# EquiLibre Technologies

> Source: https://aiwiki.ai/wiki/equilibre_technologies
> Updated: 2026-10-27
> Fact-checked: 2026-10-27
> Categories: AI Companies, Finance AI, Reinforcement Learning
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "EquiLibre Technologies." aiwiki.ai, 27 Oct 2026. https://aiwiki.ai/wiki/equilibre_technologies
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**EquiLibre Technologies** (usually shortened to EquiLibre) is an artificial intelligence company in Prague, Czech Republic, that trains [reinforcement learning](https://aiwiki.ai/wiki/reinforcement_learning) agents to trade on financial markets. Three former [DeepMind](https://aiwiki.ai/wiki/google_deepmind) researchers founded it around the turn of 2021 and 2022: Martin Schmid (CEO), Matej Moravčík (chief scientific officer) and Rudolf Kadlec (CTO).[1][2][3] Schmid and Moravčík were the joint lead authors of DeepStack, a poker program published in 2017 that beat professional players at heads-up no-limit Texas hold'em, and all three co-wrote the DeepMind game-playing algorithm first released as Player of Games.[4][5] The company calls itself a "frontier AI trading research lab" and says its models "trade billions of dollars every day."[3] TechCrunch reported in June 2026 that its algorithms trade S&P 500 and Nasdaq stocks in partnership with the quantitative trading firm Tower Research Capital.[6]

On 30 June 2026 EquiLibre announced a Series A round led by the Swedish venture firm Creandum. It did not say how much it raised; the round valued the company at about $500 million.[6][7][8] Creandum said the round was the largest single investment it had ever made in one company.[6] At that point EquiLibre had about 25 employees.[6]

## Overview

| | |
|---|---|
| Type | Private company. The Czech operating company, EquiLibre Technologies s.r.o., is wholly owned by EquiLibre Technologies, Inc., a Delaware corporation[2] |
| Founded | Czech company registered 2 December 2021 under a shelf name and renamed EquiLibre Technologies on 19 January 2022[2]; investors and later coverage give 2022 as the founding year[9][8] |
| Founders | Martin Schmid, Matej Moravčík, Rudolf Kadlec[3] |
| Headquarters | Národní 135/14, Prague 1, Czech Republic[2] |
| Focus | Reinforcement learning agents for algorithmic trading[3] |
| Trading partner | Tower Research Capital[6][10] |
| Employees | About 25 (June 2026)[6] |
| Main investors | Creandum, Blossom Capital, Credo Ventures, Miton, RockawayX, Richard Sutton[6][11][7] |
| Last reported valuation | About $500 million (Series A, closed January 2026, announced June 2026)[6][11] |
| Website | equilibre.ai[3] |

## History

### Research background

Schmid has said that he and his colleagues started building a poker AI in 2011, while they were students living in dormitories. In his account, they wanted their bots to earn money by playing online poker, which he described as illegal; the bot accounts kept being banned because they played 24 hours a day, and the project turned into the subject of their doctoral work.[12] Schmid's doctoral thesis at Charles University, "Search in Imperfect Information Games," was supervised by Milan Hladík with Michael Bowling as advisor and posted to arXiv in November 2021.[13] Kadlec earned his doctorate at Charles University in 2015 with a thesis on DyBaNeM, a Bayesian model of episodic memory for intelligent virtual agents, supervised by Cyril Brom.[14]

DeepStack came out of the University of Alberta's Department of Computing Science in Edmonton. Its paper lists Moravčík and Schmid as equal-contribution first authors, with affiliations at the University of Alberta and Charles University's Department of Applied Mathematics, alongside Neil Burch, Viliam Lisý, Dustin Morrill, Nolan Bard, Trevor Davis, Kevin Waugh, Michael Johanson and Michael Bowling. Kadlec was not an author. The paper was published in *Science* in March 2017.[4] DeepStack combined recursive reasoning to deal with hidden information, decomposition to focus computation on the current decision, and a learned "intuition" trained with deep learning from self-play. In a study of 44,000 hands, it beat professional players at heads-up no-limit Texas hold'em with statistical significance.[4] The company and outlets such as Tech.eu describe DeepStack as the first AI to beat professionals at no-limit poker.[3][8] Carnegie Mellon University's Libratus, built by [Noam Brown](https://aiwiki.ai/wiki/noam_brown) and Tuomas Sandholm, was reported in *Science* the following year as beating top professionals at the same game.[15]

