$1.1 billion seed round, $5.1 billion valuation.
That's the figure announced yesterday (April 27) by David Silver's new company, Ineffable Intelligence. No seed round this large has ever appeared in the history of European venture capital — most AI companies can't raise that much across their Series A and B rounds combined.
What matters more is who this person is and what he actually intends to do.
The man who made AI history twice
If you're not familiar with David Silver, you've certainly heard of AlphaGo. The core researcher behind the program that defeated Lee Sedol in 2016 was Silver. Later came AlphaZero — the system that, without studying any human game records, mastered chess, Go, and shogi purely through self-play within 24 hours and beat all human champions. That was also his work.
Silver spent over a decade at DeepMind leading the reinforcement learning team, while also serving as a professor at University College London (UCL). He has now left DeepMind, based the company in the UK, and assembled the team late last year.
The investor list says more than the amount
Sequoia and Lightspeed led the round, with participation from Nvidia, Google, DST, Index, the UK Sovereign AI Fund, and the British Business Bank (which invested $20 million).
Note the Google connection — the parent company of Silver's former employer DeepMind joined the investment. That suggests Google itself sees this as a bet worth hedging.
UK Science Minister Liz Kendall personally endorsed the deal, saying: "This investment will support a company at the forefront of AI with the potential to transform the entire industry." The government is treating this as a card for European AI sovereignty.
What he actually wants to do: a path completely different from today's large models
Current large models, from GPT-5.5 to Claude Opus 4.7 to DeepSeek V4, all follow the same basic approach: feed in all human-generated text, code, images, and video, then train the model to predict the next token.
Silver is not taking this route. He aims to build a "superlearner" — an AI that learns through reinforcement learning from environmental interaction, without relying on human-generated training data.
This is not a fantasy. AlphaZero proved this approach works. Starting from zero, given only the rules of a board game, it played against itself and surpassed millennia of human chess study within hours.
Silver and his team made a bold claim in their fundraising materials:
"If we succeed, this will be a scientific breakthrough on par with Darwin: his laws explained all life; our laws will explain and construct all intelligence."
Most investors would laugh at that line in a pitch deck. Sequoia and Nvidia didn't laugh. They wrote checks.
Why someone is betting on this now
Data is running out. This is a problem the LLM community has publicly acknowledged over the past six months — most of the high-quality text ever written by humans that can be used for training has already been fed in. Further performance gains will have to come from synthetic data (highly controversial), multimodality (diminishing marginal returns), or a completely different approach.
Reinforcement learning has been proven to work in closed-rule systems by AlphaZero. But applying it to the open world — the real "general intelligence" problem — has never been successfully done. That's what Silver is attempting.
Similar bets include:
| Company | Funding size | Founder background | Approach |
|---|---|---|---|
| Ineffable Intelligence | $1.1B seed | David Silver (ex-DeepMind) | RL-driven superlearner |
| AMI Labs | $1.03B | Yann LeCun (ex-Meta Chief AI Scientist) | World model, non-LLM |
| Recursive Superintelligence | $500M-$1B | Tim Rocktäschel (ex-DeepMind) | Self-improving AI |
Note a common thread: none of these three founders come from the OpenAI/Anthropic ecosystem. None of them are building "bigger language models."
What to make of the pledge to donate everything
Silver has committed to donating 100% of the proceeds from his Ineffable equity through Founders Pledge to highly effective charities that "save the most lives" — the largest commitment in Founders Pledge history. At a $5.1 billion valuation, his personal stake could be worth billions.
This move serves two purposes. If he follows through, it's genuine philanthropy. It also builds a "not here for the money" persona that helps with recruiting, fundraising, and public perception. Sam Altman's early claim of taking "zero salary" followed the same playbook — though he later changed his mind.
Silver himself described this as "his life's work."
What this actually means
Silicon Valley's "transformer + big data + compute scaling" training paradigm now has a publicly hedged counter-bet. Investors aren't betting that Silver will definitely win; they're betting that if LLM scaling laws really do hit a wall, they'll have at least one RL-based card to play.
Looking further ahead, if Silver's approach actually works, the core competitive factors in AI would shift — from "who has more and better data" to "who can design smarter environments and reward functions." The former is a compute game; the latter is a science game.
How long can $1.1 billion sustain a team? At current training compute prices, probably at least 3-4 years. What the AI landscape will look like in 3-4 years, no one can say with certainty.
But one thing is clear: every VC currently betting on the LLM route will be watching closely to see what Ineffable Intelligence does next.
Sources: DeepMind's David Silver just raised $1.1B to build an AI that learns without human data (TechCrunch); CocoLoop; Ex-DeepMind David Silver raises $1.1 billion for AI startup Ineffable (CNBC); Ineffable Intelligence raises $1.1B seed at $5.1B valuation to build 'superlearner' AI (Tech Startups)