On September 17, 2026, AWS made Elastic Beanstalk’s Cluster mode generally available, running multiple applications on EKS infrastructure fully operated by the service. For a regulated organization managing a portfolio of legacy applications, it is the promise of a single operational baseline without rewriting each workload.

On October 27, 2026, AWS raised Amazon Aurora serverless to a 384 ACU ceiling for PostgreSQL — 768 GiB of memory and 50 percent more capacity — while keeping scale-down to zero. For the bursty workloads of AI agents, this is the clearest signal yet that AWS wants the serverless database to be the default backplane, provided you treat the ACU ceiling as a real capacity decision.

On October 24, 2026, Anthropic’s draft S-1 — reviewed by The Wall Street Journal and Reuters — details $16.3 billion in first-half 2026 revenue, an $80.7 billion net loss, and roughly $450 billion in future cloud commitments. The numbers describe less a company than a compute supply chain locked in a decade ahead.

On October 21, 2026, AWS launched On-Demand Capacity Reservations for Amazon Aurora and RDS: reserve compute capacity in one or more Availability Zones, immediately or scheduled 5 to 120 days ahead. It is insurance against “insufficient capacity” errors, not a cost cut — reserved capacity is billed even when no instance is running.

On October 22, 2026, AWS made the SageMaker AI Training Session APIs generally available: you orchestrate forward_backward(), optim_step(), and sample() on stateful, serverless GPU sessions, while the service handles hardware, gradient synchronization, and isolation. If you do SFT or DPO, the high-level APIs still suffice; if you do custom RL, this removes the GPU-ops tax.

On October 20, 2026, AWS launched ECS Gateway, an account-level resource that provisions and manages Application Load Balancers, ACM certificates, WAF rules and security groups for your ECS services. A platform team defines the standard once, an application team connects a service with one line: the per-service ingress bottleneck is gone.

On October 14, 2026, AWS made S3 — including S3 Tables and S3 Vectors — provisionable from the Vercel dashboard and v0, using an OIDC token exchange that removes static access keys from the path. For front-end teams that consume object storage without ever opening the AWS console, the days of an IAM key pasted into an .env file are over.

On 16 October 2026, CloudWatch log centralization expanded to both GovCloud regions and 13 more commercial regions, bringing the covered total to fifteen. For organizations already on AWS Organizations, that is the signal to weigh your homegrown log pipeline against a native feature priced at $0.05 per gigabyte beyond the first copy.

On 16 October 2026, Amazon Braket, AWS’s quantum service, entered the FedRAMP Class C scope in the US regions where it is available. Federal agencies can now explore quantum within an authorized boundary, but the computations themselves run on third-party quantum hardware providers, outside the certified boundary.

On October 16, 2026, AWS made Azure SDK to AWS SDK code conversion generally available in AWS Transform: Azure Functions becomes Lambda, Blob Storage becomes S3, and Cosmos DB becomes DynamoDB in preview. For a .NET team weighing an exit from Azure, it is a serious starting point — not a one-click migration.

On October 8, 2026, AWS extended GuardDuty RDS Protection beyond login anomalies to detect data exfiltration and destruction in Aurora PostgreSQL and RDS for PostgreSQL, using agentless behavioral baselines. Turn on the 30-day free trial if you run PostgreSQL — but know that MySQL remains outside the scope.

On 12 October 2026, AWS added a third form factor to Lambda — Web Functions — that runs an Express, Hono or Next.js server as-is and bills for the CPU it actually uses. Try it for I/O-heavy APIs and server-side rendering; keep Event Functions for pure event processing.

On 9 October 2026, AWS detailed Amazon ECS’s automatic repair mechanisms: GPU failure detection via NVIDIA DCGM, agent-connectivity repair and Availability Zone rebalancing. If you run GPU workloads on ECS, you can stop building your own detection loops: these repairs are on by default at no extra cost.

