David Silver, the former DeepMind researcher who created AlphaZero, has raised $1.1 billion at a $5.1 billion valuation for his new startup Ineffable Intelligence. The British AI lab aims to build a "superlearner" — an AI system that discovers knowledge and skills entirely through trial and error, without ever being trained on human-generated data.
The round was led by Sequoia Capital and Lightspeed Venture Partners, with participation from Index Ventures, Google, Nvidia, and the UK's newly launched Sovereign AI fund. The company is just months old.
The AlphaZero Vision, Applied to Everything
Silver spent over a decade at DeepMind, where he led the reinforcement learning team that produced some of the most celebrated results in AI history. AlphaZero taught itself to play chess and Go at superhuman levels by playing millions of games against itself — without studying any human strategies or game records. It simply learned from raw experience.
Ineffable Intelligence aims to apply the same principle to AI models beyond board games. Where today's large language models are trained on massive datasets of human text, code, and images, Silver's superlearner would generate its own knowledge through reinforcement learning — discovering patterns, strategies, and capabilities from scratch.
The ambition is extraordinary. Ineffable's website claims that if successful, the breakthrough would be comparable in magnitude to Darwin's theory of evolution. Silver, a professor at University College London, has described the venture as his life's work and pledged to donate any money he makes from it to high-impact charities.
The Coconut Round Era
Ineffable's $1.1 billion raise at a $5.1 billion valuation makes it the latest example of what the AI industry has started calling coconut rounds — seed investments so large they have outgrown the original metaphor. The term is a tongue-in-cheek escalation of the traditional seed round, reflecting the extraordinary sums that star AI researchers can now command before shipping a single product.
The pattern is becoming familiar. Last month, Yann LeCun's AMI Labs raised $1.03 billion at a $3.5 billion valuation. Mira Murati's Thinking Machines Lab raised $2 billion at $12 billion. Recursive Superintelligence, co-founded by another former DeepMind scientist, reportedly raised $500 million with enough demand to stretch to $1 billion.
In every case, the fundraising is driven not by revenue or products but by the reputation of the founders and the conviction that the next breakthrough in AI will come from a small team with the right ideas and enough compute to test them.
London as an AI Hub
Ineffable's founding in the UK adds to growing momentum around London as a global AI center. DeepMind's continued presence since Google's 2014 acquisition has created a powerful alumni network, with several former DeepMind staffers reportedly joining Ineffable's executive team.
The UK government is also actively supporting the trend. The British Business Bank and Sovereign AI fund participated in Ineffable's round — a signal that the UK sees homegrown AI labs as a matter of national strategic importance. Jeff Bezos' AI lab, Project Prometheus, is reportedly seeking office space near Google's London AI hub, further reinforcing the city's position.
The European AI landscape is shifting. The Cohere-Aleph Alpha merger is building a sovereign Canadian-German alternative. France's Mistral continues to grow. And now London is producing billion-dollar AI labs at a pace that positions it as a serious competitor to San Francisco for frontier AI talent.
Why Reinforcement Learning Matters Now
Silver's bet on reinforcement learning comes at a moment when the AI industry is debating whether traditional training methods — scraping human data from the internet — have reached their limits. Several researchers have argued that the supply of high-quality training data is finite and that models trained exclusively on human text will eventually plateau.
Reinforcement learning offers a potential path around that wall. If AI systems can learn from experience rather than human data, the ceiling on their capabilities is no longer determined by how much training data exists — it is determined by how much compute they can access and how effectively they can explore their environment.
That is why Nvidia's participation in the round is significant. Reinforcement learning is extraordinarily compute-intensive — AlphaZero required millions of self-play games to master chess. Scaling that approach to general intelligence would require infrastructure measured in billions of dollars, which is exactly what Ineffable's funding is designed to secure.
What It Means
Ineffable Intelligence represents the purest version of the AI industry's current thesis: that the next fundamental breakthrough will come from a small team with a radical idea, a legendary founder, and a billion dollars in compute. Whether reinforcement learning can deliver the superlearner Silver envisions is an open scientific question. But the investors backing him — Sequoia, Nvidia, Google, and the UK government — are betting that if anyone can answer it, it is the person who taught a machine to beat the world at chess by playing itself.
