Last 7 days
58
Features: 56
Changes: 1
Fixes: 1
Deprecations: 0
Cloud services, infrastructure, AI, databases, and developer platforms.
Latest AWS changelog updates, official release notes, breaking changes, security patches, pricing changes, and developer reactions in one product feed.
Follow this AWS release-notes page to spot useful features, risky migrations, noisy announcements, and source links before they hit your backlog.
Changes.Watch links back to official changelog and release-note sources so summaries stay easy to verify.
Use channels to follow groups of tools around a stack, workflow, or topic.
Rolling windows show how many product updates landed in the last 7, 30, 90, and 365 days, grouped by existing changelog semantics.
58
247
651
651
AWS Local Zone in Las Vegas, Nevada is now generally available, extending core services to the metropolitan area.
Adds support for AWS KMS customer‑managed keys to encrypt data at rest for InfluxDB 2, read replicas, and InfluxDB 3 clusters.
Amazon EC2 P6-B300 instances are now available in the Asia Pacific (Seoul) region.
Introduces automated certificate authority (CA) rotation for Amazon EKS clusters, enabling proactive replacement before expiration.
CloudFront now natively signs requests to Amazon S3 Multi-Region Access Points using Origin Access Control, eliminating the need for custom Lambda@Edge SigV4a signing
Added OAuth 2.0 support via AWS Sign‑In for accessing Partner Central agents through external tools such as Amazon Quick and Kiro.
Adds an RDS Switchover Read Replica execution block to ARC Region switch for automated Oracle Data Guard recovery across regions.
Adds a low‑code/no‑code Generative AI Inference Recommendations experience in SageMaker AI Studio.
Introduces inbound prefix controls, allowing allocation of up to 1,000 IPv4 and IPv6 prefixes per VIF (up from 100).
Redshift now natively retains system table data long‑term by automatically exporting to Amazon S3 Tables in Apache Iceberg format.
Partners can configure category‑based notifications and select delivery channels (email, console mobile app, Slack or Teams via Amazon Q Developer).
Adds the external_web_access parameter to Bedrock Web Search, enabling live retrieval of public web content for up-to-date model grounding.
Web Search now supports per‑request include/exclude domain lists and published‑date range filters, with admin allowlist support and up to 100 domains per list.
EC2 now supports creating AMIs from instances on AWS Outposts using local snapshots.
Copies source log group tags to destination groups in CloudWatch Centralization and keeps them synchronized per rule settings
Added attribute‑based access control (ABAC) for DynamoDB Streams, enabling IAM policies to use stream tags in conditions
Introduces Trusted Identity Propagation (TIP) for SageMaker Notebooks, allowing per‑user IAM Identity Center identities to flow to Lake Formation for fine‑grained data access.
Added three new processors (Amazon RDS log parser, XML parser, GeoIP enrichment) to CloudWatch pipelines
Introduced budget controls allowing users to set a maximum task‑hours limit for penetration tests, with graceful stop and billing only for actual hours used.
Added configurable latency modes (balanced and low-latency) per MediaConnect Router output.
Adds support for monitoring third‑party foundation model spend on Amazon Bedrock via Cost Anomaly Detection
Amazon OpenSearch Ingestion is now generally available in AWS GovCloud (US-East) and (US-West), expanding to 19 AWS regions.
AWS Marketplace now lets customers launch eligible AMIs directly on Amazon Lightsail, adding a simple, fixed‑price deployment option alongside EC2.
Added a fourth Availability Zone (eu-west-2d) to the Europe (London) Region, expanding capacity and fault isolation.
Lambda MicroVMs now available in five new AWS regions (Mumbai, Singapore, Sydney, Frankfurt, Stockholm), expanding total coverage to ten regions.
Introduces a “deny by default” governance setting for custom permissions, automatically blocking new AI capabilities at launch
Added native in-console monitoring for WorkSpaces Applications, delivering real-time session, instance, and network metrics directly in the console.
Default managed policies per IAM role increased from 10 to 20.
Added FIPS 140‑3 validated PrivateLink endpoints for Tape and Volume Gateways, allowing private, encrypted traffic.
R8a instances with 5th‑Gen AMD EPYC processors (up to 4.5 GHz) are now available in the Asia Pacific (Taipei) region.
Bedrock adds SpaceXAI Grok 4.6, a coding‑focused frontier model, accessible via the standard Responses, Chat Completions, and Converse APIs.
Bedrock adds support for OpenAI GPT‑5.6 (Terra and Luna) models in India with in‑country inference.
Introduces outbound IAM identity federation in the AWS European Sovereign Cloud (Germany) region, letting workloads exchange IAM credentials for short‑lived JWTs to access external services.
Critical security patch released for Amazon Corretto LTS and Feature Release versions (8, 11, 17, 21, 25, 26).
GA of AgentCore payments lets AI agents autonomously discover, access, and pay for APIs, MCPs, and content via integrated Coinbase and Stripe wallets.
Introduced data profiling and anomaly detection powered by AWS Glue Data Quality in SageMaker Unified Studio.
Added PythonOperator and BashOperator support for running custom Python functions and shell scripts in Amazon MWAA Serverless.
R8i instances are now available in the Israel (Tel Aviv) region
IAM Policy Autopilot now accepts Terraform plan files as input to generate baseline IAM policies.
Adds pg_stat_autovacuum_scores view and parallel autovacuum workers for improved autovacuum management
Amazon EC2 High Memory U7i-6TB instances are now available in the Europe (Zurich) region.
Adds nested virtualization support for Amazon WorkSpaces Personal and Core Managed Bundles, allowing Docker Desktop, WSL2, KVM workloads, Android emulators, and nested containers directly on Windows and Linux WorkSpaces.
Amazon S3 Metadata and annotations are now available in AWS GovCloud (US-East) and (US-West) regions.
Added support for OpenAI GPT‑5.6 models (Sol, Terra, Luna) on the Bedrock‑runtime endpoint with Responses, Converse, and Chat Completions APIs.
Increased max replication rules per registry from 10 to 25, allowing more granular multi‑region and multi‑account image distribution
Adds batch instance termination to the TerminateInstanceInAutoScalingGroup API, allowing up to 100 instances to be terminated in a single call
Added cluster‑level custom domain name configuration for Amazon MSK Provisioned clusters (both ZooKeeper and KRaft).
R8i and R8i‑flex instances are now available in the Canada West (Calgary) region.
Adds routing steps and agent proficiency reporting to Amazon Connect Customer dashboards
Adds rule hit counts for stateful rules, providing visibility into how often each rule matches traffic.
Announces general availability of Amazon Quick Microsoft 365 extensions for Excel, PowerPoint, Word, and Outlook
Adds automatic semantic enrichment to VPC-enabled OpenSearch Service domains, bringing AI-powered search to private networks.
Added support for 26 additional AWS services, expanding total coverage to 32 services.
Introduces server‑side POI category allowlist and density filtering via the GetStyleDescriptor API.
Added a built‑in visual file editor to CloudShell, launchable with a single `edit` command
Admins can define per-user limit profiles for index storage and agent hours.
Integrated Microsoft Purview DLP into Amazon Quick to enforce sensitivity labels on chats, spaces, and knowledge bases.
Introduces approval policies allowing admins to require approvers to review and approve share requests for assets like knowledge bases, spaces, and custom chat agents
Added customizable "no data" messages for visuals with custom title, subtitle, and hyperlink
Adds support for Oracle Application Express (APEX) version 26.1 in Amazon RDS for Oracle.
Adds ses:custom-path attribute to <a> tags, preserving custom path segments in click‑tracking URLs for deep linking.
Adds Managed Dashboards – pre‑configured, read‑only cost dashboards in the Billing console.
Added full CLI support for AWS Client VPN, enabling scriptable connections and IaC integration.
Added rg.large and rg.12xlarge instance types for Redshift RG in AWS GovCloud (US) regions.
Claude Opus 5 now available via Bedrock in AWS GovCloud (US) regions, using both bedrock-runtime and bedrock-mantle endpoints.
General availability of Amazon Quick Microsoft 365 extensions for Excel, PowerPoint, Word, and Outlook, enabling AI-driven tasks directly within M365 environments
Added optional inclusion of AWS Local Zones in Spot placement score calculations, providing scores for local zones alongside regions and AZs.
Introduces Daybreak Red (GPT-5.6 Cyber) and Daybreak Blue (GPT-5.6 Sol) on Amazon Bedrock for eligible customers, focused on security workloads like vulnerability discovery and incident response.
ACM now allows existing public TLS certificates to switch from email to DNS validation without reissuance, preserving the original ARN.
S3 403 Access Denied responses now include the ARN of the IAM or Organizations policy that caused the denial.
Introduces minimum aggregation thresholds for Custom analysis rule type, letting data providers enforce a minimum distinct identity count per query result.
Added approval policies requiring designated approvers to review and approve share requests for assets like knowledge bases, spaces, and custom chat agents.
Adds per‑user limit profiles for index storage and agent hours, allowing admins to cap usage and control subscription costs.
Added interactive map view to AWS Global View for visualizing Regions and Local Zones
General availability of IAM Role Manager to auto‑create or reuse IAM roles for supported services.
Added support for configuring Kubernetes control plane components (scheduler, controller manager, API server) in Amazon EKS.
Adds a “deny by default” governance setting for custom permissions that automatically blocks new AI capabilities at launch.
Agents can now view and self-assign queued agent‑first callbacks directly from the Worklist, alongside emails, tasks, and chats.
Integrates Microsoft Purview DLP policies into Amazon Quick, enabling enforcement of sensitivity labels on files across chat, spaces, and knowledge bases.
Adds automatic tracking of Amazon EBS persistent volume costs in the Usage Explorer and statistics APIs.
