Ineffable Intelligence, the British AI lab founded months ago by former DeepMind researcher David Silver, has raised $1.1B at a $5.1B valuation to build an AI that learns without human data. Sequoia Capital and Lightspeed Venture Partners led the round, with Index Ventures, Google, Nvidia, the British Business Bank, and the UK's Sovereign AI fund also participating. The company's pitch is a reinforcement-learning "superlearner" that discovers knowledge through trial and error rather than ingesting human-generated text.
That approach is Silver's specialty. He spent more than a decade at DeepMind, where he led the reinforcement learning team and helped build AlphaZero, the program that beat the world's top chess and Go engines by learning purely from self-play. He is also a professor at University College London. Ineffable now wants to apply the same principle beyond board games to a general-purpose system.
The ambition, in the company's own words, is on the order of Darwin. "If successful, this will represent a scientific breakthrough of comparable magnitude to Darwin: where his law explained all Life, our law will explain and build all Intelligence," the lab's site reads. Silver, who has called Ineffable Intelligence "his life's work," told Wired that "any money that I make from Ineffable will go to high-impact charities that save as many lives as possible."
Key facts
- 01Ineffable Intelligence raised $1.1B at a $5.1B valuation, led by Sequoia Capital and Lightspeed Venture Partners.
- 02Founder David Silver spent more than a decade at DeepMind and led the team behind AlphaZero before leaving to start the lab.
- 03Other backers include Index Ventures, Google, Nvidia, the British Business Bank, and the UK's Sovereign AI fund.
- 04Yann LeCun's AMI Labs raised $1.03B at a $3.5B pre-money valuation last month; Tim Rocktäschel's Recursive Superintelligence raised $500M with demand stretching to $1B.
- 05Silver said any personal proceeds from Ineffable will go to high-impact charities.
The fundraise lands Ineffable straight into pentacorn territory — startups valued above $5B — without a product, customers, or any disclosed revenue plan. That fits a pattern. The market has shifted from seed rounds to what investors now jokingly call coconut rounds: nine- and ten-figure first checks written to credentialed researchers who walked out of the major labs.
“Ineffable Intelligence is betting $1.1B that an AI trained without human data can match what AlphaZero did to chess and Go — but applied to all of intelligence.”— Jaeden Schafer
Last month, AMI Labs, co-founded by Turing Award winner and former Meta AI chief scientist Yann LeCun, raised $1.03B at a $3.5B pre-money valuation. Recursive Superintelligence, founded by former DeepMind principal scientist Tim Rocktäschel and incorporated in the UK, reportedly raised $500M with enough demand to push the round to $1B. Ineffable's $1.1B at $5.1B sits at the top of that cohort.
What ties these bets together is a thesis that the next jump in capability will not come from scaling the current generation of large language models. LeCun has argued publicly that LLMs are a dead end for reasoning. Silver's bet is narrower and sharper: that reinforcement learning, the technique that produced AlphaZero, can be generalized into a system that builds knowledge from its own experience rather than from scraped text.
London is quietly emerging as the staging ground. DeepMind has anchored the city's AI talent base since Google acquired it in 2014, and the alumni network is now spinning out at scale, with several former DeepMind staffers reportedly joining Ineffable's executive team. Jeff Bezos' new AI lab, Project Prometheus, is reportedly hunting office space near Google's London AI hub. The UK's Sovereign AI fund, which co-invested in Ineffable, is a deliberate signal from the government that it wants the next frontier lab headquartered domestically.
The investor list also reads as a hedge. Nvidia and Google both backing a lab that explicitly aims to leapfrog LLMs is the same playbook those companies have run with Anthropic and others — write the check whether or not you believe the technical thesis, because the downside of being absent from the next breakthrough is worse than the cost of the bet.
The skeptical case is straightforward. AlphaZero worked because chess and Go have closed rule sets and clean reward signals. The real world does not. No reinforcement-learning system has yet shown the ability to define its own reward functions across open-ended domains, and Ineffable has not published a technical roadmap explaining how it intends to clear that bar. Silver's reputation is buying him time, not certainty, and the $5.1B valuation prices in execution that has not happened yet.
For the AI market, Ineffable is another data point that capital is rotating toward post-LLM bets faster than most public commentary suggests. Between Ineffable, AMI Labs, and Recursive Superintelligence, more than $2.6B has flowed in the last few weeks into labs whose entire premise is that the dominant paradigm is wrong. Whether they are right matters less, in the short term, than the fact that Sequoia, Lightspeed, Nvidia, and Google are all hedging that they might be.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
