Google DeepMind Takes Stake in CCP Games, Turning EVE Online Into an AGI Testbed

Technology09.May.2026 10:103 min read

Google DeepMind has acquired a minority stake in EVE Online developer CCP Games, which is rebranding as Fenris Creations after buying itself back from Pearl Abyss. The partnership aims to use EVE’s complex virtual society as a sandbox for advancing artificial general intelligence research.

Google DeepMind Takes Stake in CCP Games, Turning EVE Online Into an AGI Testbed

Google DeepMind is expanding its use of virtual worlds for artificial intelligence research, announcing a minority investment in CCP Games, the studio behind the long-running space MMO EVE Online. As part of the deal, CCP has repurchased itself from Korean publisher Pearl Abyss for $120 million and will operate under a new name: Fenris Creations.

A Living Laboratory for AGI

The strategic rationale behind the investment is clear: DeepMind wants a richer, more dynamic environment to test advanced AI systems aimed at artificial general intelligence (AGI). According to DeepMind leadership, EVE Online offers an unusually complex mix of social, economic, and political systems, making it a compelling sandbox for long-horizon learning and multi-agent coordination.

Unlike tightly bounded strategy games, EVE Online operates as a persistent universe shaped by thousands of human players. Its features include:

  • Large-scale multi-agent cooperation and conflict
  • A player-driven, highly detailed virtual economy
  • Long-term political alliances and emergent governance structures

For AI researchers, this environment presents challenges that go beyond tactical optimization. Models must handle memory, strategic planning, resource allocation, negotiation, and adaptation over extended timeframes—capabilities central to AGI ambitions.

From Board Games to Social Simulations

DeepMind has a long history of using games as AI benchmarks. Systems such as AlphaGo and AlphaStar demonstrated superhuman performance in Go and StarCraft II, respectively. However, those environments—while complex—remain constrained by formal rules and match-based objectives.

EVE Online represents a shift toward open-ended, evolving systems that more closely resemble real-world social and economic dynamics. Rather than focusing on short matches, researchers can study how AI agents behave in persistent societies where consequences unfold over weeks or months.

DeepMind is reportedly running an offline version of the game on local servers, separate from the live "Tranquility" server, allowing experimentation without disrupting the player community. This setup enables controlled testing of AI agents while preserving the integrity of the commercial game world.

CCP’s Buyback and Rebranding

Simultaneously with DeepMind’s investment, CCP Games announced it had repurchased itself from Pearl Abyss for $120 million—significantly below its reported $225 million valuation in 2018. The studio will now operate as Fenris Creations, signaling a new strategic chapter that includes closer alignment with AI research initiatives.

Further details about the collaboration are expected later this month. While financial terms of DeepMind’s minority stake were not disclosed, the partnership underscores growing convergence between frontier AI labs and large-scale virtual platforms.

Why Virtual Worlds Matter for AI

The move reflects a broader trend in AI research: shifting from narrow task benchmarks toward complex, socially embedded simulations. As large language models evolve into autonomous agents capable of planning and interaction, researchers increasingly need environments that test coordination, resilience, and long-term reasoning.

Massively multiplayer online games like EVE Online offer a rare combination of structured rules and emergent behavior. If successful, DeepMind’s experiments in this domain could inform how AI systems operate in real-world contexts involving markets, institutions, and collective decision-making.

By turning one of gaming’s most intricate virtual universes into an AGI proving ground, DeepMind is signaling that the next frontier of AI research may look less like a board game—and more like a living digital society.