Announces GA of Amazon Nova Multimodal Embeddings in AWS GovCloud (US‑West) as a unified model for text, images, video, and audio.
Adds Amazon Quick's agentic AI capabilities to AWS GovCloud (US-West) for government and regulated industries.
Amazon EC2 R8a instances are now available in the Canada (Central) region.
Added cost allocation by IAM principal (users/roles) for Bedrock‑mantle inference requests
Added LocateAnything-3B, Qwen-AgentWorld-35B-A3B, and Qwen3.5-122B-A10B to SageMaker JumpStart, expanding the catalog of foundation models.
Introduces NVIDIA Nemotron 3.5 Lightning, a 30B‑parameter MoE model (3B active) with up to 4x higher throughput and 30% faster task completion.
Added managed external secrets support for Jenkins API tokens and SonarQube tokens, enabling automatic rotation from the AWS console
Adds one‑click access from the AWS Glue console to SageMaker Unified Studio for catalog browsing, data quality checks, and pipeline building
Added a performance dashboard for Cases to monitor case volume, resolution trends, and SLA targets
Adds ability to export privacy‑enhanced analysis logs for SQL queries to an S3 bucket.
Added Amazon RDS support for MariaDB 12.3 (12.3.2) major version.
FLUX.2-small-decoder and gemma-4-12B-it added to Amazon SageMaker JumpStart catalog
Added Redis langcache-embed-v3-small, JetBrains Mellum2-12B-A2.5B-Thinking, and LightOn LightOnOCR-2-1B models to the SageMaker JumpStart catalog
Amazon EC2 High Memory U7in-24TB instances are now available in the South America (São Paulo) region.
Added GLM-5.2 FP8, NVIDIA Nemotron‑Nano‑12B‑v2, and GLM‑OCR models to SageMaker JumpStart for one‑click deployment.
Added service‑managed shader cache capture and automatic distribution for GameLift Streams sessions
Introduces category-based notification subscriptions allowing buyers to select which notification types (Products, Agreements, Private Offers, Pricing Changes) they receive.
Adds Account Access Manager, letting admins assign IAM roles to workforce users and groups from IAM Identity Center.
Introduces application status checks for EC2 instances, allowing automatic monitoring of web servers, Docker daemons, network configs, and interfaces.
MSK now supports in‑place migration from ZooKeeper to KRaft, letting existing clusters switch without provisioning new infrastructure or client changes.
Increased collection limit per group from 1,500 to 10,000 in next‑gen OpenSearch Serverless.
Adds automatic preservation of UEFI boot mode for Linux servers during disaster recovery, eliminating manual BIOS/UEFI reconfiguration after recovery.
R8i and R8i‑flex memory‑optimized instances are now available in the Europe (Milan) region
Added on‑demand and scheduled backup capabilities for InfluxDB 2 and 3 engines via console, CLI, and API.
Added Amazon Cognito as a core skill (aws-auth) in the Agent Toolkit for AWS, enabling AI coding agents to set up, configure, and troubleshoot Cognito resources.
Adds BGP route protection monitoring and delegated RPKI management to Amazon VPC IPAM for BYOIP prefixes, with a unified dashboard showing ROA status, strength, and overlap detection.
Adds a one‑click option to enable multi‑Region support when creating a new IAM Identity Center organization instance.
Introduces short‑term and long‑term memory for agents, enabling context‑aware, personalized conversations without managing memory infrastructure.
Added support for 21 new EC2 instance types across C8a/C8i/C9g and M8a/M8i/M9g families.
Introduces wall‑clock evaluation windows for CloudWatch metric alarms, letting alarms align to fixed calendar boundaries (e.g., top of the hour, midnight).
Added a 12‑month extension of security and OS patch coverage for legacy Elasticsearch (1.5‑7.8, except 5.6) and OpenSearch (1.0‑1.2, 2.3‑2.9) versions, with an Extended Support surcharge equal to instance pricing starting Nov 7 2026.
Added FedRAMP (Class C & D), SOC 1/2/3, ISO, CSA STAR, PCI, and HIPAA compliance coverage for AWS Parallel Computing Service.
Introduced historical schedule adherence metrics (scheduled, adherent, non‑adherent time, and adherence percentage) on Amazon Connect dashboards.
ECS now supports fractional GPU scheduling on G6f instances, allowing containers to request as little as 0.125 of an NVIDIA L4 GPU
Adds console-to-IDE integration for Kiro and Cursor IDEs, extending the existing VS Code support.
AWS WAF introduces the Salt Security managed rule group for out‑of‑the‑box API, AI agent, and Model Context Protocol (MCP) threat detection.
Introduces EC2 G7 instances with NVIDIA RTX PRO 4500 GPUs in the Europe (Spain) Region.
Introduces GA runtime instances for AgentCore, allowing agents to run on user‑managed EC2 instances.
Added Graviton4‑based M8g, R8g, and C8gn node families for Valkey and Memcached in Amazon ElastiCache.
Added AWS Glue Schema Registry support to ten additional AWS regions across APAC, Americas, and MEA.
Introduces network access controls for OpenSearch UI, enabling restriction of app access to approved VPCs, endpoints, and IPs.
Added one‑click drill‑down on real‑time metrics dashboards for routing profile, queue, agent, and step widgets.
AWS Transform adds automated post‑launch actions in the migration workflow, configurable at the account level and applied across all source servers, including multi‑account migrations.
Adds email‑based MFA support for penetration testing via a unique forwarding address per credential
Adds StorageOperationStatus and StorageOperationPercentProgress fields to monitor storage volume initialization after restores, read replica creation, or AZ conversion
Adds new performance and session health metrics for WorkSpaces Applications to CloudWatch at no extra cost
Adds new performance and health metrics for Amazon WorkSpaces to CloudWatch, covering network, compute, storage, and session lifecycle events.
Sellers can set custom net payment terms (Net 30/45/60/90) on private offers, visible to buyers before acceptance.
Adds Authorizer Log Delivery for provisioned MSK clusters (Standard & Express) at no extra cost
Added support for creating S3 Access Points on backup recovery points, enabling immediate read‑only access via standard S3 APIs without initiating a restore.
Added logically air-gapped vault support for Amazon Neptune in three new regions: Asia Pacific (Melbourne), Europe (Spain), and Europe (Zurich).
SES Open/Click notifications now include an isBotEvent field indicating if the event was likely triggered by a bot or a human.
Adds multi‑dataset topics that let users define relationships across datasets, removing the need for manual pre‑joins.
Introduces temporal policies for stateful agent authorization, enabling workflow sequencing, argument matching, human approval, and data freshness checks.
Amazon EC2 M8g instances (Graviton4) are now offered in the Asia Pacific (Taipei) and Mexico (Central) regions
Added scalable network bandwidth for non-VPC Lambda functions, scaling from 625 Mbps at 2 GB memory up to 3 Gbps at 10 GB.
Amazon Keyspaces (for Apache Cassandra) is now available in the Canada West (Calgary) region (ca-west-1).
Introduces a new observation mode for Glue Data Quality that reduces false positives and improves anomaly detection accuracy
Introduces AI Insights in the pricing section of AWS Marketplace listings to explain pricing in plain language.
GA of native vector search in DynamoDB with single‑digit ms latency and 99%+ recall
Added preview support to export Glue Data Catalog metadata to S3 tables in Apache Iceberg format for SQL querying and time‑travel analysis.
Adds optional AWS account access management when creating a new IAM Identity Center instance
Aurora Serverless now scales to 12 ACUs within a second and up to 256 ACUs, auto‑scaling down to zero when idle.
Adds explicit forward proxy capability to AWS Network Firewall, using the same firewall policy as transparent mode
Add CSV export capability for cases directly from the agent workspace, with filtering and field selection
Introduces GA Web Search tool on Amazon Bedrock for OpenAI GPT models, enabling server‑side search grounding with a single API parameter
Adds interactive Apache Spark sessions via Spark Connect on EMR on EC2 clusters
Added Supply Chain Security as the 10th category in AWS Security Hub Extended
C8g instances are now available in Europe (Paris), Africa (Cape Town), Israel (Tel Aviv), and Canada West (Calgary) regions.
ALB and NLB now support TLS security policies compliant with RFC 9151 for CNSA 1.0 requirements (TLS 1.2/1.3).
Added Amazon S3 Vectors availability in the AWS European Sovereign Cloud (Germany) region.
Adds interval-level capacity planning (15‑minute or 30‑minute) for Voice, Chat, Task, and Email channels
General availability of EC2 I8g storage‑optimized instances in Europe (Paris) and Asia Pacific (Jakarta) regions.
RDS for SQL Server now supports publishing native SQL Server Audit logs to CloudWatch (in addition to S3)
RDS for SQL Server Developer Edition is now available in 13 additional AWS regions.
Added BYOM support for SQL Server in 10 new AWS Regions across APAC, Europe, and Mexico
GA of AWS Transform continuous modernization across all supported regions, enabling large‑scale technical debt analysis and remediation in source code repositories.
GPT-5.6 Sol, Terra, and Luna now support up to 1 million‑token context windows on Amazon Bedrock, allowing full codebases, long documents, and extensive conversation histories in a single request.
Introduces stream URLs that let end users launch unauthenticated, temporary GameLift Streams sessions directly in a web browser.
Added support for Amazon EC2 G7 instances in SageMaker Studio notebooks, enabling up to 8 GPUs and 700 Gbps EFA‑enabled networking.
Adds omicsTimeout runtime attribute for WDL tasks to enforce per‑task execution limits.
Introduces pre‑configured resilience tests in Resilience Hub that automatically inject faults via AWS Fault Injection Service and evaluate recovery against defined policies
AWS Organizations now shows the maximum account quota and its utilization via Service Quotas
Increase max Provisioned Mode SQS event pollers from 2,000 to 10,000, enabling up to 100,000 concurrent invocations per ESM.
Introduces high‑performance Storage Optimized EC2 I7i instances in the Asia Pacific (Thailand) and Israel (Tel Aviv) regions.
Added full fine-tuning support for over 25 open‑source models (e.g., gpt‑oss, Gemma, Llama, Nemotron, Qwen) in SageMaker serverless model customization
AWS Config adds support for 15 new resource types across services such as Bedrock, OpenSearch Serverless, and SageMaker
AWS WAF adds two Miggo Security managed rule groups (High Emerging Application Threats and AI/ML Application Protection) available through AWS Marketplace.
Adds GA offline source transformation for AWS Transform, letting users upload SQL Server DDL files and convert schemas, tables, and stored procedures to Aurora PostgreSQL without a live connection.
Amazon ECR now supports image layers up to 200 GB when pushing via Docker, up from the previous limit.
Adds native support for Teradata Vantage as a data source in SageMaker Unified Studio
Added multi-Region cluster support for Aurora DSQL in four new AWS regions (Stockholm, Spain, Mumbai, Singapore).
Added 1‑year and 3‑year Reserved Instances for R8i and M8i in Amazon RDS for Oracle, delivering up to 53% cost savings.
Adds SearchNearby API for proximity-based POI discovery with GrabMaps data in AP Southeast Asia (Singapore, Malaysia) regions.
Adds managed Java 8, 11, and 17 runtimes on Amazon Linux 2023, plus matching container base images.
Introduces fully managed Prometheus collectors in CloudWatch, removing the need for self‑managed OpenTelemetry collectors.
Added AWS CodeDeploy support in five new regions (New Zealand, Thailand, Taipei, Malaysia, and Mexico Central).
Introduces Context Ontology Accelerator, an open‑source tool to auto‑generate and manage business ontologies for AI agents
C7i‑flex instances with 4th‑Gen Intel Xeon (Sapphire Rapids) are now available in the Europe (Milan) AWS region.
C7i instances (4th‑Gen Intel Xeon) are now offered in Europe (Milan) and Canada West (Calgary) regions.
Adds full Git version control across all SageMaker Studio tools (Query Editor, Visual ETL, Workflows, Notebooks)
Adds BGP route visibility for private, transit, and public virtual interfaces in Direct Connect
Added Graviton‑based rg.large and rg.12xlarge instance types to the trailing track (patch P202+).
IAM Policy Simulator is now integrated into the IAM console, replacing the standalone site
Updated pricing for GPT-5.6 Terra and Luna on Amazon Bedrock, matching OpenAI first‑party rates
General availability of Policy‑Based Routing on AWS Transit Gateway, letting you route traffic based on source/destination IP, ports, and protocol.
MSK Express can now continuously materialize Kafka topics as Apache Iceberg tables on Amazon S3 with built‑in inline compaction and conflict resolution
Adds a fully managed Amazon MSK Express capability to deliver Apache Kafka data directly to Amazon S3 buckets
Grok 4.3 model is now available on Amazon Bedrock in AWS GovCloud (US‑West), expanding model provider options.
Added Gemma 4 model family (31B, 26B‑A4B, E2B) to Amazon Bedrock in AWS GovCloud (US‑West).
Enables in‑place upgrade from Standard to Enterprise Edition for AWS Managed Microsoft AD via console, CLI, or API.
Added support for OpenSearch 3.7 on Amazon OpenSearch Service, delivering vector search performance improvements, 1‑bit scalar quantization, and doc‑values‑based vector retrieval without reindexing.
Added AutoScalingInstanceRefresh as a new CloudFormation update policy that automatically triggers an Instance Refresh when replacement properties are updated.
Added long‑polling support to ExecuteStatement‑related calls via WaitTimeSeconds, reducing client polling.
Added VPC connectivity for the REST API connector, allowing private‑subnet or PrivateLink access to data sources.
AWS Interconnect – multicloud is now generally available, enabling private, scalable connections between AWS and Oracle Cloud Infrastructure (OCI).
Added pre-parse text transformations for query arguments to normalize raw query strings before parsing, closing HTTP parameter pollution gaps.
Adds automatic surfacing of relevant agent evaluation examples when preparing coaching feedback in Amazon Connect Customer.
Adds cross‑account EFS replication in AWS GovCloud (US) allowing file system copies across accounts and regions
Adds multi‑Region replication for IAM Identity Center when using the built‑in directory as the identity source
Increases HPA sync concurrency up to 40× the default, dramatically speeding pod autoscaling
Added support for second‑generation AWS Outposts racks in the Asia Pacific (Mumbai) Region.
Adds a Cost and Usage widget to AWS Console Home in the AWS European Sovereign Cloud (Germany) region.
Enhanced mode adds agent‑based transfers to HDFS, Azure Blob Storage, and self‑managed object storage.
Adds support for Amazon EFS and Amazon FSx for Lustre as source or destination in DataSync Enhanced mode
Added Variant data type support for Apache Iceberg V3 in Amazon S3 Tables, allowing direct JSON/semi‑structured writes without a fixed schema
Neptune adds tag‑based access control (TBAC) for IAM, allowing policies to use resource and principal tags to restrict data‑plane actions.
Added ML‑powered anomaly detection for catalog‑based data quality evaluations
Adds a Distribution Analyzer to AWS Glue Data Quality for generating frequency distribution profiles (histograms for numeric columns and value distributions for categorical, date, and boolean columns) with customizable bin counts.
Added Custom Aspect Ratio support, allowing per‑session resolution settings (320‑4096px) for portrait, landscape, ultra‑wide, and square displays via the DisplayConfiguration parameter.
Amazon EKS adds support for AWS PrivateLink on the cluster OIDC discovery and JWKS endpoint, enabling private VPC access.
Introduces a preview AWS Security Hub MCP App that streams Security Hub exposure findings into Claude Desktop via a local MCP server
Added AdsPersonalizationTimeouts and AdsPersonalizationConcurrency parameters to playback configurations for custom ad request timeouts and parallel ADS requests.
Amazon RDS for SQL Server now supports native backup/restore of TDE‑enabled databases on Multi‑AZ and same‑region read‑replica instances.
Adds audio optimization for agents on Microsoft Azure Virtual Desktop and Windows 365 Cloud PC, redirecting media to the local device for better call quality.
Amazon MWAA adds support for Apache Airflow 2.11.2, enabling creation of new environments or upgrades via the console.
Introduces resource matching to automatically identify and link duplicate FHIR records (Patient, Practitioner, Organization, etc.) in HealthLake.
Customers can now create Host Resource Groups for EC2 Dedicated Hosts without needing Self‑Managed Licenses, simplifying setup especially for EC2 Mac and isolation use cases.
Adds warm throughput scale‑down for on‑demand Kinesis streams, allowing you to lower write capacity as needed
Adds automatic publishing of structured JSON logs for Lambda Managed Instances capacity providers to CloudWatch Logs
Introduces guided setup for SMTP sending via Mail Manager in the Amazon SES console
Introduces All Upfront payment option for 3‑year Redshift Serverless Reservations, delivering up to 50% discount versus on‑demand rates.
Claude Opus 5 now available on AWS via Amazon Bedrock and Claude Platform
Added Claude Opus 4.8 and Claude Sonnet 5 models to Kiro IDE and CLI for AWS GovCloud (US) with 1M context windows and credit multipliers.
Introduces aws-bench, an open‑source benchmark suite for measuring AI agents' accuracy and efficiency on real‑world AWS tasks
Added support for 32 vCPU / 244 GB worker types for SQL analyses
Added ability for EFT‑paying customers to configure automatic credit memo application preferences in the Billing console.
RDS for MySQL adds support for MySQL 9.7 in the Database Preview Environment.
AWS Lambda Durable Execution SDK for .NET is now generally available, allowing C# developers to create resilient, long‑running workflows with Lambda durable functions.
AgentCore now sends traces, prompts, structured logs, and STDOUT to a single per‑agent CloudWatch log group.
Adds native support for Application Load Balancer access, connection, and health‑check logs as vended logs in CloudWatch Logs.
I8ge storage‑optimized instances (Graviton4) are now generally available in the EU (London) and Canada (Central) regions.
Amazon Elastic VMware Service (EVS) is now available in Asia Pacific (Seoul), Europe (Zurich), and Europe (Stockholm) regions
Claude Sonnet 5 is now generally available on Amazon Bedrock in AWS GovCloud (US) regions.
G7e instances are now available in Asia Pacific (Seoul), Europe (London), and Asia Pacific (Tokyo) for SageMaker AI inference
Added Amazon EC2 G6 instance support for SageMaker AI inference in AWS GovCloud (US‑East) region.
Introduces node lifecycle actions for AWS Parallel Computing Service, enabling custom scripts to run at defined compute node stages
Introduces a managed Data Retention Service for AWS Wickr Premium, allowing secure archiving of direct, group, and federated conversations.
U7in-16TB high‑memory instances are now available in the AWS South America (São Paulo) region
Compute-optimized C7a instances are now available in the US West (N. California) region.
General‑purpose EC2 M8a instances are now available in the Asia Pacific (Hyderabad) region.
M8id instances now available in the Europe (Ireland) region, expanding instance offerings
Network Load Balancer now supports listener rules to route traffic to target groups based on source IP address type (IPv4 vs IPv6).
Introduced real‑time matching with advanced rule sets (Exact, ExactManyToMany, AND/OR) via the GenerateMatchId API.
Added support for encrypting Lambda durable execution data with customer‑managed AWS KMS keys.
Security and critical updates released for Corretto 26.0.2, 25.0.4, 21.0.12, 17.0.20, 11.0.32, and 8u502.
AWS Organizations quota for resource control policies (RCPs) increased from 1,000 to 2,000 per organization.
