such as file shares, archives, media libraries, and migrations from on-premises HDD storage. It automatically moves your data across three storage tiers (Frequent Access, Infrequent Access, and Archive) based on access patterns, and an optional SSD read cache keeps your active data fast. You get high performance for active workloads and low-cost storage for everything else, paying only for what you store with no capacity to manage. With FSx Intelligent-Tiering, you can save up to 85% compared to the FSx SSD storage class and up to 20% compared to on-premises HDD-based NAS.
With this expansion, the FSx Intelligent-Tiering storage class is now available for FSx for OpenZFS file systems in the following additional AWS Regions: US West (N. California), Europe (London, Stockholm, Spain, Zurich), Asia Pacific (Hyderabad, Seoul), and South America (São Paulo). To learn more, visit the FSx Intelligent-Tiering page and the Amazon FSx for OpenZFS product page, and see the FSx for OpenZFS Region Table for complete regional availability information.
AWS today announces that AWS Cost and Usage Report 2.0 (CUR 2.0) now supports updates to data table configurations via the AWS Management Console and SDK/CLI. This capability allows customers to modify their existing exports to take advantage of new CUR 2.0 features without having to delete and recreate their exports.
Previously, customers configured CUR 2.0 exports with specific table settings — including export content, time granularity, column selection, export format, and destination settings. When AWS introduces new features, such as additional columns and finer row-level granularity, existing export settings intentionally remained unchanged to protect ETL jobs that depended on a stable schema. However, customers who wanted to adopt these new capabilities and were ready for the new schema couldn't simply update their preference in existing export. They had to delete their existing export and create a new one with the new preference. With this launch, customers can update their table configuration directly through the AWS Management Console or SDK/CLI and begin receiving exports with their updated preferences starting from the next scheduled export delivery.
To learn more about this feature, see AWS Data Exports and AWS Billing and Cost Management in the AWS Cost Management User Guide.
Starting today, Amazon Elastic Compute Cloud (Amazon EC2) P6-B200 instances accelerated by NVIDIA Blackwell GPUs are available in AWS GovCloud (US-East) Region. These instances offer up to 2x performance compared to P5en instances for AI training and inference.
P6-B200 instances feature 8 Blackwell GPUs with 1440 GB of high-bandwidth GPU memory and a 60% increase in GPU memory bandwidth compared to P5en, 5th Generation Intel Xeon processors (Emerald Rapids), and up to 3.2 terabits per second of Elastic Fabric Adapter (EFAv4) networking. P6-B200 instances are powered by the AWS Nitro System, so you can reliably and securely scale AI workloads within Amazon EC2 UltraClusters to tens of thousands of GPUs.
P6-B200 instances are now available in p6-b200.48xlarge size in the following AWS Regions: US West (Oregon), US East (N. Virginia, Ohio), AWS GovCloud (US-West) and AWS GovCloud (US-East) Region. To learn more about P6-B200 instances, visit Amazon EC2 P6 instances.
AWS announces the availability of Claude Fable 5 on Amazon Bedrock and Claude Platform on AWS. Claude Fable 5 makes Mythos-level capabilities available to all customers, with strong safeguards designed to make it safe for broader use.
お知らせ 2026年7月からオンラインでサーバーレスに関するワークショップを4件開催します。ぜひ、ご参加くださ […]
英語版ブログ： “ Highlights from the 2026 AWS Life Sciences Sy […]
AWS オムニチャネルフォールバックソリューションに LINE Messenger を統合する方法を解説します。アーキテクチャの変更点、LINE Messaging API チャネルの作成手順、デプロイとテストの手順を説明します。LINE をプライマリまたはフォールバックチャネルとして活用することで、日本・台湾・タイの主要市場のユーザーへのリーチを拡大できます。
本稿は株式会社八十二長野銀行と AWS Japan の共同執筆により、AWS 人材育成プロジェクトを通じて得ら […]
本日、Amazon Bedrock と Claude Platform on AWS で Claude Fab […]
本日 AWS は、AWS FinOps Agent のパブリックプレビューを発表します。これは、コスト異常を調 […]
In this post, you learn how to build an automated, serverless pipeline that converts scanned PDF medical records into FHIR R4-compliant data using Amazon Bedrock Data Automation and AWS HealthLake. We walk through the architecture, explain how each AWS service connects to the next, show you what the pipeline looks like when it runs, and get you deployed in under 20 minutes.
This post shows engineering teams how to apply that principle to one of the most time-sensitive workflows in engineering: incident triage. You will build a custom incident triage assistant agent using Amazon Quick that orchestrates a response with the New Relic Model Context Protocol (MCP) Server and Asana through native integrations. From a single prompt, the Amazon Quick agent investigates the incident, assembles a root cause analysis (RCA) brief with evidence links, and creates a tracked Asana task ready for handoff.
In this post, we demonstrate how a hands-free FNOL intake system combines agents built with the Strands Agents SDK for domain reasoning with Amazon Bedrock AgentCore Browser Tool for live portal interaction. This approach preserves human expertise while removing repetitive screen work.
In this post, we show how to train robot policies for the Unitree H1 humanoid with NVIDIA Isaac Lab on Amazon SageMaker AI across two compute options: Amazon SageMaker HyperPod and Amazon SageMaker Training Jobs.