Ineffable Intelligence
Last reviewed
Jun 3, 2026
Sources
8 citations
Review status
Source-backed
Revision
v1 · 1,727 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 3, 2026
Sources
8 citations
Review status
Source-backed
Revision
v1 · 1,727 words
Add missing citations, update stale details, or suggest a clearer explanation.
Ineffable Intelligence is a London-based artificial intelligence research lab founded in late 2025 by David Silver, the DeepMind researcher who led the work on AlphaGo and AlphaZero. The company emerged from stealth on 27 April 2026 with a roughly $1.1 billion seed round at a $5.1 billion valuation, which it and its backers described as the largest seed round in European history.[1][2][3] Its stated goal is to build a "superlearner," an AI system that reaches superintelligence by learning from its own experience through reinforcement learning, rather than by training on text and other data produced by humans.[1][2]
The launch made Ineffable one of the most heavily capitalized startups ever to come out of stealth, and it placed Silver among a small group of senior lab leaders who left the large AI companies in 2025 and 2026 to chase superintelligence on their own terms. What sets Ineffable apart from most of that cohort is the bet itself. While the dominant approach in the industry is to scale up large language models trained on human data, Silver argues that getting to superintelligence will require systems that discard human knowledge and rediscover it from first principles.[2][4]
Silver registered Ineffable Intelligence in November 2025 and was appointed its director on 16 January 2026, after leaving Google DeepMind that same month.[4] He had been on sabbatical from DeepMind for several months before the departure became public.[4] Fortune first reported the move in late January 2026, framing it as one of the highest-profile exits from a frontier lab in that period.[4]
Silver, born in 1976, is one of the most decorated researchers in reinforcement learning. He studied at Christ's College, Cambridge, then completed a PhD at the University of Alberta, where he worked on some of the first master-level computer Go programs. He has held a professorship at University College London since the early 2010s and consulted for DeepMind from 2010 before joining full-time in 2013.[5] At DeepMind he led the reinforcement learning team and directed or co-directed a string of landmark systems: AlphaGo, which beat Go champion Lee Sedol in 2016; AlphaZero, which reached superhuman play in Go, chess and shogi purely through self-play; AlphaStar in StarCraft II; MuZero, which learned to plan without being told the rules of a game; and AlphaProof, which solved International Mathematical Olympiad problems.[4][5] He won the 2019 ACM Prize in Computing and was elected a Fellow of the Royal Society in 2021 for his work on Deep Q-Networks and AlphaGo.[5]
Silver has known DeepMind co-founder Demis Hassabis since university, and was among the company's earliest scientific hires.[4] In interviews around the launch he described Ineffable as his life's work and said he wanted to return to "the awe and wonder of solving the hardest problems in AI."[3][4]
Ineffable's technical premise comes directly from a 2025 position paper, "Welcome to the Era of Experience," that Silver co-wrote with Richard Sutton, the reinforcement learning pioneer who shared the 2024 Turing Award.[6] The paper argues that AI is moving past the era of training on static human data and into one where agents improve mainly by acting in the world, observing the results, and adjusting. Silver and Sutton sketch four ingredients for this shift: continuous streams of lifelong experience, actions taken through sensors and effectors rather than only through text, rewards grounded in the environment instead of in human preference judgments, and forms of reasoning that need not imitate human thought.[6]
Applied as a company mission, this becomes the "superlearner." The idea is a system that learns everything from its own experience without relying on human-generated data, the way AlphaZero learned Go and chess by playing millions of games against itself rather than studying human matches.[1][2] Silver has put the claim in stark terms: to reach superintelligence, he argues, an AI will not just have to surpass human knowledge but discard it and learn to achieve goals from scratch.[2] He has compared the potential payoff to a scientific breakthrough on the scale of Darwin, while acknowledging the approach is unproven at the frontier.[1]
This is a genuine wager on a road not currently taken. Most of the capital and attention in AI flows to ever-larger language models, and reinforcement learning is mostly used as a fine-tuning layer on top of them rather than as the engine of capability itself. Ineffable is proposing to make experience-driven RL the engine. Whether that scales the way transformer pretraining did is the open question that the whole company rests on.
