David Silver, the former DeepMind researcher behind AlphaGo, has raised $1.1 billion for a new artificial intelligence venture focused on systems that learn without human-labelled data, according to multiple reports. The funding round values the startup at $5.1 billion, representing one of the largest capital raises in AI research this year.
The financing, reported by TechCrunch and Bloomberg, comes as investors increasingly back alternatives to the data-intensive training methods that underpin current large language models. Silver’s approach centres on self-supervised learning techniques that allow AI systems to develop capabilities through interaction with environments rather than relying on vast datasets of human-annotated examples.
Silver led the team that developed AlphaGo, the system that defeated world champion Go player Lee Sedol in 2016, and later worked on AlphaZero, which mastered chess, shogi, and Go through self-play alone. His departure from Google’s DeepMind division was confirmed earlier this year, though the company has not publicly commented on the move.
The $1.1 billion raise places Silver’s unnamed startup amongst the most well-capitalised AI research ventures, alongside Anthropic, which has raised over $7 billion, and xAI, Elon Musk’s company that secured $6 billion in May. The funding reportedly includes participation from several prominent venture capital firms and strategic investors, though specific backers have not been disclosed.
Self-supervised learning represents a potential shift in how AI systems are trained. Current foundation models from OpenAI, Anthropic, and Google require enormous quantities of text, images, and other data labelled or curated by humans. This approach has produced impressive results but faces mounting challenges around data availability, copyright concerns, and computational costs that can exceed $100 million per training run.
Silver’s work at DeepMind demonstrated that reinforcement learning—where systems learn through trial and error—could produce superhuman performance in constrained domains like board games. The commercial question is whether similar techniques can scale to more general applications in robotics, scientific research, or enterprise software.
Market Implications
The substantial valuation suggests investors see autonomous learning as a credible alternative to the scaling strategies pursued by OpenAI and Google. Companies developing robotics systems, particularly in manufacturing and logistics, stand to benefit if self-learning approaches prove more efficient than supervised methods. Firms like Boston Dynamics, Figure AI, and Tesla’s Optimus programme have all emphasised learning-based approaches to robot training.
Conversely, the funding intensifies pressure on data labelling companies such as Scale AI, which has built a $7.3 billion valuation on human annotation services. If self-supervised methods gain traction, demand for large-scale data labelling could decline, though near-term impact appears limited given the continued expansion of traditional model training.
The raise also highlights a bifurcation in AI investment strategies. While some capital flows toward deploying existing models through applications and infrastructure, substantial sums continue backing fundamental research bets with uncertain timelines. Silver’s track record provides unusual credibility for such a speculative approach, but translating research breakthroughs into commercial products remains unproven.
Technical and Regulatory Context
Self-learning systems face distinct challenges from supervised models. They typically require extensive computational resources for training through simulation or real-world interaction, and their behaviour can be harder to predict or constrain. Regulatory frameworks emerging in the EU and UK have focused primarily on models trained on human data, leaving questions about how autonomous learning systems would be assessed for safety and reliability.
The timing coincides with growing concerns about the sustainability of current scaling approaches. Several AI labs have reported diminishing returns from simply increasing model size and training data, prompting exploration of alternative architectures and learning paradigms.
What to Watch
Silver’s startup has not announced a timeline for product releases or disclosed technical details about its approach. Key indicators will include whether the company pursues partnerships with robotics firms, gaming companies, or scientific research institutions—domains where self-learning has shown promise. The composition of the technical team, particularly researchers with reinforcement learning expertise, will signal the venture’s near-term focus.
The $5.1 billion valuation establishes high expectations for commercial outcomes within the typical venture capital timeframe of seven to ten years. Whether autonomous learning can deliver applications that justify such valuations will test both the technology’s maturity and investors’ patience with long-horizon AI research bets.
