Turbopuffer is a search engine architected from the ground up on object storage, an approach that lets it offer vector and full-text search at a fraction of the cost of memory-resident systems while scaling to enormous datasets. By keeping data in cheap, durable object storage and intelligently caching hot data, Turbopuffer separates storage from compute and avoids the cost cliffs that traditional vector databases hit as indexes grow.
This design has made it a backend of choice for some of the most demanding AI products. Turbopuffer manages trillions of vectors and tens of petabytes of data for customers including Anthropic, Atlassian, Cursor, and Notion, powering retrieval and search experiences at production scale.
The company was founded by Simon Eskildsen (CEO) and Justine Li (CTO), both former Shopify engineers, and is headquartered in Ottawa with additional offices in Vancouver and New York City. Its focus on cost efficiency and scale resonated as RAG and AI search workloads exploded, with the company reporting roughly 10x revenue growth and 5x headcount expansion in 2025.
Turbopuffer raised early angel backing from Lachy Groom in 2024, then closed a venture round announced in December 2025 in which Thrive Capital joined as a new investor and Lachy Groom doubled down, with additional backing from Redpoint Ventures. The company reached tens of millions in annual recurring revenue and continued rapid growth into 2026.
By treating object storage as the foundation rather than an afterthought, Turbopuffer competes with established vector databases on price and scalability, positioning itself as core retrieval infrastructure for AI search and RAG.