Amazon ECR now lets you push container images with layers up to 200 GB via Docker, making it easier to bundle large models, datasets, and dependencies without splitting across multiple layers. This eliminates the hassle of managing complex storage workarounds for demanding workloads like LLMs and genomics data. Note that AWS SDK and CLI uploads stay capped at 50 GB. The feature rolls out across all AWS regions except Middle East locations.
AWS just released Context Ontology Accelerator, an open source tool that helps you build machine-readable models of your business to make AI agents more reliable and explainable. It connects to your data sources, uses AI to draft an ontology, and lets your experts review it before storing in a knowledge graph. This cuts months of manual work down to days, making it easier to move AI agents from testing to production with full auditability.
AWS now offers 1 and 3-year Reserved Instances for R8i and M8i Oracle databases, saving you up to 53% on costs. These instances pack custom Intel Xeon 6 processors with 15% better price-performance and 2.5x faster memory bandwidth than previous generations. You get flexibility to switch between Multi-AZ and Single-AZ setups, plus size flexibility under BYOL licensing. Start purchasing through AWS Console, CLI, or SDK today.
Amazon CloudWatch now offers fully managed Prometheus collectors, eliminating the need to deploy and maintain your own agents. Simply provide a scrape configuration, and CloudWatch automatically handles provisioning and scaling across EKS, EC2, ECS, MSK, and OpenSearch workloads. Metrics arrive in OpenTelemetry format and integrate seamlessly with PromQL queries for unified dashboards and alarms.
Great news for developers! Amazon Bedrock is cutting prices for OpenAI's GPT-5.6 models starting July 30, 2026. GPT-5.6 Luna gets an 80% discount, perfect for high-volume content processing and automation, while GPT-5.6 Terra drops 20% for balanced production workloads. Best part: pricing updates automatically with zero effort needed.
Amazon MSK Express brokers can now deliver Apache Kafka data straight to S3 buckets with zero manual connector management. This fully managed feature automatically scales to 10 GB/s throughput, handles retries and backpressure on its own, and cuts costs by up to 60% compared to self-managed solutions. Perfect for log archival, compliance, and ML training pipelines.
AWS has upgraded IAM Policy Simulator with a major refresh—it's now built right into the IAM console instead of running separately. The tool can now test service control policies (SCPs) and lets you simulate "what if" scenarios by excluding specific policies. You'll get better insights into cross-account permissions and how conditions like regions and tags affect access decisions.