$100M/year PostHog of vector databases
Jan-Erik Asplund
TL;DR: With its vector database built on cheap object storage & developer-first branding, Turbopuffer won customers like Cursor, Notion, and Anthropic. Now, it’s expanded into full text search & hybrid search as it looks to build the definitive search infrastructure for AI agents. Sacra estimates Turbopuffer hit $100M in annualized revenue in March 2026, up 2,400% YoY in 2025. For more, check out our full report and dataset on Turbopuffer, as well as our interviews with AI engineers at Meta & Indeed about evaluating Turbopuffer in production.
Key points via Sacra AI:
- Where Pinecone (founded 2019) built its vector database around keeping search indexes ready on live servers for single-digit millisecond latency, Turbopuffer (founded 2023) built around cheap object storage, allowing companies to build retrieval-augmented generation (RAG) over massive datasets and roll it out to millions of users at ~10-30% the cost of Pinecone. Turbopuffer monetizes on storage, writes, and queries with prepaid minimums (from $64/month to $4,096/month), aligning pricing with AI workloads where every user codebase, workspace, document, and ticket needs to be searchable, but most data sits dormant most of the time.
- After working with Cursor ($3B annualized revenue, +1,100% YoY in 2025) and then winning top names like Notion ($500M ARR) & Anthropic ($43M annualized revenue, +800% YoY in 2025), Sacra estimates Turbopuffer hit $100M in annualized revenue in March 2026, up from $75M at the end of 2025 (up 2,400% YoY). Compare to database/search infrastructure comps Algolia at $230M of ARR in 2025, up 10% YoY, valued at $2.25B as of its July 2021 Series D for a 24x multiple on $95M ARR, MongoDB (NASDAQ: MDB) at $2.46B revenue, up 23% YoY, valued at ~$26B for a 10.6x multiple and Elastic (NYSE: ESTC) at ~$1.7B revenue, up 18% YoY, valued at ~$5.7B for a 4.6x multiple.
- Like PostHog ($58M ARR as of February 2026, +112% YoY in 2025), which turned its developer-centric brand & self-serve onboarding into a wedge against larger analytics platforms, Turbopuffer is bringing the same developer-first brand (cute animal mascot, retro-themed website) & self-serve motion to vector search with infrastructure that is easy for engineers to adopt, start using, and scale to large-scale production workloads. While Turbopuffer started as cheap serverless vector search for RAG, it has since added full-text (2024) and hybrid search (2025) to become the retrieval layer for agentic search, where agents searching codebases, tickets, docs, transcripts, and customer records need both semantic similarity and exact-match retrieval to find the right class names, error codes & customer names.
For more, check out this other research from our platform:
- Turbopuffer (dataset)
- AI engineer at Indeed on TurboPuffer vs. Vespa vs. Elasticsearch at scale
- AI engineer at Meta on evaluating Turbopuffer vs. Pinecone vs. Weaviate
- Edo Liberty, founder and CEO of Pinecone, on the companies indexed on OpenAI
- Pinecone: the MongoDB of AI
- AI program manager at AstraZeneca on running self-hosted ClickHouse
- Will Bryk, CEO of Exa, on building search for AI agents
- Product manager at Cohere on enterprise AI search infrastructure and deep research agents
- Ex-employee at Exa on building search infrastructure for AI data pipelines
- Exa at $10M growing 11x YoY
Read more from
Read more from
Read more from
+4 others
RIP AI Slides 2022-2026
Free Report
Continue Reading
Retell AI revenue, growth, and valuation
Unlocked Report
Continue Reading