Ineffable Intelligence, the London-based AI startup founded by former Google DeepMind reinforcement-learning lead David Silver, came out of stealth Monday with a $1.1 billion seed round at a $5.1 billion valuation. Sequoia and Lightspeed co-led; Nvidia, DST Global, Index Ventures, Google, and the U.K.'s Sovereign AI Fund participated. By dollar volume it is the largest seed round in European history and one of the largest seeds ever raised globally.
The size of the round is not the most interesting thing about it. The thesis is.
Silver is the researcher behind AlphaGo, AlphaZero, and AlphaStar — the line of DeepMind systems that defeated world champions in Go, chess, shogi, and StarCraft II using reinforcement learning rather than human-generated training data. AlphaZero, in particular, learned chess from nothing but the rules of the game and self-play. It did not need a corpus of grandmaster games. It did not need a transformer. It needed a reward signal and a lot of compute, and it produced a level of play that no human-trained system has matched since.
Ineffable's stated mission is to extend that lineage. The company is building what it calls a "superlearner": a system that acquires knowledge entirely through reinforcement learning and direct experience, rather than by ingesting human-written text. In Silver and Richard Sutton's recent paper, "Welcome to the Era of Experience," the argument is that the LLM paradigm — pretrain on the internet, fine-tune on human feedback — is approaching the ceiling of what human-generated data can teach a model, and that real intelligence will require systems that learn the way AlphaZero did, through interaction with environments rather than imitation of text.
This is the most credible non-LLM bet the AI market has seen since the transformer became the default. It lands in a quarter where global startup funding hit a record $297B and AI absorbed $242B (81%) of it — and almost all of that capital flowed into transformer-based labs. OpenAI just hit $25B in annualized revenue. Anthropic is approaching $19B. Google's Gemini 3.1 Pro rolled out globally to AI Pro and Ultra subscribers earlier in April. Microsoft and OpenAI publicly restructured their partnership last week. The LLM consensus is not just dominant; it is consolidating.
A $1.1B seed at a $5.1B valuation, from a researcher with Silver's résumé, is the first piece of capital large enough to credibly fund a different bet. Whether that bet pays off is a separate question — RL-only training has historically been compute-intensive in ways that make even frontier-LLM economics look efficient, and the open problem of how to design useful reward functions outside of closed games like chess remains genuinely unsolved. But the raise itself reframes the field. For two years, the conversation has been "which transformer lab wins." Ineffable is the first credible answer to "what if the answer isn't a transformer."
One personal note that is not just a personal note: Silver has pledged to give 100% of his equity proceeds to Founders Pledge. In a year stacked with founder-CEOs cashing out via secondaries, the optics of a researcher of his stature taking a zero-personal-upside posture on a billion-dollar raise reads as a signal — both about Silver and about how he wants this company perceived in the long-term safety conversation.
The deeper context for office workers watching this play out is the same one we flagged last month: AI anxiety has moved from the boardroom to the break room. A "superlearner" framing is not going to calm anyone down. But it is, finally, an interesting bet against the only AI strategy that has mattered for the last three years.
TechCrunch has the deepest write-up; CNBC frames the cap-table dynamics.
Comments