Adds support for G7 instances (ml.g7.xlarge to ml.g7.48xlarge) in SageMaker AI inference, delivering up to 4.6x faster performance than G6.
Added support for EC2 placement groups (cluster, spread, partition) in EKS Auto Mode and Karpenter node pools.
Secrets Manager now publishes secret value change events directly to Amazon EventBridge.
AWS Direct Connect adds 100 Gbps dedicated connections at the Cirion data center in Lima, Peru.
AWS Elastic Disaster Recovery is now available in six additional AWS Regions, bringing total coverage to 36 regions.
M6in and M6idn instances now available in AWS Asia Pacific (Hyderabad) and South America (São Paulo) regions, expanding to 10 regions total.
Add R6in and R6idn instances to the Europe (Paris) and Canada (Central) AWS regions
Added Amazon EC2 C6in network‑optimized instances to the Asia Pacific (Taipei) region.
Added Amazon OpenSearch as a native data source in SageMaker Unified Studio
Added built‑in blue/green, linear, and canary deployment strategies for Amazon ECS in the AWS European Sovereign Cloud.
- Amazon SES launches tiered pricing plans (Essentials, Pro, Enterprise) bundling features at discounted rates
Adds Action Logs to Amazon ECS, delivering timestamped records of service deployments and Managed Daemon updates for deeper visibility.
Supports up to 1.5 billion active metric time series per workspace
Adds support for the latest cumulative and GDR updates for SQL Server 2016‑2022 on Amazon RDS
Enables India‑based sellers to upload and manage GSTIN‑related seller signatures directly in AWS Partner Central
RDS for SQL Server now supports Microsoft SQL Server 2025 (Enterprise, Standard, Developer) with AI integration that lets T‑SQL invoke external REST endpoints.
AWS Partner Central agents now support all funding programs, adding Strategic Collaboration Agreement (SCA) and AWS Growth Initiative (AGI) to the existing four
Added AWS HealthOmics private workflow support in Asia Pacific (Tokyo) and US East (Ohio) regions.
Adds standardized Amazon Bedrock product metadata (model provider, model name, pricing unit, inference type, and feature) to AWS Data Exports (CUR)
R8i and R8i‑flex instances are now available in the Europe (Stockholm, Zurich) regions.
Added IAM user‑identity filtering for CloudTrail VPC endpoint network activity events
Added support for 50+ languages and over 100 new voice options for more natural agentic voice experiences
Introduces EC2 I8ge storage‑optimized instances in GovCloud (US‑East, US‑West) powered by Graviton4 processors.
Added real-time and historical agent‑queue metrics to Amazon Connect Customer analytics dashboards.
Added OneDrive for Business and Google Drive as persistent storage options for multi‑session WorkSpaces Applications fleets.
Adds support for Apache Flink 2.3 in Amazon Managed Service for Apache Flink
AWS launches a new Local Zone in Athens, Greece, now generally available.
Introduces CloudWatch Coding Agent Insights for visibility into AI coding tool usage and ROI across organizations.
Introduces KNFSD File Cache, an open‑source NFS cache (preview) that fronts multiple NFS sources and re‑exports them to AWS clients
Adds ability to create, share, and reuse custom visual transforms in SageMaker Unified Studio visual ETL flows
Adds IAM role credential support for GameLift Streams sessions via RoleArn parameter.
Introduces one‑click migration from legacy OpenSearch Dashboards to OpenSearch UI for both domains and serverless collections
HyperPod now auto-configures partition-level network topology for Slurm clusters, letting each partition use tree or block topology based on its instance types.
Adds support for customer‑managed KMS keys to encrypt CloudWatch Synthetics canary environment variables at rest
Amazon Managed Grafana is now FedRAMP High authorized in AWS GovCloud (US‑East and US‑West) regions.
Adds a Cost Efficiency widget to Billing and Cost Management dashboards, showing efficiency scores over time by account, region, or overall.
EC2 now includes the associated public SSM Parameter Store parameter in the metadata returned by DescribeImages for public AMIs.
PostgreSQL 19 Beta 2 is now available in the Amazon RDS Database Preview environment, enabling evaluation of the upcoming PostgreSQL 19 engine on a fully managed service.
Adds system-generated bucket tags to S3 Event Notifications across all destinations (EventBridge, SQS, SNS, Lambda).
Added water withdrawals data to AWS Sustainability, viewable by region, service, and account on an annual basis.
Added Amazon EC2 High Memory U7in-24TB (u7in-24tb.224xlarge) instances to the Europe (Paris) region.
GA of two new Graviton‑powered Redshift instance types (rg.large, rg.12xlarge) delivering up to 2.4× query speed vs RA3 and 30% lower price per vCPU.
Removed the 30‑day minimum retention for transitioning objects to S3 Standard‑IA and S3 One Zone‑IA, allowing transitions as soon as day 0.
Logically air-gapped backup vaults are now supported in six additional AWS Regions (AP‑Taipei, AP‑Malaysia, AP‑New Zealand, AP‑Thailand, Central Mexico, Canada West‑Calgary).
Added AWS Backup restore testing support to six new regions (Asia Pacific Taipei, Malaysia, New Zealand, Thailand; Mexico Central; Canada West Calgary).
AFT now auto‑reapplies an account’s customizations on OU moves via aft_customization_triggers=["account_move"]
Amazon Aurora DSQL is now in scope for FedRAMP Moderate in the US East (Ohio & N. Virginia) and US West (Oregon) regions.
Added 25 new query commands and functions to CloudWatch Logs Insights for advanced data transformation, statistical analysis, and enrichment.
Introduced Intelligent Tiering for CloudWatch Logs, automatically moving data between Standard, Infrequent Access, and Archive Instant Access tiers based on access patterns.
Added configurable EBS storage size for RabbitMQ M7g brokers, set via Console, CLI, CloudFormation, or CDK.
G7e instances with NVIDIA RTX PRO 6000 GPUs are now available in Europe (Frankfurt, Stockholm) and Asia Pacific (Mumbai) regions
Added Apache Kafka 4.2 support for MSK Express Brokers, including Eligible Leader Replicas enhancements, a new consumer rebalance protocol, and a Streams Rebalance Protocol.
Added ability to import users with pre‑computed password hashes via CSV imports
Added R8gd instances to 12 new regions and M8gd instances to 6 new regions for Aurora PostgreSQL, RDS for PostgreSQL, MySQL, and MariaDB.
Added GA support for Graviton4‑based R8g instances for Aurora and RDS in multiple new AWS regions.
Allows up to four storage modifications per DB instance within a rolling 24‑hour window, removing the previous six‑hour cool‑off restriction.
Added AWS Glue SAP OData connector and zero‑ETL integrations to AWS GovCloud (US‑West & US‑East) regions.
Integrates Amazon OpenSearch Service with the AWS Agent Toolkit, adding a curated skill that lets AI coding agents manage domains and serverless collections via natural‑language commands across migration, operations, search, log analytic...
Added Amazon RDS for Db2 support in five new AWS Commercial regions: Asia Pacific (Thailand, Malaysia, Taipei), Mexico (Central) and Canada West (Calgary).
Introduces self‑managed Amazon S3 bucket storage mode (`S3ObjectStorageMode=REFERENCE`) allowing Lambda to reference code directly without copying.
Adds support for setting an Amazon EBS volume initialization rate on DRS launch templates, enabling faster full‑performance recovery of volumes.
Introduces a lookup processor for CloudWatch Pipelines, allowing log events to be enriched by matching fields against CSV lookup tables.
AWS DRS can now skip unnecessary preparation steps for EC2‑based source servers, reducing recovery time up to 65% for Windows and 40% for Linux.
Introduces a one‑click setup prompt in the Lambda console that configures coding agents with AWS Serverless skills and the Serverless MCP server
M8in, M8idn, M8ib, and M8idb instances are now available in additional AWS regions (US East Ohio, Europe Ireland, and Asia Pacific Tokyo), extending their existing regional coverage.
AWS IAM Identity Center added to FedRAMP Class C scope in multiple US regions.
Connect Customer Outbound Campaigns (SMS, WhatsApp, email) are now available in the Africa (Cape Town) region.
GuardDuty introduces AI Protection, expanding threat detection to AWS AI services (Bedrock, SageMaker) for AI‑specific threats such as anomalous model invocations, cost harvesting, and prompt injection.
Added AI Agent Skills to Amazon Managed Service for Apache Flink, offering guided assistance for tasks like creation, troubleshooting, scaling, monitoring, networking, and cost optimization.
Amazon Aurora DSQL is now available in the Europe (Spain) region for single‑Region clusters.
Adds AI inventory to Security Hub, automatically discovering and cataloging AI workloads across AWS services
Adds automated rollback and bulk PCoIP‑to‑DCV protocol migration for stopped WorkSpaces
Added console UI to copy file shares between gateways, preserving configuration
Added cf.logCustomData() helper to CloudFront Functions for direct logging into CloudFront access logs.
GPT-5.6 Sol, Terra, and Luna are now generally available on Amazon Bedrock via the Responses API, covering three capability tiers from flagship reasoning to fast, cost‑efficient inference.
Amazon Managed Service for Prometheus is now available in the Asia Pacific (New Zealand) region.
Adds Amazon DocumentDB (MongoDB‑compatible) as a new database skill in the Agent Toolkit for AWS.
Added 46 new MongoDB aggregation operators and cursor methods across seven categories (e.g., accumulators, trigonometry, bitwise, arithmetic, data size/type, timestamp, stages).
Introduces Voxtral-Mini-4B-Realtime-2602, a multilingual real-time speech transcription model, to SageMaker JumpStart.
