Cambridge-based startup Worldmodeldata has emerged from stealth, announcing a £7 million seed funding round led by Iona Star Capital. The company is on a mission to build the world's largest AI training data library sourced from video games. Its goal is to accumulate one million hours of licensed gameplay data by the end of 2026 to power next-generation AI.
A Strategic Vision for AI's Future
Although the funding round closed last December, the company delayed the announcement for seven months. Founder Rhea Loucas explained this period was used to strategically define the data most valuable for training advanced AI systems. This calculated pause allowed the team to refine its approach before launching into the fast-paced AI sector.
Worldmodeldata is targeting the development of "world models," a new AI paradigm requiring data that demonstrates cause and effect. These systems learn from recorded actions and their outcomes to understand and predict how environments change. The company will aggregate licensed data from games built on engines like Unreal and Unity to meet this demand.
The Untapped Potential of Gaming Data
Video games provide an ideal source for this specialized data, offering safe and controllable environments rich with records of human interaction. By licensing gameplay, Worldmodeldata creates a new monetization channel for game developers and their communities. This ethical approach also addresses growing legal scrutiny over the provenance of AI training datasets.
Gerry Buggy, co-founder of lead investor Iona Star Capital, emphasized the critical market need for this data. He described the link between action and consequence as the "scarcest resource in AI today." In his view, Worldmodeldata is manufacturing an essential ingredient required by every company building physical or digital AI.
Navigating a Competitive Landscape
The startup enters a competitive field where rivals have recently secured substantial funding for similar data projects. Competitors like Origin Lab and General Intuition have raised millions, highlighting intense investor interest in the space. This dynamic landscape underscores the challenge and opportunity Worldmodeldata faces as it begins to scale its operations.
Worldmodeldata distinguishes itself by operating as a pure data provider rather than an AI model developer. Loucas clarified that unlike competitors who build proprietary models, her company's mission is to supply high-quality data to AI labs. This positions Worldmodeldata as an infrastructure partner, similar to Scale AI, instead of a direct competitor.
Building for an Ambitious Future
With the new capital, the company plans to expand its team, which currently numbers around ten people including advisors. The funding will primarily support hiring and securing data-sourcing agreements with game studios. While a customer pipeline is developing, the company has not yet finalized contracts or started generating revenue.
The one-million-hour data library goal serves as the company's "north star," signaling its ambition to become a foundational player. This target is reportedly 25 times larger than the biggest existing dataset, positioning the UK-based startup as a potential leader. The coming quarters will be critical in demonstrating whether these ambitions can translate into market traction.
Worldmodeldata's £7 million funding marks its formal entry into the AI data infrastructure market, aiming to solve a key bottleneck for next-generation AI. By focusing on ethically licensed video game data, the company is positioned to supply the fuel for advanced world models. Its success now depends on executing its vision, scaling its team, and securing partnerships in a competitive field.