Twelve Labs, a San Francisco-based company building video-native AI infrastructure, has raised $100 million in Series B funding 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 platform turns raw video into searchable, AI-ready data through embedding models like Marengo and language models like Pegasus that process up to two hours of video in one request. The capital will advance model development, expand R&D in San Francisco and Seoul, and open new offices in New York and London.
Video AI Funding Signals Infrastructure Shift
The round arrives as video understanding gains traction beyond generation tools. Synthesia raised $200M Series E in January 2026, while Deepgram secured $130M Series C the same month and Tavus closed $40M Series B in November 2025. Twelve Labs differentiates by focusing on comprehension of existing video rather than synthetic creation, addressing gaps in semantic search and temporal reasoning.
Video Analytics Market Expands Rapidly
Organizations face exploding volumes of unstructured video that current text-based or metadata systems cannot query effectively. The video analytics market stands at $14.81B in 2026 and is projected to reach $65.08B by 2034 at a 21% CAGR. Multimodal AI, which integrates vision, audio, and text, is growing from $3.23B in 2026 to $20.82B by 2033 at 36.4% CAGR. Customers such as NFL Media and Sejong City report dramatic reductions in manual review time.
Proprietary Models Enable Temporal Understanding
Twelve Labs built Marengo 3.0 for multimodal embeddings across vision, audio, and speech, achieving 78.5% composite accuracy, and Pegasus 1.5 for generating timestamped structured metadata from videos up to two hours long. These outperform general models on segmentation and prompting tasks. The approach avoids flattening video into text descriptions, preserving causality and sequence that LLMs sampling frames miss. Integrations with AWS Bedrock and Snowflake AI Data Cloud allow direct use within existing enterprise data workflows.
Strategic Investors Validate Cloud and AI Thesis
NEA and NAVER Ventures co-led the round, with Amazon participating as both investor and infrastructure partner through a multiyear Trainium commitment. Earlier backers include NVIDIA and Index Ventures. The investor mix signals conviction that video-native models represent a distinct layer in the AI stack, separate from generative tools. Total funding now exceeds $207 million.
Video Intelligence Infrastructure Attracts Capital
The broader video AI space draws funding across generation, audio, and niche search players, yet Twelve Labs remains the only horizontal player with proprietary foundation models spanning perception, memory, and reasoning. Market trends show enterprises moving from experimentation to production deployments in media, advertising, government, and defense. The company's SOC 2 Type II certification and air-gapped options further support regulated verticals.
Expansion Plans Target Global Reach
With the new capital, Twelve Labs will hire researchers, engineers, product builders, and operators while opening offices in New York and London. The company plans continued R&D on Marengo and Pegasus alongside geographic growth to serve media, sports, and public sector customers worldwide.