Databricks Vector Search (now Databricks AI Search)
https://www.databricks.com/product/artificial-intelligence/ai-search
Renamed Databricks AI Search: a fully managed search engine combining semantic vector, keyword (BM25) and hybrid search with reranking in one API. Storage-optimised index scaling to billions of vectors. Hosted MCP servers for UC functions, Genie, Vector Search. Agent framework integrates retrieval with Unity Catalog governance.
Renamed · Verified 2026-10-29 · fresh as of 2026-10-29 · 47 cited facts, 2 with a third-party source · 2 recorded changes · Suggest a correction
At a glance
- Type
- Storage-optimised + Unity Catalog governance vendor fix
- Tier
- T1
- Section
- Vector-database infrastructure
- Latest release
- not applicable — not OSS fix
- License
- not applicable — not OSS fix
- GitHub
- not applicable — no GitHub repo fix
- Pricing
- Pay-per-use DBUs — Standard $0.28/hour per unit (2M 768-dim vectors, 4 DBU/hour); Storage Optimized $1.28/hour per unit (64M vectors, 18.29 DBU/hour), effective US East prices vendor fix
- Funding
- $20.2B total over 14 rounds; $62B valuation Dec 2024 third-party fix
- Freshness
- 2026-10-29 fix
- Layer
- model-infra repo fix
- Licensing
- commercial vendor fix
- SDK languages
- python, rest vendor fix
- Protocols
- mcp-server vendor fix
- Compliance
- soc2-type2 vendor fix
When to use
Optimised for: low-latency similarity search + scale
Anti-fit: not for relational / graph-heavy queries (vector-first by design)
Pros & cons
Pros
Vector index inside the Databricks lakehouse with first-class Delta integration; strong for shops already centralized on Databricks.
Cons
Databricks-only; less mature ecosystem of agent integrations than Pinecone / Weaviate.
Claims & capabilities
7× lower cost (2026). Memory-scaling docs address multi-user agent memory patterns.
Technical surface
- API surface
- REST, SDK: Python
- Backend storage
- Delta Lake (columnar/lakehouse)
- Deployment
- Managed-only
- Embedding model
- multiple supported
- Multi-tenancy
- Workspace-level isolation; separate CMEK per business unit/environment for encryption isolation; revoking key renders data inaccessible
- MCP
- yes — Databricks managed MCP server for AI Search indexes (/api/2.0/mcp/ai-search/{catalog}/{schema}/{index})
- A2A
- no Google A2A (Agent2Agent) integration documented as of 2026-05.
- OpenTelemetry
- first-class — Unity Catalog + OTel
Alternatives to Databricks Vector Search (now Databricks AI Search) → live options ranked by shared design, licence, hosting and MCP support
Similar systems
Other vector-database infrastructure in the catalog, ranked by inbound references.
- Qdrant T1
Distribution-Based Score Fusion + RRF. Sparse vectors native; filtering via ANN graph modification.
- Chroma T1
Limited native hybrid (users build RRF custom). Fast Rust core (v2.5). Acquired by Replit (definitive agreement announced 2026-10-23); Chroma Cloud is winding down (writes disabled 2026-11-22, reads disabled 2026-12-22) while open-source ChromaDB continues under Apache-2.0.
- pgvector T1
Stores embeddings alongside relational + full-text data. HNSW + IVFFlat ANN indexes. Used as agent conversation memory via LangChain + MCP. Foundation of Supabase AI and many self-hosted RAG stacks.
- Weaviate T1
Hybrid Search 2.0 with learned ranking. BM25 + vector in unified index.
- Milvus T1
Multi-vector columns (10 simultaneous). Native hybrid search (v2.5). CAGRA + Vamana GPU/CPU (v2.6).
- Pinecone T1
Managed vector DB. Cascading sparse + dense + rerank pipeline; pinecone-rerank-v0 .
Sources (46 cited facts, 2 third-party)
Each fact links to the page it came from; the badge says who published it (how to read sources).
