AWS AI News

New Bedrock AgentCore Runtime GA with elastic memory and fast cold starts

AWS made its next-generation AgentCore Runtime generally available — the serverless microVM compute in Amazon Bedrock. It reclaims unused memory throughout a session so customers pay for actual usage rather than peak, and delivers consistent cold-start times regardless of container image size or concurrency, addressing two persistent pain points for long-running production agents.

2026-09-19

Moonshot AI's Kimi K3 now generally available on Amazon Bedrock

Kimi K3 from Moonshot AI became generally available on Amazon Bedrock as an open-weight option for coding and knowledge work. Moonshot says it is its most capable model and the first open model to reach 2.8 trillion parameters, with native vision, a 1-million-token context window, and explicit prompt caching. AWS simultaneously expanded Bedrock with OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5.1.

2026-09-19

Amazon SageMaker HyperPod Inference Gateway cuts first-token latency up to 82%

AWS launched SageMaker HyperPod Inference Gateway, a Kubernetes-native, GPU-aware routing add-on for Amazon EKS. It uses real-time GPU signals to route each inference request to the best-suited pod, cutting first-token latency by up to 82% with no changes to model servers or client applications.

2026-09-19

AWS declares Bahrain region permanently unrecoverable after Iranian strikes

AWS confirmed it cannot restore data hosted exclusively in its Bahrain ME-CENTRAL-1 region and one UAE availability zone after Iranian drone strikes in March 2026 — the first known case of physical war destroying a cloud region. AWS also revealed its AI services business surpassed a $25 billion annualized run rate and raised 2026 capital spending to $220 billion.

2026-09-19

Open-weight models on Bedrock now available in AWS European Sovereign Cloud

AWS made open-weight models on Amazon Bedrock generally available in its European Sovereign Cloud, letting EU organizations run generative AI while keeping data within the European Union to meet regulatory requirements. The move targets the same sovereignty-conscious buyers Mistral and Cohere are courting.

2026-09-19

AWS confirms permanent data loss in Bahrain and one UAE zone after drone strikes

Amazon Web Services said it cannot restore its Bahrain cloud region or one of three UAE hosting zones (mec1-az2) six months after March 2026 Iranian drone strikes, causing permanent loss of some customer data. The damage exceeded what its multi-availability-zone architecture was designed to withstand.

2026-09-18

AWS Enhances Bedrock AgentCore With Open-Source HCLS Agent Skills and Prompt Optimization

AWS released 38 open-source agent skills across 11 healthcare and life sciences domains to fix agents that cite the right guideline but apply it incorrectly, with a 410-prompt evaluation showing a 70–86% win rate. AgentCore also added a Single Agent Reflector that turns production traces into automated system-prompt recommendations.

2026-09-17

AWS Reframes AI Agents Around an Inbox With Open-Source 'Pizza Bot'

AWS open-sourced Pizza Bot, a self-hosted application giving users an inbox for managing background AI agents, with separate threads for ongoing tasks and a queue for completed work or work needing human input — arguing agents need an inbox rather than another chat window.

2026-09-17

AWS Reimagines Getting-Started Experience With $100 Free Tier Credits

AWS launched a simplified onboarding experience with smart defaults, $100 in Free Tier credits, managed project environments, and simplified permissions, lowering the barrier for new developers to ship on AWS while letting them activate advanced features as workloads evolve.

2026-09-17

Amazon Bedrock Prompt Caching Cuts Input Token Costs Up to 90%

AWS detailed how prompt caching in Amazon Bedrock can reduce input token costs by up to 90% when repeatedly sending the same context to foundation models. The post walks through six practical scenarios using the Converse API, including system prompts, tool definitions, mixed TTL, tenant isolation, and LangChain integration.

2026-09-16

AWS and Salesforce Expand AI Agent Integration Across Quick, Slack, and Bedrock

AWS and Salesforce expanded their collaboration, bringing Salesforce business context into Amazon Quick via the Model Context Protocol, AWS frontier agents into Slack, and Agentforce model choice through Amazon Bedrock. The integration lets account teams pull pipeline health, account summaries, and open cases into a single interface.

2026-09-16

AWS makes OpenAI's GPT-6 Astra generally available on Amazon Bedrock

Amazon Web Services made OpenAI's GPT-6 Astra generally available on Amazon Bedrock for business customers, letting them call the frontier model directly or point ChatGPT Work and Codex at Astra on Bedrock. The AWS Weekly Roundup also flagged Amazon Quick desktop GA and Kiro for students. Astra retains its 1-million-token context window under Bedrock's IAM, VPC, and CloudTrail controls.

2026-09-15

AWS adds managed OAuth consent for AI agents in Bedrock AgentCore

Amazon Bedrock AgentCore Identity now offers a managed Consent portal and session-binding endpoint for AgentCore Gateway, letting AI agents obtain end-user OAuth consent for third-party services. The post details provisioning a portal, configuring GitHub and Slack 3LO targets, the end-user consent flow, and reviewing agent activity in AWS CloudTrail.

2026-09-15

SageMaker HyperPod adds model caching to cut inference cold starts to seconds

Amazon SageMaker HyperPod now supports model caching, pre-loading model weights and container images onto cluster nodes so pods start in seconds instead of minutes. It targets cold-start bottlenecks in LLM inference, agentic pipelines, and RAG workloads.

2026-09-14

SageMaker Inference adds prefix-aware routing, cutting LLM latency up to 77%

AWS introduced prefix-aware routing on Amazon SageMaker Inference, which sends requests sharing the same prompt prefix to the same instance so the KV cache stays warm. In Llama 3.1 70B benchmarks it cut P50 time-to-first-token by up to 77% and raised cache hit rates from about 25% to over 80%, targeting agentic pipelines, RAG and high-throughput chat workloads.

2026-09-13

AWS pairs DevOps Agent with AgentCore Evaluations for production agent monitoring

AWS detailed a dual-layer approach to monitoring production AI agents: Bedrock AgentCore Evaluations for continuous quality scoring and AWS DevOps Agent for autonomous infrastructure investigation, demonstrated on a four-agent airline reservation system. The DevOps Agent also gained bidirectional Slack communication, letting engineers run the full investigation lifecycle across AWS, multicloud and on-premises environments without leaving Slack.

2026-09-13

Bedrock Knowledge Bases gains multimodal video search via TwelveLabs Marengo 3.0

AWS made TwelveLabs Marengo 3.0 available as an embedding model in Amazon Bedrock Managed Knowledge Bases — the first video-understanding model integrated into the service — enabling natural-language search across video, image and audio stored in S3. Unlike prior transcription-based search, Marengo 3.0 encodes visual scenes directly.

2026-09-12

OpenAI's GPT-6 Astra reaches general availability on Amazon Bedrock

AWS announced GPT-6 Astra, OpenAI's latest and most capable model, is generally available on Amazon Bedrock, callable via APIs or through ChatGPT Work and Codex. Astra features deeper reasoning, professional writing, advanced computer/browser use, and a 1-million-token context window, priced at $10 per M input tokens and $50 per M output. New enterprise plugins extend Astra's browser-use across business workflows.

2026-09-11

Amazon Quick AI assistant reaches general availability on desktop

The Amazon Quick desktop app is now generally available on macOS and Windows, offering an AI assistant that handles real work while keeping data in the customer's environment. A new mobile activity feed consolidates email, calendar, and CRM, positioning Quick as an enterprise productivity agent that competes with Meta Muse and ChatGPT Work.

2026-09-11

AWS and Azure launch multicloud interconnect preview; AWS expands NVIDIA GPU deal

AWS announced a public preview of AWS Interconnect–multicloud with Microsoft Azure, enabling private, encrypted Layer-3 connectivity across US East, US West, Sydney, and Frankfurt with paths to 100 Gbps at GA. AWS also expanded its NVIDIA partnership to deploy an additional 2 million GPUs across its infrastructure in 2027-2028.

2026-09-11

AWS Bedrock Knowledge Bases add Marengo 3.0 for video and image search

TwelveLabs Marengo Embed 3.0 is now generally available in Amazon Bedrock Knowledge Bases, bringing fully managed natural-language semantic search across video, image, and audio content. Developers can build a knowledge base and run semantic queries against media, extending RAG beyond text into multimodal enterprise content.

2026-09-11

AWS Lambda adds 90-minute function timeout on Managed Instances

AWS Lambda now supports a 90-minute function timeout for asynchronous and event-source-mapping invocations on Lambda Managed Instances — a 6x jump from the prior 15-minute cap. Data processing, media transcoding, financial calculations, AI inference, and batch workloads can now run on Lambda without re-architecting.

2026-09-10

AWS expands free Kiro AI agent access to 132 universities with $500M commitment

AWS expanded free access to its Kiro AI agent to 132 universities across 16 countries, giving eligible students a full year of free access with 1,000 monthly credits and no credit card required. The move accompanies a commitment of over $500 million to student cloud and AI training, as CEO Andy Jassy framed the need for more skilled builders.

2026-09-10

Bedrock AgentCore Memory adds direct ingestion to long-term memory

Amazon Bedrock AgentCore Memory introduced a new IngestData API that lets developers submit content directly for long-term memory extraction, without first persisting it as a short-term memory event. It supports both conversational and JSON payloads, fanning content out to configured long-term strategies so it's retrievable like any other memory record. The change removes the prior requirement that all content pass through short-term storage first.

2026-09-09

AWS named a Gartner Magic Quadrant Leader for the 16th consecutive year

Gartner named AWS a Leader in its 2026 Magic Quadrant for Strategic Cloud Platform Services for the sixteenth straight year, again placing it highest on the Ability to Execute axis. That makes AWS the longest-running Leader in the category. The recognition lands the same week AWS reported $42.2 billion in quarterly revenue and a $496 billion backlog.

2026-09-09

AWS wires Bedrock AgentCore evaluations into GitHub Actions CI

AWS published a pattern for embedding Bedrock AgentCore Evaluations into a GitHub Actions pipeline. It deploys an AI agent and an OAuth-protected MCP server to AgentCore runtime, invokes the agent with test prompts, scores the responses, and automatically blocks pull requests when agent behavior regresses. The approach brings CI/CD-style regression gating to non-deterministic agents.

2026-09-09

SageMaker Feature Store adds UpdateRecord for feature-level writes

Amazon SageMaker Feature Store introduced a new UpdateRecord API enabling feature-level writes, letting data scientists update one or more feature values in a single call without rewriting the entire record. It works across both the DynamoDB Standard and ElastiCache In-Memory online store tiers, lowering write latency for high-frequency feature updates. The change targets real-time ML pipelines where only a subset of features changes.

2026-09-09

Pathway scales brain-inspired BDH post-transformer architecture on SageMaker HyperPod

AWS detailed Pathway's Baby Dragon Hatchling (BDH), a brain-inspired, post-transformer architecture that reasons in latent space instead of emitting chain-of-thought tokens. Pathway scales BDH on Amazon SageMaker HyperPod, and its BDH-CQ variant set a new cost-efficiency mark on the ARC-AGI-1 benchmark. The work represents a rare non-transformer bet at meaningful scale.

2026-09-09

AWS Agent Registry hits general availability to govern enterprise AI agents

AWS launched the AWS Agent Registry, generally available since August 31, to help enterprises track and manage AI agents across their systems. It automatically detects agents running on Amazon Bedrock AgentCore and consolidates them into a central registry with lifecycle management, cost allocation and audit capabilities, priced at $0.40 per 1,000 records. Notably, it does not enforce security policies automatically.

2026-09-08

AWS deploys a multimodal WhatsApp ordering assistant on Bedrock AgentCore

AWS detailed how to build a multimodal WhatsApp ordering assistant handling text, voice notes and real-time voice calls on a single business number. Built on Bedrock AgentCore with Amazon Nova 2, it keeps channel and ordering layers separate while sharing one memory that recognizes each customer across channels.

2026-09-07

Bedrock Managed Knowledge Base adds ServiceNow connector and self-serve setup

AWS added a native ServiceNow connector to Amazon Bedrock Managed Knowledge Base, crawling knowledge articles and catalog items into its managed RAG service. It also introduced user-managed (3LO) setup for SharePoint, OneDrive and Confluence — removing the need for admin-level 2LO credentials — plus automatic sync scheduling so knowledge bases stay current without manual triggers.

2026-09-07

AWS details lifecycle policies for AgentCore memory

AWS showed how to design memory lifecycle policies for Amazon Bedrock AgentCore — scoring, consolidating, and pruning agent memories to avoid quality degradation and compliance risk. The approach runs on a nightly Step Functions workflow with a deployable CDK stack, tackling the growing problem of unbounded agent memory.

