AWS News - 2026-07-23

Amazon SageMaker AI inference now supports G7 instances powered by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, enabling you to deploy machine learning models with up to 4.6x AI inference performance compared to previous-generation G6 instances. Customers deploying generative AI models for production inference need high GPU throughput and memory capacity to serve medium-to-large models cost-effectively, but previous-generation instances often required over-provisioning expensive compute or quantizing models to fit within memory constraints.

G7 instances provide 32 GB of GPU memory per GPU with 5th Generation Tensor Cores, up to 700 Gbps of EFA-enabled networking (7x compared to G6), and up to 7.6 TB of local NVMe SSD storage for keeping large models close to compute. These capabilities make G7 instances well suited for serving models in the 7B–30B parameter range, image and video generation workloads, and multi-model inference endpoints that benefit from higher memory bandwidth and throughput. You can deploy models on G7 instances using the SageMaker AI Inference console, API, or SDK by specifying G7 instance types (such as ml.g7.xlarge through ml.g7.48xlarge) in your endpoint configuration.

G7 instances for SageMaker AI inference are available in US East (N. Virginia, Ohio) and US West (Oregon). For pricing information on these instances, please visit our pricing page.

AWS Organizations now supports up to 2,000 resource control policies (RCPs) per organization, doubling the previous limit of 1,000. RCPs enable centralized management of the maximum permissions available to resources across member accounts in your organization. This quota increase helps customers managing large, complex, multi-account environments define more granular resource access controls without encountering policy limits. 

With RCPs, you can restrict which external principals can access resources across your organization's member accounts, helping enforce organization-wide access control guidelines at scale — without updating individual resource-based policies. This is especially valuable for organizations that require fine-grained, centralized permission management across hundreds of accounts, where the previous 1,000-policy limit created barriers to sufficiently detailed access control configurations. 

The increased quota of 2,000 RCPs per organization is available at no additional cost in all AWS Regions where AWS Organizations is supported. No action is required — existing organizations automatically have access to the higher limit. 
 To learn more about resource control policies and managing quotas in AWS Organizations, visit the AWS Organizations documentation.

On July 22, 2026, Amazon announced quarterly security and critical updates for Amazon Corretto Long-Term Support (LTS) and Feature Release (FR) versions of OpenJDK. Corretto 26.0.2, 25.0.4, 21.0.12, 17.0.20, 11.0.32, and 8u502 are now available for download. Amazon Corretto is a no-cost, multi-platform, production-ready distribution of OpenJDK.

Starting with this release, the default Amazon Corretto Docker images are based on Amazon Linux 2023. Amazon Linux 2 images will continue to be provided as non-default options for customers who cannot yet migrate.

JavaFX binaries are no longer included with Corretto 8 starting from this release. You can learn more about the migration recommendations at Corretto 8 GitHub.

Visit Corretto home page to download Corretto 26, Corretto 25, Corretto 21, Corretto 17, Corretto 11, or Corretto 8. You can also get the updates on your Linux system by configuring a Corretto Apt, Yum, or Apk repo.

Feedback is welcomed!

AWS Lambda durable functions now supports encryption of durable execution data with an AWS Key Management Service (AWS KMS) customer managed key. Lambda durable functions lets you build long-running, reliable workflows directly in your Lambda function code with automatic state management. Lambda encrypts execution state at rest by default with an AWS owned key. Now with support for AWS KMS, you can choose and manage the encryption key yourself.

If you operate in regulated industries such as financial services or healthcare, your data governance policies may require customer-owned encryption keys. You can now configure a customer managed key for durable execution data, giving you control over key rotation and who can access execution history and state. The durable execution key operates independently of the function-level key that protects environment variables and SnapStart snapshots, so you can manage access to execution data separately from function configuration.

This feature is available in all AWS Regions where Lambda durable functions is available. Standard AWS KMS charges apply for customer managed keys. There are no additional Lambda charges for this feature.

To learn more, see Encrypting Lambda durable execution data in the AWS Lambda Developer Guide. 

AWS Entity Resolution now supports real-time matching with advanced matching workflows, enabling customers to match records in milliseconds using complex rulesets through the GenerateMatchId API. Previously, real-time matching was limited to simple rule-based workflows, while advanced rulesets—which support operators like Exact and ExactManyToMany combined with AND/OR logic—could only be used for batch processing that took minutes to hours. This created a critical gap for customers needing real-time entity resolution with sophisticated matching logic.

With this launch, customers performing fraud detection, real-time account lookup, or website personalization can define advanced matching rules and get results in real-time without maintaining separate matching infrastructure or re-architecting applications. To enable advanced real-time matching, customers set the enableRealTimeMatching parameter to true on their matching workflow, then call the existing GenerateMatchId API—no new endpoints or migration required.

