Google DeepMind Takes Minority Stake in EVE Online Studio Fenris Creations

The space simulation MMO EVE Online is about to become a laboratory for artificial intelligence. Google DeepMind has acquired a minority stake in the studio formerly known as CCP Games, which is now rebranding as Fenris Creations following a management buyout from Pearl Abyss.

The deal cost somewhere in the millions of dollars, though neither party disclosed the exact figure or the specific percentage of ownership. Hilmar Veigar Pétursson, the studio's CEO, confirmed the investment range in an interview with Bloomberg. He also noted that the financial transaction was only one component of a broader research agreement.

According to reporting from Gizmodo, the partnership will focus on studying how intelligence functions within complex, dynamic systems. DeepMind will conduct experiments using an autonomous version of EVE Online running on a local server, isolated from the live player environment.

This arrangement makes practical sense. EVE Online has been running since 2003, and its universe contains more than 7,000 star systems where players engage in mining, piracy, trading, combat, and politics. The game's emergent economy and player-driven politics create a testing ground far more intricate than most AI training environments. (It's also a game where people have been known to spend decades building empires that can be destroyed in a single afternoon.)

The collaboration will explore specific technical areas including long-horizon planning, memory, and continual learning. These are persistent challenges in AI development. Most models struggle to maintain context over extended periods or adapt to genuinely novel situations without retraining. EVE Online's persistent universe offers a controlled setting to evaluate these capabilities.

Demis Hassabis, CEO and co-founder of DeepMind, has a personal connection to this type of work. He began his career designing and programming complex AI simulation games like Theme Park in 1994. He has cited video games as the perfect training ground for developing and testing AI algorithms, pointing to DeepMind's previous breakthroughs including AlphaGo defeating world champion Lee Sedol in 2016 and AlphaStar reaching Grandmaster level in StarCraft II in 2019.

The timing of this announcement coincides with Fenris Creations regaining independence. The studio was bought out by its own management for $120 million, even though Pearl Abyss had reportedly paid $425 million for it at the time of the original acquisition. The rebranding reflects a shift in ownership and governance only. The leadership team, studios, products, and ongoing development plans remain unchanged.

Independent reporting from Pocket Gamer corroborates the financial details and confirms that no layoffs or office closures are expected. Fenris Creations ended 2025 with a record-breaking month in November and reported over $70 million in revenue for the year. The game is profitable.

There's a physical reality to how this research will actually work. DeepMind will not be testing AI against live players. The experiments will run on an offline version of EVE Online on a local server. This means the AI agents will interact with simulated economies, simulated politics, and simulated player behaviors. The friction of real human unpredictability gets removed, but the complexity of the systems remains.

Other AI companies have used games for similar purposes. In 2019, an AI built by OpenAI defeated world champions in Dota 2. The success with Dota 2 has resurfaced during ongoing legal battles involving Elon Musk and Sam Altman. The pattern is clear: complex multiplayer games provide measurable benchmarks for AI progress.

The partnership also includes plans to explore AI-powered gameplay experiences. This could mean anything from smarter NPCs to automated economy balancing to new player tools. Fenris Creations stated they will continuously evolve a living universe and actively explore what it can become. The language is deliberately vague, which is typical for early-stage research partnerships.

There are concerns within the AI industry about how these technologies get deployed. DeepMind workers have reportedly voted to unionize, citing concerns about AI being used to empower authoritarianism through military or surveillance applications. One anonymous DeepMind worker told The Guardian they joined the union due to these concerns. Google has also reportedly provided the Israeli military with access to AI tools during the conflict in Gaza.

Whether this particular partnership will produce breakthrough AI research or just another corporate press release remains to be seen. The offline testing environment removes the most interesting variable: actual human players. Real intelligence emerges from unpredictable interactions, not controlled simulations. (Though controlling the variables does make the results easier to publish in academic papers.)

The financial structure also raises questions. A minority stake in the millions suggests DeepMind is treating this as a research investment rather than a strategic acquisition. They're buying access to a unique testing environment, not the company itself. Fenris Creations retains full control over creative direction and operations.

For EVE Online players, the immediate impact is minimal. The game continues operating as it always has. The AI experiments run separately on local servers. But the long-term implications could be significant if the research produces AI systems capable of genuine long-horizon planning or adaptive learning in complex environments.

Whether users actually pay for AI-enhanced gameplay features remains the real question. The technology might advance, but the business case for monetizing it in a subscription MMO is less clear. Players have historically resisted changes that feel like they're optimizing the experience for algorithms rather than for human enjoyment.

The deal represents another step in the convergence of gaming and AI research. Games provide the complexity; AI labs provide the capital and technical expertise. The question isn't whether this will work technically. It's whether the resulting products will serve players or just serve as benchmarks for AI progress.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn