← back to blog
cloudFeatured·Mar 2, 2026·11 min read

What's New with AWS — February 2026: The Month in Review

A themed roundup of every headline launch AWS shipped in February 2026 — new EC2 families, database and security defaults, a wave of Bedrock/agentic AI releases, and a landmark Amazon–OpenAI partnership. Every item links back to the original announcement.

SJ
Sabin Joshi
DevOps Engineer
#aws#whats-new#bedrock#ec2#databases#security#roundup

February 2026 was a busy month across AWS — a steady stream of new compute instances, database and security enhancements, and (no surprise) a wave of AI and agentic-development launches. It closed with one of the biggest pieces of industry news of the year: a multi-year strategic partnership between Amazon and OpenAI.

Below is a roundup of the month's most notable launches, grouped by theme. Every item links straight to its original announcement.

ℹ️
A note on scope. AWS pushes far more to its What's New feed than any single post can capture — often dozens of items a week, many of them regional expansions or incremental availability updates. This covers the headline, developer-relevant launches AWS itself highlighted in its weekly roundups for the month. For the exhaustive, item-by-item list, the canonical source is the What's New with AWS page.
5
new EC2 families
8+
AI / Bedrock launches
1,160+
EC2 instance types now
$50B
Amazon → OpenAI

Artificial Intelligence & Agentic Development

AI was, once again, the busiest category of the month.

Claude Opus 4.6 in Amazon Bedrock bedrock

Anthropic's most capable model at the time landed in Bedrock, targeting coding, enterprise agents, and demanding professional workloads that need deep reasoning and reliability.

Claude Sonnet 4.6 in Amazon Bedrock bedrock

A lower-cost sibling that approaches Opus 4.6-level intelligence, tuned for high-volume coding and knowledge work where throughput matters.

Structured outputs in Amazon Bedrock bedrock

Bedrock can now force model responses to conform to a defined JSON schema — no more prompting for valid JSON and bolting on extra validation in application code.

Six new open-weights models in Amazon Bedrock bedrock

Fully managed support for DeepSeek V3.2, MiniMax M2.1, GLM 4.7, GLM 4.7 Flash, Kimi K2.5, and Qwen3 Coder Next — spanning frontier reasoning and agentic-coding workloads, powered by Project Mantle with out-of-the-box OpenAI API compatibility.

Bedrock PrivateLink for the bedrock-mantle endpoint bedrock

AWS PrivateLink support extended to the OpenAI-API-compatible Mantle endpoint (in addition to bedrock-runtime), across 14 AWS Regions.

Amazon SageMaker Inference for custom Amazon Nova models sagemaker

Tune instance types, auto-scaling policies, and concurrency settings when deploying customized Nova models.

Kiro in AWS GovCloud (US) Regions kiro

AWS's agentic AI development tool arrived in GovCloud, bringing spec-driven development to teams working under stricter government security controls.

Agent Plugins for AWS dev-tools

A set of open-source plugins that extend coding agents with AWS deployment skills. The deploy-on-aws plugin generates architecture recommendations, cost estimates, and infrastructure-as-code straight from your coding agent.

Amazon Location Service adds "LLM Context" location

Curated AI-agent context for location features shipped as a Kiro power, a Claude Code plugin, and an agent skill in the open Agent Skills format.

Compute & Networking

AWS noted it now offers over 1,160 EC2 instance types — and February added five new families alone.

Amazon EC2 C8id, M8id, and R8id instances ec2

Powered by custom Intel Xeon 6 processors — up to 43% higher performance and 3.3× more memory bandwidth than the previous generation.

Amazon EC2 M8azn instances (GA) ec2

General-purpose, high-frequency instances on 5th-gen AMD EPYC with the highest maximum CPU frequency in the cloud at 5 GHz — up to 2× the compute, 4.3× the memory bandwidth, and a 10× larger L3 cache versus M5zn.

Amazon EC2 Hpc8a instances ec2

Also on 5th-gen AMD EPYC — up to 40% higher performance and 300 Gbps Elastic Fabric Adapter networking for tightly coupled HPC and engineering simulation workloads.

Nested virtualization on virtual EC2 instances ec2

Run KVM or Hyper-V inside virtual EC2 instances — useful for mobile-app emulators, in-vehicle hardware simulation, and Windows Subsystem for Linux on Windows workstations.

Amazon ECS adds NLB support for linear & canary deployments ecs

Managed, incremental traffic shifting now works natively from ECS for Network Load Balancer workloads (TCP/UDP, low-latency, static-IP use cases).

Amazon EKS Auto Mode enhanced logging eks

Auto Mode capabilities (autoscaling, storage, load balancing, pod networking) can now be configured as CloudWatch Vended Logs delivery sources at a reduced price, delivering to CloudWatch Logs, S3, or Data Firehose.

Amazon EKS Node Monitoring Agent goes open source eks

The agent is now on GitHub for visibility, customization, and community contributions.