Gemma-4-E2B-it multimodal, instruction‑tuned model added to Amazon SageMaker JumpStart
Introduces OpenAI privacy‑filter model for PII detection and masking in SageMaker JumpStart
Added Qwen3-VL-Embedding-2B and Qwen3-Reranker-4B models to SageMaker JumpStart for multimodal retrieval pipelines
Restoring a Redshift Serverless namespace from a snapshot now retains zero‑ETL and S3 event integrations within the same namespace.
Adds support for custom Amazon Machine Images (AMIs) on SageMaker HyperPod Slurm clusters, letting customers embed security agents, compliance tools, and proprietary libraries.
R8i family (R8in, R8ib, R8idn, R8idb) EC2 instances are now launched in Asia Pacific (Tokyo) and Europe (Frankfurt, Ireland) regions, expanding availability beyond existing US and EU locations.
Introduces an Apache Spark troubleshooting agent for EMR on EKS, providing AI‑driven root‑cause analysis and PySpark code recommendations for failed jobs.
Added AddressNamesMode and AddressTranslations parameters for customizable address formatting and multilingual support across Places APIs.
Added AMI‑based node lifecycle configuration for SageMaker HyperPod Slurm clusters using continuous provisioning.
EKS Auto Mode now integrates with Amazon Application Recovery Controller zonal shift and autoshift
Amazon EC2 G7 instances with NVIDIA RTX PRO 4500 GPUs are now available in the US East (N. Virginia) region (also Ohio and US West).
Adds offline source conversion for SQL Server in DMS Schema Conversion, allowing metadata extraction locally and upload without direct DB connectivity.
Introduced Amazon EC2 I7ie instances in the Asia Pacific (Hyderabad) region.
Adds support for R8g.24xlarge and R8g.48xlarge instances powered by Graviton4 with DDR5 memory.
Added AI agent automation to DMS Schema Conversion via AWS MCP Server, enabling IDE-driven natural language migration workflows.
New organizations created via the AWS Organizations console automatically have SCPs that prevent member accounts from leaving or closing the organization.
Enable partners to share Associate qualifications (specializations, certifications, program enrollments) across connected subsidiary accounts
Added OAuth 2.0 support for connecting AI agents to the AWS MCP Server using AWS Sign‑In.
C8id, M8id, and R8id instances are now available in additional AWS regions (Mumbai, London, Stockholm).
Introduces M8gd and R8gd instances in AWS GovCloud (US-East and US-West) with up to 11.4 TB NVMe SSD storage and Graviton4 CPUs, offering ~30% better performance than Graviton3.
General availability of EC2 Storage Optimized I8g instances in AWS GovCloud (US East & West) regions.
Added ability to replace a running EC2 instance’s root volume using an existing EBS volume
Adds EventBridge integration that emits lifecycle and state change events for Timestream for InfluxDB instances and clusters
Added AWS Client VPN availability in four new regions: Canada West (Calgary), Mexico Central, New Zealand, and Taipei.
Introduces AI‑powered Partner Lead Prospecting for ACE‑eligible AWS Partners to auto‑generate sales plays, call scripts, and email outreach
Adds support for custom asset types in IAM‑based SageMaker Unified Studio domains, allowing any file format to be cataloged.
Adds deep health checks for Slurm clusters with continuous provisioning, enabling proactive GPU health validation on new and existing nodes.
Added BatchWriteRecord API for high‑throughput, multi‑record writes with per‑record error handling and TTL support across online/offline stores
MSK Replicator now supports replication from external Apache Kafka clusters to Amazon MSK Standard brokers (previously limited to MSK Express).
Added 191 new AWS Config managed rules across services like Bedrock, SageMaker, ECS, EKS, RDS, Redshift, S3, and CloudTrail
Introduces free, time‑limited sandbox environments for Builder Center workshops, removing the need for a personal AWS account or credit card.
Added 19 new operators for Bedrock, S3 Tables, S3 Vectors, Glue Data Catalog, and MWAA Serverless to SageMaker Unified Studio Workflows.
Adds Network Scanning to Security Hub for probing resources from the internet and confirming actual public reachability.
M8gd and R8gd Graviton4 instances launched in additional AWS regions (EU Paris, AP Mumbai, EU Stockholm/Milan, AP Osaka/Melbourne/Singapore/Taipei)
Adds Graviton-based rg.4xlarge and rg.xlarge instance types to the trailing track (P201) starting July 7, 2026.
Aurora DSQL change data capture (CDC) is now generally available, streaming real-time DB changes to Kinesis Data Streams.
Adds forecasting, capacity planning, and scheduling for Tasks and Emails alongside voice and chat workloads
Adds NKI 0.5.0 with FP8 scale dtype support, tensor indirection, and zero‑cost NkiTensor view APIs.
Introduced Amazon EC2 High Memory U7i-12tb instances (12 TiB RAM, 896 vCPUs) in the AWS Europe (Zurich) region.
Introduces Stream Session Admin Shell, a secure terminal for live stream session debugging via the new CreateStreamSessionAdminShell API
Amazon S3 Vectors is now available in AWS GovCloud (US‑East) and GovCloud (US‑West) regions.
Adds Azure resource monitoring to AWS Security Hub, delivering unified CSPM, vulnerability management, and response across AWS and Azure environments.
Management fees for GPU‑accelerated ECS Managed Instances are cut by 35% for G‑series and 60% for P‑series and Trainium, effective July 1 2026.
Added a 32 vCPU, 244 GB worker size, doubling the previous maximum of 16 vCPU/120 GB
C8ine instances (sixth‑gen Intel Xeon with Nitro) are now available in the Europe (Frankfurt) region.
Added OpenLineage‑compatible data lineage support for IAM‑based SageMaker Studio domains, covering EMR Spark, Glue, Visual ETL, and notebooks.
RG instances (Graviton-powered) are now available in AWS GovCloud US-West and US-East regions
Adds support for Oracle Database 26ai on Amazon RDS, including new AI capabilities (Select AI, AI Vector Search, RAG) and advanced data features (JSON Relational Duality Views, SQL Property Graphs).
Amazon S3 Express One Zone storage class is now available in the Europe (Frankfurt) Region.
EKS Auto Mode management fees for GPU and accelerated instances are reduced – G-series by 35% and P-series/Trainium by 60% effective July 1 2026.
Added native Azure VM support in Systems Manager via Cloud Connector, auto‑deploying the SSM Agent and showing Azure VMs alongside EC2 instances for unified Session Manager, Automation, Run Command, etc.
Adds support for connecting existing Amazon MWAA environments to SageMaker Studio projects.
Added impact analysis to exposure findings, mapping downstream resources vulnerable via permission paths.
Introduces Disaggregated Prefill and Decode (DPD) for SageMaker HyperPod, separating LLM prefill and decode onto dedicated GPU pools via EFA RDMA
Introduces a self‑service capability to increase or decrease Amazon Cognito provisioned API rate limits on demand via console or new API operations
Adds one‑click integration from Hugging Face to SageMaker Studio, allowing immediate model customization or deployment with a pre‑loaded workflow.
- Introduces declarative policies to centrally enable VPC Encryption Controls in monitor or enforce mode across all VPCs, including future ones
Adds support for VMware Cloud Foundation 9.0 and 9.1 on Amazon EVS
ACM now provides a managed ACME server endpoint for automated issuance of public TLS certificates (45‑day validity) via any ACMEv2 client.
Introduces Service Events for CloudWatch Application Signals, automatically capturing exception, latency, and deployment events without code changes.
AWS CodePipeline is now available in the Asia Pacific (New Zealand) ap‑southeast‑6 region.
Added managed external secrets support for Paddle API Keys and GitLab access tokens in AWS Secrets Manager.
Introduced host kernel selection to choose Amazon Linux 2023 (kernel 6) or Amazon Linux 2 (kernel 4) for on‑demand CodeBuild builds.
Added Terraform module (terraform-aws-sagemaker-unified-studio) for provisioning SageMaker Unified Studio domains via infrastructure-as-code pipelines
X8i instances (custom Intel Xeon 6) are now available in AP Seoul, AP Malaysia, and AP Tokyo regions
Added support for AMD SEV‑SNP on EC2 Dedicated Hosts, enabling confidential‑computing workloads on dedicated physical servers.
Adds AMI versioning to HyperPod, letting admins view exact AMI versions per node and roll back via UpdateClusterSoftware API
Adds support for 8 new AWS resource types across services like API Gateway, EC2, and S3 Vectors.
Increased default runtime quota limits for AgentCore, allowing up to 5,000 concurrent sessions in US East/West regions and 2,500 in other regions
CloudWatch now lets you create alarms directly from log query results, removing the need for intermediate metric filters or custom metrics.
ECS Service Connect now routes traffic to same‑AZ endpoints first, cutting cross‑AZ data transfer and latency.
Adds real-time deployment observability in the Amazon ECS console with a live timeline, health monitoring, and automatic refresh.
Introduces Assurance Assistant in AWS Artifact, an AI‑driven tool that generates citation‑backed answers to security and compliance queries.
Partners can now link existing AWS Marketplace solutions and product listings to co‑sell opportunities in Partner Central.
Added configurable deployment circuit breaker settings for ECS services, allowing custom failure thresholds by count or percentage.
Added cross‑region automated backup replication for Amazon RDS in four new AWS regions, expanding geographic disaster‑recovery options
Added Amazon Bedrock AgentCore support in four new AWS regions: AP Bangkok, AP Malaysia, EU Milan, and EU Spain.
Added three Runtime Monitoring detections (Persistence, PrivilegeEscalation, DefenseEvasion) that trigger on sensitive file modifications in EC2, EKS, and ECS workloads.
Introduces managed A/B testing and multivariate experiment capabilities directly in AWS AppConfig.
Express Mode now accepts custom ECS task definitions, enabling reuse of existing configurations and advanced task-level settings.
Adds ability to rollback an EKS cluster to the previous Kubernetes minor version within 7 days of an upgrade.