- What it is Storage-optimised + Unity Catalog governance vendor
- What it does & what's distinct Renamed Databricks AI Search: a fully managed search engine combining semantic vector, k… vendor
- Claims / benefits 7× lower cost (2026). Memory-scaling docs address multi-user agent memory patterns. vendor
- Modalities text vendor
- Created 2013 third-party
- Funding $20.2B total over 14 rounds; $62B valuation Dec 2024 third-party
- Customers / scale 9000+ companies including AT&T, Shell, Burberry, Toyota, Adobe; 500+ with >$1M ARR vendor
- Pricing Pay-per-use DBUs — Standard $0.28/hour per unit (2M 768-dim vectors, 4 DBU/hour); Storag… vendor
- Compliance SOC 2 Type II; ISO 27001; FedRAMP High; HIPAA; GDPR-compliant vendor
- Data handling Never trains on customer data vendor
- Deployment Managed-only vendor
- HQ US vendor
- Founders / pedigree Ali Ghodsi (CEO/co-founder), Ion Stoica, Matei Zaharia, Patrick Wendell, Reynold Xin, An… vendor
- Memory volume / scale billions of vectors (vendor claim) vendor
- Performance Databricks reports 7× lower TCO and faster index builds vs comparable managed vector ser… vendor
- API surface REST, SDK: Python vendor
- Backend storage Delta Lake (columnar/lakehouse) vendor
- Multi-tenancy Workspace-level isolation; separate CMEK per business unit/environment for encryption is… vendor
- Encryption AES at rest with CMEK across AWS/Azure/GCP; encryption at vector index level via managed… vendor
- SSO / RBAC SSO + RBAC vendor
- Embedding model multiple supported vendor
- Versioning immutable append-log vendor
- Tombstoning soft-delete only vendor
- Schema evolution schema with migration vendor
- Namespace primitives per-tenant vendor
- Contradiction handling always overwrite (last-write-wins) vendor
- Forgetting policy none (permanent) vendor
- MCP support yes — Databricks managed MCP server for AI Search indexes (/api/2.0/mcp/ai-search/{catal… vendor
- OTel first-class — Unity Catalog + OTel vendor
- Webhooks / events webhooks + Lakeflow events vendor
- Import / export Delta Lake / Parquet / CSV / JSON vendor
- Validated verticals RAG developers, agent builders (broad horizontal); financial-services, healthcare, legal… vendor
- Time to running <10min (managed cloud) to <1hr (self-host Docker) vendor
- Anti-fit not for relational / graph-heavy queries (vector-first by design) vendor
- Optimised for low-latency similarity search + scale vendor
- Adjacent infrastructure BYO embedding model + LLM; this IS the vector store vendor
- Pros Vector index inside the Databricks lakehouse with first-class Delta integration; strong … vendor
- Cons Databricks-only; less mature ecosystem of agent integrations than Pinecone / Weaviate. vendor
- Project & sources Databricks Vector Search vendor
- architecture-deployment-locus cloud vendor
- licensing-bucket commercial vendor
- positioning-platform-vs-product platform vendor
- taxonomy-three-layer-rollup model-infra repo
- compliance-certifications soc2-type2 vendor
- sdk-languages python,rest vendor
- protocol-support mcp-server vendor
What changed
Changes to this record's lifecycle, licence, release, pricing and hosting facts, from the catalog's git history. Subscribe (Atom)
- 2026-10-29
- Name “Databricks Vector Search” → “Databricks Vector Search (now Databricks AI Search)”
- Homepage “https://www.databricks.com/product/machine-learning/vector-search” → “https://www.databricks.com/product/artificial-intelligence/ai-search”
- 2026-10-27
- Pricing “Pay-per-use” → “Pay-per-use DBUs — Standard $0.28/hour per unit (2M 768-dim vectors, 4 DBU/hour); Storage Optimized $1.28/hour per unit (64M vectors, 18.29 DBU/hour), effective US East prices” source
- MCP support “via official adapter — Databricks MCP” → “yes — Databricks managed MCP server for AI Search indexes (/api/2.0/mcp/ai-search/{catalog}/{schema}/{index})” source
Related systems
Referenced by (3)
- LangChain (framework) integrates with — LangChain docs have a Databricks Vector Search vector store integration page. (verified 2026-10-29)
- LlamaIndex (framework) integrates with — LlamaIndex has an official vector store integration package llama-index-vector-stores-databricks in its repo. (verified 2026-10-29)
- Mem0 integrates with — Mem0 docs have a Databricks vector store page for Mem0 open source. (verified 2026-10-29)