Funding & VC

LanceDB Lands $57.5M Series B to Scale Its AI Training Lakehouse

LanceDB raised $57.5M led by Bessemer and Innovation Endeavors, after tenfold revenue growth and 120M open-source downloads, to expand its AI training lakehouse.

By Nathan Brooks

4 min read

LanceDB Raises $57.5M Series B to Expand Its Multimodal AI Lakehouse
LanceDB Raises $57.5M Series B to Expand Its Multimodal AI Lakehouseblickpixel / pixabay

What's News

  • LanceDB raised $57.5M in Series B funding announced November 17, 2026, bringing total funding to $98.5M.
  • The company reports more than tenfold year-over-year revenue growth since its June 2025 Series A.
  • Its open-source library has been downloaded over 120 million times, six times the year-earlier figure.
  • Customers include Netflix, Microsoft AI, Runway, Luma AI and Midjourney.
  • The platform supports 130,000+ requests per second, tables of 40 billion+ vectors, and training across 8,000 GPUs.

LanceDB has raised $57.5 million in Series B funding to expand its multimodal lakehouse, pushing the company's total funding to $98.5 million. Bessemer Venture Partners and Innovation Endeavors led the round, announced November 17, 2026, with participation from Theory Ventures, Dimension, Swift Ventures, CRV and Y Combinator.

The numbers behind the round explain why investors moved. LanceDB reports more than tenfold year-over-year revenue growth since its Series A in June 2025 — roughly 16 months ago. Its open-source library has passed 120 million downloads, six times the total at the same point a year earlier. The customer list includes Netflix, Microsoft AI, Runway, Luma AI and Midjourney.

What does the company actually sell?

LanceDB's pitch targets a bottleneck AI practitioners know well: data preparation. The company says researchers can lose weeks or months obtaining and curating training data, even though frontier AI advances happen in days and weeks. That gap is the business.

The multimodal lakehouse combines curation, search and retrieval, feature engineering and training for frontier models, world models, physical AI and enterprise applications. It handles videos, images and sensor data alongside large-scale retrieval and training workloads.

The technical specs are the kind infrastructure buyers ask for directly:

  • More than 130,000 requests per second, with latency below 210 milliseconds for 95% of requests
  • Individual tables holding hundreds of petabytes or more than 40 billion vectors
  • Training across 8,000 GPUs in parallel

Enterprise features address a harder problem: updating tables while they're in use. A durable, memory-accelerated write path handles real-time ingest. Distributed indexing scales across large datasets. Managed compaction continuously optimizes tables in the background — LanceDB says it supports up to 100 billion multimodal rows while multiple researchers concurrently read, write and query the same table.

What do customers say changed?

Runway, the video-model developer, offers the clearest picture of how training data behaves in practice. Runway CTO Kamil Sindi described how one training example evolves: "A training example at Runway doesn't stay still. It starts as a video and a prompt, then we add embeddings, resolution and duration metadata, whatever the next experiment needs."

Runway adopted the Lance data format in 2024, then became a LanceDB Enterprise customer. It now runs Lance and LanceDB Enterprise together so researchers can work through large, changing datasets faster.

NVIDIA Cosmos, an open-source user and partner, gives a deployment at world-model scale. Technical lead Max Li said the team curated roughly one million hours of video with Lance for Cosmos 3 and is moving toward training directly from the format. He flagged the object-storage-native architecture as useful for streaming, filtering and inspecting large datasets as they're revised while feeding large GPU clusters.

Microsoft AI carries the enterprise angle. Suqiang (Jack) Song, the company's data-platform head, said Microsoft AI uses LanceDB as its training-data layer for billions of rows spanning images and code, with distributed inference jobs reading Lance directly on thousands of nodes. He was blunt about the deployment condition: operating within Microsoft's own tenant and security policy was non-negotiable for production.

Where does the Series B money go?

LanceDB plans to spend across the full model-development cycle — training-data management, feature engineering, experiment execution and data loading. The strategic goal is bringing training data, context, logs and rollouts into one system at petabyte scale.

One planned capability matters for AI-agent work in particular: research teams could run large numbers of agent experiments in parallel without contaminating shared datasets. That directly attacks a workflow problem — manual pipeline oversight and data-loading friction — that slows research teams today.

Feature engineering gets deeper investment too. A team would define a transformation once; the platform would then apply it across the dataset, backfill existing rows and process new data as it arrives. Further work targets more pre-training preparation plus better data loading and caching during training runs.

What does this change for practitioners?

The Series B moves LanceDB beyond vector retrieval — where it built its reputation — into the continuing work of changing datasets, running experiments and supplying training jobs. For teams currently stitching together pipelines by hand, that's an argument for consolidating infrastructure on one platform.

The bet carries execution risk. LanceDB's next phase will test whether a single system can handle both frontier-scale data volumes and the constant, experiment-by-experiment churn that customers like Runway describe. Winning that test would position the company as the data layer for AI development; failing it would leave the lakehouse competing on retrieval alone, in a market that's crowding fast.

Source: unite.ai

Share this article:

More from Nathan Brooks

Nathan Brooks

Show full bio

News editor covering regulation, policy and their impact on entrepreneurship, startups and venture funding at Business Bearings.

2,352 articles

Related articles

« Previous articleNext article »