Amazon Managed Service for Prometheus is now FedRAMP High and DoD CC SRG Impact Level 4/5 authorized in AWS GovCloud (US), enabling use in sensitive federal and public sector workloads.
AWS Security Agent is now available in Mumbai, Singapore, and São Paulo regions
RDS for Db2 now supports joining self‑managed Microsoft AD domains (on‑prem, AWS, or other clouds) using Kerberos for single sign‑on.
Introduces a new log‑analytics‑optimized engine for OpenSearch (v3.5+) delivering up to 4× price‑performance and up to 70% lower storage.
Added rule‑based redaction for agent screen recordings, allowing specified URLs or applications to be automatically removed.
CloudWatch Logs now automatically enriches log events with AWS resource tags at ingestion time
CloudFormation now performs automatic pre‑deployment validation on CreateStack and UpdateStack operations, catching errors before resources are provisioned
Adds CloudFormation and CDK "express" mode that completes stack operations as soon as resources are configured, bypassing long stabilization checks.
Added dynamic connection scaling for IAM database authentication, letting auth request rates grow with instance resources
Introduces managed in‑place Slurm major version upgrades for existing PCS clusters, allowing jumps up to three versions ahead.
Added dual‑stack mode enabling Neptune clusters to accept IPv4, IPv6, or both connections
Amazon ElastiCache adds T4g node types in Africa (Cape Town), APJ (Jakarta, Osaka), and AWS GovCloud (US‑East, US‑West) regions.
Increases per‑user security profile limit from 2 to 7, enabling agents to have distinct permission sets for multiple business lines.
Added SendRcsMessage API enabling rich media (cards, carousels, images, videos) and interactive suggestion buttons for RCS in all 22 supported countries.
Added microsecond‑accurate time on 26 additional EC2 instance types across all commercial regions via Amazon Time Sync Service.
Announces general availability of Amazon WorkSpaces for AI agents, letting agents securely access and operate legacy desktop applications without code changes.
Introduce "reservations‑then‑balanced" AZ distribution strategy for EC2 Auto Scaling groups
Introduces Capability Insights, an open‑source, self‑hosted dashboard for viewing AWS regional capabilities within your own VPC.
Added container attribute‑based firewall rules for Amazon EKS and Amazon ECS, allowing policies based on namespace, cluster name, labels, and container instance attributes.
Introduces container image caching for SageMaker Inference, pre‑pulling large model images to halve scale‑out startup latency.
Introduces the AI Security Best Practices standard in AWS Security Hub CSPM with 31 automated controls for AI workloads.
Adds Dynamic Instrumentation to CloudWatch Application Signals, enabling live capture of variables, arguments, return values, and stack traces without restarting or redeploying applications.
IAM Identity Center now lets customer‑managed applications act as trusted token issuers to discover user accounts/roles and retrieve temporary credentials programmatically.
Claude Sonnet 5 is now available on AWS through Amazon Bedrock and Claude Platform, offering unified access with AWS billing and authentication.
Added serverless model customization (SFT, DPO, RFT) for Gemma 4 E4B and 31B models
Announces GA of EC2 C9g and C9gd compute‑optimized instances powered by 5th‑gen Graviton5 processors
AWS launches Interconnect - last mile, a fully managed service for linking branch offices, data centers, and remote sites to AWS with a few clicks.
Claude Opus 4.8 is now offered on AWS GovCloud (US) through Amazon Bedrock.
Added OpenAI GPT-5.4 (272K context, 1.2x credit multiplier) to Kiro in AWS GovCloud US-West for complex reasoning, coding, and multi-step agentic workflows.
Adds DDoS Protection client SDKs for C# and Unity, offering token‑based authentication and per‑player traffic limits to safeguard multiplayer sessions.
CloudWatch pipelines now can process and enrich OpenTelemetry metrics during ingestion.
Multiple AWS services (e.g., Bedrock Agents Classic, Cognito Sync, SageMaker AI features) are moving to maintenance mode and will be inaccessible to new customers after July‑August 2026.
Added support for writing SQL query results to intermediate tables within a collaboration
Added GA support for AWS WAF protection on Amazon Bedrock AgentCore Gateway.
MWAA Serverless now validates and accepts subnets shared via AWS RAM, enabling shared VPC usage across accounts
S3 server access logs can now be delivered to CloudWatch Logs with KMS encryption, cross‑account/region aggregation, and alarm support.
Amazon EC2 R8g (Graviton4) instances are now available in five new regions (Thailand, New Zealand, Cape Town, Milan, Calgary)
Compute‑optimized C7a instances (4th Gen AMD EPYC) are now available in the Asia Pacific (Singapore) region.
Amazon EC2 M8a general‑purpose instances are now available in the Asia Pacific (Mumbai) region.
Added FedRAMP High and DoD CC SRG IL‑4/5 approved OpenAI GPT, OpenAI GPT OSS, and NVIDIA Nemotron models to Amazon Bedrock in GovCloud (US).
Adds two VisionHeight managed rule groups—Zero-Day Threat Protection and Tor/Scanner Protection—to AWS Network Firewall.
Amazon EC2 C8in instances are now available in the US East (Ohio) and Europe (Ireland) regions
Kiro now authorized for FedRAMP High and DoD CC SRG Impact Levels 4‑5 in AWS GovCloud (US) regions
Introduced All Upfront and Partial Upfront payment options for 1‑year and 3‑year Redshift Reserved Instances (RG).
Technical support cases for AWS GovCloud (US) are now automatically routed to US‑based, US‑citizen engineers without requiring opt‑in
AWS Backup now copies S3 backups up to 8× faster for large, low‑change buckets using enhanced change tracking
Added support for SQL Server 2019 CU32+GDR (KB5090407) and SQL Server 2022 CU25 (KB5081477) on Amazon RDS Custom.
Add support for the latest GDR updates across SQL Server 2016‑2022 on Amazon RDS
Added Amazon OpenSearch Ingestion availability in the Europe (Paris) region (eu-west-3).
Adds live updates to EMR Serverless application settings (max capacity, custom images) without stopping the app.
Introduces the AWS IoT Device SDK for Swift, now generally available for native development on Apple platforms and Linux.
Adds support for sharing Route 53 Global Resolver DNS views across AWS accounts using AWS RAM.
Adds AWS CloudFormation support via a new AWS::Neptune::GlobalCluster resource to provision and manage Neptune global databases as code.
Added support for tagging CloudWatch dashboards via new Tags parameter in PutDashboard and TagResource/UntagResource/ListTagsForResource APIs.
Add Amazon EC2 High Memory U7in-24TB (u7in-24tb.224xlarge) instances in the Asia Pacific (Seoul) region
Introduces managed syslog ingestion for CloudWatch Logs, enabling firewalls, routers, switches, and Linux servers to send syslog over TCP, TLS, or UDP to a VPC endpoint without agents.
Introduces preview AI‑powered investigations in GuardDuty to auto‑analyze findings and account activity
Added support for EC2 G6e instances on SageMaker notebook instances, offering up to 8 NVIDIA L40 GPUs with 48 GB GPU memory each.
Added support for Valkey 9.1 in Amazon ElastiCache node‑based clusters
Introduced cross‑account access for Amazon Bedrock AgentCore Memory, allowing memory resources and agents to span multiple AWS accounts.
Introduces 16 GiB of encrypted, automatically cleaned‑up ephemeral storage per workflow task, mountable at /tmp.
Cognito user pools now support customer‑managed AWS KMS keys for at‑rest encryption
Introduces two new policy refinement workflows (iterative improvement and ambiguity reduction) for Automated Reasoning checks in Amazon Bedrock Guardrails.
Introduces CloudWatch OpenTelemetry Container Insights for Amazon EKS, collecting infrastructure metrics at 30‑second intervals via cAdvisor, Kube‑State‑Metrics, and NVIDIA DCGM.
Introduces Claude Tag beta for Slack, allowing teams to tag @Claude in channels and delegate tasks
Introduces AI‑assisted Migration Assistant to simplify moving self‑managed Solr, Elasticsearch, or OpenSearch deployments to OpenSearch Serverless or Managed Clusters
Added support for Amazon EC2 G7e instances in SageMaker Studio notebooks, offering up to 8 RTX PRO 6000 GPUs, 96 GB GPU memory, and up to 192 vCPUs.
Adds support for Nextflow profiles, letting users activate predefined execution settings at run time.
Introduces Lambda MicroVMs, a serverless primitive that gives each user or job a dedicated VM‑level isolated environment with near‑instant launch and up to 8 hour stateful resume.
Adds support for all AWS commercial regions as migration targets in AWS Transform for migrations
Default stateful action for new firewall policies changed to "Application drop established (server‑directed only)".
Introduces the Agentic CX Designer (NLX) in preview, a no‑code canvas for building AI‑powered voice and digital self‑service experiences.
Adds Best Fit Progressive Ordered (BFPO) and Spot Capacity Optimized Prioritized (SCOP) allocation strategies for AWS Batch compute environments.
IAM Identity Center now has independent quotas for AWS accounts and applications (default 7,000 each).
Introduces AI Agent Skills for Amazon MSK, offering AI coding assistants expert, up‑to‑date guidance on troubleshooting, sizing, configuration, monitoring, and migration tasks.
MSK Replicator now adds mutual TLS (mTLS) authentication for connecting to external Kafka clusters when replicating to MSK Express brokers.
Adds self‑service lifecycle management for Outposts, including configuration, quoting, ordering, subscription handling, renewal, and decommissioning via Console, CLI, and API.
General availability of a new AWS Local Zone in Hanoi, Vietnam, extending core services closer to end users.
Introduces multilocation canaries, enabling a single canary to run concurrently across multiple AWS regions
Adds Intelligent Rebalancing support to all existing MSK Provisioned clusters with Express brokers at no extra cost.