AWS Network Firewall price reductions networking

Added hourly and data-processing discounts for NAT Gateways service-chained with Network Firewall secondary endpoints, and removed extra data-processing charges for Advanced (TLS) Inspection.

Amazon CloudFront mutual TLS (mTLS) for origins cloudfront

Certificate-based authentication so only verified CloudFront distributions can connect to backend origins — AWS, on-premises, third-party cloud, or external CDNs.

Databases

Amazon DynamoDB global tables — cross-account replication dynamodb

Global tables can now replicate across AWS accounts (as well as Regions), improving resiliency, workload isolation, and account-level security/governance.

Amazon RDS enhanced console connect experience rds

Ready-made, auth-aware code snippets (Java, Python, Node.js, psql, and more) plus integrated CloudShell access directly from the RDS console.

RDS/Aurora backup configuration on snapshot restore rds

View and set the backup retention period and preferred backup window before and during a restore, instead of only after it completes. Available across all RDS engines and Aurora editions, at no extra cost.

Server-side encryption by default in Amazon Aurora aurora

All new Aurora clusters are now automatically encrypted at rest using AWS-owned keys — fully managed, transparent, and with no cost or performance impact.

Amazon OpenSearch Serverless Collection Groups opensearch

Share OpenSearch Compute Units across collections that use different KMS keys — reducing OCU cost while preserving collection-level security and adding guaranteed minimum-OCU capacity.

Security, Identity & Management

AWS Builder ID now supports Sign in with Apple identity

Adds Apple as a social login option (alongside Google) for AWS Builder Center, Training & Certification, re:Post, Startups, and Kiro.

AWS STS validation of identity-provider-specific claims iam

Validate select custom claims from Google, GitHub, CircleCI, and OCI as condition keys in IAM role trust policies and resource control policies, for finer-grained federated access control.

AWS Management Console shows Account Name in the nav bar console

A simple but welcome quality-of-life change that makes it far easier to tell accounts apart at a glance.

AWS Config adds 30 new resource types config

Broader coverage across services including Amazon EKS, Amazon Q, and AWS IoT for discovery, assessment, audit, and remediation.

AWS Security Hub Extended security-hub

A new plan that simplifies procurement, deployment, and integration of full-stack partner security solutions (CrowdStrike, Okta, Splunk, Zscaler, SailPoint, Proofpoint, and more) with AWS as seller of record, pay-as-you-go pricing, a single bill, and unified operations in Security Hub.

Media, Developer Tools & Operations

AWS Elemental Inference media

A fully managed AI service that transforms live and on-demand video for mobile and social platforms in real time — AI-powered vertical cropping for TikTok, Instagram Reels, and YouTube Shorts, plus automatic highlight extraction at 6–10 second latency. Fox Sports was an early adopter.

AWS Elemental MediaConvert video Probe API media

A new free API that reads media header metadata to return codec, pixel-format, and color-space details without processing the video.

AWS AppConfig integrates with New Relic appconfig

Integration with New Relic Workflow Automation enables automated, intelligent rollbacks during feature-flag deployments, cutting detection-to-remediation time from minutes to seconds.

Strands Labs open-source

A new open-source Git organization for experimental agentic-AI projects, launching with three projects: Robots, Robots Sim, and AI Functions.

The Big One: Amazon & OpenAI Announce a Strategic Partnership

Late in the month, Amazon and OpenAI announced a multi-year strategic partnership to accelerate AI innovation. Amazon committed to invest $50 billion in OpenAI — an initial $15 billion, followed by a further $35 billion once certain conditions are met.

The two companies are co-creating a Stateful Runtime Environment powered by OpenAI models and delivered through Amazon Bedrock, and AWS will act as the exclusive third-party cloud distribution provider for OpenAI Frontier. They also expanded an existing $38 billion multi-year compute agreement by a further $100 billion over eight years, with OpenAI committing to roughly 2 gigawatts of AWS Trainium capacity spanning Trainium3 and next-generation Trainium4 chips.

⚠️
The partnership was announced in the final week of February. A few closely related Bedrock features — such as the OpenAI-compatible Projects API — followed in early March, so they sit just outside this month's window.

The Month in a Paragraph

February 2026 was one of AWS's busiest months in a while, and it ended on the biggest note: a multi-year strategic partnership between Amazon and OpenAI — a landmark deal that will shape a lot of what comes next. Around that headline, the month leaned hard into AI everywhere, with Claude Opus 4.6 and Sonnet 4.6 landing in Amazon Bedrock, six new open-weights models, custom Amazon Nova inference on SageMaker, and agentic tooling via Kiro (now in GovCloud) and Agent Plugins. On the infrastructure side came more and faster compute — three new Intel-based EC2 families, two new AMD EPYC families, and nested virtualization — alongside a clear push toward security and resilience by default, including Aurora encryption-at-rest, DynamoDB cross-account replication, and Security Hub Extended. A dense, AI-heavy month that set the tone for the rest of the year.