LanceDB raises $57.5M to build the data layer for AI training
What's the deal? LanceDB has raised $57.5 million in a Series B round led by Bessemer Venture Partners and Innovation Endeavors. The company builds a multimodal data lakehouse for researchers, model builders and platform teams, and the round brings its total funding to $98.5 million.
Who else is in? Theory VenturesDealroom has a profile for this one. Try Dealroom →, Dimension, Swift VenturesDealroom has a profile for this one. Try Dealroom →, CRV, and Y Combinator also participated. The raise lands just five months after a $30 million Series A in June 2025.
What's the endgame? LanceDB wants to fix the infrastructure problem behind AI training, where teams scatter datasets across systems and lose weeks preparing them for a single run. "There's a missing data layer for AI today, and that's what we're trying to build," co-founder and chief executive officer Chang She said.
How it works: Its technology is built on Lance, an open-source format that combines structured data with images, video and other large files. The lakehouse lets developers search, organise and prepare multimodal datasets in one place, connecting to existing cloud storage without moving the data.
Why now? Multimodal AI — spanning lidar clouds, sensor data and video — can fill a single table with 10 to 20 petabytes, and sometimes approach 100. LanceDB says customers include Microsoft AI, Netflix, Runway and Midjourney, and that revenue has grown more than 10x year over year since its Series A.
The signal: At $57.5 million, the round sits in the top 9% of all US enterprise software Series B deals on record, a sign of investor appetite for data preparation as a bottleneck in scaling AI. LanceDB's pitch — managing and preparing training data, not just retrieving it — differentiates it from vector databases like Pinecone, though its performance and customer claims remain self-reported.
Read more: siliconangle.com
Image credit: "Virginia Tech - data center" by cbowns, licensed under CC BY-SA 2.0