[For more diverse corporate information on the startups mentioned in this article, please visit the Unicorn Factory big data platform 'Data Lab'.]
Twelve Labs is a startup leveraging AI technology to understand video, targeting global markets including the United States. In 2023, it gained attention as the first Korean startup to receive investment from NVIDIA.
The person who recognized Twelve Labs' potential even before NVIDIA was Kim Yoon, Chief Strategy Officer (CSO) of Twelve Labs. He met the founding team, including Lee Jae-seong (CEO), during the company's seed investment stage in 2022 and decided to make an angel investment. As both an investor and advisor, he observed the company from the outside for two years. When the founding members proposed that he take on the CSO role, he began overseeing all aspects of strategy—from research and development (R&D) to commercialization, product development, and business development—starting in 2024.
He is a graduate of Stanford University's Gong Hak (Dr.) program and led the development of Apple's voice assistant service "Siri." In Korea, he served as Chief Technology Officer (CTO) at SK Telecom before joining Sahang Venture Partners, where he began his involvement in the startup industry.
During his time at Sahang Venture Partners, Kim CSO initially intended to invest in Twelve Labs using company funds. However, since the firm was then investing only in mid-to-late stage startups with established scale, he decided to make an angel investment using personal funds. What potential did he see in Twelve Labs during its early days? Above all, it was the keyword "video understanding" that Twelve Labs sought to solve that drew him in.
"90% is dark data"... New market opened by video understanding gains attention
He stated, "Even during my tenure at Apple, I considered the possibility of AI expanding beyond voice into multiple modalities such as images and videos." He added, "While many companies were focusing on generative AI that creates information, Twelve Labs' unique initiative to build models capable of understanding vast amounts of data caught my attention."
Twelve Labs aims to create a foundation that enables the utilization of all video data existing in the world. Kim CSO explained, "Approximately 90% of all data in existence is video data, but most remains stored as 'dark data' because machines cannot yet understand its content." He further noted, "With advancements in platforms like YouTube and OTT, global video data volume is projected to surge from the current 180 zettabytes (ZB—1 billion times a terabyte) to 600 zettabytes in the future."
Twelve Labs has built a foundation model called "Marengo" that extracts meaning from videos. It addresses the challenge of combining this extracted information with linguistic knowledge, enabling users to ask and receive answers about video content or automatically generate metadata, through its "Pegasus" solution.
Using these technologies, Twelve Labs is currently challenging market validation in the United States. Major domestic sports leagues such as American football and leading media studios have already joined as early customers. For example, Maple Leaf Sports & Entertainment (MLSE), a major Canadian sports company, adopted Twelve Labs' technology to drastically reduce personalized highlight video production time from 16 hours to just 9 minutes.
He emphasized, "This not only saves production personnel and time but also provides opportunities for clients to rediscover valuable content within their vast archives of past videos, thereby generating new licensing revenue."
From B2B to B2C... Demand will grow as more 'robots' with cameras appear
Kim CSO announced that Twelve Labs' revenue last year grew sevenfold compared to the previous year. He expects similar high growth this year. While short-term revenue expansion is an important goal, he anticipates that the arrival of the "physical AI" era will open up a much larger market in the medium to long term.
He explained that once environments where over 5 billion smart devices, robots, and cameras replace humans to capture video become widespread, demand for analyzing vast video data and converting it into knowledge and insights will explode.
In the short term, the company plans to accelerate market expansion. Beyond its core target of the United States, it is preparing to target major European sports leagues and global media companies. It is also expanding application areas to include compliance solutions that review cultural context and regulatory violations in corporate advertising, as well as government public safety sectors. Additionally, it is considering entry into the Japanese market, which holds the world's second-largest media market and vast intellectual property (IP) archives.
Kim CSO stated, "In the second half of this year, we plan to launch an agent product that analyzes and infers video content while simultaneously deriving insights and taking actions." He added, "While we are currently focused on B2B business, our long-term goal is to expand into the B2C sector that supports all creators."
The ultimate goal is to grow into the first "global AI decacorn (startup valued over $10bn) (private company with a corporate valuation of over 10 trillion won)" founded by Korean entrepreneurs. Kim CSO said, "We aim to overcome the limitations faced by domestic companies operating in the U.S., leap forward as an excellent enterprise, and become a company that paves the way for future Korean entrepreneurs to challenge global stages without fear."
[MoneyToday startup media platform Unicorn Factory]
in FUTURE
