Pomegra
UK startup Emulate, founded just weeks ago by ex-DeepMind Genie researchers, is raising $700M at a $3.7B valuation led by Index Ventures and Lightspeed.
- Emulate was founded in August 2026 by Jack Parker-Holder, Matthew McGill, and Philip Ball, all former Google DeepMind researchers on the Genie world-model project.
- Index Ventures and Lightspeed are co-leading the $700M seed round, with Creandum also participating; the company has no product, customers, or public website.
- Emulate is the third DeepMind London spinout this year to close a nine- or ten-figure raise, following Ineffable Intelligence ($1B) and Recursive Superintelligence ($600M).
Lead
Emulate, a London-based AI startup incorporated in August 2026, is in advanced talks to raise $700M at a pre-money valuation of roughly $3.7B in what would rank among the largest seed rounds in venture history. Index Ventures and Lightspeed Venture Partners are co-leading, with Creandum participating. The company has no product, no customers, and no public website. Its three founders - Jack Parker-Holder, Matthew McGill, and Philip Ball - left Google DeepMind earlier this year after building core components of Genie, the lab's generative world-model series.
What Is Emulate Actually Building?
Emulate's focus is world models: AI systems trained to simulate how physical environments behave over time. The core research premise, carried from the Genie project at DeepMind, is that a model capable of accurately predicting physical dynamics - how objects deform, fall, and collide - can underpin robotics, simulation, and scientific computation. Parker-Holder was a principal researcher on Genie 3, the version DeepMind released last year that generated interactive 3D environments from text prompts. The team has published no technical roadmap and released no demo.
The distinction from earlier generative video models matters here. World models are designed to be interactive and physically consistent, not merely visually plausible. Whether Emulate is pursuing a general-purpose physical simulator, a narrower robotics foundation model, or something oriented toward scientific simulation has not been disclosed.
Why Are Investors Paying $3.7B for a Month-Old Startup?
The valuation is almost entirely a bet on the founding team and the perceived trajectory of world-model research. No revenue exists. No product has shipped. The $3.7B figure implies a price-to-revenue multiple that is, by definition, infinite at this stage. This pattern has become routine in the current funding cycle: investors competing to back researchers before they produce anything are willing to price in years of unverified output upfront.
Genie's public reception sharpened that investor attention. When DeepMind released Genie's interactive 3D environment capabilities earlier in 2026, it triggered a sharp sell-off across gaming and simulation stocks. Take-Two, Roblox, and Unity collectively shed billions in market capitalization within days. That reaction - regardless of Genie's actual commercial readiness - told venture investors that physical simulation AI was a credible near-term disruption vector. Parker-Holder, McGill, and Ball built the thing that moved those markets. That is what is being priced.
DeepMind's London Spinout Pattern
Emulate is the third significant DeepMind spinout to raise at scale in 2026. Ineffable Intelligence raised $1B at a $5B valuation; Recursive Superintelligence followed with $600M at a $4B valuation. The sequence points to a broader structural shift: as frontier AI research concentrates inside a small number of large labs, the financial incentive for senior researchers to exit and raise independent capital has grown substantially. Each spinout monetizes the researcher's institutional knowledge and reputation before any proprietary product exists.
The terms under which the Emulate founders departed DeepMind - including any IP arrangements, data access agreements, or non-compete periods - have not been disclosed. Those terms will matter considerably once Emulate begins publishing research or shipping a product, since the line between "carrying forward expertise" and "carrying forward intellectual property" is one that DeepMind's legal team will draw carefully.
Creandum's Participation
The inclusion of Creandum alongside Index and Lightspeed is a small but notable data point. Creandum is a Stockholm-based early-stage fund with a track record in European software. Its presence suggests Emulate is positioning partly as a European AI infrastructure company, which carries implications for UK and EU research grant eligibility, regulatory posture under the EU AI Act, and potential government contracts in defense-adjacent simulation.
Outlook
Emulate will need to convert $700M in runway into something demonstrable before a Series A. The competitive field for world models includes well-funded internal efforts at DeepMind, OpenAI, and several other stealth startups. The founders' Genie lineage is a genuine technical credential, not marketing copy. But the gap between world-model research and production-grade simulation infrastructure has proven wide at every organization that has attempted the crossing. If Emulate ships a compelling foundation model for physical dynamics, the $3.7B entry price will look conservative in retrospect. If it joins the longer list of well-capitalized research shops that never reached the product threshold, this round will serve as a data point in any future accounting of peak-cycle seed valuations.