2026-09-06

Bedrock Managed Knowledge Base adds ServiceNow connector and user-managed 3LO setup

AWS added a native ServiceNow data source connector to Amazon Bedrock Managed Knowledge Base, letting customers crawl knowledge articles and catalog items directly into RAG. It also introduced user-managed (3LO) setup for SharePoint, OneDrive, and Confluence, removing the need for admin-generated 2LO credentials and simplifying secure connection to enterprise content.

2026-09-05

AWS publishes Bedrock AgentCore migration path for production agentic workloads

AWS published a guide for migrating agentic workloads to Amazon Bedrock AgentCore, walking through moving a LangGraph customer-support agent onto AgentCore Runtime, Gateway and Memory, then to model-driven planning on Strands Agents. It also enables server-side tool use like web search and database updates within AWS security boundaries, and extended prompt caching for long-running agent conversations. Web Search on Bedrock reached AWS GovCloud (US-West) for compliance-sensitive workloads.

2026-09-04

AWS expands NVIDIA Blackwell EC2 capacity with P6-B200 and P6-B300 in new Asia-Pacific regions

AWS made EC2 P6-B200 Blackwell instances available in Asia Pacific (Hyderabad) — 8 Blackwell GPUs, 1,440 GB GPU memory, up to 2x P5en performance — and launched EC2 P6-B300 Blackwell Ultra instances in Asia Pacific (Jakarta) with 8 GPUs, 2.1 TB GPU memory and 6.4 Tbps EFA networking. The rollout roughly doubles AI training and inference performance over the prior generation in these regions.

2026-09-04

AWS opens GPT-5.6 access on Amazon Bedrock from Australian regions; invests $5.3B in Saudi Arabia

AWS announced Australian teams can now access OpenAI's GPT-5.6 models — Sol, Terra and Luna variants — on Amazon Bedrock with global cross-Region inference from the Sydney and Melbourne regions. Separately, Amazon is investing $5.3 billion in Saudi Arabia cloud and AI infrastructure, combining dedicated AWS AI servers, Trainium chips and the latest NVIDIA hardware.

2026-09-04

AWS aggregates six model families on Bedrock and expands GovCloud access

AWS expanded Amazon Bedrock in GovCloud (US) to offer government agencies models from OpenAI, Anthropic, Meta, Nvidia, xAI, and its in-house Nova series in isolated, compliant environments. The move consolidates six major model families behind a single procurement and security boundary for public-sector workloads.

2026-09-03

Web Search on Amazon Bedrock lands in AWS GovCloud (US-West)

Amazon Bedrock's server-side Web Search tool is now available in AWS GovCloud (US-West), bringing grounded web results with citations to compliance-sensitive government and public-sector workloads. It supports OpenAI GPT models including GPT-5.4, GPT-5.6 Terra and Luna while keeping data within AWS boundaries.

2026-09-03

AWS makes agentic CX designer generally available in Amazon Connect

Amazon Connect announced general availability of agentic CX designer, a no-code canvas for building AI-powered self-service voice and digital experiences. It combines agentic and deterministic AI to give enterprises control and reliability when deploying customer-facing automation.

2026-09-03

AWS details agentic security: detection and response at machine speed

AWS published guidance arguing that autonomous AI agents authenticating on behalf of users represent the biggest security posture shift since cloud adoption. The post outlines detection and response approaches to keep pace with agents executing multistep workflows and making autonomous decisions.

2026-09-03

AWS adds Claude Fable 5.1 and third-party models to Bedrock on GovCloud

Amazon widened AI access for U.S. government customers via AWS GovCloud, offering third-party model families from Anthropic, Meta, OpenAI, xAI and NVIDIA through Amazon Bedrock, and made Claude Fable 5.1 available on Bedrock. AWS's AI and custom-chip businesses have each surpassed $25 billion in annualized run rate, with backlog reaching $496 billion.

2026-09-02

Amazon Bedrock AgentCore Payments reaches general availability

AWS moved AgentCore payments to general availability, enabling agents to autonomously discover, access, and pay for APIs and MCPs via x402 protocol support and Coinbase credential provisioning. Partner t54 has already built a trust layer on top that has governed more than 20 million agent-initiated transactions with no human in the loop.

2026-09-02

Amazon Quick becomes generally available for natural-language app building

Amazon Quick is now generally available, letting users build custom applications by describing them in natural language. Product managers, finance leads, HR partners, and ops analysts can create trackers, dashboards, and portals without writing code, with a companion guide on securing Quick from POC to production.

2026-09-02

AWS GovCloud adds Claude, Llama, OpenAI, Grok and Nova to Bedrock for government

Amazon expanded AI model choice in AWS GovCloud (US) via Bedrock, adding Amazon Nova, Anthropic Claude, Meta Llama, OpenAI, xAI Grok and NVIDIA Nemotron. It is part of an up-to-$50B AWS commitment to deploy integrated AI and HPC infrastructure for the US Government.

2026-09-01

AWS Agent Registry reaches general availability

AWS made Agent Registry generally available — a single, searchable, governed catalog for the agents, tools, skills and MCP servers across an organization, with org-wide auto-detection of agents running on AgentCore. Registry resources are also discoverable inside Amazon Quick.

2026-09-01

Amazon EC2 R9g and R9gd instances on Graviton5 reach general availability

AWS made memory-optimized EC2 R9g and R9gd instances generally available, running on fifth-generation Graviton5 processors and delivering up to 25% better compute performance than R8g. They target memory-intensive workloads including databases, in-memory caches and real-time analytics.

2026-09-01

Kinesis Data Streams adds streaming tables for Apache Iceberg on S3 Tables

Amazon Kinesis Data Streams now supports streaming tables, a fully managed capability that continuously delivers streaming data as queryable Apache Iceberg tables on Amazon S3 Tables. Inline compaction cuts delivery costs up to 25% and query costs up to 30%, with no infrastructure to operate.

2026-09-01

AWS AgentCore Memory adds fine-grained access control and flexible namespaces

Amazon Bedrock AgentCore Memory now supports fine-grained access control, enforcing per-user and per-tenant memory isolation through AgentCore Gateway without custom authorization logic, using OAuth (JWT) auth and Cedar policies. It also added flexible namespace variables to scope long-term memories along dimensions like organization, tenant, team, or environment.

2026-08-31

Bedrock AgentCore Memory adds flexible namespace variables for multi-tenant scoping

AWS added flexible namespace variables to Bedrock AgentCore Memory, letting developers scope long-term memories along any application-specific dimension — organization, tenant, team or environment — without duplicate strategies or overloaded built-in variables. The feature gives multi-tenant and complex-hierarchy applications fine-grained control over how memory is organized.

2026-08-30

AWS and NVIDIA plan 2 million-GPU AI expansion; Anthropic commits $100B+ for 5GW

AWS and NVIDIA announced a 2 million-GPU AI expansion, integrating NVIDIA Nemotron models via Amazon Bedrock and SageMaker plus cuDF with EMR and cuVS with OpenSearch. Anthropic committed over $100 billion over 10 years for up to five gigawatts of AWS capacity, while OpenAI expanded its AWS deal by $100 billion over eight years for roughly two gigawatts of Trainium chips.

2026-08-30

Amazon EC2 P6-B300 instances with NVIDIA Blackwell Ultra GPUs expand to new regions

EC2 P6-B300 instances are now available in Asia Pacific (Hyderabad) and South America (Sao Paulo). Each instance offers 8 NVIDIA Blackwell Ultra GPUs with 2.1 TB of high-bandwidth GPU memory, 6.4 Tbps EFA networking and 4 TB of system memory, delivering 2x networking bandwidth and 1.5x GPU memory over the prior generation.

2026-08-30

Amazon Bedrock AgentCore Memory Adds Fine-Grained Access Control

Amazon Bedrock AgentCore Memory now supports fine-grained access control (FGAC), enabling per-user and per-tenant memory isolation through AgentCore Gateway without custom authorization logic. It enforces access policies based on authenticated caller identities and Cedar policies, exposing 12 Memory operations as Cedar actions via the AgentCore Memory connector.

2026-08-29

SpaceXAI Grok 4.6 Now Available on Amazon Bedrock in GovCloud (US)

SpaceXAI's flagship Grok 4.6 is now available on Amazon Bedrock in AWS GovCloud (US), bringing a frontier coding and agentic model with a 500K context window and configurable reasoning efforts (low, medium, high, xhigh) to regulated government workloads. It supports Responses, Chat Completions, and Converse APIs with cross-region inference routing.

2026-08-29

Amazon triples Nvidia chip order and adopts full physical-AI stack for robotics

Amazon tripled its Nvidia chip order amid surging demand and plans to adopt Nvidia's full physical-AI stack — Omniverse simulation, Cosmos world models, Isaac robotics, and Jetson edge hardware — to power its robot fleet. AWS also moved to acquire Amsterdam-based DuckDB maker DuckLabs.

2026-08-28

AWS makes OpenAI GPT-5.6 Terra and Luna generally available on Bedrock in GovCloud

AWS made OpenAI's GPT-5.6 Terra and Luna models generally available on Amazon Bedrock in AWS GovCloud (US-West and US-East), bringing OpenAI's model family to government-cloud customers via Bedrock's inference engine built for high performance and security. The move accompanies an expanded AWS–NVIDIA partnership involving 2 million additional GPUs.

2026-08-27

AWS Lambda previews Node.js 26 and Python 3.15 managed runtimes before GA

AWS Lambda introduced managed runtimes in public preview, starting with Node.js 26 and Python 3.15, letting customers, partners, and language communities test upcoming runtimes before general availability and share feedback — a departure from Lambda's prior practice of launching runtimes directly as GA.

2026-08-26

AWS Introduces Agentic Resource Discovery Open Spec and Agent Registry

AWS introduced Agentic Resource Discovery (ARD), an open specification, alongside an Agent Registry that provides a centralized, searchable catalog for agents, tools, and skills, enabling cross-environment discovery and governance at scale.

2026-08-25

Amazon SageMaker HyperPod Adds Managed Ray Support on EKS

SageMaker HyperPod now offers managed Ray support on Amazon EKS with built-in observability, resilient training, and accelerated inference, letting users create and monitor Ray clusters and run distributed workloads from SageMaker Studio using open-source KubeRay and standard Ray APIs.

2026-08-25

AWS Glue 6.0 launches with 30% price cut and Apache Iceberg v3 support

AWS Glue 6.0 reached general availability with a 30% price reduction and full Apache Iceberg v3 support, including the new VARIANT data type with automatic shredding. It is built on Apache Spark 4.1, Python 3.13 and Scala 2.13, and adds newer Apache Hudi and Delta Lake versions.

2026-08-24

OpenAI's GPT-5.6 Terra and Luna Go Live on Amazon Bedrock in India

AWS made OpenAI's GPT-5.6 Terra and Luna models available on Amazon Bedrock in India with in-country inferencing, giving regulated industries like financial services local access to frontier models. AWS also committed over $500 million to student cloud and AI training and partnered with Novo Nordisk on an AI drug discovery hub.

2026-08-23

AWS Bedrock AgentCore Gateway governs AI agent tool access

AWS detailed a four-scope maturity model—Connect, Control, Catalog, Harden—for building a governed, auditable tool gateway with Amazon Bedrock AgentCore. The approach gives AI agents access to enterprise tools without consolidating infrastructure, advancing governance only as needs demand.

2026-08-22

AWS Adds Natural-Language Dogwood Policy Authoring to Bedrock AgentCore

AWS launched natural-language Dogwood policy authoring in Bedrock AgentCore, letting teams enforce agent controls—including new time-based constraints—directly from policy documents. The feature converts natural-language policy text into correct Dogwood policies, part of a broader push to govern autonomous agents. Separately, AWS detailed propagating user authorization context through agents to enforce data-access controls.

2026-08-21

Cross-Region Inference for OpenAI GPT-5.6 Models Arrives on Amazon Bedrock

Amazon Bedrock now offers OpenAI GPT-5.6 models (Sol, Terra, and Luna) across more than 25 AWS Regions with cross-Region inference. US geographic and global inference profiles route requests for higher throughput, callable via both the OpenAI and Converse APIs with configurable IAM, quotas, and monitoring.

2026-08-21

AWS Makes Bedrock AgentCore Payments GA with Coinbase Quick Create

AWS announced general availability of Payments in Amazon Bedrock AgentCore, letting agents transact safely and autonomously at scale. The GA includes a 'Quick Create' flow for Coinbase so developers can provision credentials from the console or CLI without leaving AgentCore, with developer credentials held in AgentCore identity so the agent never sees them, using short-lived tokens.

2026-08-20

Amazon Bedrock adds OpenAI GPT-5.6 models with in-country inference in India

Amazon Bedrock now supports OpenAI GPT-5.6 models (Terra and Luna) in India with India Geo cross-Region inference, letting customers with in-country requirements run OpenAI models at scale while keeping processing within India and automatically routing across AWS Regions for higher throughput.

