November 19, 2026Funding-Series BInfrastructureOpen Source

LanceDB Raises $57.5M Series B for the Missing Data Layer in AI Training

LanceDB raised a $57.5M Series B co-led by Bessemer Venture Partners and Innovation Endeavors, with Theory Ventures, Dimension, Swift Ventures, CRV and Y Combinator participating. Total funding is now $98.5M; no valuation was disclosed. The round was announced on Nov 17 and picked up widely on Nov 18. SiliconANGLE: https://siliconangle.com/2026/11/17/lancedb-raises-57-5m-to-accelerate-ai-model-training-with-multimodal-data-platform/ , site: https://lancedb.com

The problem it solves is boring and expensive. A modern training set is not a table of text. It is video, images, lidar point clouds, audio, embeddings and annotations, often 10 to 20 petabytes in one table and sometimes 80 to 100. Research teams scatter those pieces across different systems and, in CEO Chang She's words, "lose weeks or months" assembling one dataset for one training run.

LanceDB's answer is Lance, an open-source file format built for AI data that can fetch and modify individual records fast (analytics formats like Parquet are tuned for scanning whole tables), and a multimodal lakehouse on top of it that works on the customer's own cloud storage without copying the data. Runway, Netflix and Microsoft are named customers; Runway's CTO says a training example "doesn't stay still", it keeps gaining embeddings and metadata with every experiment, which is exactly the column-append workload Lance was built for.

The new money goes to curation, feature engineering and data loading that keeps thousands of GPUs fed, with explicit focus on video generation, world models and robotics. Put it next to Humyn Labs raising the same day to collect first-person robot footage: physical AI is producing data faster than the old storage stack can hand it to a trainer, and the picks-and-shovels layer is getting funded.
← Previous
VitriVax Raises $87M Series C for a Lyme Disease Vaccine
Next β†’
Humyn Labs Raises $47M Series A to Film Human Chores for Robot Labs
← Back to all articles

Comments

Loading...
>_