Sequoia and Nvidia back David Silver's startup Ineffable Intelligence at a $5.1 billion valuation

London-based startup Ineffable Intelligence, founded by former Google DeepMind researcher David Silver, has raised $1.1 billion and been valued at $5.1 billion before launching a product. The company was only incorporated in November 2025, and it has no revenue and no public roadmap yet. It is one of the largest early-stage bets in European AI, and a rare case of investors putting a billion into a scientific hypothesis and a founder's reputation rather than a finished service.

Silver is one of the key figures in the history of reinforcement learning (RL). At DeepMind he spent more than ten years working on AlphaGo, AlphaZero and AlphaStar, and also contributed to the development of Gemini. AlphaGo became the symbol of a breakthrough in 2016, when the system beat world Go champion Lee Sedol and showed that learning from its own experience can produce strategies no human programmed in advance. AlphaZero later went further: it mastered Go, chess and shogi from scratch, without a database of human games, purely through self-play.

Silver's new company is built around the same idea, on a far more ambitious scale. Ineffable Intelligence wants to create a so-called superlearner: a system that discovers knowledge and skills not by reading human text but by interacting with an environment. The bet is on reinforcement learning: the model should learn by trial and error, gradually accumulating experience and improving its behavior. In Silver's logic, this is an alternative to the current mainstream, in which almost all leading labs rely first and foremost on large language models and masses of human data.

Silver's argument is that LLMs, however strong, remain limited by the corpus of human knowledge. They can scale, combine and speed up what is already known, but are less suited to discovering something fundamentally new. He earlier laid out this position together with Richard Sutton, one of the main theorists of RL, in a paper on the "era of experience."

The authors argued that the next source of progress will be not a static dataset but continuous learning from one's own experience. For that, according to Silver, you need complex simulations in which agents can explore the world, cooperate and develop new strategies without human hints. In essence, this is not about a chatbot conversation partner but about a system that produces new knowledge itself.

Sequoia Capital and Lightspeed Venture Partners co-led the round, and participants include Nvidia, Google, DST Global, Index Ventures, EQT Ventures and several British development institutions. The British side separately highlights the participation of the Sovereign AI initiative and the British Business Bank, which reported a $20 million investment. Amid the general excitement around superintelligence, this looks like an attempt not only to back a star scientist but also to cement London as one of the key hubs of the global AI industry.

Two things matter to investors here: Silver's track record and a counterintuitive bet on a path that runs not alongside LLMs but, to a noticeable degree, against them. For the market, it is a signal that the race for the next generation of AI is less and less about building yet another language model. Money is starting to flow into riskier, more fundamental approaches that have no revenue, no finished product and no clear path to monetization.

If Silver is right, Ineffable Intelligence could become one of the companies that define the architecture of the post-LLM era. If not, this round will remain an expensive reminder that in AI today, people are willing to pay a lot not only for results but also for a convincing scientific vision of the future.