2026-08-19

AWS IAM identity federation arrives in the European Sovereign Cloud

AWS IAM now lets workloads in the AWS European Sovereign Cloud (Germany) Region authenticate with external services using short-lived JSON Web Tokens. The independent, EU-located cloud is designed to help customers meet evolving data-sovereignty requirements.

2026-08-19

Amazon Bedrock adds OpenAI GPT-5.6 models with cross-Region inference

Amazon Bedrock now supports OpenAI's GPT-5.6 models (Sol, Terra and Luna) on its bedrock-runtime endpoint, with Responses, Converse and Chat Completions API support. New cross-Region inference lets customers use Global and Geo routing for higher throughput, lower inference costs and data-residency compliance, expanding deployment flexibility for generative-AI workloads.

2026-08-18

AWS adds agent-payment rails: OpenClaw agents transact via Bedrock AgentCore and x402

AWS published a guide showing how to give autonomous agents a wallet and spending guardrails so they can pay for paywalled APIs, MCP servers and web content. It connects OpenClaw agents to Bedrock AgentCore payments and the x402 protocol using the aws-agents-pay plugin, enabling bounded, human-approved testnet payments — a step toward agent-marketplace economics.

2026-08-18

AWS Lambda adds public preview runtimes with Node.js 26 and Python 3.15

AWS Lambda introduced public preview runtimes, letting developers try upcoming language versions before general availability. Developers can start using Node.js 26 and Python 3.15 today and provide feedback to shape runtime quality. Amazon separately raised its 2026 capex forecast to $220 billion tied to AI infrastructure buildout.

2026-08-17

Aurora serverless adds faster scaling for spiky agentic AI workloads

Aurora serverless now automatically adds 12 Aurora Capacity Units within a second and continues scaling to 256 ACUs as workloads grow. AWS benchmarks show how the cluster responds to sudden spikes versus a provisioned db.r8g.xlarge instance, targeting the bursty demand patterns typical of agentic AI applications.

2026-08-17

Amazon Quick Microsoft 365 extensions reach general availability

AWS made Amazon Quick's Microsoft 365 extensions generally available for Word, Excel, PowerPoint, and Outlook, bringing connected data access and agentic document editing directly into M365. The extensions can redline documents, build financial models, create decks, and manage Outlook inboxes without leaving Microsoft's apps.

2026-08-16

AWS Certificate Manager to discontinue email validation, adds DNS switch

AWS Certificate Manager will discontinue support for email-validated public certificates by September 30, 2027, aligning with the CA/B Forum's industry-wide deprecation of email-based domain validation. In tandem, ACM now lets customers switch existing public TLS certificates from email to DNS validation without reissuing the certificate or changing its ARN.

2026-08-14

AWS and OpenAI put Daybreak Red and Daybreak Blue cyber models on Amazon Bedrock

AWS and OpenAI made OpenAI's Daybreak Red (GPT-5.6 Cyber) and Daybreak Blue (GPT-5.6 Sol with defensive safeguards) available on Amazon Bedrock for eligible customers. The collaboration targets cyber-defense workloads such as vulnerability research and incident response, delivered under existing AWS governance. AWS also announced a strategic partnership with Novo Nordisk naming AWS its preferred cloud and AI partner for drug discovery.

2026-08-13

AWS IAM role manager reaches GA, auto-configuring roles for services

AWS announced general availability of role manager, an IAM capability that automatically sets up the roles AWS services need. When configuring a supported service in the console, role manager creates a default role or reuses an existing matching one, rethinking the starting point for IAM permissions and reducing setup friction. A companion security-blog post details the design.

2026-08-13

Amazon EKS adds advanced Kubernetes control-plane configuration

Amazon EKS now supports configuring control-plane components—including the scheduler, controller manager, and API server—directly through EKS APIs. Teams can tune pod placement, adjust autoscaling responsiveness, and set event retention, with a hands-on guide covering MostAllocated bin-packing and event-retention tuning. EKS 1.34 also makes the Kubelet Checkpoint API functional for forensic container checkpointing.

2026-08-13

AWS adds web search on Bedrock, open-sources Dogwood agent governance, launches Kiro Crew

Amazon Bedrock now lets OpenAI models GPT-5.4, 5.5 and 5.6 browse real-time web content with zero data egress. AWS also open-sourced Dogwood, a governance language for AI agents supporting Cedar policies and temporal conditions, and introduced Kiro Crew, a persistent multi-agent workspace inside the Kiro IDE.

2026-08-12

AWS becomes Novo Nordisk's preferred cloud provider for drug discovery, opens London AI hub

AWS became Novo Nordisk's preferred cloud provider and strategic AI partner for drug discovery, launching an AI co-innovation hub in London that uses Amazon Bio Discovery and Bedrock to analyze genomic, imaging and clinical data. The partnership aims to compress the path from drug target to first human dose.

2026-08-12

AWS partners with Novo Nordisk on AI drug discovery and retires four Nova models

AWS formed a strategic partnership with Novo Nordisk, becoming its preferred cloud provider and launching a London AI co-innovation hub using Amazon Bio Discovery, Bedrock, and Bedrock AgentCore to compress drug discovery timelines. Separately, Amazon began winding down Nova Premier, Nova Omni, Nova Reel, and Nova Canvas as part of a frontier-chasing overhaul.

2026-08-11

AWS expands SageMaker JumpStart with GLM-5.2, Nemotron-Nano-12B, and gemma-4-12B

AWS added a wave of foundation models to SageMaker JumpStart, including Z.ai's GLM-5.2 FP8, NVIDIA's Nemotron-Nano-12B-v2, GLM-OCR, Black Forest Labs' FLUX.2-small-decoder, Google's gemma-4-12B-it, plus specialized OCR, embedding, and code-reasoning models for scalable AWS deployment.

2026-08-11

Amazon Bedrock AgentCore Runtime Instances Reach GA — Persistent Compute for Production Agents

AWS made Amazon Bedrock AgentCore runtime instances generally available, providing persistent, managed EC2 infrastructure for production AI agents with sessions up to 14 days, GPU support, and multi-agent collaboration. The launch targets teams moving agentic workloads from prototype to production.

2026-08-09

AWS Adds Temporal Policies and Rate Limiting to Bedrock AgentCore Gateway

AWS introduced new AgentCore capabilities to control agent behavior and cost beyond a single action: temporal policies powered by Dogwood, a new open-source policy language for AI agents, plus gateway rate limiting. The features target 'sequence-of-permitted-actions' failures where individually-legitimate calls collectively breach budgets or policy.

2026-08-09

Cohere Health Digitizes Clinical Policies on Amazon Bedrock AgentCore

Cohere Health built a multi-tenant agentic architecture on Amazon Bedrock AgentCore, using Runtime's secure MicroVM isolation, unified tool access via Gateway, AgentCore Memory, and the Agent Skills standard. The system scales clinical-policy digitization while preserving transparency, version control, and human oversight — a concrete production deployment of AgentCore.

2026-08-09

AWS open-sources Dogwood runtime verification and ships AgentCore governance controls

AWS open-sourced Dogwood, a policy language and runtime verification system that enforces checks on AI-agent tool calls before they execute. It added Dogwood-powered temporal policies to Bedrock AgentCore — enforcing workflow sequencing, preventing data fabrication, and requiring human approval for high-value actions — plus gateway rate limits capping per-user token and connection traffic for deterministic control over agent behavior and cost.

2026-08-07

AWS adds built-in Web Search to Amazon Bedrock for foundation-model grounding

AWS made Web Search generally available on Amazon Bedrock, letting foundation models ground answers in current web knowledge without third-party search providers. The server-side tool is enabled via a single parameter in an existing OpenAI-compatible API call and returns responses with structured citations, expanding a capability previously limited to Bedrock AgentCore.

2026-08-07

Amazon ECS adds fractional GPU scheduling with EC2 G6f instances

Amazon ECS now supports fractional GPU scheduling with EC2 G6f instances, enabling GPU partitions as small as one-eighth of an NVIDIA L4 with 3GB of memory. The feature lets teams right-size containers for small-model inference, experimentation, and graphics rendering, improving GPU utilization and cost efficiency.

2026-08-07

AWS Q2 Revenue Hits $42.2B (+37% YoY), Fastest Growth in 18 Quarters; Capacity 'Spoken For' Through 2028

Amazon Web Services reported $42.2 billion in Q2 revenue, up 37% year-over-year and its fastest growth since late 2021. CEO Matt Garman said much of AWS capacity is 'spoken for through 2027 and into 2028,' with AI and Chips businesses each eclipsing $25 billion run rates at triple-digit growth. AWS also cut on-demand inference prices for GPT-5.6 Luna 80% on Bedrock effective July 30.

2026-08-06

AWS Security Hub adds Supply Chain Security as its 10th category with Chainguard and Socket

The AWS Security Hub Extended plan now includes Supply Chain Security as its 10th category, with Chainguard and Socket as curated partners, letting security teams detect and block malicious open-source dependencies before they enter their environments. The launch responds to escalating software supply-chain attacks.

2026-08-05

Amazon Bedrock Cuts OpenAI GPT-5.6 Prices by Up to 80%, Luna Drops to $0.20/M Input

Amazon announced price reductions for OpenAI's GPT-5.6 models on Bedrock effective July 30: GPT-5.6 Luna dropped 80% to $0.20/M input and $1.20/M output tokens, while GPT-5.6 Terra fell 20%. AWS also added automatic policy refinement for Automated Reasoning checks in Bedrock Guardrails and launched CloudWatch managed collectors for Prometheus metrics.

2026-08-04

AWS Transform Continuous Modernization Goes GA for Autonomous Tech-Debt Remediation

AWS Transform continuous modernization is now generally available in all supported regions, letting engineering teams autonomously analyze and remediate technical debt across GitHub organizations, GitLab groups, and Bitbucket workspaces. Teams can run analyses on demand or on a recurring schedule and prioritize findings, part of a broader batch of AWS agentic tooling this week.

2026-08-04

Amazon raises 2026 capex to $220B as AWS growth accelerates to 37% YoY

Amazon raised its 2026 capex guidance from $200B to $220B, citing AI demand and elevated memory costs. AWS revenue grew 37% YoY to $42.2 billion in Q2 2026 — its fastest in 18 quarters — reaching a $169B annualized run rate. Graviton and Trainium each crossed $25B annualized run rates, and the backlog reached $496 billion, with CEO Andy Jassy floating a $1 trillion AWS ambition.

2026-08-03

Amazon Bedrock AgentCore Observability launches to optimize production AI agents

AWS introduced Bedrock AgentCore Observability, integrated with CloudWatch, to find performance bottlenecks and diagnose memory issues in long-running agent sessions. The tooling targets teams moving AI agents from prototype to production while keeping them fast and cost-efficient.

2026-08-03

AWS booms as Amazon lifts AI capex to $220B and launches Forward Deployed Engineering

Amazon reported AWS at a ~$169B annualized run rate with a nearly $500B cloud backlog, prompting a $220B 2026 capital-spending plan that CEO Andy Jassy said still won't meet AI demand. Bedrock spending in Q2 topped all prior quarters combined. Amazon will also invest $1B in an AWS Forward Deployed Engineering team, embedding AI engineers inside customers like the NBA, NFL, and Southwest.

2026-08-02

AWS adds Web Search and Managed Knowledge Base to Bedrock AgentCore

At the AWS Summit, AWS announced general availability of Amazon Bedrock Managed Knowledge Base — with native connectors, Smart Parsing, and an agentic retriever — plus Web Search on AgentCore for current, cited web information inside secured AWS environments. Swami Sivasubramanian cited a 15x increase in tasks completed by AgentCore in six months.

2026-08-02

Amazon Q2: AWS AI tops $25B run rate with $496B backlog as Bedrock catalogue expands

Amazon reported AWS AI revenue running above $25 billion annually with a $496 billion backlog — its fastest cloud growth since 2021. Bedrock added 10+ managed models including GPT-5.6, Claude Opus 5, Gemma 4, and Grok 4.3, and SageMaker HyperPod added disaggregated prefill/decode inference. CEO Andy Jassy argued Amazon can win through platform breadth, not owning the single best model.

2026-08-01

Amazon scales back Nova AI models as Premier nears end of life

Amazon is scaling back development on several Nova AI models, including Nova Premier, Nova Canvas, and Nova Reel, which are scheduled to reach end-of-life in September 2026. The models are being classified as Legacy — new customers denied access, existing users limited — signaling a strategic shift toward a new frontier-model research effort.

2026-08-01

Amazon Aurora DSQL adds multi-Region clusters in four more Regions

Amazon Aurora DSQL now supports multi-Region clusters in Stockholm, Spain, Mumbai, and Singapore. The serverless distributed SQL database offers active-active high availability and multi-Region strong consistency with writable endpoints in each Region, alongside a wave of other AWS infrastructure updates.

