Ex-DeepMind researcher launches British AI startup, raises $1.1 billion

Former DeepMind researcher David Silver has turned one of AI’s best-known reinforcement learning names into one of the sector’s most expensive bets. His new British startup, Ineffable Intelligence, raised $1.1 billion at a $5.1 billion valuation, a seed round that instantly places the company among the most heavily financed early-stage AI labs in Europe.

The backing is striking not just for its size, but for what it is meant to support: AI systems that learn without relying on large pools of human-generated data. In practice, that means investors are betting that the next performance leap may come from reinforcement learning, self-play, synthetic environments or other methods that reduce dependence on the internet’s text supply. That matters because human-curated training data is getting scarcer, more expensive and more legally fraught, especially as the industry runs up against the limits of scraping ever more text to improve large language models.

Silver brings unusual credibility to that pitch. University College London identifies him as a professor of computer science, and the Royal Society lists him as a former lead of DeepMind’s reinforcement learning team. DeepMind’s own account of AlphaGo shows the logic behind the new company’s thesis: the system first learned from human games and then improved through repeated self-play. Ineffable Intelligence is extending that idea into a broader commercial wager that experience-based learning can produce models that outperform today’s dominant systems.

The financing also reflects how little the market has cooled on frontier AI. Sequoia Capital and Lightspeed Venture Partners led the round, with participation from Nvidia, Google, Index Ventures, DST Global, BOND, EQT, Evantic, Flying Fish, the British Business Bank and the UK government’s Sovereign AI Fund. The British Business Bank said it invested $20 million, while the government said its Sovereign AI backing is aimed at scaling a UK-built, self-learning AI that can generate new knowledge and drive breakthroughs.

Ineffable Intelligence was founded in late 2025 and came out of stealth on April 27, 2026. Its rapid rise underscores a central tension in the AI race: the biggest checks are still going to teams promising a route beyond large language models, but the technical road map remains more ambition than proof. Whether that proves to be a true breakthrough or a repackaging of older reinforcement learning dreams, the money shows that investors still believe the next edge in AI may come from systems that learn by doing, not by reading more human text.