LanceDB, which sells a data platform for training AI models on images, video and other multimodal data, said on 17 November that it has raised $57.5 million in a Series B led by Bessemer Venture Partners and Innovation Endeavors. Chief executive Chang She wrote in the announcement that the round brings total funding to $98.5 million and that customers include Netflix, Microsoft AI, Runway, Luma AI and Midjourney.
Our number: the Series B is 1.9 times the $30 million Series A that LanceDB announced on 24 June 2025, 511 days earlier, by our arithmetic. Subtracting both rounds from the $98.5 million total leaves about $11 million raised before the Series A, also our arithmetic; LanceDB did not give that figure.
Business Pill · Physical AI
An evergreen Business Engineer explainer on physical AI, the robotics and world-model work that LanceDB names among its use cases. It teaches the general idea only and does not discuss LanceDB.
The key insight: LanceDB raised $57.5 million, 1.9 times its Series A of 511 days earlier, by our arithmetic, and says revenue grew more than 10x year over year without giving a dollar figure; its paid platform sits on an open-source format it says has been downloaded over 120 million times.
What LanceDB Announced
Theory Ventures, Dimension, Swift Ventures, CRV and Y Combinator also took part, according to the post. LanceDB did not disclose a valuation.
She wrote that “Since our Series A last June, revenue has grown by more than 10x year over year”. The post gives no revenue figure in dollars.
On open source, the post says the library has been “downloaded over 120 million times”, “a 6x growth in total downloads compared to this time last year”. The Series A post, in June 2025, put downloads at “more than 20 million times”.
LanceDB lists what the product supports today: “130K+ requests per second, under 210ms P95 latency”, “Over 40 billion vectors in a single table” and “8K GPUs training in parallel”. These are the company’s own figures.

Who Uses It, in Their Words
Runway chief technology officer Kamil Sindi is quoted saying the company “first started using the Lance format in 2024” and has “since become a LanceDB Enterprise customer”. Runway also open-sourced a robotics model this month.
The post describes Max Li, technical lead for NVIDIA Cosmos, as “a LanceDB open source user and partner”. He is quoted: “With Cosmos 3, we curated roughly a million hours of video using Lance; now we’re taking the next step and training directly from Lance”. Nvidia has been pushing robot models trained on video.
Suqiang (Jack) Song, head of data platform at Microsoft AI, is quoted: “We rely on LanceDB as the data layer for training across billions of rows of image and code data”.
An Investor’s Account
Tomasz Tunguz of Theory Ventures wrote about the round in an X article. Theory, he wrote, “led their $30m Series A” and joins the Series B; LanceDB’s own Series A post also says Theory led that round. Tunguz wrote that Exa “stores 35pb in Lance tables across 100b semantic web pages” and that Adobe “manages 40pb to 50pb across 800k tables”.
Those Exa and Adobe figures appear in his article, not in LanceDB’s announcement, and Tunguz is an investor in the company.
The Business Lever
The lever is turning an open-source file format into a paid enterprise product. Lance, the format, is open source; LanceDB Enterprise is the paid platform, and Runway says it became an Enterprise customer after starting with the format. As we read it, the download count is the funnel and the 10x revenue claim is the conversion, as LanceDB tells it.
The Structural Read
As we read the announcement, the round rests on growth LanceDB does not size: more than 10x on an undisclosed base, with no valuation given.
As we read the customer list, the named buyers are model builders such as Runway, Luma AI, Midjourney and Microsoft AI, which matches how LanceDB pitches the product: data for training runs rather than business analytics.
Kamil Sindi, CTO, Runway, in LanceDB’s announcement, 17 November 2026
“We first started using the Lance format in 2024 because it was the only option that handled unstructured data, column appends, distributed reads, and ad-hoc SQL all at once”
Three Implications
FOR MODEL BUILDERS LanceDB says the platform supports 8K GPUs training in parallel and over 40 billion vectors in a single table, figures it gives for its own product.
FOR INVESTORS LanceDB gave no valuation; it says total funding is $98.5 million after the $57.5 million round.
FOR DATA PLATFORMS Runway says it started on the open-source Lance format in 2024 and later became a LanceDB Enterprise customer.
The Business Engineer Lens
This story maps onto the Business Engineer framework The Four Intelligence Moats.
The framework says: “Each moat is built by a different data pipeline.”
As we read it, LanceDB sells tooling for the pipeline that prepares training data, the layer the framework says each moat is built on. The framework predates the round and does not discuss LanceDB.
What Is Not Established
LanceDB gave no valuation, no revenue in dollars and no customer count, so the 10x growth cannot be sized. The quotes from Runway, Nvidia and Microsoft AI are as LanceDB published them. The Exa and Adobe figures come from an investor’s article and were not confirmed by those companies. The $11 million before the Series A is our subtraction from LanceDB’s total. We did not contact LanceDB.
The Bottom Line
LanceDB raised $57.5 million, 1.9 times its Series A of 511 days earlier, by our arithmetic, taking total funding to $98.5 million. It names Netflix and Microsoft AI among its customers and says revenue grew more than 10x year over year, without giving a dollar figure.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. Round size, investors, total funding, customers, growth claims and the customer quotes are from LanceDB’s Series B announcement of 17 November 2026; the Series A details are from LanceDB’s post of 24 June 2025. The Exa and Adobe figures are from an X article by Tomasz Tunguz of Theory Ventures, an investor in LanceDB. The 1.9 times, the roughly $11 million and the 511 days are our arithmetic. We did not contact LanceDB.
Sources: LanceDB, LanceDB Raises $57.5M as AI Research Outgrows the Modern Data Stack, 17 Nov 2026 · LanceDB, LanceDB Raises $30M Series A to Build the Multimodal Lakehouse, 24 Jun 2025 · Tomasz Tunguz (Theory Ventures, an investor), The Lakehouse for AI, X article, 17 Nov 2026









