AlphaGo Researcher Raises Funding for AI Reasoning Startup Metis

Thore Graepel, one of the researchers who helped build AlphaGo, is in preliminary talks to raise tens of millions of dollars for a new artificial intelligence company called Metis Reasoning, according to Bloomberg. Graepel departed Alphabet's Google DeepMind research division earlier this summer.

AlphaGo made headlines in 2016 when it defeated a world champion at the ancient board game Go by training against itself across millions of simulated matches and calculating future moves. The breakthrough demonstrated a new approach to machine intelligence — learning through self-play rather than human instruction.

Metis Reasoning is developing artificial intelligence capable of responding to novel problems and selecting actions, with applications spanning robotics, scientific research, and engineering, the company's website states. Graepel describes the project as applying AlphaGo-style reasoning to cutting-edge AI systems, enabling machines to plan and make decisions when outcomes are uncertain.

The technology differs fundamentally from large language models, which generate the next word based on patterns in human-written text. Metis Reasoning and similar ventures are building systems that learn from their own trial-and-error experiences and evaluate potential actions before committing to them.

The fundraising follows a wave of departures from DeepMind that have attracted massive venture capital commitments. David Silver, who directed DeepMind's reinforcement learning research, secured $1.1 billion at a $5.1 billion valuation in April for his startup Ineffable Intelligence. Sequoia and Lightspeed co-led that round, with participation from Nvidia and Google. Reinforcement learning trains AI through experimentation rather than studying human-generated examples.

In March, Yann LeCun's AMI Labs raised $1.03 billion at a $3.5 billion pre-money valuation to develop world models — systems that simulate how environments will react to actions, allowing robots to test different strategies before executing them physically.

Emulate, launched in August by three DeepMind alumni, is negotiating to raise up to $700 million at a $3.7 billion valuation. It represents the third London-based lab to spin out of DeepMind this year with nine-figure funding.

AI startups are increasingly adopting tranched funding rounds, in which capital is deployed in stages. This structure allows companies to establish high valuations that signal market confidence and help recruit top talent, while giving early investors entry at lower prices before subsequent participants join at elevated terms.

Metis Reasoning may pursue additional funding totaling hundreds of millions of dollars at a higher valuation in a later tranche, sources told Bloomberg. The exact structure and total amount remain subject to change as negotiations continue.