2026-08-01

AWS Q2 revenue jumps 37% to $42.2B on AI demand; Claude Opus 5 lands on Bedrock

Amazon's AWS posted Q2 revenue of $42.23 billion, up 37% year-over-year — its fastest growth since 2021 — with an AI revenue run rate now exceeding $25 billion at triple-digit growth. CEO Andy Jassy credited Bedrock and SageMaker adoption and proximity of inference to customer data. AWS also signed a $410 million multi-year deal with Recursive for its self-improving AI system, and Anthropic's Claude Opus 5 became available on Amazon Bedrock.

2026-07-31

OpenAI GPT-5.6 models arrive on Amazon Bedrock with explicit prompt caching

OpenAI's GPT-5.6 Sol, Terra, and Luna became generally available on Amazon Bedrock, and AWS introduced explicit prompt caching for them — co-written with OpenAI — giving developers precise control over which prompt segments are cached and reused to cut inference cost. Bedrock now offers models from OpenAI, Anthropic, DeepSeek, Meta and others.

2026-07-31

AgentCore Gateway adopts major MCP 2026-07-28 spec revision

The Model Context Protocol published its largest revision since launch, making MCP stateless with a governed extensions system and hardened authorization. Amazon Bedrock AgentCore Gateway can adopt the new version with a single UpdateGateway call, and AWS demonstrated cross-system autonomous business insights via pre-built MCP connectors, fine-grained access control and persistent memory.

2026-07-30

Amazon ties North Korean hacker group to open-source supply-chain attacks

Amazon Threat Intelligence linked recent compromises of popular NPM libraries to a DPRK-affiliated threat actor targeting open-source software, spotlighting risks in the shared building blocks companies use to build applications. AWS also detailed protections for the riskiest first hours after a package is published in Amazon Linux.

2026-07-30

AWS details 'Beyond RAG' task-aware knowledge compression for enterprise AI

AWS detailed task-aware knowledge compression (TAKC), which pre-compresses entire knowledge bases into task-specific representations cached at multiple fidelity tiers. It addresses RAG's ceiling on analytical tasks spanning hundreds of documents, and ships with an open-source implementation.

2026-07-28

AWS moves ~20 AI services into maintenance mode to consolidate on agent-centric successors

AWS placed roughly 20 AI services — including Amazon Kendra, Q Business, and Bedrock Agents (now 'Classic') — into maintenance mode, halting new sign-ups, to consolidate its portfolio around agent-centric successors like Bedrock Knowledge Bases and Bedrock AgentCore. The move streamlines AWS's sprawling AI lineup but imposes a migration burden on early adopters.

2026-07-27

AWS releases aws-bench, an open-source benchmark for measuring AI agents

AWS announced a research preview of aws-bench, an open-source benchmark measuring how accurately and efficiently AI agents complete real-world AWS tasks. Derived from real usage analysis, it gives model providers and researchers a reproducible way to measure agent performance and diagnose failures — addressing the industry's lack of standardized agent evaluation.

2026-07-27

AWS moves ~20 AI services to maintenance mode and cuts AGI-group jobs

AWS moved approximately 20 AI services — including Amazon Kendra, Q Business, and Bedrock Agents (now Bedrock Agents Classic) — into maintenance mode, halting new customer sign-ups while letting existing users continue. The consolidation pushes users toward agent-centric successors like Bedrock Knowledge Bases and Amazon Quick Suite. Amazon also cut jobs in its AGI group.

2026-07-25

Amazon cuts AGI group jobs as Bezos pushes AI-focused Prime Video redesign

Amazon cut jobs in its artificial general intelligence group, the latest reduction since a larger January round, while maintaining AI is a top long-term priority. Jeff Bezos is separately pushing Prime Video head Mike Hopkins toward an AI-focused redesign to showcase Amazon's hundreds of billions in AI investment, part of a broader wave of AI-driven tech-workforce reshuffling.

2026-07-24

Amazon Bedrock AgentCore adds optimization to detect silent agent failures

AWS introduced AgentCore optimization, which surfaces silent behavioral failures in production AI agents — the ones that pass every health check but still deliver wrong outcomes. It discovers, explains, and ranks failure patterns across sessions so teams can fix the highest-impact issues first. AWS separately unified AgentCore observability by routing traces and prompts into the same CloudWatch log group.

2026-07-24

OpenAI's GPT-5.6 models land on Amazon Bedrock as SageMaker inference scales 2x faster

AWS made OpenAI's three-tier GPT-5.6 models — Sol (flagship reasoning), Terra (balanced), and Luna (cost-efficient) — available on Amazon Bedrock via a next-generation inference engine. Amazon SageMaker Inference now supports container image caching for up to 2x faster scaling during scale-out, and AWS expanded Kiro and Strands Agents tooling for enterprise agentic development.

2026-07-23

AWS launches GuardDuty investigation agent for AI-powered threat assessment

AWS launched the GuardDuty investigation agent in public preview, an AI-powered tool that autonomously investigates security findings across an AWS environment. It reduces investigation time from hours to minutes by assessing suspicious and potentially malicious activity detected by the managed threat-detection service.

2026-07-21

AWS ships self-managed S3 buckets for Lambda function code, removing 75GB quota

AWS now lets Lambda read function code directly from customer-controlled S3 buckets, eliminating the 75GB code storage quota. The change helps teams managing functions at scale and addresses security governance concerns about where deployment artifacts live.

2026-07-20

AWS launches Amazon Quick, an agentic AI sales teammate that works the full cycle

AWS introduced Amazon Quick, an agentic AI assistant spanning the full sales cycle — identifying high-priority prospects, contacting them, working deals to close, and keeping the CRM updated automatically. Quick was also named a Leader in the 2026 Gartner Magic Quadrant for Analytics and BI Platforms.

2026-07-19

AWS brings xAI's Grok 4.3 to Amazon Bedrock

AWS added xAI's Grok 4.3 to Amazon Bedrock, positioning it for agentic and enterprise workloads. The managed offering supports configurable reasoning effort, tool calling, structured output, image input, and stateful multi-turn conversations, expanding the model roster available to Bedrock customers.

2026-07-18

AWS launches Amazon Bedrock Managed Knowledge Base for enterprise agent search

AWS detailed Amazon Bedrock Managed Knowledge Base, built around three pillars — simplified setup, smarter retrieval, and production readiness — with code examples for standing up and querying a knowledge base for agents.

2026-07-17

AWS brings open agent protocols to the Strands Agents SDK

AWS detailed how the Strands Agents SDK now supports a converging set of agent protocols — MCP, A2A, UTCP, AG-UI, and x402 — to help agents reach internal tools, data, and external APIs more easily.

2026-07-17

AWS ships CloudWatch Logs intelligent tiering, DynamoDB Bulk Executor, and G7e in new regions

AWS released a wave of infrastructure updates: CloudWatch Logs intelligent tiering that auto-classifies logs across Standard, Infrequent Access, and Archive tiers to cut cost; an open-source Bulk Executor for DynamoDB for large-scale table operations without custom code; and EC2 G7e GPU instances (NVIDIA RTX PRO 6000 Blackwell) now in Frankfurt, Stockholm, and Mumbai with up to 2.3x inference over G6e.

2026-07-16

OpenAI GPT-5.6 Sol, Terra, and Luna reach general availability on Amazon Bedrock

OpenAI's GPT-5.6 models — Sol, Terra, and Luna — are now generally available on Amazon Bedrock. Sol targets advanced reasoning for coding and cybersecurity, Terra offers GPT-5.5-level performance at half the cost, and Luna provides fast, cheap high-volume inference, all on Bedrock's next-gen engine with in-region processing and 90% cached-input discounts.

2026-07-16

Amazon adds GuardDuty AI Protection and Security Hub AI inventory for Bedrock/SageMaker

AWS launched GuardDuty AI Protection, extending threat detection to Bedrock and SageMaker with monitoring for anomalous model invocations, cost-harvesting attacks and prompt injection. Security Hub added a continuously updated org-wide AI inventory of models, agents and pipelines, plus multicloud support for Microsoft Azure.

2026-07-15

AWS adds Gemma-4, Voxtral, Qwen3 and OpenAI privacy-filter to SageMaker JumpStart

AWS expanded SageMaker JumpStart with a batch of new models: Google DeepMind's multimodal gemma-4-E2B-it, Mistral's Voxtral-Mini-4B-Realtime for real-time transcription, Qwen3 embedding and reranking models for retrieval, and OpenAI's privacy-filter for PII detection and masking. The additions broaden AWS's managed catalog for building AI applications on its infrastructure.

2026-07-14

AWS launches Lambda MicroVMs: serverless compute with VM-level isolation for AI-agent code

AWS introduced Lambda MicroVMs, a new serverless primitive providing VM-level isolation, near-instant startup, and state retention, letting each user or job run in its own execution environment to securely run just-in-time code — including AI-generated code — without managing virtualization. A companion guide details how AI coding agents get an isolated sandbox with controlled tool access.

2026-07-13

AWS shows how KTern.AI built agentic AI for SAP on Bedrock AgentCore

AWS detailed how KTern.AI evolved from a SaaS platform into an agentic AI platform orchestrating multiple specialized agents across enterprise SAP programs, built on Amazon Bedrock AgentCore with the Strands Agents SDK for persistent context and production reliability. It's one of several new AgentCore case studies from AWS this week.

2026-07-12

Amazon plans $25B+ bond sale to fund heavy AI infrastructure investment

Amazon aims to raise at least $25 billion through a U.S.-dollar bond sale, per Bloomberg, to fund its massive AI infrastructure investments. The move underscores the enormous capital demands of the AI buildout, with the final offering size still being determined.

2026-07-12

Amazon EC2 G7 instances with NVIDIA RTX PRO Blackwell GPUs launch in N. Virginia

AWS made EC2 G7 instances, powered by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs, available in US East (N. Virginia). G7 delivers up to 4.6x AI inference performance and 2.1x graphics performance over the prior G6 generation, plus faster GPU-accelerated data analytics.

2026-07-11

Darktrace finds AI gateway with Amazon Bedrock access hijacked for cryptomining

Darktrace reported a cloud intrusion where an Amazon EC2 instance running a LiteLLM proxy for Amazon Bedrock was compromised and used to mine cryptocurrency. The incident highlights a new security risk as AI gateways — which centralize access to cloud identities, permissions, and foundation models — become high-value targets.

2026-07-11

AWS raises EC2 GPU Capacity Block prices ~20%, wins Warner Bros. Discovery and Fox

AWS raised EC2 Capacity Block prices for Nvidia P5 (Hopper) and P6 (Blackwell) instances by ~20% effective July 1, citing tight supply — its second hike in six months after January's 15% increase. Separately, Warner Bros. Discovery and Fox named AWS their preferred AI cloud provider for agentic ad-tech.

2026-07-11

OAuth support arrives for the AWS MCP Server

AWS now lets AI agents connect directly to the AWS MCP Server using AWS Sign-In and industry-standard OAuth, with no extra authentication software. Existing IAM permissions, identities, and governance controls continue to apply, with support for IAM federation and IAM Identity Center.

2026-07-10

AWS introduces Loom for building secure AI agents at scale

AWS introduced Loom, a framework for building and running AI agents within enterprise security and governance requirements. It targets organizations adopting agentic capabilities while needing compliant, serverless-friendly controls.

2026-07-10

Amazon raising at least $25 billion in bond sale to fund AI buildout

Amazon is seeking to raise at least $25 billion through a US dollar bond sale, its latest push to fund heavy AI infrastructure investments, adding to an industry-wide AI debt boom as tech companies tap debt and equity markets. Bloomberg first reported the offering as Amazon returns to the US bond market.

2026-07-09

Claude Sonnet 5 lands on Bedrock alongside Graviton5 instances

AWS made Claude Sonnet 5 available on Amazon Bedrock as its most capable Sonnet model for coding and agentic work. New EC2 C9g and C9gd instances powered by Graviton5 deliver 25% better compute and 5x larger cache than Graviton4, and SageMaker Inference added container image caching for up to 2x faster scale-out.

2026-07-08

AWS ships AgentCore image-editing and support-companion agents

AWS detailed two serverless agents built on Bedrock AgentCore: an image-editing agent that lets users describe edits in plain English, and a Support Companion that analyzes CloudWatch logs, searches AWS docs and creates support cases. Both run on AgentCore harness with minimal custom orchestration.

2026-07-08

Amazon ECS Managed Instances cuts GPU management fees up to 60%

Starting July 1, Amazon ECS Managed Instances reduced G-series management fees by 35% and P-series and Trainium fees by 60%. The reductions apply automatically, lowering costs for customers running GPU and accelerated AI workloads on ECS. AWS also expanded S3 Vectors to GovCloud (US) regions.

