Live

Coming soon

Agriculture Tech
Autonomy
Avatars & Digital Humans
Biotech / Synthetic Biology
Blockchain / Crypto
Brain-Computer Interfaces
Climate Tech
Cloud & Edge Computing

Commerce
Cybersecurity
Defense
Digital Identity
Education Tech
Energy
Fashion & Textiles
Food Tech

Healthcare Systems
Longevity & Human Enhancement
Materials Science
Mobility
Quantum Computing
Smart Homes
Space Tech
Wearables

LanceDB — Lakehouse Platforms · fobi

Live

Coming soon

Agriculture Tech
Autonomy
Avatars & Digital Humans
Biotech / Synthetic Biology
Blockchain / Crypto
Brain-Computer Interfaces
Climate Tech
Cloud & Edge Computing

Commerce
Cybersecurity
Defense
Digital Identity
Education Tech
Energy
Fashion & Textiles
Food Tech

Healthcare Systems
Longevity & Human Enhancement
Materials Science
Mobility
Quantum Computing
Smart Homes
Space Tech
Wearables

LanceDB

LanceDB was founded in 2022 in San Francisco by Chang She, a co-author of the pandas library, and Lei Xu, an Apache Hadoop committer who led ML infrastructure at Cruise. Its open-source Lance format rethinks columnar storage for multimodal AI — fast random access and scans over images, video, embeddings, and text — and had passed 20 million downloads by mid-2025, with the Rust, Python, and Java SDKs reaching 1.0.0 and a community governance structure landing in early 2026. The company raised a $30M Series A led by Theory Ventures in June 2025 (joined by CRV, Y Combinator, Databricks Ventures, and RunwayML), bringing total funding to roughly $41M, and launched LanceDB Enterprise as a 'multimodal lakehouse' spanning search, SQL exploration, feature engineering, and training-scale data management. Generative AI leaders including Midjourney, Runway, Character.AI, and Netflix run it in production, searching tens of billions of vectors and managing petabytes of training data.

LanceDB is the AI-native generation of the lakehouse: where Databricks and Snowflake built for analytics on tables, Lance is becoming the default storage and retrieval layer for multimodal training data, already running underneath Midjourney, Runway, Character.AI, and Netflix-scale AI workloads.

lancedb.com Lakehouse Platforms

United Statesprivate

At a glance

Total raised

$41M

last round series a

Headcount

11-50

approximate · public range

Headquarters

San Francisco

United States

Founded

2022

4 years active

Status

Private

Lakehouse Platforms

Subsector

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Already subscribed? Sign in →

AI bull / bear

Bull case

LanceDB is running the play that won the last data era — own the format, then sell the platform — for the multimodal one. The open-source Lance format passed 20 million downloads by mid-2025, shipped 1.0.0 SDKs in Rust, Python, and Java, and stores embeddings, images, video, and text in one table on cheap object storage: exactly the shape of data AI companies actually train on. The customer list is the proof — Midjourney, Runway, Character.AI, and Netflix run it in production, searching tens of billions of vectors and managing petabytes of training data. Founder credibility runs deep: Chang She co-authored pandas, and Lei Xu was an Apache Hadoop committer who led ML infrastructure at Cruise. The $30M Series A led by Theory Ventures, with CRV, Y Combinator, Databricks Ventures, and RunwayML participating, funds the Enterprise multimodal lakehouse push. If Lance becomes the Parquet of multimodal AI, LanceDB holds the commercial high ground above it.

Bear case

The format-first playbook cuts both ways. Lance moved under community governance in early 2026, which means hyperscalers and lakehouse vendors can adopt the format without paying LanceDB anything — Databricks sits on the cap table and could absorb the idea as easily as champion the company. The commercial layer competes in the most crowded aisle of data infrastructure: Pinecone, Weaviate, Qdrant, Chroma, and Zilliz sell vector retrieval, while Redis, Elastic, MongoDB, and Postgres bolt vector search onto databases enterprises already run. With a sub-fifty-person team and roughly $41M raised, LanceDB is structurally outgunned on go-to-market. Its marquee revenue concentrates in generative AI companies whose own economics remain unproven; if model-training budgets contract, petabyte-scale training-data spending contracts first. Winning format adoption while losing platform monetization is the canonical open-source failure mode, and nothing yet proves LanceDB escapes it.

Generated by editorial-sourced · June 12, 2026Refreshes quarterly · see disclosures