Skip to content
OddBrief
Business & Startups4 minTraced to the primary source

A six-week-old DeepMind spinout is already shopping a near-$700M seed

Emulate, founded in August by DeepMind world-model researchers behind Genie-style systems, is in advanced talks to raise up to about $700 million. Index Ventures and Lightspeed are reportedly leading, with Creandum joining. Bloomberg and the Financial Times disagree on whether the pre-money ask is closer to $3 billion or $3.7 billion.

AI-assisted, human-reviewed

Official Genie 3 announcement hero from Google DeepMind — world-model scene gridBusiness & Startups
Google DeepMind

Key facts

Round
Up to ~$700M seed, terms not yet final
Valuation
Bloomberg ~$3B pre-money; FT ~$3.7B pre-money
Lead investors
Index Ventures and Lightspeed; Creandum participating
Founders
Jack Parker-Holder, Matthew McGill, Philip Ball (DeepMind world-model / Genie work)

A UK AI company that did not exist two months ago is already negotiating a seed round large enough to buy a mid-size public software firm. Emulate, founded in August by three Google DeepMind researchers who worked on world models, is in advanced talks to raise up to about $700 million, according to Bloomberg and the Financial Times. Index Ventures and Lightspeed Venture Partners are said to lead, with Creandum participating. The valuation numbers diverge: Bloomberg puts the pre-money ask near $3 billion, while the FT cites about $3.7 billion. Either way, the deal would rank among the richest early checks ever written for a lab spinout.

World models, not another chatbot race

Emulate is chasing systems that simulate and predict how the physical world behaves, the same research vein DeepMind has pushed with Genie-style models that generate interactive 3D environments from text. Founders Jack Parker-Holder, Matthew McGill, and Philip Ball all came out of DeepMind's world-model work. Parker-Holder was a principal researcher on Genie 3. Investors see that lane as less crowded than the large-language-model arms race around ChatGPT-class products, and more useful for robots, agents, games, and virtual worlds that have to obey something like physics.

The company has stayed quiet. Emulate did not comment to Bloomberg. Index and Lightspeed also declined comment; Creandum refused to speak. The FT noted that terms are not locked and could still move. That caveat matters for a round this size: a "seed" label with near-unicorn math is less a stage label than a signal that pedigree and scarcity now price like late-stage growth.

Third DeepMind mega-raise of the year

Emulate would be the latest in a streak of outsized checks for researchers leaving frontline labs. David Silver's Ineffable Intelligence raised about $1.1 billion in April at a $5.1 billion valuation. Yann LeCun's AMI Labs, also chasing world models, raised about $1.03 billion in March. Emulate is being cast as another UK neo-lab born from DeepMind's London orbit, arriving only weeks after a leadership shuffle in which Demis Hassabis moved from chief executive to chairman and Koray Kavukcuoglu took day-to-day control, with more senior researchers peeling off to found companies.

That pattern is the odd part of the story. Capital is no longer waiting for a product demo or a Series A narrative. It is bidding on the people who trained the last generation of frontier systems, on the assumption that world models are the next scarce skill. If Emulate closes near the top of the reported range, a company younger than a school semester will sit at a multi-billion-dollar pre-money valuation before most startups finish their first pitch deck. For competitors still hunting Series B dollars to ship chatbots, that is the new market reality: DeepMind alumni are raising nation-state money before they publish a landing page.

Sources

Related reading