2026-07-08

AWS Raises Q3 Trainium 3 Server Shipment Targets Up to 30% on AI Chip Demand

Amazon AWS is increasing its Q3 ASIC server shipment targets by up to 30% for its 3nm Trainium 3 chips, driven by strong demand from Anthropic, OpenAI, and its own Bedrock platform. The chip reportedly supports trillion-parameter models and significantly outperforms its predecessor, positioning AWS to compete more aggressively with Google's TPU.

2026-07-07

AWS Adds One-Click Hugging Face to SageMaker Studio Integration

AWS announced a deep-link integration between Hugging Face and Amazon SageMaker AI, letting developers go from model discovery on Hugging Face to hands-on experimentation in SageMaker Studio with a single selection, cutting friction from discovery to deployment.

2026-07-07

AWS Introduces rDPO Selective Unlearning for Amazon Nova Content Moderation

AWS introduced Reverse Direct Preference Optimization (rDPO), the unlearning technique behind Amazon Nova's Customizable Content Moderation Settings. It teaches models to selectively 'forget' content while reducing over-deflection and preserving overall model quality.

2026-07-07

SageMaker HyperPod Adds Disaggregated Prefill and Decode for Faster LLM Serving

SageMaker HyperPod added Disaggregated Prefill and Decode (DPD), separating LLM inference's prefill and decode phases onto dedicated GPU pools. The KV cache transfers over Elastic Fabric Adapter using GPU-Direct RDMA, optimizing production chat and agentic workloads.

2026-07-07

AWS commits $1B to a Forward Deployed Engineering org to co-build agentic systems

AWS announced a $1 billion investment in a new Forward Deployed Engineering organization that will embed thousands of engineers directly with customer teams to co-build agentic systems in their own environments. The move targets the 'last mile' gap between AI pilots and production business processes.

2026-07-06

AWS commits $1B to Forward Deployed Engineers and raises Bedrock AgentCore quotas

AWS announced a $1B Forward Deployed Engineering organization to embed thousands of engineers directly with customers to co-build agentic systems, aiming to close the enterprise 'last mile.' It also raised Bedrock AgentCore quotas to 5,000 concurrent sessions and 200 interactions/second, and added Terraform provisioning in SageMaker Unified Studio.

2026-07-05

Amazon Bedrock adds detection for AI-generated phishing

AWS detailed how Amazon Bedrock helps detect AI-generated phishing emails, an escalating threat as attackers use generative AI and OSINT to craft thousands of unique, hard-to-detect messages. The capability targets the new asymmetry where defenders face industrialized, personalized phishing at scale.

2026-07-05

AWS Invests $1B in Forward-Deployed Engineers; Debuts Continuum Security and Context Knowledge Graph

At AWS Summit NYC, Amazon announced a $1 billion investment to embed forward-deployed engineers with enterprise customers, and launched AWS Continuum, an AI-native security tool that remediates code vulnerabilities at machine speed, and AWS Context, a knowledge graph inferring relationships across data for agents. AWS also raised GPU EC2 rates on P4/P5/P6 instances effective July 1.

2026-07-04

AWS commits $1 billion to new forward-deployed AI engineer unit

Amazon Web Services launched a $1 billion internal organization of forward-deployed engineers who embed within companies to build and deploy purpose-built AI agents, explicitly embracing the FDE model pioneered by OpenAI and Anthropic. Separately, Anduril and Amazon unveiled a mobile data center venture to bring edge computing to military frontlines.

2026-07-03

AWS raises AgentCore runtime quotas up to 5x to help enterprises scale AI agents

AWS increased Amazon Bedrock AgentCore runtime quotas by up to fivefold, letting enterprises run more concurrent AI agent sessions without requesting increases — up to 5,000 active concurrent sessions in major US regions and 2,500 in others. Separately, Aily Labs announced a partnership making its AI Decision Intelligence agents available on AWS Marketplace and Bedrock.

2026-07-03

AWS raises EC2 GPU prices ~20% on July 1; Graviton5 CPUs reach general availability

Amazon raised EC2 Capacity Blocks for ML prices ~20% starting July 1, with P6-B300 at $14.04/hr and P5 at $5.19/hr, citing supply and demand. Separately, AWS launched Graviton5 CPUs with 25% better performance and 5x larger cache than Graviton4, plus new tools including Continuum, Context, and Amazon Quick agent enhancements.

2026-07-03

AWS ships Lambda MicroVMs and Graviton5-powered EC2 C9g instances

AWS introduced Lambda MicroVMs for isolated stateful execution with near-instant launch, and EC2 C9g/C9gd instances on Graviton5 delivering 25% better compute than Graviton4. It also expanded Bedrock availability of Claude Fable 5, Gemma 4, and OpenAI GPT-5.4/5.5 with enterprise safeguards.

2026-07-02

Amazon Bedrock AgentCore expands to four new regions and raises runtime quotas to 5,000 sessions

AWS scaled Bedrock AgentCore with higher default runtime quotas — up to 5,000 active concurrent sessions in US East (N. Virginia) and US West (Oregon), and 2,500 in other regions — and expanded availability to Asia Pacific (Bangkok, Malaysia) and Europe (Milan, Spain). The changes give developers more capacity and geographic reach to scale agent workloads.

2026-07-02

Amazon EKS adds Version Rollback for safer Kubernetes cluster upgrades

Amazon EKS Version Rollback lets administrators safely reverse Kubernetes version upgrades within seven days with no cluster rebuilds required, turning upgrades into a reversible, low-risk operation and providing a safety net for fleet-wide rollouts.

2026-07-02

New log analytics engine for Amazon OpenSearch Service cuts cost up to 70%

AWS introduced a purpose-built log analytics engine for Amazon OpenSearch Service delivering up to 4x price performance, 2x faster ingestion, up to 2x faster analytical queries, and up to 70% lower storage costs — without sacrificing search capabilities on the same data.

2026-07-02

AWS commits $1 billion to embed forward-deployed AI engineers with customers

AWS launched a Forward Deployed Engineering (FDE) organization backed by $1 billion, seeding it with thousands of engineers who embed within customer companies to build purpose-built AI agents. VP Francessca Vasquez said AWS aims to send five to six pods of engineers to customers for 45-day periods, compressing agentic AI deployment from months to days. The move follows similar FDE units from OpenAI ($4B) and Anthropic ($1.5B), and echoes Palantir's decade-old model.

2026-07-01

AWS CloudFormation Express mode speeds deployments up to 4x

AWS launched CloudFormation and CDK Express mode, cutting infrastructure deployment time by up to 4x for developers and AI agents. Express mode completes stack operations once resource configuration is applied, skipping extended stabilization checks like traffic readiness and region propagation. It's available in all commercial Regions at no additional cost.

2026-07-01

Amazon WorkSpaces for AI agents reaches general availability

Amazon WorkSpaces for agents is now generally available, enabling AI agents to securely access and operate desktop applications through managed WorkSpaces environments. It targets critical business processes running on ERP systems, CRMs, mainframes, and proprietary tools where customization and compliance make them hard to replace.

2026-07-01

AWS Summit NY debuts Lambda MicroVMs, EC2 G7 Blackwell instances, and Bedrock Managed Knowledge Bases

At the AWS Summit in New York, AWS launched Lambda MicroVMs (a serverless primitive with VM-level isolation, rapid launch/resume, and 8-hour state preservation) and EC2 G7 instances with NVIDIA RTX PRO 4500 Blackwell GPUs delivering up to 4.6x inference performance over G6. AWS also made Bedrock AgentCore generally available and introduced a fully Managed Knowledge Base that streamlines enterprise RAG pipelines with native connectors, smart parsing, and an Agentic Retriever, plus Continuum for vulnerability detection and S3 annotations for agents.

2026-06-30

AWS adds WAF protection for Amazon Bedrock AgentCore Gateway

AWS announced general availability of AWS WAF protection for Amazon Bedrock AgentCore Gateway, letting security and platform teams apply consistent, customizable web protections to agentic AI workloads moving from prototype to production, guarding against common web exploits and abuse.

2026-06-30

Anthropic and Amazon clash publicly over token-based Claude pricing

Anthropic publicly fired back at Amazon's CEO amid tensions over a renegotiated contract that ties Amazon's payments to token usage, leading to much higher bills as Claude consumption grows. The Information reports Amazon is now looking for alternatives to Claude, exposing strain in one of AI's most important cloud-model partnerships.

2026-06-30

AWS pairs Amazon Nova 2 Lite with Claude for cost-optimized document processing

AWS detailed a two-model Bedrock pipeline pairing Amazon Nova 2 Lite with Anthropic's Claude Sonnet 4.6 to digitize scanned documents at scale. Nova 2 Lite handles native multimodal extraction — detecting photos and extracting names with coordinates — while Claude adds reasoning, delivering a cost-efficient workflow.

2026-06-30

AWS launches Lambda MicroVMs with VM-level isolation and near-instant starts

AWS introduced Lambda MicroVMs, a serverless primitive offering VM-level isolation and near-instant launch speeds, among 41 announcements. Developers are flagging it as a game-changer for safely running AI-generated code and agent workloads, while EC2 R8g (Graviton4) instances expanded to new regions and GovCloud Bedrock models hit FedRAMP High.

2026-06-29

Amazon raises EC2 Capacity Block prices ~20% as GPU demand and memory costs soar

Amazon raised prices for EC2 Capacity Blocks for ML — which let companies reserve GPUs in advance — by roughly 20% starting in July, after a ~15% hike in January. AWS framed it as a supply-and-demand response amid soaring memory-chip costs, a sign of the industry-wide AI affordability squeeze.

2026-06-29

AWS details Stripe's production-grade AI agents for financial compliance

AWS published a detailed account of how Stripe built a production-grade AI agent system for financial compliance using a ReAct agent framework and a dedicated agent service. The post covers human oversight, task decomposition, orchestration and cost optimization — a concrete reference architecture for regulated-industry agents.

2026-06-29

Amazon adds $13B to lift India AI and cloud investment to $48B, raises AI cloud prices

Amazon announced a fresh $13 billion investment to expand AI and cloud infrastructure in India, lifting its total commitment to $48 billion, with CEO Andy Jassy pledging access to custom AI chips and managed AI services. Separately, Amazon raised AI cloud prices as memory chip costs surged, and AWS signed a multi-year agentic AI collaboration with healthcare firm Innovaccer.

2026-06-28

Amazon GuardDuty adds AI-powered investigations in preview

AWS previewed AI-powered investigations in Amazon GuardDuty that automatically analyze findings and accounts to separate true threats from benign alerts. The feature targets alert fatigue and faster incident response for security operations teams.

2026-06-25

AWS builds healthcare appointment voice agent on Amazon Nova 2 Sonic

AWS detailed building a voice agent on Amazon Nova 2 Sonic and Bedrock AgentCore that authenticates patients by voice, manages appointments, collects pre-visit information, and escalates to human staff. The pattern targets reducing no-shows and scaling routine call handling.

2026-06-25

ArcelorMittal taps AWS for industrial AI automation and lower-carbon steel

ArcelorMittal announced a strategic collaboration with AWS to bring cloud, AI, and edge technologies to its manufacturing for predictive maintenance, computer-vision quality control, and energy optimization. The deal includes a multi-year framework for ArcelorMittal to supply lower-carbon XCarb steel to Amazon facilities and AWS data centers.

2026-06-25

Amazon Bedrock AgentCore Memory adds cross-account access for multi-account agents

AWS enabled cross-account access for Bedrock AgentCore Memory, allowing memory resources and consuming agents to span multiple AWS accounts. Resource-based policies grant principals in one account permission to call memory data-plane APIs in another, with configurable delivery destinations like S3—targeting enterprises running agents across organizational boundaries.

2026-06-24

Bedrock Guardrails adds automated reasoning policy-refinement workflows

AWS announced new automated refinement workflows for Automated Reasoning checks in Bedrock Guardrails. The checks use formal logic to mathematically validate generative AI responses against defined policies, helping detect hallucinations and providing verifiable explanations—now with workflows to iteratively refine those policies.

2026-06-24

ArcelorMittal partners with AWS to drive AI-powered industrial automation across global steelmaking

ArcelorMittal entered a strategic collaboration with AWS to bring cloud, AI, and edge technologies to its steel manufacturing across 14 countries for predictive maintenance, quality control, and energy optimization, converging OT and IT on AWS. The companies also signed a multi-year framework to supply lower-carbon XCarb steel to Amazon facilities and AWS data centers.

2026-06-24

AWS Lambda introduces MicroVMs for isolated execution of user and AI-generated code

AWS launched Lambda MicroVMs, a serverless compute primitive offering VM-level isolation with no shared kernel between sessions, near-instant launch/resume, and state preservation up to 8 hours. The primitive targets securely running user- or AI-generated code without managing virtualization infrastructure.

