Thore Graepel, a former Google DeepMind researcher and co-creator of AlphaGo, has launched a London-based AI company called Metis Reasoning, Tech.eu and Sifted reported on 20 October 2026. According to Tech.eu, the company is pursuing a “compute-efficient” alternative to the frontier models built by Anthropic and OpenAI.
Metis Reasoning, which Tech.eu says is “understood to be initially raising tens of millions of dollars”, has not confirmed any funding. It joins a growing group of London labs started by DeepMind alumni who are betting that the next leap in AI will come from better reasoning rather than ever bigger language models. Graepel left Google DeepMind this summer, according to Tech.eu.
In brief
- Thore Graepel, an AlphaGo co-creator, has launched Metis Reasoning in London with a 12-strong team, Tech.eu and Sifted report.
- Tech.eu says the company is understood to be initially raising tens of millions of dollars; co-founder Piers Ingram would not comment on the funding reports.
- The company says it is building AI for high-stakes fields such as embodied AI, science and engineering; it is unclear when it will launch its first model.
Who is behind Metis
According to Tech.eu, Metis’s founding team is Graepel; Piers Ingram, a serial technology and biotech entrepreneur who has raised more than $350m; Michael Warren, a robotics and autonomous systems expert with experience at Boeing; and David Berman, head of AI at Cambridge Consultants and a UK government and NATO AI adviser. The team of 12 includes researchers, mathematicians and robotics experts, the outlet says.
The company’s website gives few launch details. It says: “Metis Reasoning is developing a new approach to artificial intelligence.” It adds: “Our ambition is to expand what can be achieved in high-stakes fields, such as embodied AI, science and engineering, where correctness is critical.” The site also links to a white paper on AI reasoning.
The pitch: reasoning that can be checked
Metis argues that large language models have made significant progress but have a fundamental limitation in how they reason, and can be unreliable and lack an audit trail, Tech.eu reports. That, the company says, makes their work hard to verify and holds back adoption in many industries. It says its approach builds on lessons from Graepel’s work on AlphaGo.
“When you think about higher-stakes cases in engineering or science, or medical or legal reasoning, the arguments that the models come up with basically can’t be trusted,” Ingram told Tech.eu. “I don’t think that’s a good foundation for what we want our AI to be.”
Ingram said the company’s architecture was “much much more” compute-efficient than frontier models. “We are very much not going in this direction of ever larger models and ever larger training budgets,” he said. He told Tech.eu that Metis had achieved some “encouraging” results on frontier reasoning problems at a fraction of the cost. Those claims are the company’s own and have not been independently shown.
Funding: reported, not confirmed
Ingram would not comment on reports that Metis was raising tens of millions of dollars, according to Tech.eu, but said: “We have been able to fund the company more than adequately to date.”
The funding talks were first reported by Bloomberg on 24 September, citing people familiar with the matter, according to a summary by Investing.com. Bloomberg said Graepel was seeking tens of millions of dollars from a small group of backers and could later raise a further tranche of hundreds of millions at a higher valuation. The discussions were preliminary, and the size and structure could change, the people said.
Metis Reasoning is not connected to Metis TechBio, the drug-delivery AI company that recently signed a deal with Eli Lilly.
Our read
DeepMind alumni are now among the most sought-after founders in AI, and Tech.eu notes Metis joins London rivals such as David Silver’s Ineffable Intelligence. Metis’s focus on verifiable reasoning addresses a genuine weakness of today’s chatbots. But it has no public model, no confirmed funding and no date for either, so its claims will need to be tested once it ships something others can examine.
Sources
- Tech.eu: “Former Google DeepMind top researcher launches London ‘compute-efficient’ AI company” (John Reynolds, 20 October 2026).
- Sifted: “Ex-DeepMind researcher Thore Graepel launches new startup Metis Reasoning” (Freya Pratty, 20 October 2026).
- Metis Reasoning: company website.
- Investing.com: “Startup Metis Reasoning seeks tens of millions of dollars in funding – Bloomberg” (24 September 2026).
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X