In late September 2026, AWS launched Amazon Bedrock Managed Agents powered by OpenAI in public preview — an AWS-native version of OpenAI’s Agents API that runs agents inside your account with your identities and guardrails. If you want OpenAI agents without letting your data leave your AWS perimeter, this is the shortest path.

As of October 5, 2026, Redshift can write the results of its materialized views as Apache Iceberg tables in S3, queryable by Athena, Spark, Glue, and SageMaker with no copy in between. For a team that wants its precomputed aggregates open to every engine instead of locked inside Redshift, this is a concrete pivot.

On 2 October 2026 AWS Health launched a version catalog that gives a centralised view of the software-version lifecycles across AWS services, starting with RDS, EKS and Lambda. Teams on Business Support Plus, Enterprise Support or Unified Operations can query it through an API to build upgrade plans before end of support, instead of absorbing lifecycle events one by one.

Six months after Iranian drone strikes hit its me-central-1 (UAE) and me-south-1 (Bahrain) data centers, AWS confirms that some data hosted exclusively in those zones is unrecoverable — a first for a global cloud provider. Audit your data placement now and replicate any critical workload out of the region.

Amazon S3 Vectors now evaluates the metadata filter before the similarity search, returning up to 5 times more relevant vectors on highly selective filters. Enabling it is free, in place and re-ingestion-free: switch your indexes to ENHANCED mode.

On 2 October 2026, AWS shipped four fixes for Loom, its open-source AI agent orchestration platform, and for SageMaker Unified Studio. The most severe, CVE-2026-103956, gives any network client full administrative control of the agent control plane when no identity provider is configured — a lesson that applies to every MCP or A2A deployment.

AWS embeds DuckDB inside Aurora PostgreSQL to query operational data and the data lake in a single query, with no ETL pipeline. A turning point for teams that replicated the lake into Postgres.

AWS launches a preview AI agent that analyzes an environment against 65+ services and produces dollar-quantified recommendations, with Terraform or CloudFormation fixes ready to apply. Read it like a junior architect, not an oracle.

The SageMaker HyperPod Inference Gateway replaces round-robin with real-time signal-driven routing, cutting first-token latency by up to 82% with no application code changes. Deploy it as an EKS add-on if you serve multiple models or a heterogeneous GPU fleet behind a single endpoint.

On September 15, 2026, AWS and Salesforce expanded their partnership to push CRM data and AI agents into the tools teams already use every day: Amazon Quick, Slack, and voice. Access is zero-copy — no migration, no duplication — with model choice through Amazon Bedrock.

On September 24, 2026, AWS announced an enhanced custom event bus in Amazon EventBridge: one centralized bus shared across every account in an AWS organization, with ordering guarantees, a Subscriber resource, and a new ingress/egress pricing model. For a multi-account platform still juggling cross-account rules, the workaround is officially over.

On September 22, 2026, Amazon CloudWatch launched Omni, a unified observability experience for applications and AI agents, delivered inside VS Code, Kiro, and a standalone web console. Adopt it to trace, compare, and evaluate your agents before a prompt regression silently degrades production responses.

Amazon EC2 T8i instances, powered by custom sixth-generation Intel Xeon Scalable (Granite Rapids) processors and the Nitro System, became generally available on September 17, 2026 with up to 30% better price performance than T3. Low-CPU workloads should move from T3 to T8i without changing their credit model.

On September 21, 2026 AWS released Strands Harness, an open-source AI agent that bundles tools, context, and memory, and claims a 45% cost saving over Claude Code and Codex. Teams that pay for tokens by volume can evaluate it without depending on AWS infrastructure.

On September 15, 2026 AWS acknowledged it cannot restore access to data hosted in availability zones hit in Bahrain and the United Arab Emirates. Replicate across regions and reread what your durability commitments actually guarantee.

On August 31, 2026, AWS launched streaming tables for Kinesis Data Streams, turning a stream into a queryable Apache Iceberg table with no pipeline to operate. The vendor claims up to 50% delivery savings and 30% query savings — at the cost of delegating compaction.

