AWS AI Updates: August 6, 2026

1. Amazon DynamoDB Adds Native Real-Time Vector Search

AWS. Amazon DynamoDB is now generally available with native vector search, indexing and querying vector embeddings in real time at single-digit millisecond latency while maintaining 99%+ recall and scaling to trillions of vectors. The feature uses approximate nearest neighbor search with configurable vector index partition keys and attribute filtering, letting developers store embeddings generated by Amazon Bedrock models alongside standard item attributes for RAG, agent memory, and semantic retrieval without managing separate infrastructure. It is available in all commercial AWS Regions and AWS GovCloud (US) Regions under DynamoDB’s serverless, consumption-based model. Source

2. Aurora Serverless Scales Faster for Agentic AI Workloads

AWS. Amazon Aurora Serverless now reaches up to 12 Aurora Capacity Units within a second during scale-up and continues expanding to 256 ACUs as demand grows, then scales back to zero when idle, targeting the bursty, unpredictable traffic patterns of AI agent applications. The faster initial ramp is enabled by default on Aurora Serverless v2 platform versions 3 and 4 with no configuration required, and clusters on versions 1-2 can upgrade to version 4. Customers can verify their platform version through the RDS console or the ServerlessV2PlatformVersion API parameter. Source

3. AWS Marketplace Adds AI Insights for Pricing Explanations

AWS. AWS Marketplace introduced AI Insights, which generates plain-language explanations of how a product’s pricing works directly on listing pages, covering what a pricing unit maps to, how bills change as usage scales, how multiple pricing dimensions combine, and what is and is not included. The system draws on seller-published pricing data plus additional context from sellers’ public websites and cites its sources so buyers can trace where each explanation originates. It is live on most listings with available external pricing context across all commercial AWS Regions where Marketplace operates, and sellers can review and request edits to explanations on their own listings. Source

4. Amazon EMR on EC2 Adds Spark Connect for Interactive Development

AWS. Amazon EMR on EC2 now supports interactive Apache Spark sessions through Spark Connect, whose client-server architecture decouples the application client from the Spark driver and maintains a persistent Spark context that spans across notebook cells and scripts. Data engineers and scientists can develop and debug from managed notebooks in Amazon SageMaker Unified Studio or their own IDEs such as Jupyter and VS Code, blending local Python with remote Spark operations while monitoring via Spark UI and the History Server. The capability requires the AWS runtime for Apache Spark (emr-spark-8.0) or later and is available in all AWS Regions with EMR support, excluding GovCloud and China. Source