Advanced real-time matching is available in all AWS Regions where AWS Entity Resolution is available. 

To get started, see Using GenerateMatchId in the AWS Entity Resolution User Guide.  

For more information about AWS Entity Resolution, visit the product page.

Network Load Balancer (NLB) now supports listener rules allowing you to route connections to different target groups based on the source IP address type. With listener rules, a single dual-stack NLB sends IPv6 client traffic to IPv6 targets and IPv4 client traffic to IPv4 targets, preserving the original client IP address end to end for both address families.

Previously, serving both IPv4 and IPv6 clients from one NLB meant accepting a tradeoff: either run two separate load balancers (one per IP version) and split clients with DNS, or send all traffic to one target group and lose the original client IP through protocol translation. Listener rules remove that tradeoff by enabling conditional routing at Layer 3, directing each connection to a same-family target group with no translation and no additional infrastructure.

You can add listener rules to existing dual-stack NLBs without recreating them. Rules are supported on TCP, UDP, TCP_UDP, and TLS listeners and work alongside existing NLB features including connection draining, target group stickiness, cross-zone load balancing, weighted target groups, and client IP preservation.

Listener rules for Network Load Balancer are available in all AWS commercial Regions and the AWS GovCloud (US) Regions at no additional charge. Standard NLB pricing for load balancer hours and LCUs applies. To get started, see this AWS blog, and the Network Load Balancer User Guide.

Amazon EC2 M8id instances are now available in (Ireland). These instances are powered by custom Intel Xeon 6 processors and deliver up to 43% higher performance and 3.3x more memory bandwidth compared to previous generation M6id instances.

M8id instances offer up to 384 vCPUs, 1.5TiB of memory, and 22.8TB of NVMe SSD storage, 3x more than previous generation instances. These instances deliver up to 46% higher performance for I/O intensive database workloads, and up to 30% faster query results for I/O intensive real-time data analytics than previous sixth-generation instances. Additionally, these instances support Instance Bandwidth Configuration, allowing 25% flexible allocation between network and EBS bandwidth, allocating resources optimally for each workload.

M8id instances are well-suited for balanced workloads including application servers, microservices, enterprise applications, and small to medium databases.

Customers can purchase these instances via Savings Plans, On-Demand instances, and Spot instances. For more information visit the Amazon EC2 instance type page.

Starting today, the general-purpose Amazon EC2 M8a instances are available in AWS Asia Pacific (Hyderabad) region. M8a instances are powered by 5th Gen AMD EPYC processors (formerly code named Turin) with a maximum frequency of 4.5 GHz, deliver up to 30% higher performance, and up to 19% better price-performance compared to M7a instances.

M8a instances deliver 45% more memory bandwidth compared to M7a instances, making these instances ideal for even latency sensitive workloads. M8a instances deliver even higher performance gains for specific workloads. M8a instances are up to 60% faster for GroovyJVM benchmark, and up to 39% faster for Cassandra benchmark compared to Amazon EC2 M7a instances. M8a instances are SAP-certified and offer 12 sizes including 2 bare metal sizes. This range of instance sizes allows customers to precisely match their workload requirements.

M8a instances are built using the latest sixth generation AWS Nitro Cards and ideal for applications that benefit from high performance and high throughput such as financial applications, gaming, rendering, application servers, simulation modeling, mid-size data stores, application development environments, and caching fleets.

To get started, sign in to the AWS Management Console. Customers can purchase these instances via Savings Plans, On-Demand instances, and Spot instances. For more information visit the Amazon EC2 M8a instance page.

Starting today, the compute optimized Amazon EC2 C7a instances are now available in AWS US West (N. California) Region. C7a instances, powered by 4th Gen AMD EPYC processors (code-named Genoa) with a maximum frequency of 3.7 GHz, deliver up to 50% higher performance compared to C6a instances.

C7a instances offer new processor capabilities such as AVX-512, VNNI, and bfloat16. They feature Double Data Rate 5 (DDR5) memory to enable high-speed access to data in memory and 2.25x more memory bandwidth compared to C6a instances, making these instances ideal for even latency sensitive workloads. C7a instances offer 12 sizes from medium to 48xlarge, including a bare-metal size. And with the launch of C7a instances, customers can attach up to 128 EBS volumes to an EC2 instance — by comparison, C6a instances allow up to 28 EBS volume attachments to an EC2 instance. These instances are built on the AWS Nitro System and ideal for high performance, compute-intensive workloads such as batch processing, distributed analytics, high performance computing (HPC), ad serving, highly-scalable multiplayer gaming, and video encoding.

C7a instances are available through On-Demand, Spot Instances, and Savings Plans. To get started, visit the AWS Management Console, AWS Command Line Interface (CLI), and AWS SDKs. To learn more, see C7a instances.

