Thursday, November 19, 2026 Updated Briefings Saved 中文
AI Daily Radar

AI and the frontier tech around it, explained every morning

Announced

LanceDB raises $57.5M Series B for multimodal AI training lakehouse

Photo: kewl / Pixabay

LanceDB Inc., which sells what it calls a “multimodal lakehouse” for AI training, announced a $57.5M Series B led by Bessemer and Innovation Endeavors. The platform is built on Lance, an open data format, and is aimed at storing and querying the mixed media — images, video, audio, embeddings — that multimodal models train on, rather than the text-heavy tables a conventional data warehouse handles. The announcement, reported by SiliconANGLE, gives no valuation, total amount raised to date, customer counts, or a timeline for how the new capital will be spent.

Verified and unconfirmed

  • Verified: LanceDB announced a $57.5M Series B led by Bessemer and Innovation Endeavors, and describes its product as a multimodal lakehouse built on the open Lance format, according to the company’s announcement as reported by SiliconANGLE.
  • Not yet confirmed: The company’s valuation, total funding raised, revenue or customer adoption figures, how the capital will be allocated, and any product roadmap or release dates — none of these appear in the brief report, and no independent estimate of the business is provided.

Background

LanceDB is trying to address a specific data problem in model development: training corpora now mix images, video, audio and text, while conventional analytics stacks were designed mainly for tables. A “lakehouse” combines inexpensive file storage with database-style management, and LanceDB says its lakehouse is built on Lance, an open format for that mixed-media data. The company announced the round on its own blog, and Bessemer published a companion post describing LanceDB as “the data layer for multimodal AI.”

Why it matters

For teams assembling training data for multimodal models, the relevant detail is the open format: because the lakehouse sits on Lance rather than a proprietary store, data written to it is not tied to LanceDB’s own software, which is the lock-in question worth testing before a purchase. The funding also puts another venture-backed vendor alongside established table-format and vector-database options that buyers will compare it against.

Further reading

Comments

Comments

Comments use a GitHub account and are moderated. Comment rules

More in AI

See all