Google DeepMind co-creator Thore Graepel raises funds for AI reasoning startup
The researcher behind AlphaGo's historic victory is betting that structured reasoning, not bigger language models, is the future of artificial intelligence
Thore Graepel, one of the key minds behind AlphaGo, has left Google DeepMind to build a startup focused on AI reasoning. He’s currently raising initial funding in the tens of millions, with expectations that future rounds could reach the hundreds of millions, according to Bloomberg.
Graepel’s bet is straightforward: the next leap in AI won’t come from making language models bigger. It’ll come from teaching machines to actually think through problems the way AlphaGo once demolished the world’s best Go player.
From board games to boardrooms
Graepel originally joined DeepMind in 2015, about a year before AlphaGo made global headlines by defeating world champion Lee Sedol. That 2016 match wasn’t just a milestone for games. It was the moment much of the world realized AI could handle problems with near-infinite complexity.
His new venture’s architecture reportedly emphasizes structured search methods and probabilistic planning. Think of it as the difference between memorizing every answer in a textbook versus learning how to solve new problems from scratch. LLMs are exceptionally good at the former. Graepel wants to crack the latter.
Between his stints at DeepMind, Graepel spent time at Altos Labs from 2021 to 2025, working on biology-focused research. He returned to DeepMind before departing again in summer 2026 to launch this startup.
The great DeepMind diaspora
Graepel isn’t leaving in isolation. The year 2026 has seen a notable exodus of senior talent from DeepMind, and the pattern is starting to look less like coincidence and more like a trend.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
David Silver, another architect of AlphaGo’s success, left to found Ineffable Intelligence back in January 2026. Silver’s departure was itself a significant signal, given his central role in developing AlphaGo, AlphaZero, and AlphaFold’s predecessors.
Leadership changes at DeepMind, combined with the gravitational pull of startup economics in a red-hot AI funding environment, appear to be accelerating the brain drain.
Why reasoning matters now
LLMs can write a convincing essay about supply chain logistics. They struggle to actually optimize a supply chain when conditions change unpredictably. That gap between fluency and competence is exactly where Graepel’s startup appears to be aiming.
Probabilistic planning, one of the technical pillars of his approach, deals explicitly with uncertainty. Rather than generating the most statistically likely next word in a sequence, systems built on this framework evaluate possible futures and assign probabilities to different outcomes.
What investors are watching
The fundraising target tells an interesting story about market appetite. Initial rounds in the tens of millions with subsequent rounds potentially reaching the hundreds of millions suggests Graepel is building something capital-intensive, likely requiring significant compute resources and specialized talent.