News
TwelveLabs Raises $100 Million to Build Video Superintelligence
Video-AI startup TwelveLabs has closed a $100 million Series B funding round and signed a multi-year deal making Amazon Web Services its preferred cloud partner, with new models running first on Trainium chips.
Contents
TwelveLabs, a startup building AI models specialized in video understanding, announced the close of a $100 million Series B funding round. The round was co-led by NEA and NAVER Ventures, with participation from Amazon, Radical Ventures, Korea Investment Partners, Index Ventures, Quadrille Capital, and Red Bull Ventures. The company's total funding has now surpassed $200 million.
Video as a Distinct Class of AI
From the outset, TwelveLabs has run counter to the mainstream trajectory of large language model development, betting that video, not text, should be the primary substrate of machine intelligence. The company argues that video footage is the closest digital record of reality we have, and that models trained on text alone will never be able to fully understand the world as it appears in image and sound.
On this premise, the company has built two flagship products. Marengo 3.0 is an embedding model that converts image, sound, speech, and text into searchable representations, which the company itself describes as the most powerful video embedding model on the market. Pegasus 1.5 turns raw footage into structured data, detecting scene boundaries, objects, and events. More recently the company added Rodeo, a new application-layer product aimed at content creators and video platform operators.
Amazon as Investor and Infrastructure Partner
Particularly notable is the scope of the partnership with Amazon, which plays a dual role here as both an investor in the round and an infrastructure provider. Amazon Web Services has signed a multi-year agreement with TwelveLabs making AWS the company's preferred cloud partner, with new TwelveLabs models set to run first on dedicated Trainium chips before appearing anywhere else. This illustrates how major cloud providers are working to lock in promising AI startups through a combination of capital and infrastructure.
TwelveLabs CEO Jae Lee explained investors' rationale by saying that models are becoming a commodity, but the intelligence layer that connects and organizes them is not. That statement captures the company's strategic pivot, from building individual foundation models toward a full cognitive system for video that unites perception, memory, and reasoning into a single architecture.
Expansion and Spending Plans
The funds from the round will go primarily toward research and development, but also toward geographic expansion. The company currently has offices in San Francisco and Seoul, and is now opening new locations in New York and London to serve growing demand from global clients, including media, advertising, and security firms, for whom searching massive video archives is a daily challenge.
Significance for the Market and Poland
TwelveLabs' funding round fits into a broader investment trend of 2026, in which, following a wave of megarounds for language models, capital has begun flowing toward specialized layers of AI infrastructure, from chips to specific data modalities such as video. For the Polish market, it signals that narrow, deeply specialized AI companies can attract capital on a scale comparable to general-purpose chat models, much as Poland's own ElevenLabs previously did in the audio segment.
For media, advertising, and content moderation companies in Poland, tools in the TwelveLabs class mean, in practice, the ability to automatically search and tag video archives without manually describing every piece of footage, something that today remains one of the most time-consuming parts of the work at newsrooms and streaming platforms.
Sources: GlobeNewswire (globenewswire.com), SiliconANGLE (siliconangle.com), TwelveLabs (twelvelabs.io).
