[Edaily Reporter Kim Hyun-ah ] Global video artificial intelligence (AI) startup TwelveLabs has successfully secured $100 million (approximately 150 billion won) in Series B funding. Its cumulative funding has surpassed $207 million (approximately 300 billion won).

Founded in 2021 in Silicon Valley by Korean co-founders, TwelveLabs develops ultra-large multimodal AI models that understand and analyze media (video, images, and audio) in a human-like manner. Rather than competing in the text-centric large language model (LLM) space, the company has focused on building visual AI infrastructure and is gaining recognition for its technological capabilities in the global market. Notably, it garnered attention as the first Korean AI startup to receive a direct investment from NVIDIA, and strategic investors both in Korea and abroad—including Samsung Next, Intel, Korea Investment Partners, and SKTelecom—also participated.

This investment round was co-led by U.S. venture capital firm NEA and Naver Ventures, with Amazon participating as a strategic investor.

Existing investors Radical Ventures, Korea Investment Partners, and Index Ventures participated in the follow-on investment, while Quadrille Capital and Red Bull Ventures joined as new investors.

Twelve Labs plans to allocate the funds raised primarily to research and development (R&D) and global business expansion. The company intends to accelerate the development of a “Video Cognition System” that goes beyond simple video search and analysis to enable videos to understand, remember, and reason on their own.

Video AI: Moving from ‘Search’ to ‘Reasoning’

Twelve Labs is developing a “full-stack video cognition system” that integrates perception, memory, and reasoning into a single framework, going beyond the search and analysis capabilities offered by existing video AI.

While existing large language models (LLMs) have limitations—such as extracting only certain frames during video analysis or having to reanalyze the video from the beginning each time a question is asked—Twelve Labs’ approach involves understanding the video once, storing it as structured memory, and subsequently performing reasoning based on the accumulated information. The company envisions implementing a “video intelligence infrastructure” through which AI performance improves as video data accumulates.

This technological capability is based on the company’s proprietary video AI models, “Marengo” and “Pegasus.” Marengo comprehends visual information, audio, and motion within videos to quickly search for specific scenes, while Pegasus uses this information to summarize video content or infer context.

Competing with B2B APIs… Expanding into Public Sector, Security, and Mobility

Twelve Labs has adopted a B2B platform strategy targeting enterprise customers rather than general consumer AI services. The company provides APIs to enable developers to easily integrate video AI capabilities into their own services, and its models are also integrated into Amazon Web Services’ (AWS) generative AI platform, “Bedrock.”

Building on this foundation, the company is expanding its customer base beyond the media and entertainment sectors into various industries, including public institutions, advertising, security, sports, and automotive. The company explained that the video AI market has entered a phase of full-scale growth, as global companies have recently begun applying video AI to actual business operations and services, moving beyond the pilot project stage.

Strengthening Collaboration with AWS… Optimizing for AI Chips

Along with this investment, the company is also expanding its strategic partnership with Amazon Web Services (AWS).

Twelve Labs has designated AWS as its preferred cloud provider and plans to optimize its video inference models for AWS’s AI chip, “Trainium.” Future video foundation models will also be made available first through AWS.

Expanding Global Presence… Launching AI Services

Twelve Labs plans to expand its teams in San Francisco and Seoul, while establishing new offices in New York and London to strengthen its global operations, including in Los Angeles (LA).

The company is also expanding its business beyond the supply of AI models to include applications. Recently, it launched a closed beta of its AI video analytics service, “Rodeo.” This service enables users to search, analyze, and utilize large-scale video data without requiring separate AI development capabilities.

“Video Is the Core Data for AI”

Lee Jae-sung, CEO of Twelve Labs, said, “Five years ago, we chose a different path based on our conviction that the foundation of AI lies not in language but in video.” He added, “Going beyond foundation models, we will complete a full-stack video perception system that can be utilized by everyone—from developers and enterprises to AI agents—and usher in the era of ‘Video Superintelligence.’”

Tiffany Luck, a partner at NEA, commented, “At a time when video understanding is establishing itself as the new AI infrastructure, I am confident that Twelve Labs is the company that will usher in the next chapter of the market.”

Park Yong-jeong, Head of North American Investments at Naver Ventures’ D2SF, remarked, “In the era of Agent AI, video will become the most critical form of data,” adding, “Twelve Labs is a team with world-class technical capabilities in this field.”