Elastic N.V. has introduced a serverless vector database designed for AI infrastructure, specifically targeting retrieval-augmented generation and AI agent workloads. The product, announced on September 11, 2026, operates on Elastic Cloud Serverless and is capable of managing hundreds of billions of vectors.
The platform features pre-optimized configurations for indexing, embedding, and hybrid retrieval. It supports multi-modal data, including text and images, within a single index. Additionally, the Elastic Inference Service allows for managed GPU inference, supporting both native and third-party models such as those from Jina AI.
Elastic utilizes proprietary Better Binary Quantization and DiskBBQ technologies, which the company claims reduce memory requirements by up to 32x. The underlying VectorDB index mode and automatic calibration features were released in August 2026.
The company has adopted a pricing model based on data volume and search capacity rather than compute units. This approach aims to provide more predictable costs for enterprise finance teams compared to traditional compute-based billing.
Elastic is positioning its solution as a consolidated platform for keyword and vector search, competing against specialized startups like Pinecone, Weaviate, Qdrant, and Milvus, as well as incumbents like MongoDB and PostgreSQL. In mid-September 2026, ESTC stock traded at approximately $85.
