Emulate (AI lab)

Emulate (registered as Emulate AI Ltd) is a London-based artificial intelligence company that is developing world models, AI systems that simulate how environments behave and how actions change them. It was incorporated in the United Kingdom on 4 August 2026 by Jack Parker-Holder, Matthew McGill and Philip Ball, three researchers from Google DeepMind who worked on the lab's Genie series of world models.^[1][2][11] Parker-Holder led the development of Genie 2, and all three contributed to Genie 3.^[4][19][20]

In September 2026 the Financial Times and Bloomberg reported that Emulate was in advanced talks to raise up to $700 million in a seed round to be co-led by Index Ventures and Lightspeed Venture Partners, with Creandum also taking part. The Financial Times put the valuation at about $3.7 billion including the new money, while Bloomberg reported that the company was seeking $3 billion before the investment.^[4][5][6][8] In October 2026 Bloomberg reported that investors were buying into the same round at rising prices; the round had not been announced as closed.^[14][15] The company had not publicly described its technology or released a product by late October 2026, and its website carried only the line "World models to model the world."^[4][14][16]

The company is unrelated to Emulate, the maker of organ-on-a-chip systems for biological research.^[30]

Overview

Background: the Genie world models

The founders worked at Google DeepMind on Genie, a line of generative models that turn a prompt into an interactive environment a person or an AI agent can act in.^[17][19][20] The work belongs to a broader research program at DeepMind that treats world models as a way to generate unlimited training environments for agents. DeepMind described them in 2025 as "a key stepping stone on the path to AGI, since they make it possible to train AI agents in an unlimited curriculum of rich simulation environments."^[20]

Google opened access to Genie 3 through Project Genie, an experimental web prototype, for Google AI Ultra subscribers in the United States on 29 January 2026.^[21] The Financial Times, as summarized by Dealroom, noted that the January release unsettled shares of video game companies including Take-Two, Roblox and Unity.^[7]

History

Founding

On 12 August 2026 Sifted reported that Parker-Holder, whom it described as a Google DeepMind researcher who "led the company's world models development", was targeting a $500 million fundraise for a new London lab, that Index Ventures was in talks to lead the round, and that six other DeepMind researchers were involved with the startup.^[9] The company had been incorporated on 4 August 2026 under the name Tea Mule Ltd, with Parker-Holder, McGill and Ball appointed as its directors on the same day.^[1][2] It adopted the name Emulate AI Ltd by a resolution dated 21 September 2026, and Companies House issued the change-of-name certificate on 22 September.^[3]

City AM noted that none of the three founders had previously run a company.^[4] The Financial Times, as summarized by Dealroom, said Emulate was founded by Parker-Holder, McGill, Ball "and other former DeepMind researchers who helped create Genie".^[7]

Seed financing talks

On 17 September 2026 the Financial Times reported, citing people familiar with the talks, that the month-old company was in advanced talks to raise as much as $700 million in a seed round. Index Ventures and Lightspeed Venture Partners were expected to lead it jointly.^[4][5][7] The FT put the valuation at about $3.7 billion including the new capital, and said the terms had not been finalized and could change.^[4][5][6] Bloomberg reported the same day that the company was seeking a valuation of $3 billion before the new money, and that Creandum would also take part, according to one person familiar with the matter.^[6][8] Emulate did not comment on the reports.^[4][6]

Secondary coverage differed on how to read the FT figure. The Next Web described it as a pre-money valuation,^[6] while Silicon UK and City AM, reporting on the FT story, described it as the valuation including the new capital.^[4][5]

The FT described Emulate as the third AI lab to spin out of DeepMind's London offices in 2026 with hundreds of millions of dollars in financing.^[13] Silicon UK named two other labs founded by former DeepMind researchers that year: Ineffable Intelligence, founded by David Silver, and Recursive Superintelligence.^[5] The talks came about six weeks after Demis Hassabis moved from chief executive of Google DeepMind to chair, with Koray Kavukcuoglu taking over day-to-day running of the lab.^[6] The Next Web reported that several senior researchers had since left DeepMind to start their own companies.^[6]

Rising prices within the round

On 22 October 2026 Bloomberg reported that venture firms were competing to invest in Emulate, then about three months old and still without a product, and that investors were willing to buy into the same round at progressively higher prices, which produced different valuations for the company within a single financing.^[14] Startup Fortune, in an article on AI funding rounds split into tranches priced at different valuations, said the Bloomberg report centered on Emulate.^[15] At the time of that report the round had not been announced as closed, and no final size or valuation had been made public.^[14][15]

