LanceDB Inc. said it has closed a $57.5 million early-stage round to expand a multimodal data lakehouse aimed at researchers, model builders and platform teams, bringing its total funding to $98.5 million. Bessemer Venture Partners and Innovation Endeavors led the Series B. Theory Ventures, Dimension, Swift Ventures, CRV and Y Combinator also participated.
The company argues that AI teams lose time because training data is scattered across separate systems. Co-founder and Chief Executive Chang She said research groups typically store different portions of a dataset in different places and spend weeks or months assembling a single training run. He described a missing data layer for AI as the gap LanceDB is trying to fill.
LanceDB’s platform is built on Lance, an open-source data format for AI workloads that holds structured information alongside images, video and other large files. Unlike analytics formats built to process whole tables, Lance is designed for fast retrieval and modification of individual records, a property the company says suits searching and preparing very large training sets. LanceDB has built that format into a large-scale lakehouse for retrieval, combining data lake storage with database-style organization and querying.
She said the size of multimodal data is part of what makes it harder to manage. Sensor data from autonomous vehicles, robotics and world models adds three-dimensional information, geospatial coordinates, time-series data and annotations to the mix. Large training workflows can average 10 to 20 petabytes in a single table, with some approaching 80 to 100 petabytes, according to the company.
LanceDB connects to customers’ existing cloud storage without copying or moving data, keeping it in place while distributing computation across the infrastructure needed to process it. She said the objective is to let teams reach data without converting or transferring it, which the company says accelerates what he calls the AI flywheel.
Named customers include Runway AI Inc., which expanded onto LanceDB’s enterprise platform as its training needs grew, and Netflix Inc., which built a media data lake on the technology for content creation. Microsoft Corp. uses the platform as a data layer for training models on tens of billions of rows of image and code data. Runway Chief Technology Officer Kamil Sindi said Runway began using the Lance format in 2024 because it handled unstructured data, column appends, distributed reads and ad-hoc SQL together, and that the company has since become a LanceDB Enterprise customer.
LanceDB said the funding will go toward expanding the multimodal platform and speeding AI model development, including large-scale data curation, feature engineering and data loading to keep thousands of graphics processing units supplied during training. The company also plans deeper support for video generation, world models and robotics.
More company and startup news from TechManNews.





