DeepMind’s David Silver Raises $1.1B for Self-Learning AI
The former DeepMind reinforcement learning lead has raised Europe’s largest seed round for Ineffable Intelligence, a startup developing AI systems that learn from experience rather than human data.
On April 27, 2026, David Silver announced that his London-based startup, Ineffable Intelligence, closed a $1.1 billion seed round at a $5.1 billion valuation.
The round was co-led by Sequoia Capital and Lightspeed Venture Partners, with participation from NVIDIA, Google, Index Ventures, and the UK government. It is the largest seed financing ever raised by a European company, according to TechCrunch and Bloomberg.
The company is not building another large language model. Its goal is to create a “superlearner” that acquires knowledge and skills entirely through reinforcement learning and self-generated experience, without using human-generated training data.
David Silver’s Background
Silver led reinforcement learning at Google DeepMind for over a decade. He contributed to AlphaGo (2016), which defeated world champion Lee Sedol, and AlphaZero, which taught itself to master Go, chess, and shogi at superhuman levels using only self-play and no human data or strategies.
He left DeepMind in late 2025 to found Ineffable. In a January 2026 note published on the company’s site, he described it as “my life’s work” and said the reinforcement learning paradigm needed a dedicated environment free from short-term product pressures.
The Company’s Approach
Ineffable’s mission, stated on its website i.e ineffable.ai, is to “make first contact with superintelligence” by building systems that “discover all knowledge from its own experience.”
The company believes superintelligence will come from learning through interaction rather than human data, and that it can be achieved in years rather than decades.
The approach mirrors AlphaZero but at much larger scale: AI agents learn inside simulations by setting goals and observing results.
Silver has argued that large language models are limited by a finite supply of human data, while reinforcement learning offers an unlimited, self-generated source of experience. He discussed this view in detail with Wired.
Funding and Investor Backing
The $1.1 billion round is notable because Ineffable has no product or revenue. Investors are backing Silver’s track record and the technical thesis.
Sequoia’s Sonya Huang told Wired the round reflected Silver’s “purity of vision” and “truly foundational work.” Lightspeed’s Ravi Mhatre described his career as “a single, coherent argument for scaling intelligence without human priors.”
The UK government participated through the British Business Bank, which invested $20 million, and the Sovereign AI Fund. Science Secretary Liz Kendall said the investment supports a company “at the very frontier of AI” and reinforces Britain’s goal of being an “AI maker, not taker,” according to the UK Sovereign AI announcement.
Also read: Vercel Breach Explained
Silver’s Personal Stance
In his founding note, Silver acknowledged the high risk of failure but said success could “positively transform the course of AI and thereby humanity.”
He has committed all personal financial gains from the company to high-impact charities.
Implications and Challenges
The funding highlights growing interest in alternatives to pure large language model scaling. Silver’s approach could enable AI systems that make genuine discoveries and operate more robustly in new environments. It also offers potential safety benefits, as behaviors can be studied in simulation.
However, reinforcement learning remains computationally expensive and historically sample-inefficient. Scaling it from games to open-ended scientific or real-world discovery presents major technical hurdles. The company’s timeline of “years, not decades” will be tested against these realities.
Ineffable represents one of the best-funded bets yet on a post-LLM path to advanced AI. Its success or failure will provide important data on whether experience-based learning can deliver more capable systems than continued scaling of current methods.
Also read: $292 Million KelpDAO Heist Explained