ECS service auto scaling now supports 20‑second high‑resolution metrics for faster scaling signal detection
GA of EC2 G7 instances with up to 8 NVIDIA RTX PRO 4500 GPUs, 32 GB GPU memory, custom Intel Xeon 6 CPUs, and up to 700 Gbps EFA bandwidth
Adds private networking connectivity for Amazon MQ for RabbitMQ brokers using VPC Lattice, RAM, and PrivateLink
AWS Parallel Computing Service now supports EC2 P6e-GB200 and P6e-GB300 UltraServer instances for large‑scale GPU workloads.
Added nested virtualization support to many new Intel instance families (C7i, R7i, M7i, etc.) and their flexible variants
Adds urgent‑contact interruption so a contact can ring an agent even if they’re on another call, overriding normal routing.
Introduces Ministral-3-14B-Instruct multimodal model in SageMaker JumpStart for edge‑optimized AI assistants and vision‑enabled apps
Compute Optimizer now incorporates two new CloudWatch metrics (VolumeIOPSExceededCheck and VolumeThroughputExceededCheck) for EBS volume rightsizing
Added all-MiniLM-L12-v2 model to SageMaker JumpStart for semantic search and sentence similarity applications.
Adds customer‑routed control‑plane egress, letting API server traffic flow through your VPC
Adds native observability for SageMaker inference endpoints, exposing token latency, GPU health, scaling events, and other performance metrics in real time.
Added SMS publishing capability for Amazon SNS in the Asia Pacific (Seoul) region
Adds ability to customize Linux capabilities for containers in fleet definitions, enabling specialized networking and debugging features.
General Purpose (gp3) volumes now support up to 64 TiB, 80,000 IOPS, and 2,000 MiB/s throughput (4‑8× previous limits).
Added Ubuntu 24.04 LTS bundle for Amazon WorkSpaces Personal in the AWS China (Ningxia) region.
Added Spark Connect support to AWS Glue Interactive Sessions, enabling development and execution of Spark jobs from notebooks, IDEs, and APIs.
Added real-time streaming of HealthOmics workflow engine logs to Amazon CloudWatch.
Extended support for Aurora MySQL 5.7 and RDS MySQL 5.7 extended to June 30 2029 (previously Feb 2027).
Introduces Graviton5‑based M9g instances for RDS PostgreSQL, MySQL, and MariaDB, offering up to 30% higher performance.
Adds a preview release management capability to AWS DevOps Agent, providing release readiness reviews and autonomous release testing.
AWS Outposts second‑generation racks now support new bmn‑cx3a instances with 5th‑gen AMD EPYC CPUs and NVIDIA ConnectX‑7 NICs delivering up to 800 Gbps accelerated networking.
Introduces autonomous agents that execute natural‑language described tasks with adjustable autonomy for continuous workflow automation (e.g., follow‑ups, regulatory summaries, PO processing).
AgentCore policy now integrates Bedrock Guardrails, enabling real‑time evaluation of inputs and outputs to block prompt injection, harmful content, and sensitive data exposure.
Adds continuous optimization loop for Bedrock agents, surfacing failure, intent, and trajectory insights across production sessions.
General availability of Amazon Bedrock AgentCore managed harness, letting teams define agents via configuration and get production‑grade orchestration, tool execution, state persistence, and isolation in minutes.
Preview of business context and semantic search for AWS Glue Data Catalog, enabling enrichment of tables with glossary terms and custom metadata fields.
Introduces Oracle Autonomous AI Database Serverless (ADB‑S) on Oracle Database@AWS, a fully managed serverless Oracle DB on Exadata with automatic patching, tuning, and scaling.
Announces AWS Continuum, a preview service that automates discovery, prioritization, validation, and remediation of security risks at machine speed.
Adds a secret safety skill to the aws-core plugin in the Agent Toolkit for AWS, letting agents use Secrets Manager without exposing raw secret values.
Amazon Bedrock Managed Knowledge Base is now generally available as a fully managed RAG service
Added Kiro and Claude Code plugins, enabling developers to trigger security scans and build threat models directly from their IDEs.
Introduces AI‑powered threat modeling in AWS Security Agent (now part of AWS Continuum) that auto‑generates STRIDE‑based threat models from design documents or source code.
Adds Automated Reasoning checks using formal verification to validate Bedrock Guardrails outputs with up to 99% accuracy.
Adds a fully managed Web Search capability to AgentCore, letting agents ground responses with up-to-date web data without external integrations.
Adds 16 new connectors (e.g., Adobe, Figma, WhatsApp, Snowflake) expanding Quick’s integration ecosystem across design, analytics, finance, commerce, and communication tools
Adds AWS Management Console Private Access via VPC endpoints using PrivateLink, removing the need for internet connectivity.
Introduces a preview continuous modernization capability in AWS Transform that automatically detects, prioritizes, and remediates tech debt across large codebases.
Adds STRICTLY_CONSISTENT extraction type allowing apps to attach exact metadata values to long‑term memory records without LLM inference
Adds on‑demand FSx for OpenZFS snapshot replication to opt‑in AWS Regions
Introduces Log Analytics console that unifies CloudWatch Logs Insights, Live Tail, and Contributor Insights in a single UI.
Integrated Palo Alto Networks Advanced DNS Security (preview) into Route 53 Resolver DNS Firewall, enabling DNS threat categories to be added directly to firewall rules.
Amazon ECS Express Mode is now available in AWS GovCloud (US‑East and US‑West) regions.
Added Cost Explorer historical data retention for accounts in billing groups, preserving original billable rates
Adds tag propagation for Amazon EC2 instances, EBS volumes, and ENIs created by Lambda Managed Instances
Added support for custom SRE agents and sub‑agents, enabling scheduled automated workflows within Agent Spaces.
Amazon EC2 C7i instances with 4th‑Gen Intel Xeon (Sapphire Rapids) are now available in the Israel (Tel Aviv) region
Launch of Amazon EC2 C8i instances in the Europe (Stockholm) region, adding 13 sizes including new 96xlarge and bare metal options.
Introduces AI traffic monetization in AWS WAF, enabling publishers to price, collect payments, and grant scoped edge access to AI bots and agents
Introduces Amazon CloudWatch Query Studio, a unified UI for querying and visualizing metrics via PromQL and Metrics Insights.
Introduces CloudWatch Metrics Centralization for cross‑account and cross‑region metric replication
Free inbound/outbound network bandwidth now included for all GameLift Servers instance types generation 6+ (On‑Demand & Spot).
Added MariaDB 12.3 Long‑Term Support to the Amazon RDS Database Preview Environment for early evaluation.
Amazon RDS for MariaDB now supports community minor versions 10.6.27, 10.11.18, 11.4.12, and 11.8.8.
Introduces Amazon RDS Extended Support minor version 5.7.44‑RDS.20260521 for MySQL
Added Amazon Lightsail availability in three new AWS Regions: Hong Kong, São Paulo, and Spain
Added serverless model customization for NVIDIA Nemotron 3 Nano (30B) with supervised and reinforcement fine‑tuning on SageMaker AI.
Adds high‑performance storage‑optimized EC2 I7i instances to the AWS Europe (Paris) region.
EC2 Capacity Blocks for ML now available in AWS GovCloud (US-West and US-East) regions
Launch of U7i-8TB (u7i-8tb.112xlarge) high‑memory EC2 instances in the AWS Europe (Paris) region
Amazon Quick now integrates with Snowflake Cortex AI via the Model Context Protocol (MCP) using OAuth, enabling natural‑language queries over structured data and unstructured documents.
Adds support for Amazon EKS local clusters on first‑ and second‑generation Outposts racks using EC2 instance‑store backed instances.
Added service health ranking on the application map and new infrastructure, logs, and traces tabs in CloudWatch Application Signals.
Amazon Aurora PostgreSQL now supports PostgreSQL major version 18 (starting with Aurora version 18.3).
MWAA Serverless now emits workflow and task state change events to Amazon EventBridge
Adds native CloudWatch Logs integration to the Elastic Beanstalk console Logs tab, allowing users to view and filter log groups and streams without leaving the console
Lake Formation now grants read/write access to underlying S3 data for tables in the Glue Data Catalog via existing table permissions.
Added support for ingestion, storage, and querying of Prometheus native histograms
Adds out-of-order sample ingestion with a configurable time window per workspace
Introduces AWS Workload Credentials Provider for automated export of ACM certificates and local caching of Secrets Manager secrets.
GPT-5.4 and GPT-5.5 are now available in the US East (N. Virginia) region on Amazon Bedrock.
Amazon OpenSearch Service adds MCP Apps, enabling agentic IDEs (e.g., Claude Desktop, VS Code) to run observability workflows directly against OpenSearch and Amazon Managed Service for Prometheus data.
Adds the ability to cancel future-dated EC2 Capacity Reservations, with optional cancellation fees based on timing
Added pidMode and ipcMode settings to ECS Managed Daemon definitions for shared process and IPC namespaces
Added Gemma 4 model family (31B, 26B‑A4B, E2B) to Amazon Bedrock
VPC Flow Logs now embed EC2 resource tags (ENI, instance, Auto Scaling group) directly in log entries
Introduce EC2 M9g and M9gd instances powered by Graviton5, offering up to 25% higher compute performance versus M8g/M8gd
P6‑B200 instances with 8 Blackwell GPUs are now available in the AWS GovCloud (US‑East) region (p6‑b200.48xlarge size).
Adds ability to update existing Cost and Usage Report 2.0 table configurations (columns, granularity, format, destination) via console or SDK/CLI.
Added Intelligent‑Tiering storage class for FSx for OpenZFS in 8 new AWS regions (US West, Europe, APAC, South America).
Adds Amazon EMR Serverless as a selectable Spark runtime for SageMaker Unified Studio notebooks, enabling PySpark and Spark SQL execution.
