Amazon CloudFront Origin mTLS for WebSocket & gRPC: CloudFront now supports mutual TLS authentication from edge to origins for HTTP, WebSocket, and gRPC protocols, enabling certificate-based trust with ALB, API Gateway, on-premises, and external origins. Introduced to replace less-secure shared secrets or IP allow-lists, this strengthens end-to-end encryption and zero-trust architectures while maintaining compliance with standards requiring strong client authentication. No additional charges apply. *Terraform Support*: Supported in Terraform: aws_cloudfront_distribution (via origin block with custom_origin_config and connection_attempts arguments) in v5.70.0+.
Amazon CloudFront Standard Logging with CMCD & Extended Fields: Enhanced standard logs now capture the handling CloudFront server IP, staging vs. primary distribution identifiers, media quality confidence scores, and comprehensive Common Media Client Data (CMCD) fields such as bitrate, buffer length, and session IDs. These enrich observability for troubleshooting, security forensics, and media workload optimization; logs are available in all edge locations to S3, CloudWatch Logs, or Firehose. *Terraform Support*: Supported in Terraform: aws_cloudfront_distribution (via logging_config block and related default_cache_behavior settings) in v5.80.0+.
Analytics
Amazon OpenSearch Service Memory Stores (GA): Fully managed memory stores for AI agents that capture interactions verbatim, derive structured memories asynchronously, and handle the complete lifecycle of extraction, embedding, indexing, consolidation, and retrieval. Supports keyword, semantic, and hybrid search with IAM-based access control (including actor/session scoping), configurable TTL, and auto-scaling. Now generally available in select regions; critical for secure, auditable AI memory with strong identity controls. *Terraform Support*: Pending provider release (use AWS CLI/SDK or custom provisioner in the interim).
AWS Glue ETL Apache Spark Optimization Agent: A new intelligent agent that analyzes completed Spark jobs and provides targeted cost or performance optimization recommendations within minutes, backed by deterministic reasoning, telemetry evidence, and best-practice checks (e.g., Graviton usage, Glue version upgrades). Recommendations span infrastructure, Spark config, code, and data layout while respecting trade-off limits. IAM-authenticated and CloudTrail-logged, available at no extra charge for Glue 4.0+. Security insight: ensures least-privilege job configurations and auditability of optimization actions. *Terraform Support*: Pending provider release (use AWS CLI/SDK or custom provisioner in the interim).
Machine Learning
Amazon SageMaker AI Cross-Region Inference Routing: Enables creation of inference routers that aggregate real-time, asynchronous, and component endpoints across regions. The router intelligently prioritizes local endpoints, fails over on error thresholds, and safely restores traffic, with CloudWatch metrics, response metadata, endpoint policies for cross-account access, and optional capacity caps. No extra charge; simplifies resilient, multi-region AI architectures while maintaining security boundaries. *Terraform Support*: Pending provider release (use AWS CLI/SDK or custom provisioner in the interim).
Database
Amazon RDS for MySQL 9.7: Introduces MySQL 9.7 with Replication Applier Metrics (Performance Schema tables for lag diagnosis), JSON Duality Views (bidirectional relational-to-JSON access), and the opt-in Hypergraph Optimizer for complex multi-join queries. These features improve observability, developer agility, and query performance for modern applications. Security note: continue leveraging RDS encryption, IAM auth, and audit logging. *Terraform Support*: Supported in Terraform: aws_db_instance (via engine_version argument) in v5.82.0+.
Amazon RDS for MySQL 8.4.13 Minor Version: Support for the latest community minor release 8.4.13, with seamless in-place upgrades or snapshot restore during maintenance windows. Can be orchestrated via AWS Organizations Upgrade Rollout Policies for phased, compliant rollouts across accounts. *Terraform Support*: Supported in Terraform: aws_db_instance (via engine_version argument) in v5.82.0+.
Compute
AWS Lambda GPU Instances on Managed Runtimes: Lambda now supports GPU-accelerated workloads (AI inference, media processing, 3D rendering, scientific computing) without managing drivers or scaling. Execution environments scale on GPU/CPU utilization; includes pre-built base images with vLLM, SGLang, and PyTorch, up to 32 GB NVMe storage, SOCI lazy loading, and integration with 200+ event sources. Uses EC2 G-series instances with On-Demand/Savings Plans pricing plus a management fee. Security insight: GPU functions inherit Lambda's strong IAM execution roles, VPC isolation, and ephemeral storage encryption. *Terraform Support*: Supported in Terraform: aws_lambda_function (via image_uri, ephemeral_storage, and compute configuration blocks) in v5.82.0+.
Security, Identity, & Compliance
Amazon Cognito Point-in-Time Recovery (PITR): Enables restoration of user pool data (users, groups, MFA settings, passkeys) to any point within the recovery window while preserving identifiers and credentials. PITR is auto-enabled on Essentials (1-day window) and Plus (14-day window) plans at no extra charge, with preview of change impact before execution. Enhances compliance, disaster recovery, and data protection posture for identity systems. *Terraform Support*: Pending provider release (use AWS CLI/SDK or custom provisioner in the interim).
Containers
Amazon EKS Hybrid Nodes Pricing Update: Billed vCPU per hybrid node is now capped at 32 vCPU for large servers, delivering up to 88% cost reduction on high-core-count on-premises or edge GPU hosts. Smaller nodes retain existing tiered pricing. Improves cost efficiency for security-sensitive or latency-critical workloads that must remain outside the cloud. *Terraform Support*: Supported in Terraform: aws_eks_node_group (via hybrid node configuration and IAM role settings) in v5.76.0+.