Founders

Jack Parker-Holder

Parker-Holder joined DeepMind as a research scientist in the Open-Endedness team in July 2022, and in October 2023 he was appointed an honorary lecturer in computer science at University College London, in the UCL DARK group.^[22] On his research page he describes his interest as "training world models from Internet scale data, providing unlimited training environments for embodied AGI."^[22] He completed a DPhil at St Peter's College, Oxford, in the university's Machine Learning Research Group, advised by Stephen Roberts, and interned at Facebook AI Research in London with Tim Rocktäschel and Edward Grefenstette.^[22] Before his research career he spent seven years at J.P. Morgan in New York as a quantitative researcher and ETF trader, studied part-time for a master's degree at Columbia University, and earlier studied mathematics at the University of Exeter.^[22]

His earlier research covered reinforcement learning methods for open-ended learning and automated hyperparameter tuning. With Minqi Jiang and others he wrote "Evolving Curricula with Regret-Based Environment Design" (ACCEL), presented at ICML 2022, which builds training curricula of environment levels that start simple and become increasingly complex.^[22][27] His population-based bandits method for online hyperparameter tuning (PB2, NeurIPS 2020) was added to the Ray Tune library.^[22]

Philip Ball

Ball describes himself as a core contributor to Genie 2 and Genie 3, interested in "generative video models, high fidelity world models, and machine learning that survives the test of scale."^[23] He did his PhD at Oxford, also supervised by Stephen Roberts, interned at Waymo and Microsoft Research, and worked as a visiting student with Sergey Levine at the RAIL lab at the University of California, Berkeley. He was previously an AI resident at Microsoft Research Cambridge, holds master's degrees in machine learning from Cambridge and in engineering science from Oxford, and earlier worked as a management consultant at Accenture.^[23] With Levine and others he wrote "Efficient Online Reinforcement Learning with Offline Data" (ICML 2023).^[28]

Ball and Parker-Holder, both supervised by Stephen Roberts at Oxford, co-wrote several papers on learned world models. They include "Ready Policy One: World Building Through Active Learning" (ICML 2020), which treats model-based reinforcement learning as an active learning problem aimed at improving the world model with as few samples as possible, and "Augmented World Models Facilitate Zero-Shot Dynamics Generalization From a Single Offline Environment" (ICML 2021).^[22][25][26]

Matthew McGill

The Leverhulme Centre for the Future of Intelligence (LCFI), based in Cambridge, lists McGill as a senior research scientist at DeepMind "working on video models as world models". Before DeepMind he was a research fellow on the centre's Kinds of Intelligence project, based at Imperial College London in Murray Shanahan's lab.^[24] He is named among the key research and engineering contributors to Genie 3.^[20]

Research focus

Emulate had not unveiled its technology or a commercial product by late October 2026.^[4][14] Press coverage describes it as building world models: AI systems that simulate and predict how physical environments behave and how actions affect them, as distinct from the large language models behind ChatGPT and Gemini.^[4][6] The Next Web wrote that such models are useful for training robots and AI agents and for creating games and virtual worlds.^[6] City AM gave the example of a robotics company testing thousands of movements in a simulated warehouse before deploying a robot.^[4] TechFundingNews noted that the founders were best known for research on open-ended simulation rather than for selling commercial products.^[12]

Coverage placed the company alongside other well-funded world-model developers such as World Labs and AMI Labs.^[4][6][12]

Reception

The size and timing of the reported round drew comment. Charlie Dai, a vice president and principal analyst at Forrester, told City AM that the deal reflected investors betting on "future capability creation, not present revenue or commercial traction", and that investors were "increasingly underwriting elite AI research teams before product-market fit".^[4] The Cap Table calculated that, on the reported figures, a $700 million investment at a $3.7 billion post-money valuation would give new investors close to 19 percent of the company, a dilution it called unusually low for a seed round.^[11] TechFundingNews wrote that the main test would be whether DeepMind spin-offs could turn research demonstrations into products that robotics companies would buy.^[12]

Sifted described Emulate as a "neolab".^[10] Debarghya Das's "Ultimate Neolab List" includes it under world models and classifies it as a clean example of the category, citing top researchers, no revenue, compute-heavy world-model work, a software-only business and the $700 million seed.^[29] See Neolab for the use of the term.