2026-06-23

Ampersend builds pay-per-intelligence routing on Bedrock AgentCore Payments

AWS detailed how Ampersend built a pay-per-intelligence routing layer on Amazon Bedrock AgentCore Payments, where AI agents autonomously route tasks to the most effective model, pay per request, and operate within spending budgets. The post walks through the two-hop payment pattern end-to-end.

2026-06-23

Grok 4.3 from xAI now available in Amazon Bedrock

AWS announced that xAI's Grok 4.3 is now generally available on Amazon Bedrock, with a 1-million-token context window, tool calling, structured output, and low hallucination rates for enterprise generative AI. The addition broadens Bedrock's multi-model lineup alongside Claude and other frontier options.

2026-06-23

SageMaker Data Agent adds Snowflake SQL analytics and interactive charting

Amazon SageMaker Data Agent gained three capabilities in Unified Studio notebooks: SQL analytics on Snowflake sources, materialized view management, and interactive charting from natural-language prompts — demonstrated via cross-platform fraud detection across Snowflake and AWS data.

2026-06-23

AWS DevOps Agent reaches GA with PagerDuty, Datadog and LaunchDarkly integrations

AWS DevOps Agent is now generally available, integrating with Datadog's MCP Server, PagerDuty, and LaunchDarkly to push toward autonomous incident resolution in minutes. The agent correlates monitoring data with infrastructure, surfaces root causes, and can coordinate feature-flag changes during outages.

2026-06-22

Web Search on Amazon Bedrock AgentCore reaches general availability

Web Search on Amazon Bedrock AgentCore is now generally available, giving agents built-in live web retrieval. Developers can wire it into agent workflows with a few lines of code to ground responses in current information.

2026-06-22

AWS unveils Continuum AI-powered vulnerability management platform

AWS launched Continuum, an AI-native security service that discovers, prioritizes, validates, and remediates code vulnerabilities at machine speed. Separately, an Amazon executive said the company expects the first commercially useful small-scale quantum computers within five-to-seven years, framing quantum as a moat to outlast the AI boom.

2026-06-22

Amazon in talks to sell custom AI chips, challenging Nvidia dominance

Amazon is reportedly in talks to sell its custom AI chips to other companies in a bid to cut into Nvidia's ~90% accelerator dominance amid shortages and high costs. SVP Peter DeSantis said Amazon expects its own AI models to be competitive with frontier models from Anthropic and OpenAI within the next year.

2026-06-21

Amazon EC2 G7 instances with NVIDIA RTX PRO 4500 Blackwell GPUs now GA

AWS launched EC2 G7 instances accelerated by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. They deliver up to 4.6x AI inference performance and 2.1x graphics performance versus the prior G6 generation, targeting inference, graphics, and analytics workloads.

2026-06-21

AWS adds Web Search and continuous learning to Bedrock AgentCore

AWS launched a fully managed Web Search tool for Bedrock AgentCore that grounds agent responses in current, cited web knowledge with zero data egress from the customer's secured environment. AgentCore also gained features connecting agents to organizational, web, and paid knowledge, plus production debugging and scalable controls.

2026-06-18

AWS unveils Continuum, an AI-native 'security at machine speed' platform

At Summit NYC, AWS announced Continuum, a security platform that shifts from the collect-store-query-dashboard model toward telemetry, context, reasoning, and autonomous action. AWS also added STRIDE-based threat modeling and a Claude Code plugin to its Security Agent, enabling repo and PR scanning with remediation.

2026-06-18

Amazon Bedrock Guardrails launches InvokeGuardrailChecks API for agentic AI

AWS launched a resourceless InvokeGuardrailChecks API for Amazon Bedrock Guardrails, letting developers apply individual safety checks at any point in agentic AI applications without creating guardrail resources. It offers per-request control over which safeguards run at each step of an agent loop, returning numeric severity and confidence scores.

2026-06-17

Amazon S3 annotations attach up to 1GB of queryable context to objects for AI agents

At AWS Summit New York, Amazon S3 launched annotations letting users attach up to 1GB of rich, mutable, queryable context directly to objects. The feature is purpose-built for AI agents and autonomous workflows that need to discover, understand, and act on data at scale without maintaining separate metadata systems.

2026-06-17

Amazon S3 Vectors now returns up to 10,000 similarity-search results per query

Amazon S3 Vectors can now return up to 10,000 similarity-search results per query, a 100x increase from the prior limit. The higher cap helps multi-stage retrieval pipelines that need a larger candidate set for reranking and aggregation.

2026-06-17

Amazon's Jassy reportedly flagged Fable jailbreak to White House before crackdown

Reuters and The Information report Amazon CEO Andy Jassy first raised concerns about a Fable jailbreak with the White House, preceding the government directive ordering Anthropic to block foreign nationals from Fable 5 and Mythos 5. The episode complicates the Amazon-Anthropic strategic partnership and, per the Washington Post, sets a precedent for intervening in deployed models via export-control authorities.

2026-06-16

AWS WAF launches AI traffic monetization to let publishers charge AI bots

AWS WAF announced AI traffic monetization, a new Bot Control capability that lets content owners and publishers price, meter, and collect payment from AI bots and agents accessing their content and APIs. Owners can set a price, accept payment through third-party providers, and grant scoped access directly at the edge.

2026-06-16

AWS Nitro Isolation Engine formally verifies the Nitro hypervisor

AWS introduced the Nitro Isolation Engine, applying formal verification to the hypervisor at the heart of the AWS Nitro System — a significant confidential-computing advance strengthening guarantees that customer data stays secure and isolated for sensitive workloads.

2026-06-14

SageMaker AI adds serverless fine-tuning for NVIDIA Nemotron 3 Nano

Amazon SageMaker AI now supports serverless model customization for NVIDIA's open-weight Nemotron 3 Nano via supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT), letting teams adapt the 30B model to custom domains and workflows without managing infrastructure.

2026-06-14

Amazon secures $17.5B loan for AI capex as Bezos's Prometheus hits $41B valuation

Amazon secured a $17.5 billion delayed-draw term loan facility to fund its AI-driven capital-expenditure ramp, following a five-part C$14 billion debt offering in Canada. Separately, Jeff Bezos's new AI startup Prometheus was valued at $41 billion in a blockbuster raise, and AWS signed a $4 billion AI deal with Pinterest for visual search.

2026-06-13

Moonshot's Kimi K2.7-Code draws developer buzz for token efficiency

Moonshot AI released Kimi K2.7-Code, a coding-focused agentic model built on Kimi K2.6, on Hugging Face. It improves real-world long-horizon coding and reduces thinking-token usage by ~30% versus its predecessor. The community framed it as a serious open-source challenger to leading proprietary coders, with its Hugging Face page hitting 406 points on Hacker News.

2026-06-13

AWS launches Graviton5-powered M9g/M9gd EC2 instances; Claude Fable 5 GA on Bedrock

AWS made Amazon EC2 M9g and M9gd instances generally available on the new Graviton5 processor, delivering up to 25% faster compute, 35% faster ML inference, and 30% faster database performance versus Graviton4. Claude Fable 5 simultaneously became generally available on Amazon Bedrock for agentic workloads.

2026-06-12

Apache Spark 4.0 reaches general availability on Amazon EMR

AWS announced general availability of Spark 4.0 across EMR Serverless, EMR on EC2 and EMR on EKS, adding Spark Connect, the Variant data type, SQL scripting, Python API improvements and streaming enhancements. SageMaker Unified Studio Notebooks also gained EMR Serverless support.

2026-06-11

AWS launches Neuron Agentic Development to automate Trainium kernel tuning

AWS announced Neuron Agentic Development, a collection of AI agents and skills for developers building on AWS Trainium and Inferentia, aiming to replace manual hand-tuning of kernels with an agent-driven workflow. The release deepens AWS's effort to make its custom AI silicon more accessible to developers.

2026-06-11

Claude Fable 5 reaches general availability on Amazon Bedrock

AWS announced the general availability of Anthropic's Claude Fable 5 on Amazon Bedrock and the Claude Platform on AWS, bringing Mythos-level capabilities to all customers with built-in safeguards. AWS VP Swami Sivasubramanian said the model can sustain complex tasks for days without intervention and understands diagrams and tables nested in documents, while running under the security, identity and data controls enterprises already use on Bedrock.

2026-06-10

Amazon taps Corning to power US AI data centers; Corning shares jump 9%

Corning shares rose 9% after striking a deal to supply Amazon's US AI data centers, which Corning's CEO called a milestone for American manufacturing. Amazon's largest AI data center has come online powering Anthropic's workloads without Nvidia chips, underscoring Amazon's push to build a resilient domestic AI infrastructure base.

2026-06-10

AWS Lambda Managed Instances expand to all commercial Regions

AWS Lambda Managed Instances (LMI) — which run Lambda functions on managed EC2 instances for specialized compute and EC2 pricing advantages — are now available in all commercial AWS Regions except Tel Aviv, Bahrain, UAE, and Auckland, while retaining Lambda's operational simplicity.

2026-06-09

PostgreSQL 19 Beta 1 lands in Amazon RDS Preview with native graph queries

Amazon RDS for PostgreSQL 19 Beta 1 is now testable in the RDS Database Preview Environment. PostgreSQL 19 introduces native graph query support via SQL Property Graph Queries, letting developers evaluate the pre-release with full managed-database benefits ahead of GA.

2026-06-09

AWS adds AI-powered root-cause investigations for cost anomalies via Amazon Q

AWS Cost Anomaly Detection now uses Amazon Q to analyze the root cause of detected anomalies, correlating cost data with CloudTrail events and resource activity to return a plain-language explanation in minutes — cutting work that traditionally took FinOps and engineering teams hours.

2026-06-09

New Amazon Bedrock console optimized for Anthropic- and OpenAI-compatible APIs

AWS launched a redesigned Bedrock console that lets users browse and compare the latest models side by side, organize work into projects with streamlined evaluation workflows, and get project-aware live documentation with auto-prefilled, copy-ready code snippets. The experience is optimized for the bedrock 'Mantle' endpoint supporting recent GPT, Claude and open-weight models via OpenAI- and Anthropic-compatible APIs.

2026-06-08

Amazon commits $50B to government AI and supercomputing capacity

Amazon announced a $50 billion investment to build 1.3 gigawatts of AI and supercomputing capacity across AWS Top Secret, AWS Secret and GovCloud regions. The move reflects a strategic focus on federal AI adoption and national-security positioning, alongside new agentic Amazon Connect solutions for supply chain, hiring, customer experience and healthcare.

2026-06-08

Bedrock AgentCore Runtime adds interactive terminal shells into agent sessions

Amazon Bedrock AgentCore Runtime now supports interactive shells via a new InvokeAgentRuntimeCommandShell API, opening a persistent, PTY-backed terminal into a running agent session over WebSocket. It complements one-shot execution with a full terminal experience—colors, tab completion, Ctrl sequences—inside an isolated microVM. AWS also extended its MCP Server to support cross-account and cross-role access.

2026-06-08

Amazon unveils natural-language Proteus warehouse robot in $12B Europe push

Amazon revealed an upgraded Proteus autonomous mobile robot that workers can direct with conversational prompts — no programming required — as part of a €10 billion ($11.6B) European fulfilment investment. Unlike the current Proteus (deployed at 25 US sites, limited to dock areas, ~400kg loads), the new model operates across warehouse floors and figures out priorities and routes itself, due in Europe in the first half of 2027.

2026-06-07

AWS makes OpenAI GPT-5.5, GPT-5.4 and Codex generally available on Bedrock

Amazon Bedrock now offers OpenAI's GPT-5.5, GPT-5.4 and the Codex coding agent at first-party pricing with no added fees, running on Bedrock's next-generation inference engine with isolated queues, automated capacity management, and unified IAM/VPC/KMS controls. GPT-5.4 reached AWS GovCloud (US-West) on June 3. Separately, Anthropic signed a $30B infrastructure deal to deploy Claude on AWS.

2026-06-05

AWS open-sources ExtendDB, a DynamoDB-compatible database backed by PostgreSQL

AWS open-sourced ExtendDB, a Rust implementation of the DynamoDB wire protocol that stores data in PostgreSQL while supporting the full API surface — offering a path to DynamoDB compatibility without AWS lock-in.

2026-06-03

AWS brings Claude Opus 4.8 to Bedrock and previews OpenAI-powered Managed Agents

Amazon Web Services made Anthropic's Claude Opus 4.8 available on Amazon Bedrock and the Claude Platform on AWS, and separately announced Bedrock Managed Agents powered by OpenAI in preview for faster long-running agentic tasks. The dual move consolidates frontier models from both leading labs onto Bedrock as AWS deepens its Anthropic bet.

