# Worldmodeldata

Startup profile on Capital Atlas: https://capitalatlas.ai/startups/worldmodeldata

- Industry: Software
- Headquarters: Cambridge, UK
- Website: www.worldmodeldata.com
- Total disclosed funding: $19M
- Latest round: Seed (Nov 3, 2026)
- Investors on record: 2

## About

Worldmodeldata, based in the UK, specializes in creating high-fidelity datasets derived from video games for training advanced AI models, including world models and embodied AI, supporting cutting-edge research and development in the AI sector.Worldmodeldata was founded in 2025 and is headquartered in Cambridge, UK. The company focuses on providing precise datasets for AI training through innovative methodologies. In July 2026, Worldmodeldata raised GBP 7 million in Seed funding led by STAR Capital, marking a significant step in its growth strategy. The firm has completed a total of 1 deal to date, with its most recent funding round reflecting continued investor interest.Worldmodeldata specializes in the provision of high-fidelity, action-conditioned datasets derived from video games, designed specifically for training advanced AI models, including world models and embodied AI. Their core product offering includes synchronized datasets that feature temporally aligned video, telemetry inputs, and ground-truth 3D state, enabling researchers and developers to train AI systems without the noise and inaccuracies typically associated with generic video corpora. These datasets are engineered to provide extensive volume, diversity, and quality, extracted from a wide range of commercial game titles, environments, and gameplay scenarios. Worldmodeldata targets various sectors including academic research institutions, AI development firms, and technology companies focused on AI advancements across North America, Europe, and parts of Asia. By leveraging their proprietary capture stack, they ensure that the data delivered is immediately usable for training, facilitating a robust understanding of dynamic environments and player interactions.Worldmodeldata generates revenue through partnerships with academic institutions and commercial companies that require high-quality training data for AI applications. The transaction structure typically involves framework agreements where clients gain access to the comprehensive datasets tailored to their specific needs, including additional property extraction or custom environments. Pricing models are established based on the bespoke requirements of clients, offering services such as data delivery in popular formats like Parquet, HDF5, and JSON-L. Their flagship datasets, which include rigorous action and state annotations, support a variety of AI research and applications, making them invaluable to organizations aiming to advance their AI capabilities. The company’s focus on delivering precise and coherent datasets underpins its revenue generation strategy, facilitating ongoing collaborations within the AI research community.In July 2026, Worldmodeldata raised GBP 7 million in Seed funding led by STAR Capital. The company plans to utilize this funding to develop its platform, expand its team, and secure additional data licensing agreements for artificial intelligence training datasets. Worldmodeldata aims to introduce new products designed to enhance its dataset offerings and expand its market reach into additional regions, particularly in Europe and Asia, by the end of 2027.

## Funding rounds

- Seed, Nov 3, 2026, $9.4M — Iona Star Capital
- Seed, Jul 6, 2026, $9.3M — STAR Capital

## Investors

- STAR Capital — Seed
- Iona Star Capital — Seed