Ineffable raised what it reported as approximately $1.1 billion (around £830 million, or roughly €937 million) in a seed round announced on 27 April 2026, at a $5.1 billion valuation.[1][2][3] The round was co-led by the U.S. venture firms Sequoia Capital and Lightspeed Venture Partners.[1][2] Reported participants included NVIDIA, Google, DST Global, Index Ventures, EQT, BOND, Flying Fish, the U.K. government's Sovereign AI Fund, and the British Business Bank.[2][3] Press accounts noted that the company had earlier targeted around $1 billion at a $4 billion pre-money valuation before the round grew.[3]
| Detail | Reported figure |
|---|---|
| Announced | 27 April 2026 |
| Round | Seed |
| Amount | ~$1.1 billion (≈£830M / €937M) |
| Valuation | $5.1 billion |
| Co-leads | Sequoia Capital, Lightspeed Venture Partners |
| Other investors | NVIDIA, Google, DST Global, Index, EQT, BOND, Flying Fish, UK Sovereign AI Fund, British Business Bank |
| Headquarters | London, United Kingdom |
The size is the headline. A $1.1 billion seed is larger than the entire early funding history of most well-known startups, and the company and several outlets called it the largest seed round ever raised in Europe.[1][2][3] It edged out another billion-dollar European seed from earlier in 2026, a Paris frontier lab staffed by former Meta researchers, in a quarter when AI accounted for more than half of all venture funding raised in Europe.[7] Observers read the round less as a bet on a product, of which there is none yet, and more as a bet on Silver and the team he can assemble. As one account put it, scarce top AI research talent has become the real collateral behind these mega-rounds.[7]
Silver has said he intends to give away the money he personally makes from his Ineffable equity, directing it to high-impact charities aimed at saving as many lives as possible, a pledge made through the Founders Pledge initiative.[1]
On 13 May 2026, about two weeks after the funding announcement, NVIDIA disclosed an engineering partnership with Ineffable to co-design infrastructure for large-scale reinforcement learning.[8] NVIDIA framed the collaboration as a way to work out the hardware and software needed as AI shifts away from human data and toward systems that learn through simulation and experience.[8] The work begins on NVIDIA's Grace Blackwell systems, and Ineffable is set to be among the first labs to test the upcoming Vera Rubin platform.[8] NVIDIA chief executive Jensen Huang said the company was partnering with Ineffable to "codesign the infrastructure for large-scale reinforcement learning" as the startup "pioneer[s] a new generation of intelligent systems."[8] The deal is notable because training agents through trial and error in rich simulated environments places very different demands on compute than running a single forward pass through a language model.
Ineffable is based in London, which makes it one of the larger AI ventures attempting to build a frontier lab in Europe rather than in the San Francisco Bay Area.[2][3] Beyond Silver, the company has not published a detailed roster, though reporting at the time of the funding said several former DeepMind staffers were set to join the executive team, and that Ineffable was actively recruiting AI researchers.[1][4] The London base, combined with backing from the U.K. Sovereign AI Fund and the British Business Bank, gave the launch a national dimension. British coverage cast it as a test of whether the country can keep and fund world-class AI research at home rather than watch the talent and the value migrate to U.S. labs.[3]
Ineffable arrived during a stretch in which several senior researchers left the big labs to start their own superintelligence companies with very large rounds. The pattern recalls Safe Superintelligence, founded by Ilya Sutskever after he left OpenAI, and Thinking Machines Lab, founded by former OpenAI executive Mira Murati, both of which raised enormous sums on reputation well before shipping anything.[2] Ineffable fits that mold on the financing side. On the research side it is the outlier, betting on experience-driven reinforcement learning rather than scaling language models, which is a contrarian position relative to most of the field.
It is worth being precise about identity here, since the 2026 launch cycle produced several similarly themed companies and names. Ineffable Intelligence is David Silver's London RL lab and is not the same entity as any startup marketed under names such as "Ricursive Intelligence" or "Recursive Superintelligence." They are distinct companies and should not be conflated.
Whether the superlearner thesis pays off will take years to judge, and the lab has no public model or benchmark results as of mid-2026. What is already clear is that a large pool of investors was willing to fund the attempt at a scale normally reserved for companies with revenue, on the strength of a research idea and the person behind it.