Wednesday, November 18, 2026
AI 인프라 · 뉴스 & 분석
홈 › 자본시장 › 리포트
자본시장 · 리포트

LanceDB has raised nearly $58 million to manage AI training data for Microsoft, Netflix and Runway.

Funding for training-data infrastructure software shows continued investment in the data layer beneath model training, though it is a software business rather than compute capacity.
업계 전문지Slicast · 2026년 11월 17일 18:31 UTC · 미국 · 출처: MLQ.ai
중요도 30

LanceDB, which builds a data platform for training AI models on images, video and sensor data, said on Tuesday, November 17, that it had raised $57.5 million in a Series B round. Bessemer Venture Partners and Innovation Endeavors led the round, with Theory Ventures, Dimension, Swift Ventures, CRV and Y Combinator participating. With this raise, total funding reaches $98.5 million. The company did not disclose a valuation.

LanceDB's customers include Netflix, Microsoft AI, Runway, Luma AI and Midjourney. The company said revenue has grown more than tenfold year over year since its Series A last June, and its open-source library has been downloaded more than 120 million times, six times the total a year earlier. It did not disclose revenue figures.

**What the product does**

The platform is built on Lance, an open-source data format that stores structured fields alongside images, video and other large files, and supports fast retrieval and changes to individual records. LanceDB adds a lakehouse layer on top for search, curation, feature engineering and training. It reads data in place within customers' existing cloud storage rather than copying it.

Co-founder and chief executive Chang She told SiliconANGLE that research teams lose "weeks or months" assembling a training dataset spread across separate systems. SiliconANGLE reported that large multimodal training tables commonly hold 10 to 20 petabytes, with some approaching 80 to 100 petabytes. LanceDB Enterprise now offers managed compaction, which the company says supports tables of up to 100 billion rows with multiple researchers reading and writing at once.

Suqiang (Jack) Song, head of data platform at Microsoft AI, said the company uses LanceDB as the data layer for training across billions of rows of image and code data. Max Li, technical lead for Nvidia Cosmos, an open-source user and partner, said the team curated roughly a million hours of video with Lance for Cosmos 3 and is now training directly from it.

LanceDB said the new funding will go toward data curation, feature engineering and data loading that keeps large GPU clusters fed during training, as well as expanded support for video generation, world models and robotics.

원문 보기
LanceDB has raised nearly $58 million to… · Slicast