The Creator of AlphaGo Just Raised $1.1 Billion on One Radical Thesis — And It Could Redefine the Entire Future of AI

Old-school AI watchers still get chills remembering 2017. That was the year DeepMind dropped AlphaZero — the spiritual successor to the legendary AlphaGo that had already shocked the world by beating the best human Go players. But AlphaZero was something far more audacious.

The system was given nothing but the raw rules of chess: how knights move in L-shapes, how rooks slide, what checkmate looks like. No grandmaster games, no opening books, no endgame tables — just the rules. Then it played against itself. Millions of games. Non-stop.

Four hours later, it crushed Stockfish, the reigning non-neural chess engine that had spent decades absorbing every piece of human chess wisdom ever written. The same trick worked for Go and shogi. AlphaZero didn’t just win.

It played in ways no human had ever imagined — sacrificing pieces with what looked like suicidal abandon, making moves that violated every textbook principle, inventing entirely new strategic concepts that grandmasters are still studying today.

The machine had stepped outside the entire human frame of reference. It didn’t care about “common sense,” tradition, or what the books said was “correct.” It only cared about winning.

That was the prelude.

The real story is about the man who built it: David Silver, the British researcher who led AlphaGo, AlphaZero, and much of DeepMind’s reinforcement-learning breakthroughs. A year ago, Silver and Richard Sutton — widely called the godfather of reinforcement learning — published a short but explosive paper titled *Welcome to the Era of Experience*. Its core thesis was blunt:

The LLM era is already over.

We have scraped nearly every scrap of human-generated text on the internet. The data that powers GPTs, Claude, and Gemini is running out. The next leap in intelligence won’t come from bigger models trained on more human words. It will come from systems that learn the way AlphaZero did — through relentless self-generated experience, trial, error, and reinforcement, in environments with clear rules and unambiguous feedback.

In late 2025, Silver quietly left DeepMind and founded Ineffable Intelligence in London. He brought a handful of top DeepMind talent with him. No product. No revenue. No flashy demo. Just the thesis — and an audacious mission statement: “Make the first contact with superintelligence.”

Last week the world found out just how seriously investors took that bet.

On April 27, 2026, Ineffable Intelligence emerged from stealth and announced it had raised $1.1 billion in a seed round — one of the largest ever — at a staggering $5.1 billion valuation.

The money came from some of the heaviest hitters in tech and venture capital: Sequoia Capital and Lightspeed led, with Nvidia, Google, Index Ventures, DST Global, the British government’s Sovereign AI Fund, and others piling in.

Yes, a company with zero shipped products just became one of Europe’s most valuable AI startups overnight. Because the smartest money in the room is betting that Silver is right again.

If he is, the implications are seismic.

Current frontier models from OpenAI, Anthropic, Google, and Meta are, in this view, the “feature phones” of AI — incredibly sophisticated at imitating human language and knowledge, but fundamentally limited by the ceiling of human data.

They are prediction machines trained to autocomplete the internet. Brilliant at what they do. But they cannot invent truly new knowledge the way AlphaZero invented new chess strategies.

Silver’s bet is that the next intelligence explosion will come from “experience engines” — AI systems that generate their own data through interaction with simulated or real environments, learn from their own mistakes at superhuman speed, and gradually discover truths that no human has ever articulated.

The name of the company is no accident. *Ineffable* means “too great or extreme to be expressed or described in words.” It’s the perfect banner for an effort to teach machines what humans literally cannot put into text.

Of course, skeptics have a point. Chess, Go, and shogi are closed systems with perfect information, clear rules, and a single objective: win. The real world is messy, open-ended, and full of ambiguous goals. Turning reinforcement learning into a general superintelligence engine is an unsolved problem — one that even Silver and his team haven’t cracked yet.

Still, the sheer scale of the bet signals a philosophical shift in the industry. While Sam Altman continues to tell the world that scaling today’s large language models will get us to AGI any day now, some of the deepest thinkers in AI are quietly walking away from that narrative and raising record sums to pursue something radically different.

It’s the ultimate irony of the current moment: the very people who helped create the LLM revolution are now saying the revolution needs a new chapter — one that moves beyond human language and into pure, self-directed experience.

Whether Ineffable Intelligence succeeds or not, the $1.1 billion vote of confidence matters.

It tells us that the smartest investors no longer believe the path to superintelligence runs exclusively through bigger training runs on human text. They believe the future belongs to systems that can teach *themselves*.

The Era of Experience isn’t coming. According to David Silver and the people who just wrote the biggest check in European AI history, it has already begun.

Also read:

• Polymarket Partners with Chainalysis to Deploy Advanced Tools Against Insider Trading
• What Will Be Scarce in the Age of AI
• Neuralink Scales Up: Mass Production of Brain-Computer Interfaces Set for 2026

Thank you!