On September 10, AWS open-sourced Pizza Bot, an app that replaces chat with an email-style inbox for tracking AI agents working in the background. For any team running autonomous agents in the cloud, the thesis fits in one sentence: the interface must assume no one is watching.

AWS extends Amazon EBS Volume Clones to copy volumes across accounts, with optional re-encryption in the destination account. Multi-account teams can now refresh test environments with current production data, provided they work within the encryption and Availability Zone constraints.

On September 9, 2026, AWS made the Amazon Quick desktop app generally available on macOS and Windows and added a mobile activity feed that consolidates email, calendar, CRM, and messaging. For an organization already on AWS, it is a lever to bring generative AI back under governance: CloudTrail audit, HIPAA, FedRAMP, SOC 2 and ISO 27001 compliance, without moving data out of your environment.

On September 8, 2026, Amazon Web Services announced a partnership with Qualcomm for custom AI inference silicon and 1.6 Tbps optical interconnect gear, under a commercial commitment that could reach $60 billion through September 2036. A strong signal about the diversification of AWS’s silicon supply chain — and about the real bottleneck of AI clusters: the network.

Aurora MySQL 8.4.8, available in early September 2026, adds delayed replication: a replica deliberately applies each change with a configurable lag, making it the only copy that does not reproduce an accidental DROP TABLE or DELETE. Configure it by stored procedure and write the recovery runbook before you need it.

On August 31, 2026, AWS Interconnect multicloud opened to Azure in public preview: a VPC and a VNet linked over AWS’s private backbone, no third-party carrier, but capped at 1 Gbps with no availability commitment. Test it for shared AWS-Azure workloads, but keep your existing circuits until general availability.

On September 3, 2026, AWS opened programmatic management of CloudFront flat-rate plans through the new PricingPlanManager API, the CLI, CloudFormation and the CDK. Teams can now codify subscribing, changing tiers and cancelling a no-overage monthly price, with a two-phase approval that prevents unintended billing.

Announced at LEAP in Riyadh, AWS’s first infrastructure region in Saudi Arabia will go live in December 2026, bringing the global network to 40 regions on a planned investment of over $5.3 billion. For teams serving the Gulf, it is the long-awaited answer to data-residency requirements, doubled with a 50 MW AI Zone planned for 2028.

On September 1, 2026, AWS opened GuardDuty to custom detection rules, associable per account and controllable at the organization level with a dry-run mode and an expiry date. Security teams can now code their own detection scenarios instead of waiting for vendor findings.

Announced on August 26, 2026 and confirmed in the August 31 AWS weekly roundup, the acquisition of DuckLabs by AWS brings the DuckDB maker in as a subsidiary, without touching the MIT license or the DuckDB Foundation governance. For data teams, DuckDB remains a safe bet; the open question is whether the server-side roadmap tilts toward AWS.

On August 28 2026 AWS made EC2 C8gn instances, powered by Graviton4 and capable of 600 Gbps of network bandwidth, available in the Europe (Paris) region. For teams running virtual network appliances or high-throughput analytics, this European expansion reshapes the cost-performance case for arm64.

On 26 August 2026, AWS and NVIDIA announced two million additional GPUs to be deployed by 2028, NVIDIA Vera CPUs, NVHBM memory, and 100,000-GPU federal AI factories. This is no longer a silicon supply deal: AWS is now co-engineering interconnect and memory with NVIDIA. Here is what it changes for a cloud engineer.

On August 19, 2026, AWS added an external_web_access parameter to Web Search on Amazon Bedrock, letting agents pull content live from the public web. The lock is an IAM permission, and the default is open.

AWS announced general availability of a Las Vegas Local Zone (us-west-2-las-2a) on August 20, 2026, with EC2 C7i/M7i/R7i/C8gn, ECS, EKS, the Application Load Balancer, and Direct Connect. Teams under latency or data-residency pressure should weigh it against a full region before committing.