2026-06-02

Claude Opus 4.8 now available on Amazon Bedrock

AWS made Claude Opus 4.8 available on Amazon Bedrock and the Claude Platform on AWS, exposing the model's 1M-token context window and multi-step workflow capabilities. AWS separately expanded Service Quotas visibility for its bedrock-mantle endpoint, which serves OpenAI and Anthropic APIs.

2026-06-01

Amazon shuts down internal token leaderboard, warns against 'tokenmaxxing'

Amazon SVP Dave Treadwell told staff to stop 'tokenmaxxing,' shutting down an internal token-usage leaderboard and urging employees to use AI to solve real business problems rather than for its own sake. The move reflects a broader corporate pivot away from free-wheeling AI spending.

2026-06-01

AWS launches Bedrock AgentCore Payments so agents can transact via micropayments

Amazon Bedrock launched AgentCore Payments, a suite letting AI agents autonomously make micropayments for web content, APIs and other agents. Built with Coinbase and Stripe, it aims to make agents economically active by letting them discover, evaluate and pay for resources within their operational loops.

2026-05-31

AWS Brings Claude Opus 4.8 to Bedrock, Launches Native Claude Platform Inside AWS Accounts

AWS made Claude Opus 4.8 generally available on Bedrock and became the first cloud to host Anthropic's native Claude Platform inside customer AWS accounts with unified credentials and billing. AWS also announced Bedrock AgentCore Payments with Coinbase and Stripe.

2026-05-29

AWS Unveils Random Network Graph (RNG) Datacenter Network Design

AWS detailed a new datacenter network architecture called Random Network Graph (RNG) that boosts throughput and reliability via a power-free fiber-optic cable management system and a custom network protocol — engineered for AI-scale traffic patterns.

2026-05-29

AWS revamps Partner Central with Bedrock AgentCore agents, claims 40% admin-time reduction

AWS revamped its Partner Central hub with AI agents powered by Amazon Bedrock AgentCore that automate administrative tasks, funding workflows and co-sell opportunity management. The company claims a 40%+ reduction in partner administrative time. The hub also connects to MCP servers so partners can integrate Partner Central with their existing IT tools.

2026-05-27

Uber reportedly burns full-year AI budget in 4 months after Claude Code rollout

Energy News Beat reports Uber exhausted its full-year AI budget in roughly four months after rolling out Claude Code across engineering, with token-based billing breaking enterprise forecasting. The story is becoming the canonical example developers and CFOs cite when arguing that current AI pricing models are unsustainable at production scale.

2026-05-27

ServiceNow crosses $1B in AWS Marketplace transactions as enterprise AI procurement consolidates

ServiceNow surpassed $1B in cumulative AWS Marketplace transactions, underscoring how enterprises are concentrating AI procurement through hyperscaler marketplaces for governance and compliance. A separately disclosed (then redacted) SpaceX S-1 detail also reveals underlying economics of the Anthropic–AWS infrastructure relationship.

2026-05-25

Bedrock becomes operational layer for enterprise AI as Q1 token volume exceeds all prior years combined

AWS is repositioning Bedrock from a managed-model marketplace into the operational layer for enterprise AI — hosting models, managing inference, supporting agent runtimes, enabling orchestration, and providing cost, security and governance controls. AWS disclosed Q1 2026 tokens processed on Bedrock exceeded all previous years combined, marking an inflection point from isolated experimentation to integrated AI operational systems.

2026-05-22

AWS Hits GA on Claude Platform, Expands OpenAI Partnership Bringing OpenAI Models to Bedrock

AWS reached general availability for Claude Platform on AWS — letting customers access Anthropic's native Claude experience through their AWS accounts with no separate credentials, a cloud-first. Andy Jassy also announced an expanded OpenAI partnership bringing OpenAI models, Codex, and Bedrock integrations to AWS, and AWS released Advanced Prompt Optimization in Bedrock with built-in evaluation feedback loops across up to five models.

2026-05-21

AWS explains why Bedrock AgentCore chose Cedar to secure agentic workflows

AWS's security blog published a deep-dive on why Bedrock AgentCore adopted the Cedar policy language for agent authorization. The team argues LLM non-determinism, prompt injection, and confidently-hallucinated harmful actions need declarative, auditable policy — not guardrails baked into prompts.

2026-05-20

Amazon Bedrock launches Advanced Prompt Optimization for cross-model prompt migration

AWS announced Amazon Bedrock Advanced Prompt Optimization (launched May 14), a tool that optimizes prompts for any Bedrock model and lets users compare original and optimized prompts across up to five models simultaneously. It supports multimodal inputs (PNG, JPG, PDF) and includes built-in evaluation feedback loops for prompt migration.

2026-05-20

AWS Bedrock AgentCore expands: OpenSearch Agent Skills, custom code evaluators, Nova 2 moderation

AWS shipped a stack of Bedrock AgentCore expansions: OpenSearch Agent Skills (a repository of composable skills for agentic IDEs), custom Lambda-based code evaluators registered with AgentCore, and structured prompting recipes for Amazon Nova 2 Lite content moderation aligned with MLCommons' AILuminate standard. Bedrock Q1 token processing exceeded all previous years combined.

2026-05-19

AWS positions Bedrock as the operational hub for enterprise agentic AI; Q1 2026 token volume tops all prior years combined

AWS is repositioning Amazon Bedrock from model marketplace to comprehensive operational hub for enterprise AI — hosting models, managing inference, supporting agent runtimes, and orchestrating workflows. Q1 2026 Bedrock token processing surpassed all previous years combined, and Amazon's market cap is nearing $3 trillion on AI momentum.

2026-05-17

AWS brings OpenAI models to Bedrock; publishes AI Security Framework; sets Q Developer EOL

AWS expanded its OpenAI partnership, bringing the latest models to Amazon Bedrock with Managed Agents in limited preview — one day after Microsoft's exclusivity ended. It also published its AI Security Framework with a free SHIP assessment, launched Bedrock Advanced Prompt Optimization, and set Amazon Q Developer EOL for April 30, 2027.

2026-05-16

SageMaker JumpStart adds FLUX.2, Qwen3 speech/embedding models, GLM-5.1, and Phi-4-mini

AWS expanded SageMaker JumpStart with Black Forest Labs' FLUX.2-klein-base-4B for image generation, Qwen3-Embedding-0.6B, three Qwen3 speech models (TTS and ASR), Z.ai's GLM-5.1-FP8, and Microsoft's Phi-4-mini-instruct for agentic and efficient inference.

2026-05-15

Amazon scraps API fees, launches Claude Platform on AWS, plans $200B AI capex

Amazon reversed planned API fees after tech-firm backlash and launched a new Claude Platform on AWS sitting alongside Bedrock, deepening its Anthropic relationship. AWS is also racing to future-proof AI data centers with $200B in 2026 capex including the modular 'Houdini' initiative.

2026-05-15

Bedrock AgentCore adds managed payments via Coinbase and Stripe

Amazon Bedrock AgentCore introduced managed payment capabilities, letting AI agents autonomously access and pay for APIs and services through Coinbase and Stripe integrations. AWS also released an Agent Toolkit and made AWS MCP Server generally available.

2026-05-14

Amazon merges Rufus and Alexa+ into 'Alexa for Shopping' on Amazon.com

Amazon is dropping the Rufus brand and embedding its Alexa+ LLM assistant directly into Amazon.com search and the Shopping app. Queries now return agentic AI responses alongside traditional results, unifying Amazon's two parallel shopping-AI efforts into one product.

2026-05-13

AWS publishes end-to-end agentic SRE and expands WorkSpaces AI Agents to legacy Windows

AWS detailed how to build an end-to-end agentic SRE using AWS DevOps Agent for incident correlation across observability tools and microservice architectures. Separately, Amazon WorkSpaces AI Agents (public preview) lets autonomous agents drive full desktop environments via IAM, MCP, and computer vision — including legacy Windows apps without APIs. Service Catalog also expanded to Calgary and New Zealand.

2026-05-10

S3 Vectors + Aurora PostgreSQL: SQL over billion-scale embeddings

AWS detailed how to query Amazon S3 Vectors directly from Aurora PostgreSQL using standard SQL, combining vector similarity with relational filters in a single statement. The integration targets cost-efficient, large-scale semantic search inside existing relational databases.

2026-05-09

AWS MCP Server hits general availability with full-service auth for AI agents

AWS announced GA of the AWS MCP Server, a managed remote Model Context Protocol server giving AI agents and coding assistants authenticated access to all AWS services. It ships as part of the broader Agent Toolkit for AWS, alongside skills and plugins for coding agents.

2026-05-06

Tomofun deploys vision-language models on AWS Inferentia2 for pet behavior

Taiwan-based pet-tech startup Tomofun (Furbo Pet Camera) moved its vision-language inference to EC2 Inf2 instances on AWS Inferentia2 chips. AWS detailed the architecture for cost-effective VLM deployment while maintaining accuracy.

2026-05-06

GPT-5.5, GPT-5.4, and Codex land on Amazon Bedrock with Managed Agents in preview

AWS expanded its OpenAI partnership, bringing GPT-5.5 and GPT-5.4 to Amazon Bedrock in limited preview alongside Bedrock Managed Agents powered by OpenAI for production agentic deployments. Codex is also coming to Bedrock with OpenAI's harness integrated for faster execution and reasoning.

2026-05-06

Amazon WorkSpaces gives AI agents their own cloud desktops to drive legacy apps

AWS launched a preview that lets AI agents securely operate legacy desktop applications — mainframes, ERP systems, proprietary tools — via managed WorkSpaces environments. The service uses IAM authentication, MCP support, and computer vision so agents can tackle the 'last-mile' problem of apps without modern APIs, all inside existing security frameworks.

2026-05-06

AWS brings GPT-5.5 and GPT-5.4 to Bedrock with Managed Agents

AWS announced availability of OpenAI's GPT-5.5 and GPT-5.4 on Amazon Bedrock alongside new Bedrock Managed Agents powered by OpenAI. AWS also overhauled Amazon Connect into four vertical agentic modules and launched Amazon Quick Desktop, an AI assistant for work, in preview.

2026-05-05

SageMaker adds agentic fine-tuning for Llama, Qwen, DeepSeek and Nova

Amazon SageMaker AI now ships an AI agent that helps developers customize open-weight language models including Llama, Qwen, DeepSeek and Amazon's own Nova. The feature targets teams that want guided fine-tuning workflows rather than hand-rolled pipelines.

2026-05-05

Pentagon strikes classified-network AI deals with eight providers including AWS, OpenAI, Google, NVIDIA

The Department of Defense announced agreements with AWS, Microsoft, Google, OpenAI, NVIDIA, SpaceX, Reflection and Oracle to deploy advanced AI on classified IL6 and IL7 networks. The pact is framed as the operational foundation for an 'AI-first' U.S. military, covering data synthesis, decision support and warfighter tooling.

2026-05-05

Amazon rolls out Claude Code and Codex to all employees after internal pushback against Kiro

Amazon formally rolled out Anthropic's Claude Code and OpenAI's Codex to all corporate employees, going beyond its in-house Kiro tool after staff pressure. The move is a notable validation of third-party coding agents at hyperscaler scale and intensifies competition in the enterprise developer-tools market — even as Amazon is itself one of Anthropic's largest investors.

2026-05-05

Pentagon expands classified AI deployments with AWS, Microsoft, Nvidia

Bloomberg reports AWS and Microsoft are giving the Pentagon more direct control over AI systems on classified, closed networks, alongside expanded Nvidia hardware contracts. The deals widen frontier-AI deployment inside national-security infrastructure.

2026-05-04

Amazon Quick launches as stateful enterprise agent with M365, Workspace, Salesforce hooks

AWS unveiled Amazon Quick, a stateful agent that maintains context across enterprise apps, learns user preferences, and autonomously executes tasks. Native integrations with Microsoft 365, Google Workspace, and Salesforce are pitched as the differentiator from point-in-time chatbots, backed by AWS's $50B OpenAI tie-up.

2026-05-03

Bedrock AgentCore launches in São Paulo region; CloudFront adds WebSocket VPC origins

Bedrock AgentCore — AWS's framework-agnostic platform to build, connect, and optimize agents — is now available in South America (São Paulo). CloudFront separately added WebSocket support through VPC origins for real-time apps in private subnets.

2026-05-03

AWS–OpenAI pact: GPT-5.5, Codex land on Bedrock with $100B commitment

AWS and OpenAI announced an expanded partnership making OpenAI's frontier models (including GPT-5.5 and 5.4), Codex, and Bedrock Managed Agents available on Amazon Bedrock in limited preview. OpenAI is expanding its existing $38B AWS commitment by $100B over eight years; Amazon plans to invest $50B in OpenAI. The deal lands a day after OpenAI ended Microsoft cloud exclusivity.