Amazon EC2 High Memory U7in-16TB instances (u7in-16tb.224xlarge) are now available in AWS South America (São Paulo) region, and U7in-24TB instances (u7in-24tb.224xlarge) are now available in AWS Europe (Ireland) region. U7i instances are part of the AWS 7th generation and are powered by custom fourth-generation Intel Xeon Scalable processors (Sapphire Rapids). U7in-16TB instances offer 16 TiB of DDR5 memory, and U7in-24TB instances offer 24 TiB of DDR5 memory, enabling customers to scale transaction processing throughput in a fast-growing data environment.

Both U7in-16TB and U7in-24TB instances deliver 896 vCPUs and support up to 100 Gbps of Amazon EBS bandwidth for faster data loading and backups, 200 Gbps of network bandwidth, and ENA Express. U7i instances are ideal for customers running mission-critical in-memory databases like SAP HANA, Oracle, and SQL Server.

To learn more about U7i instances, visit the High Memory instances page.

AWS Wickr now offers a managed Data Retention Service feature for Premium users, enabling organizations to retain conversations across their network for data archiving purposes. AWS Wickr is an enterprise-grade, secure collaboration product that provides end-to-end encrypted messaging, file management, screen sharing, and voice/video conferencing capabilities. The Data Retention Service feature provides a cloud-native alternative to traditional container-based data retention methods.

The Data Retention Service can retain conversations in your network, including direct messages and conversations in Groups or Rooms between internal members and external federated teams. The serverless architecture offers simplified deployment, managed infrastructure, automatic scaling, and comprehensive monitoring while maintaining Wickr's end-to-end encryption standards. This feature is particularly valuable for organizations that require comprehensive data archiving capabilities and maintaining audit trails.

AWS Wickr Premium customers can opt in to enable data retention for their networks. To learn more, visit the AWS Wickr documentation.

Today, AWS announces the general availability of node lifecycle actions in AWS Parallel Computing Service (PCS). With node lifecycle actions, you can run custom scripts automatically at defined points in a compute node's lifecycle. You can use them to prepare your nodes for work. For example, you can mount shared storage, join a directory service, install software, or set up monitoring.

You define node lifecycle actions in your PCS compute node group configuration when you create or update the group, and you can reuse the same script across multiple compute node groups and clusters. For each script, you set its location as an Amazon S3 or HTTPS URI, the arguments to pass, which lifecycle stage it runs in, whether it re-runs on reboot, and its error-handling behavior. AWS PCS writes the output to a dedicated log file, giving you visibility into what ran.

AWS PCS is a managed service that simplifies running and scaling high performance computing (HPC) workloads on AWS using Slurm. You can build complete, elastic environments that integrate compute, storage, networking, and visualization tools, and the service manages cluster updates and provides built-in observability.

Node lifecycle actions are available in all AWS Regions that support AWS PCS. To learn more, see the AWS PCS User Guide.

We are pleased to announce the availability of Amazon EC2 G6 instances in the AWS GovCloud (US-East) region on Amazon SageMaker AI inference. G6 instances are powered by up to 8 NVIDIA L4 Tensor Core GPUs, each with 24 GB of memory, and third-generation AMD EPYC processors, delivering up to 2x the deep learning inference performance compared to G4dn instances.

With this region expansion, government agencies and organizations operating in GovCloud can deploy inference endpoints on G6 instances to serve generative AI workloads—including small-to-medium language models, image generation, and computer vision tasks—while meeting strict compliance and data residency requirements. G6 instances offer strong price-performance for production inference workloads that fit within 24 GB of GPU memory.

G6 instances for SageMaker AI inference are now available in AWS GovCloud (US-East), in addition to previously supported regions. For pricing information on these instances, please visit our pricing page.

AWS Summit Japan 2026 で展示した「スマートグラス × 音声 AI エージェント」による店舗業務支援デモを紹介します。Amazon Nova 2 Sonic と Amazon Bedrock AgentCore Gateway を中核に、声で話しかけるだけで AI が音声と AR 表示で即答する体験を、システム構成や技術的なポイントとともに解説します。

Healthcare organizations face a critical scaling challenge – how to maintain deeply personalized patient interactions as member bases grow, without overwhelming care teams or compromising quality. At Pelago, a digital health company specializing in substance use disorder support, the engineering team found a way to build an AI-powered solution to address this challenge using AWS […]

AI Teammates are agentic AI on Amazon Bedrock, and few engineering organizations run them in production at the scale that monday.com does. Nine in ten Builders use AI coding tools every month, up from roughly half a year ago. Per-engineer PR throughput is up by more than half. Every figure in this post comes from monday’s own internal production data. In this post, we share the architecture behind those numbers, the retrofits that made it work in a decade-old code base, and the confidence-scored merge play closing the gap to full autonomy.