On August 14, 2026, a RIPE NCC member published a field report on an IPv6-first EKS deployment on AWS: roughly $143 a month per environment saved, against a dated list of dependencies still stuck on IPv4. Teams standing up a new cluster now have an objective criterion for choosing IPv6-first over dual-stack.

On August 20, 2026, Amazon EKS announced certificate authority rotation with a managed lifecycle and automated safeguards for every cluster. Clusters created since 2018 are approaching the ten-year validity limit of their CA: rotation is a shared responsibility, and the worker-node and external-client side remains the operator’s job.

On August 19, 2026, AWS added a fourth Availability Zone (eu-west-2d) to the Europe (London) Region, carrying Trn3 and P6 capacity for training and inference. For architects it is both a four-zone resilience win and a clear signal: cloud expansion now runs on AI silicon.

Between May and August 2026, AWS suffered four notable reliability incidents, two of them on the same network path linking US-West-2 to the Seattle metro area — with the company still declining to confirm a shared root cause. Teams single-homed in us-west-2 need to audit their single points of failure before next quarter.

Amazon EC2 introduces application status checks that probe your HTTP or HTTPS endpoints every 60 seconds and flag an impaired application even when the instance itself is healthy. Wired into Auto Scaling, they trigger automatic instance replacement the moment the application stops responding.

AWS has announced general availability of native vector search in DynamoDB, where embeddings live alongside operational data and a SearchVectors API returns nearest neighbors. If your data already lives in DynamoDB, this is the end of the sync pipeline to a dedicated vector store.

On August 14, 2026, AWS added preconfigured, read-only managed dashboards to Billing and Cost Management, at no extra cost. It is the baseline FinOps visibility the vendor is finally offering natively — and direct pressure on third-party tools whose entry price starts at 5% of spend.

On August 13, 2026 AWS shipped version 6.0 of its VPN client, rebuilt on OpenVPN3 with a full CLI and centralized administrative controls. Teams automating remote access can now script connections without third-party tooling.

Generally available since August 12, 2026, Role Manager automatically creates or reuses the IAM roles AWS services need, from AWS Lambda to Amazon EventBridge. The time savings are real, but a default role is not a least-privilege role — here is how to use it without eroding your least-privilege posture.

On August 12, 2026, AWS and Google Cloud unveiled an open-source multicloud interoperability framework that eliminates egress fees and standardizes identity across both platforms. Azure will join before the end of the year. For CIOs, this marks the end of forced infrastructure duplication — and the beginning of genuinely agnostic cloud architecture.

A threat actor compromised a European Commission AWS account in early August 2026 and exfiltrated over 350 GB of data, including databases and an internal email server. The incident is a reminder that the weakest link in cloud security is not the provider's infrastructure — it's client-side identity and access management.

Announced at AWS Summit New York on August 7, 2026, AgentCore Runtime Instances bring persistent, stateful compute to Bedrock agents, removing the cold start penalty that plagued real-time deployments. If your AI agents take more than three seconds to respond, the bottleneck is your infrastructure — and AWS just fixed it.

AWS announced general availability of vector search in DynamoDB on July 30, 2026. If your operational data already lives in DynamoDB, you can drop your dedicated vector database — and the synchronization pipeline that comes with it.

On July 31, 2026, AWS announced general availability of AWS Interconnect for Oracle Cloud Infrastructure (OCI). For the first time, two competing hyperscalers offer native private interconnection without traversing the public internet. A pivotal shift for multicloud architectures.

Idle resources, overprovisioned instances, orphaned volumes, and missing Reserved Instances: cloud waste has known, quantified causes — and they’re all fixable within a month. FinOps isn’t a CFO buzzword — it’s the only IT practice that generates real, measurable cash without touching revenue.

In July 2026, an 8 vCPU/16 GB server costs €15.99/month on Hetzner versus $248/month on AWS on-demand. The gap isn’t a promotion — it’s a cost structure hyperscalers cannot replicate.