Research Finder
Find by Keyword
AWS Rebuilds OpenSearch Serverless for the Agentic Era
A proprietary shared storage layer, default GPU acceleration, and a co-developed integration with Anthropic position OpenSearch Serverless as retrieval infrastructure for enterprise AI workloads.
5/29/2026
Key Highlights
- Amazon OpenSearch Serverless introduces a rebuilt architecture with a proprietary shared storage layer that decouples compute from storage, improving elasticity and enabling consumption-based pricing.
- GPU acceleration is now enabled automatically for large-scale vector indexing workloads, delivering up to 10x faster indexing performance at approximately one quarter of previous cost levels.
- OpenSearch Agent Skills for Claude, co-developed and co-announced with Anthropic, integrates Claude Code as a natural language interface for OpenSearch workflows.
- AWS redesigned OpenSearch Serverless around the bursty, unpredictable demand patterns that agentic AI workloads produce, with a 5-second cold start including networking and security provisioning.
The News
AWS announced a rebuilt Amazon OpenSearch Serverless architecture designed around retrieval-intensive AI workloads and changing demand behavior. The announcement introduces a proprietary shared storage architecture, default GPU acceleration, usage-based billing, and OpenSearch Agent Skills for Claude. AWS also expanded hybrid retrieval capabilities, simplified connectivity across environments, and natural language-driven development workflows for OpenSearch applications. For more information, see the AWS News Blog.
Analyst Take
OpenSearch has evolved from an open source search project into a large-scale managed service supporting search, analytics, and AI workloads. Built on Apache 2.0 licensing and now operating under the Linux Foundation through the OpenSearch Software Foundation, the broader OpenSearch ecosystem extends beyond AWS while allowing organizations flexibility in how and where deployments occur.
AWS stated that Amazon OpenSearch Service now supports more than 100,000 monthly active customers processing over 10 trillion requests each month, with examples including Intuit using OpenSearch as a vector store within its AI platform, Adobe supporting Acrobat AI Assistant workloads, and DoorDash enabling high-volume AI-driven contact center experiences.
HyperFRAME Research Lens data (1H 2026) indicates that 60.7% of organizations identify infrastructure as a very significant challenge in scaling AI initiatives. Retrieval infrastructure increasingly contributes to that friction.
This announcement focuses on infrastructure behavior beyond search functionality. AWS framed the redesign around three challenges: unpredictable demand patterns, cost inefficiencies created by overprovisioning, and workflow complexity introduced by fragmented development environments.
Retrieval environments frequently require substantial pre-provisioning because scaling actions can take minutes rather than seconds. Short-lived AI workloads often remain idle between requests, creating inefficiencies when infrastructure remains allocated continuously. Development teams also frequently move across multiple tools and services to configure environments and connect workflows.
The architectural center of the new OpenSearch Serverless is a proprietary shared storage layer built exclusively for this service. Previous serverless implementations attached EBS volumes directly to individual workers, requiring data movement between workers whenever new capacity was added. The new platform gives both search and indexing resources access to a common storage tier, eliminating that movement and allowing compute to scale independently. AWS claims autoscaling up to 20x faster than the previous generation, with cold starts including networking and security provisioning completing in approximately 5 seconds. Writes are committed simultaneously to the shared storage layer and Amazon S3 before acknowledgement, maintaining service reliability while separating storage from compute.
The Anthropic co-announcement warrants attention, as OpenSearch Agent Skills for Claude represents a deeper integration between AWS retrieval infrastructure and Anthropic’s developer environment. AWS stated that OpenSearch should become a preferred way for developers to build search-powered AI applications, and the Launchpad Skills framework is the first implementation of that direction. In our view, the template-based experience is a strong starting point; enterprises with complex data environments and non-standard access control requirements will need a clear path from Launchpad templates to custom configuration.
The competitive context is relevant for enterprise buyers. Elastic remains the most direct competitor in managed search and observability. The proprietary shared storage architecture, purpose-built for OpenSearch, represents a differentiated infrastructure investment relative to Elastic and dedicated vector database vendors including Pinecone and Weaviate. Microsoft Azure AI Search paired with Copilot Studio represents the most comparable integrated alternative at the platform level.
What Was Announced
AWS rebuilt Amazon OpenSearch Serverless around a new shared storage architecture that replaces the previous EBS-per-worker design. Search and indexing resources now access a common storage layer instead of relying on worker-attached EBS storage. AWS stated that the redesign removes the need to move data between workers during scaling operations and allows compute resources to expand independently from storage.
Writes are committed simultaneously to both the shared storage layer and Amazon S3 before completion acknowledgements are returned. According to AWS, this approach maintains service reliability while supporting the redesigned storage architecture. In addition to autoscaling up to 20x faster, AWS stated that cold starts with networking and security provisioning complete in approximately 5 seconds and cited cost reductions of up to 60% compared to provisioned or managed OpenSearch environments.
GPU acceleration for large-scale vector ingestion workloads is now enabled automatically within serverless environments. The capability can deliver up to 10x faster indexing performance at approximately one quarter of previous costs.
OpenSearch Agent Skills for Claude, co-developed with Anthropic, integrates Claude Code as a natural language interface for OpenSearch workflows. The hybrid retrieval capability combines lexical, semantic vector, and agentic search capabilities within OpenSearch Serverless. AWS also expanded development capabilities through integrations with environments including Claude Code, Vercel, and Kiro.
Networking updates introduce collection-level and regional endpoints along with EC2 API support for VPC connectivity across cloud and on-premises environments. AWS also introduced OpenSearch UI as the default experience for new serverless deployments.
Looking Ahead
AWS now offers multiple capabilities spanning retrieval, search, and AI application development. Enterprise buyers will need to understand where and how these capabilities fit into broader application architectures. Reducing that complexity and giving development teams a clearer path to a working agentic application will be central to how broadly AWS captures the market this announcement targets.
The alignment with Anthropic reflects a partnership depth that goes beyond a standard ecosystem connector. Launching OpenSearch Agent Skills with Claude Code as the first integration points to a pattern AWS is likely to extend across additional model providers and development environments.
Consumption-based pricing removes the barrier of paying for idle infrastructure, making agentic application development accessible to a much wider range of teams. As OpenSearch Serverless connects more deeply with other AWS services, costs will accumulate across more billing dimensions. Enterprise buyers will want a clear picture of what a production agentic application costs end to end, and AWS will need to provide that clarity as the integrated stack matures.
In our opinion, the most important variable to watch is enterprise adoption. Multiple approaches to building agentic search applications exist today, and AWS is making a direct claim that OpenSearch Serverless is the preferred foundation for that workload. Real-world deployments will determine whether the architecture, the developer experience, and the pipeline integrations hold up against the demands of production enterprise environments. Examples provided indicate a strong start.
Don Gentile | Analyst-in-Residence -- Storage & Data Resiliency
Don Gentile brings three decades of experience turning complex enterprise technologies into clear, differentiated narratives that drive competitive relevance and market leadership. He has helped shape iconic infrastructure platforms including IBM z16 and z17 mainframes, HPE ProLiant servers, and HPE GreenLake — guiding strategies that connect technology innovation with customer needs and fast-moving market dynamics.
His current focus spans flash storage, storage area networking, hyperconverged infrastructure (HCI), software-defined storage (SDS), hybrid cloud storage, Ceph/open source, cyber resiliency, and emerging models for integrating AI workloads across storage and compute. By applying deep knowledge of infrastructure technologies with proven skills in positioning, content strategy, and thought leadership, Don helps vendors sharpen their story, differentiate their offerings, and achieve stronger competitive standing across business, media, and technical audiences.