2026-05-01

Bedrock AgentCore reaches GA with managed memory, identity and observability

AWS made Bedrock AgentCore generally available, providing managed agent memory, OAuth-based identity, gateway and observability primitives. Pricing is consumption-based; integrates with Claude, Llama, Nova, and the newly added OpenAI models.

2026-05-01

AWS unveils agent-centric stack at What's Next 2026: Connect Decisions, QuickSight chat, HyperPod auto-topology

At What's Next with AWS 2026, Matt Garman highlighted agent-based AI for business operations. New releases include Amazon Connect Decisions for supply chain, QuickSight chat-based document and visual creation, SageMaker HyperPod automatic topology management, AgentCore CLI, Lambda S3 Files, and managed federated connectors for Athena.

2026-04-30

Amazon Quick desktop app — knowledge-graph 'always-on' assistant

AWS launched the Amazon Quick desktop app, a knowledge-graph-based assistant that stays connected to the user, accesses local files and apps, and learns from each session. AWS is positioning Quick as a transformative agentic platform.

2026-04-29

Bedrock AgentCore adds 3-API-call agent setup, VPC egress, and a managed harness

AWS shipped a managed agent harness in Amazon Bedrock AgentCore that lets developers stand up an autonomous agent in just three API calls, plus VPC egress support, an AgentCore CLI, and Claude Cowork inside Bedrock. SageMaker added unified-studio VPC support and automatic Slurm topology, S3 added five new checksum algorithms, and Connect added AI agent performance metrics. Accenture also committed to deploying Microsoft Copilot to all 743,000 employees.

2026-04-28

AgentCore goes GA — three API calls to deploy a production agent

AWS announced general availability of Bedrock AgentCore with managed runtime, memory, identity, and observability, and a new harness that cuts autonomous-agent deployment to three API calls. The release adds a CLI, persistent agent filesystem, and prebuilt skills, with consumption-based pricing and no minimums. The managed harness preview is live in initial regions.

2026-04-27

AWS Transform custom adds Learn-Scale-Improve flywheel for enterprise modernization

AWS extended Transform custom to handle modernization at the 50–200+ repository scale, where coordinating people and capturing learnings matters as much as code transformation. The Learn-Scale-Improve flywheel formalizes feedback loops to industrialize repo-by-repo migrations across an enterprise.

2026-04-27

Bedrock AgentCore: 3-API-call agent setup, VPC egress, plus Athena and SageMaker shipments

AWS introduced a managed agent harness in Bedrock AgentCore that cuts autonomous agent deployment to just 3 API calls, added secure VPC egress to AgentCore Gateway and Identity, and shipped Athena managed federated connectors, S3 checksum algorithms, SageMaker Unified Studio VPC support, and AI agent metrics in Connect.

2026-04-26

Amazon Connect adds eight new AI agent performance metrics including goal success rate and tool selection accuracy

Amazon Connect launched eight new metrics for measuring AI agent performance in customer interactions, including goal success rate, faithfulness score, and tool selection accuracy. The metrics enable enterprises to measure and continuously improve AI-driven customer interaction quality, addressing growing enterprise demand for AI agent observability.

2026-04-25

Meta signs deal for millions of AWS Graviton CPUs to power agentic AI

Meta agreed to use AWS Graviton CPUs — not GPUs — for its agentic AI workloads, signaling that real-time reasoning agents need fundamentally different infrastructure than training. The deal kicks off a new chip race centered on CPU inference.

2026-04-24

Bedrock AgentCore streamlines agent building in three decisions

AWS announced new Bedrock AgentCore features letting builders declare and run a working AI agent in just three decisions. Claude Cowork and Claude Code Desktop now run natively on Bedrock.

2026-04-24

AWS Expands Agentic AI Focus with Claude Opus 4.7 and Inline Agents

Amazon Web Services is heavily investing in agentic AI workflows, supporting Anthropic's Claude Opus 4.7 on Amazon Bedrock to automate complex business processes. AWS is promoting multi-agent workflows and expanding partnerships to help customers implement generative AI solutions beyond pilots, using supervisor agents to break down large tasks.

2026-04-23

AWS Launches 46 Major Service Updates Including Amazon Bio Discovery and Enhanced Bedrock

AWS announced 46 significant updates spanning compute, containers, data, analytics, and developer tools, including the launch of Amazon Bio Discovery, an AI-powered drug discovery application. Additional updates include EC2 X8i expansion to Europe, EKS auto-scaling improvements, S3 Express One Zone inventory support, and AWS IoT Greengrass v2.17 enhancements.

2026-04-22

Jeff Bezos' Project Prometheus AI Lab Nears $38B Valuation in $10B Funding Round

Jeff Bezos' secretive AI startup Project Prometheus is close to raising $10 billion in funding at a $38 billion valuation, according to Financial Times reports. The lab is focused on developing advanced AI models capable of understanding the physical world, representing Bezos' major bet on next-generation artificial intelligence beyond his Amazon and space ventures.

2026-04-22

Amazon commits $200B in 2026 AI capex and expands OpenAI partnership to $100B exclusive deal

AWS announced $200 billion in 2026 capital expenditure focused on AI infrastructure buildout. The company expanded its OpenAI relationship to a $100 billion, eight-year agreement making AWS the exclusive third-party cloud distribution provider for OpenAI Frontier. Custom Trainium and Graviton chips surpassed $10 billion annual run rate.

2026-04-21

Amazon Expands OpenAI Partnership to $100B+ and Commits Record $200B 2026 Capex

AWS deepened its strategic partnership with OpenAI to over $100 billion over eight years, with AWS serving as the exclusive third-party cloud distribution provider for OpenAI Frontier models. Amazon announced a record $200 billion capital expenditure plan for 2026, primarily directed at AWS AI infrastructure to support enterprise demand.

2026-04-19

Amazon Launches AI Bio Platform for Drug Discovery and Antibody Modeling

Amazon introduced AI Bio Discovery program providing cloud-based AI systems for creating digital antibody models and accelerating drug discovery workflows. The service aims to augment rather than replace scientists and contract research organizations by broadening access to computational biology tools.

2026-04-19

AWS introduces granular cost attribution for Bedrock and Model Distillation achieving 95% cost reduction

AWS launched granular cost attribution for Amazon Bedrock enabling detailed tracking of inference costs by use case, alongside Model Distillation capabilities that achieve 95% cost reduction and 50% latency reduction while maintaining quality. These features target enterprises seeking to optimize large-scale generative AI deployments.

2026-04-18

AWS launches Nova Multimodal Embeddings for video semantic search applications

AWS released Nova Multimodal Embeddings enabling developers to build video semantic search solutions that understand user intent and retrieve accurate results across multiple signal types simultaneously. The solution is available on Amazon Bedrock with reference implementations for rapid deployment with custom content.

2026-04-18

AWS Launches Automated Reasoning Checks for Formal Verification in Bedrock

Amazon Bedrock introduced Automated Reasoning checks using formal verification to provide mathematically proven AI validation for regulated industries. The tool addresses compliance gaps where probabilistic validation falls short, enabling auditable outputs across finance, healthcare, and other regulated sectors requiring formal proof of AI decision-making correctness.

2026-04-17

Spring AI SDK for Amazon Bedrock AgentCore Now Generally Available for Java Developers

AWS announced the general availability of the Spring AI AgentCore SDK, an open-source library that integrates Amazon Bedrock AgentCore capabilities into Spring AI applications. This SDK allows Java developers to build and deploy production-ready AI agents at scale using familiar Spring patterns. It significantly reduces infrastructure work, handling scalability, reliability, security, and memory management for complex multi-step AI tasks.

2026-04-16

AWS Releases Bedrock AgentCore Browser and Spring AI SDK for Production AI Agents

AWS launched Bedrock AgentCore Browser enabling AI agents to interact at the OS level for autonomous workflow automation, alongside the generally available Spring AI SDK for Amazon Bedrock AgentCore. The platform also introduced SageMaker HyperPod Gang Scheduling for coordinated distributed training jobs and Amazon WorkSpaces Advisor for AI-powered troubleshooting.

2026-04-15

Amazon Bedrock Adds GPT-4.1 and Llama 4 Models with Inline Agents Feature

AWS updated Amazon Bedrock to include GPT-4.1 via a cross-cloud model access preview and Llama 4 Scout and Maverick as fully managed models. A new Inline Agents capability allows developers to define agent instructions and tools dynamically at runtime without pre-configuring agents in the console. The update also expanded Bedrock Guardrails with contextual grounding checks for RAG pipelines and features Claude Mythos Preview as a gated research preview.

2026-04-14

Amazon Reports $15B AI Revenue Run-Rate as CEO Jassy Outlines AI-First Company Transformation

Amazon CEO Andy Jassy revealed AWS is generating a $15 billion AI revenue run-rate, marking strong validation of its AI infrastructure strategy. In his third consecutive AI-focused shareholder letter, Jassy outlined plans to reinvent Amazon around AI, highlighted upcoming products including Amazon Leo internet access and Zoox autonomous ride-hailing, and emphasized AWS's position as an AI infrastructure leader. Amazon has also launched 'Project Houdini,' a modular data center initiative aimed at eliminating build delays to meet surging AI compute demand. AWS Bedrock expanded to nearly 100 supported models with multi-agent orchestration and cross-account guardrails.

2026-04-12

Amazon AI Revenue Exceeds $15B Run Rate as Jassy Defends $200B Capex with Customer Commitments

AWS CEO Andy Jassy disclosed that AWS AI revenue has exceeded a $15B annual run rate and is 'ascending rapidly' in his annual shareholder letter. He defended Amazon's $200B capital expenditure plan by noting substantial portions are backed by existing customer commitments, with monetization expected in 2027-2028. Jassy declared a 'new shift' away from Nvidia chips is underway, with Amazon's Trainium3 and Trainium4 chips nearly sold out. AWS also expanded Amazon Bedrock with multimodal retrieval capabilities for Knowledge Bases, enabling RAG applications to search across text, images, audio, and video, alongside stateful MCP client capabilities for interactive multi-turn agent workflows.

2026-04-11

AWS Bedrock Expands to ~100 Models with Cross-Account Guardrails and OpenAI-Compatible API Endpoints

Amazon Bedrock grew from ~60 to nearly 100 models spanning language, vision, audio, safety, and code workloads, including models from OpenAI, Google, and NVIDIA. The platform introduced Bedrock Guardrails Cross-Account Safeguards for centralized safety enforcement across AWS accounts, and added OpenAI-compatible API endpoints powered by Project Mantle, enabling developers to use existing OpenAI SDK code without modification against Bedrock's full model catalog. Additionally, Bedrock AgentCore Runtime now supports stateful Model Context Protocol (MCP) client capabilities, enabling interactive, multi-turn agent workflows with features like Elicitation, Sampling, and Progress notifications.

2026-04-10

Amazon Bedrock Expands to ~100 Models with Cross-Account Guardrails and OpenAI Partnership Growth

AWS announced the general availability of cross-account safeguards in Amazon Bedrock Guardrails, enabling a single management-account guardrail policy to automatically enforce safety controls across every model invocation organization-wide, reducing administrative burden while maintaining account-level flexibility. This follows a major 2026 model refresh that expanded Bedrock from approximately 60 to nearly 100 models spanning language, vision, audio, safety, and code workloads from vendors including Mistral, Google, NVIDIA, OpenAI, MiniMax, Moonshot, and Qwen. AWS also expanded its OpenAI partnership to a reported $138 billion, eight-year deal, with JP Morgan projecting AWS revenue growth of 29–30% across 2026.

2026-04-07

Iranian Missile Strikes Take Down AWS Data Centers in Bahrain and Dubai, Amazon Declares Hard-Down Status

Reports indicate Iranian missile strikes resulted in AWS data center outages across Bahrain and Dubai regions, with Amazon declaring hard-down status for multiple availability zones. The incident generated 149 Hacker News points and triggered serious conversation about infrastructure resilience and geopolitical risk management in cloud deployments. The event highlights the vulnerability of cloud infrastructure to regional military escalations and raises urgent questions for enterprises with workloads concentrated in Middle Eastern AWS regions.

2026-04-05

AWS Availability Zones in Bahrain and Dubai Go Offline Following Iran Military Strikes

AWS availability zones in Bahrain and Dubai experienced significant outages after Iran carried out military strikes in the region, leaving infrastructure 'hard down' according to reports. The incident underscores geopolitical risks to cloud infrastructure in politically volatile regions and highlights the critical importance of multi-region redundancy strategies for enterprise workloads. AWS had separately announced cross-account safeguards in Amazon Bedrock Guardrails (April 3) enabling centralized enforcement of safety controls across multiple AWS accounts, and has expanded Amazon Bedrock's model portfolio to nearly 100 models covering language, vision, audio, safety, and code workloads.

2026-04-04