See also

• List of neolabs
• Neolab
• World model
• Genie 3
• Ineffable Intelligence
• Recursive Superintelligence
• General Intuition

References

1. ^1 ^2 ^3 ^4 ^5Companies House, "EMULATE AI LTD: overview (company number 17379902)", accessed October 27, 2026.
2. ^1 ^2 ^3Companies House, "EMULATE AI LTD: officers", accessed October 27, 2026.
3. ^Companies House, "EMULATE AI LTD: filing history", accessed October 27, 2026.
4. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15Saskia Koopman, "Month-old Deepmind spinout nears $4bn on 'future capability' bet", City AM, September 17, 2026.
5. ^1 ^2 ^3 ^4 ^5 ^6Matthew Broersma, "Month-Old UK Start-Up Achieves Nearly $4bn Valuation", Silicon UK, September 21, 2026.
6. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13Cristian Dina, "Former DeepMind world-model researchers' startup Emulate in talks to raise $700M", The Next Web, September 17, 2026.
7. ^1 ^2 ^3Dealroom News, "Google DeepMind offshoot nears $4bn valuation just a month after founding" (summary of the Financial Times report), September 17, 2026.
8. ^1 ^2 ^3Briefs, "Ex-DeepMind Team Seeks $700 Million Seed for Emulate" (summary of the Bloomberg report), September 17, 2026.
9. ^1 ^2Maya Dharampal-Hornby and Freya Pratty, "Index in talks to lead $500m round for DeepMind researcher's new AI lab", Sifted, August 12, 2026.
10. ^Maya Dharampal-Hornby, "DeepMind researcher's new AI lab nears $4bn valuation, reports say", Sifted, September 17, 2026.
11. ^1 ^2 ^3The Cap Table, "Emulate nears a $700m seed at a $3.7bn valuation one month after leaving DeepMind", September 18, 2026.
12. ^1 ^2 ^3Sofia Chesnokova, "DeepMind's Genie creators launch Emulate, seeking $700M to build world models for robotics and simulation: Report", TechFundingNews, September 17, 2026.
13. ^PYMNTS, "Weeks-Old AI Startup Emulate Approaches $4 Billion Valuation", September 17, 2026.
14. ^1 ^2 ^3 ^4 ^5 ^6 ^7ChainCatcher, "AI startup Emulate reached a valuation of several billion dollars within three months of its establishment, with investors competing to raise the price in the same round of financing" (summary of a Bloomberg report published the same day), October 22, 2026.
15. ^1 ^2 ^3Janet Harrison, "VCs are splitting AI funding rounds into tranches to inflate valuations", Startup Fortune, October 22, 2026.
16. ^1 ^2 ^3Emulate, "Emulate" (company website), accessed October 27, 2026.
17. ^1 ^2 ^3Jake Bruce, Michael Dennis, Ashley Edwards, Jack Parker-Holder, Yuge Shi et al., "Genie: Generative Interactive Environments", arXiv:2402.15391, February 23, 2024.
18. ^International Conference on Machine Learning, "ICML 2024 awards", accessed October 27, 2026.
19. ^1 ^2 ^3 ^4 ^5 ^6Jack Parker-Holder, Philip Ball, Jake Bruce et al., "Genie 2: A large-scale foundation world model", Google DeepMind, December 4, 2024.
20. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9Jack Parker-Holder and Shlomi Fruchter, "Genie 3: A new frontier for world models", Google DeepMind, August 5, 2025.
21. ^Diego Rivas, Elliott Breece and Suz Chambers, "Project Genie: Experimenting with infinite, interactive worlds", Google, January 29, 2026.
22. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9Jack Parker-Holder, "Jack Parker-Holder" (personal research page), accessed October 27, 2026.
23. ^1 ^2 ^3 ^4Philip J. Ball, "Philip J. Ball: Research Scientist @ Google DeepMind" (personal research page), accessed October 27, 2026.
24. ^1 ^2 ^3Leverhulme Centre for the Future of Intelligence, "Matthew McGill", accessed October 27, 2026.
25. ^Philip Ball, Jack Parker-Holder, Aldo Pacchiano, Krzysztof Choromanski and Stephen Roberts, "Ready Policy One: World Building Through Active Learning", arXiv:2002.02693, February 7, 2020.
26. ^Philip J. Ball, Cong Lu, Jack Parker-Holder and Stephen Roberts, "Augmented World Models Facilitate Zero-Shot Dynamics Generalization From a Single Offline Environment", arXiv:2104.05632, April 12, 2021 (ICML 2021).
27. ^Jack Parker-Holder, Minqi Jiang, Michael Dennis, Mikayel Samvelyan, Jakob Foerster, Edward Grefenstette and Tim Rocktäschel, "Evolving Curricula with Regret-Based Environment Design", arXiv:2203.01302, March 2, 2022.
28. ^Philip J. Ball, Laura Smith, Ilya Kostrikov and Sergey Levine, "Efficient Online Reinforcement Learning with Offline Data", arXiv:2302.02948, February 6, 2023 (ICML 2023).
29. ^Debarghya Das, "The Ultimate Neolab List", accessed October 27, 2026.
30. ^Emulate, Inc., "Emulate: Organ-Chips for Research & Development", accessed October 27, 2026.