Introduces Amazon S3 Access Grants support in the AWS European Sovereign Cloud (Germany) region.
Introduces AWS FinOps Agent preview for cost queries, optimization recommendations, and automated anomaly investigation.
GA of metal-48xl and metal-96xl sizes for M8 and R8 instance families, using 6th‑gen Intel Xeon and Nitro cards.
Claude Fable 5 is now generally available on AWS via Amazon Bedrock and Claude Platform, offering Mythos‑class capabilities with built‑in safety guardrails.
Added native OpenTelemetry metric ingestion (OTLP) and PromQL query support in CloudWatch with a Prometheus‑compatible query API.
AWS Backup for Amazon EKS is now available in the AWS European Sovereign Cloud (Germany) Region.
EMR Serverless adds interactive Spark Connect sessions, enabling notebook/IDE‑based Spark development with a persistent context
Adds 'Analyze with Amazon Q' to Cost Explorer, providing AI‑generated cost explanations for any configured report.
Added idle resource detection for DynamoDB, ElastiCache (Redis/Valkey), MemoryDB, DocumentDB (provisioned & serverless), WorkSpaces, and SageMaker endpoints.
Add automatic topic creation for Kafka Streams on MSK Express Brokers
Adds support for engine minor version 5.0.1 with new aggregation operators ($rand, $pow, $dateToParts, $dateFromParts).
Introduced 23 new query commands and functions (hash, string, conditional, conversion, IP, analytics, parse) to CloudWatch Logs Insights.
Added target coverage analysis to Savings Plans Purchase Analyzer, enabling users to set desired On‑Demand spend coverage percentages.
RDS now offers a preview environment for PostgreSQL 19 Beta 1, enabling evaluation of the upcoming version on a fully managed service.
Added support for PostgreSQL JSONB data type in Aurora DSQL.
AWS Application Migration Service rebranded to AWS Transform MGN, preserving all existing functionality.
AWS Lambda Managed Instances is now available in all commercial AWS Regions except Israel (Tel Aviv), Middle East (Bahrain), Middle East (UAE), and Asia Pacific (Auckland).
Adds AI‑powered cost investigation to Cost Anomaly Detection, delivering plain‑language root‑cause analysis in minutes.
Adds AI agent trace UI for self-service voice interactions, showing step‑by‑step reasoning, tool calls, and timings.
Redshift now bills manual snapshot storage based on unique data blocks across snapshots, not total snapshot size.
Introduces TCO assessment for Amazon RDS for SQL Server in AWS Transform, enabling AI‑driven migration cost estimates and optimal instance recommendations.
Introduces InvokeAgentRuntimeCommandShell API for PTY‑backed interactive shells over WebSocket, providing persistent terminal access to running agent sessions
Adds IAM‑based authorization for Amazon S3 tables and Iceberg materialized views in the Glue Data Catalog.
Added OpenSearch UI availability in AWS GovCloud (US-East and US-West) for managed domains and serverless collections.
C8in instances are now available in additional regions: AP (Sydney, Singapore, Malaysia) and EU (Frankfurt).
Adds support for 32 vCPU task sizes on AWS Fargate with memory options of 60 GiB, 120 GiB, or 244 GiB for both x86 and ARM Linux workloads.
Added an interactive `aws configure agent-toolkit` command to install and manage the Agent Toolkit across multiple coding agents via the AWS CLI
Adds multi‑profile support to the AWS MCP Server, letting AI agents specify an AWS profile per command without restarting
Integrated SageMaker Data Agent with SageMaker Catalog metadata, enabling business‑term based data discovery and accurate SQL/Python generation
Added Aurora PostgreSQL, Aurora DSQL, and DynamoDB Serverless to Vercel Marketplace/v0 in additional AWS regions (16‑17 regions).
Introduces multi‑Region replication for Cognito user pools, allowing near real‑time sync to a standby pool in another region.
EKS Capabilities can now be configured as log delivery sources via Amazon CloudWatch Vended Logs for Argo CD, ACK, and kro.
Introduced plugin sync for service-managed fleets, enabling plugins uploaded to a queue's S3 bucket to auto‑sync to workers at job start.
Amazon MQ for RabbitMQ is now available in the AWS European Sovereign Cloud (Germany) Region.
Introduces a refreshed Amazon Bedrock console with project‑based workflow and built‑in model catalog for fast evaluation to production
Adds MQTT session data (timeout, expiry, socket IP/port info) to the IoT Device Management connectivity status API.
Introduces conversation history to Amazon SageMaker Data Agent, allowing users to retain and resume analytical sessions.
Introduces notebook scheduling, parameterization, and orchestration directly in SageMaker Unified Studio.
Adds a new AgentCore-powered agentic reasoning step to AWS Step Functions, enabling AI agents in workflows via a managed Bedrock harness (preview).
GPT-5.4 is now generally available on Amazon Bedrock in AWS GovCloud (US‑West) for government and regulated customers.
Introduces an optional 32‑day lookback period for EBS volume and ECS service rightsizing recommendations, extending the default 14‑day window.
Added Aurora serverless scaling execution block to ARC Region switch for automated secondary cluster right-sizing during failover
GetRecords now includes an iteratorPosition (AT_TIP or BEHIND_TIP) indicating stream tip status
Adds UI localization support for twelve languages, enabling users to work in their preferred language.
Introduces serverless multi‑turn reinforcement learning for fine‑tuning agent models on SageMaker AI
Added support for 9 new AWS resource types across Amazon Bedrock, Bedrock Agent Core, and SageMaker, expanding Config coverage.
Adds support for AWS Trainium and Inferentia AI accelerator instance types in Amazon ECS Managed Instances.
Added support for IBM Db2 v12.1, including Standard, Advanced, and new Community Edition.
Introduced Ping log events for MQTT keep-alive messages to spot devices that lose connectivity.
Adds real-time dashboard alerts for supervisors when specific keywords, phrases, or sentiment patterns are detected in live calls and chats
EKS and EKS Distro now support Kubernetes version 1.36 across all AWS regions, including GovCloud.
Introduces internal service‑linked rules for AWS Config, allowing AWS services like Security Hub CSPM to run managed rule evaluations independently of customer recorders
Added support for attaching persistent EBS volumes to Service‑Managed Fleet workers, preserving environments and assets across lifecycle events
Quick setup for SageMaker Studio now finishes in under 20 seconds, down from over 2 minutes.
Added durability support for Amazon ElastiCache Valkey with synchronous and asynchronous write options across multiple AZs
Added Transit and Intermodal travel modes to CalculateRoutes, enabling public transport and multi‑modal journeys.
CUR 2.0 now exports directly to Athena or Redshift in optimized Parquet/GZIP format
RDS for SQL Server now supports Bring Your Own Media (BYOM), allowing reuse of existing SQL Server licenses via Microsoft’s License Mobility program.
Added support for Nextflow v26.04 across all AWS HealthOmics regions, unlocking record types, strict syntax parsing, JSON workflow output summaries, and agent logging mode.
Add ability to specify Nextflow engine version at run time via the StartRun API using a new engine-settings parameter.
Introduces automated placement of ad‑hoc activities (training, meetings) in agent schedules to minimize impact on service levels
Increases Amazon Connect Customer agent scheduling capacity to 5,000 agents per schedule.
Added support for encrypting Quick Research data with customer-managed KMS keys
Adds automated email/text notifications for schedule changes and leave request status via EventBridge.
Adds VPC connectivity for MCP servers, allowing Quick to link to privately hosted MCP instances without exposing them to the public internet.
SageMaker Unified Studio now supports custom IAM permissions boundaries via Tooling blueprint, enabling SCP compliance without changing security posture.
Introduces AWS PCS‑ready Deep Learning AMI built on the Deep Learning Base GPU AMI (Ubuntu 24.04) with pre‑installed PCS Agent, Slurm, and EFS utilities.
M8i and M8i‑flex general‑purpose instances are now available in the Asia Pacific (New Zealand) region.
Adds support for referencing existing AWS Secrets Manager secret ARNs in AgentCore Identity Credential Providers.
Introduces HyperPod troubleshooting skills for AI coding assistants (e.g., Claude Code, Cursor, Kiro) to perform natural‑language diagnostics of cluster health, GPU faults, NCCL communication, and performance bottlenecks.
- New M8azn EC2 instances (5 GHz AMD EPYC) are now available in the Europe (Ireland) region.
Add support for EFA‑only network interfaces in SageMaker HyperPod, enabling dedicated EFA devices without attached ENA IP networking
Adds VIF Rate Limiters for dedicated connections to cap bandwidth per VIF and prevent congestion
Adds Amazon CloudWatch metrics for the bedrock-mantle endpoint, covering inference counts, token usage, and error counts
GPT-5.5 and GPT-5.4 are now generally available on Amazon Bedrock for production workloads.
Added tenant-level suppression lists, isolating bounces and complaints per tenant rather than sharing a single list across an account.
Introduces the Inspector VM Scanner for agent-based EC2 scanning, expanding coverage to more software stacks.
Added inbox placement metrics showing spam‑folder rates per sending domain and campaign using industry sample data
Adds RCS for Business messaging support in 20 new countries, bringing total coverage to 22.
Adds native support for CMS Interoperability and Prior Authorization Final Rule with new FHIR‑based Patient, Provider, Payer‑to‑Payer, and Prior Authorization APIs.
Adds ability to schedule tasks up to 90 days in advance via API, flows, or agent workspace
Adds DDoS attack flow logs to Shield Advanced, providing packet‑level visibility during attacks
Adds a 4‑RPU minimum capacity option for Redshift Serverless in 7 additional AWS regions, down from the previous 8‑RPU minimum.
Introduces a free 500 Mbps AWS Interconnect‑multicloud tier, giving customers a fully managed private connection to another CSP at no AWS charge.