London-based artificial intelligence company Metis Reasoning is emerging from stealth, co-founded by Thore Graepel, a former Google DeepMind research lead and AlphaGo co-creator. The company is understood to be raising tens of millions of dollars in initial funding, joining a wave of alternative AI startups in London. Its founding team of 12 researchers, mathematicians, robotics experts, and entrepreneurs aims to pursue a compute-efficient approach to AI reasoning.
Founding Team and Leadership
Metis is led by Graepel alongside three co-founders with complementary expertise. Piers Ingram is a serial technology and biotech entrepreneur who has raised more than US$350 million. Michael Warren brings robotics and autonomous systems experience from Boeing, while David Berman serves as head of AI at Cambridge Consultants and advises the UK government and NATO.
The company is named after Metis, the Greek goddess of wisdom, reflecting its focus on advanced reasoning rather than scale. Its 12-strong team includes researchers, mathematicians, robotics specialists, and technology entrepreneurs, and the venture is headquartered in London. Metis believes its thesis will lend itself well to industries such as robotics and engineering.
An Alternative Approach to AI Reasoning
Metis argues that large language models have made significant progress but remain limited in how they reason. The company contends that these models can produce fluent answers yet often lack reliability and a clear audit trail. This limitation, according to Metis, constrains AI adoption in high-stakes fields such as engineering, science, medicine, and law.
Ingram, who met Graepel at a Cambridge barbecue, said arguments produced by current models in higher-stakes cases cannot be trusted. He explained that this is not a good foundation for what AI should become, and Metis instead builds on lessons from AlphaGo. The company applies a similar architectural insight to AI reasoning, describing its method as more verifiable, robust, and safer than current frontier systems.
Compute Efficiency and Market Position
Ingram would not confirm reports that Metis was raising tens of millions of dollars, though he said the company has been funded more than adequately to date. He described Metis's architecture as much more compute-efficient than frontier models. The company is deliberately avoiding the direction of ever larger models and training budgets.
Metis has achieved encouraging results on frontier reasoning problems at a fraction of the cost of frontier models, according to Ingram. It is unclear when the company will publicly launch its first AI model. The startup is positioning itself as a more efficient and verifiable alternative to established players such as Anthropic and OpenAI.
London's Expanding AI Ecosystem
Metis joins other notable London-based AI companies, including Ineffable Intelligence, founded by Graepel's former AlphaGo colleague David Silver, and Recursive Superintelligence, co-founded by Richard Socher. Ingram said he does not know exactly what those companies are doing, but he considers Metis's approach to be a very different, almost philosophical one. London continues to attract investment and talent pursuing alternatives to large language models.
Metis Reasoning is entering a competitive market with a distinct focus on verifiable, compute-efficient AI reasoning. Its founders believe that addressing the auditability and reliability gaps in current large language models could unlock adoption in demanding technical and scientific fields. While the company has not announced a public launch date, its early progress and funding position suggest that it may become a notable player in London's AI sector.