TwelveLabs secures $100M Series B to scale video foundation models on AWS
TwelveLabs has secured $100 million in Series B funding, co-led by NEA and NAVER Ventures, to scale its video foundation models and expand international operations. The company provides video intelligence through its API and AWS Bedrock, focusing on semantic search and structured metadata generation.
Key Takeaways
- Series B funding totaling $100 million was co-led by NEA and NAVER Ventures, with strategic participation from Amazon and Red Bull Ventures.
- The Pegasus 1.5 model now converts raw footage into structured data including scene boundaries, entities, and temporal segments for automated reasoning.
- AWS remains the preferred cloud provider under a multiyear agreement to optimize video inference workloads specifically on AWS Trainium chips.
- TwelveLabs recently launched Rodeo, its first application-layer product designed for end-user video cognition workflows.
- Expansion plans include new research and development offices in New York and London to augment existing operations in San Francisco and Seoul.
Why It Matters
TwelveLabs is shifting the video AI focus from generative 'hallucination' toward verifiable intelligence and structured metadata. By providing the 'visual cortex' for AI agents, the company allows streaming platforms and sports broadcasters to turn petabytes of stagnant footage into searchable, liquid assets. The multiyear commitment to AWS Trainium indicates a strategic move to lower inference costs, which currently represent a primary barrier to deploying video AI at production scale. As hyperscalers like Amazon look for specialized models to compete with OpenAI's Sora, TwelveLabs’ integration into Bedrock cements its role as a core infrastructure layer for enterprise video applications. Watch for the performance benchmarks of its next-generation models running natively on Trainium silicon as a signal of its competitive moat against generic LLMs.
Additional Context
The Series B funding brings TwelveLabs' total capital raised to over $207 million, following a significant 2025 expansion of its model family. Per Amazon, July 2025, the company first integrated its Marengo and Pegasus models into Amazon Bedrock, allowing developers to execute video search and classification tasks without managing underlying GPU clusters. By December 2025, the release of Marengo 3.0 introduced capabilities for composed multimodal queries, enabling users to search using a combination of text, images, and audio segments across four-hour video files in 36 languages.
This funding arrives as the video AI market enters a specialized phase. While consumer-facing tools like OpenAI’s Sora and Runway Gen-3 Alpha focus on high-fidelity generation, TwelveLabs has targeted the 'understanding' layer critical for B2B workflows. In April 2026, the company launched Pegasus 1.5 on Product Hunt, highlighting its ability to transform raw video into a queryable data asset similar to how markup languages structure text documents.
Strategic partnerships have been a cornerstone of their growth strategy. Beyond the AWS relationship, TwelveLabs partnered with VAST Data in February 2026 to integrate video intelligence directly into large-scale storage environments. Earlier, in October 2025, Firstman Studio—founded by the creator of Squid Game—invested in the company, signaling a deepening tie between TwelveLabs’ AI and the global creative production industry. These moves position the company to compete for enterprise contracts against both general-purpose model providers and niche startups aiming to automate broadcast and security workflows.
Read full article at sportsvideo.org
Enjoy our coverage?
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source