Recursive emerges from stealth at £3.5bn valuation with £480m round

TL;DR:

  • Recursive Superintelligence, founded in 2025 and headquartered in London with a San Francisco office, has emerged from stealth on a £480 million round at a £3.5 billion valuation.
  • The round was led by GV (Google Ventures) and Greycroft, with AMD Ventures and NVIDIA participating — a strategic-capital pattern that ties Recursive’s compute supply to its largest investors.
  • The valuation places Recursive among the largest UK AI raises of 2026 so far, and the team profile — former research leaders from OpenAI, DeepMind, Meta AI, Salesforce and Uber — represents one of the most concentrated UK-domiciled AGI bets to date.

The 25-person team is pursuing “open-ended” self-improving AI through automated scientific discovery, described in the firm’s own statement as “likely the fastest path to superintelligence”. The framing is unusually direct — most well-funded AGI labs have moved away from explicit superintelligence rhetoric — and is more aligned with the founders’ published research backgrounds in quality-diversity algorithms, AI-generating algorithms, and self-improving coding agents.

A UK-headquartered AGI bet

The valuation matters because Recursive is being headquartered in London rather than the Bay Area. UK AI lab announcements in 2025 and early 2026 have typically been frontier-lab research offices, sales offshoots, or compute-infrastructure plays rather than full-stack research labs founded on UK soil. Recursive is closer to the DeepMind template — a London base built around research talent and compute relationships, with a US satellite office for investor and customer proximity.

UK angle: AGI ambition lands in the Knowledge Quarter cluster

The investor mix is as significant as the valuation. AMD Ventures and NVIDIA both participating means Recursive starts with privileged access to both major frontier-compute vendors at a time when capacity is the binding constraint on AGI-style research. For UK AI policy, Recursive is the kind of firm the AI Opportunity Action Plan’s compute-allocation framework was designed to attract — though whether the firm will use the AI Research Resource or rely on its commercial-cloud partnerships is unconfirmed.

Looking forward

Three questions will define whether the valuation holds: whether Recursive publishes peer-reviewable evidence of self-improvement gains within 12 months; whether its team retention rate survives the inevitable poaching attempts from frontier labs; and whether the firm avoids the “safety theatre” criticism that has dogged some recent superintelligence-focused fundraises. The London office leasing surge documented this week in separate CoStar data gives Recursive a market it can grow into without immediately relocating.