The three founders had also all worked at IBM.[1] At IBM Watson in Prague, Kadlec, Schmid, Ondřej Bajgar and Jan Kleindienst wrote "Text Understanding with the Attention Sum Reader Network" (ACL 2016), a reading-comprehension model that picks its answer directly from the context document; the authors reported that an ensemble of the models set a new state of the art on every dataset they tested.[16] A Czech Technical University biography of Schmid says he worked at IBM during his studies, interned at the University of Alberta, and then spent five years at Google DeepMind.[17] On the company's website, Schmid is listed as a former project lead and senior research scientist at DeepMind, Moravčík as a former DeepMind research scientist, and Kadlec as a former DeepMind tech lead and senior research engineer.[3]

At DeepMind the three co-authored more game-playing research with Bowling and others, including a 2018 DeepMind paper on variance reduction for Monte Carlo counterfactual regret minimization (VR-MCCFR).[18] Their best-known DeepMind project was Player of Games, posted to arXiv in December 2021. It was published in *Science Advances* in November 2023 under the title "Student of Games." The algorithm combines guided search, self-play learning and game-theoretic reasoning. According to the paper, it reached strong performance in chess and Go, beat the strongest openly available agent in heads-up no-limit Texas hold'em, and defeated the state-of-the-art agent in the hidden-information board game Scotland Yard. The published version lists EquiLibre Technologies, Prague, as an affiliation of Schmid, Moravčík and Kadlec.[5]

### Founding (2021 to 2022)

CNBC reported in April 2022 that Schmid, Kadlec and Moravčík had left DeepMind that January to set up EquiLibre, moving from Edmonton to Prague. "Our idea is that rather than playing poker, our algorithms will play algorithmic trading," Schmid told CNBC, adding: "We are also looking into crypto." He said the company would compete by building better algorithms than those already used in largely algorithmic markets. CNBC wrote that in the long run, the founders hoped either to base a new hedge fund on the AI or to sell it to a large institutional bank or another investor. The advisory board at the time included Michael Bowling, then head of DeepMind's Edmonton office, and [Richard Sutton](https://aiwiki.ai/wiki/richard_sutton). Schmid said EquiLibre had raised the largest seed round ever in the Czech Republic but would not give the amount.[1]

According to the Czech commercial register, the company's Czech entity was registered on 2 December 2021 as a shelf company called Pontymax Group s.r.o. On 19 January 2022 it was renamed EquiLibre Technologies s.r.o., EquiLibre Technologies, Inc. (Delaware) became its sole shareholder, and Schmid became its managing director (jednatel). Its registered office moved to Národní 135/14 in central Prague in February 2022.[2] Other sources give different founding dates: Tech Funding News gives December 2021 and The Recursive gives late 2021, while Creandum, Tech.eu and a Czech Technical University event page give 2022.[19][20][9][8][17]

Schmid told TechCrunch that the founders moved back to Czechia because they had many former colleagues there and a large Czech diaspora at Google and other companies: "These were our friends, so we told them, 'Hey, guys, we are moving back to Prague, do you want to join us?'" He also said it is "much easier to keep the good people here" than in San Francisco, "because there's not a new sexy AI thing happening every two months."[6] The company's March 2025 announcement said that in 2022 it had raised $6 million from the Czech investors Credo Ventures, Miton and RockawayX, which it and CzechCrunch described as the largest pre-seed round in Czech history.[21][10]

### Crypto prototypes and FTX

EquiLibre built its first prototype for cryptocurrency trading, and the company has said that the results gave it the confidence to move to traditional exchanges.[21] In an interview on CzechCrunch's Money Maker podcast, excerpted by CzechCrunch in August 2026, Schmid said the early prototypes were tested on the FTX crypto exchange and that FTX then offered to invest. He said he flew to the Bahamas in autumn 2022 and agreed terms in person with [Sam Bankman-Fried](https://aiwiki.ai/wiki/sam_bankman_fried), FTX's chief executive at the time, including a partnership to deploy EquiLibre's models across FTX. According to Schmid, after FTX collapsed EquiLibre quickly bought back the stake FTX had taken, and the collapse sped up the company's move from crypto to US stock exchanges.[12]

### Tower Research Capital and US equities

Moving to US markets required a partner that could connect EquiLibre's technology with US regulatory requirements for trading.[10] "Several of the largest trading firms took turns coming to Prague to convince us to work with them. In the end we chose Tower Research Capital from New York," Schmid said in March 2025 (translated from Czech). He told CzechCrunch that all of the interested firms were American.[10][21] In the same announcement, EquiLibre said it had raised another $10 million from the London venture firm Blossom Capital. It also said it had 20 employees, all working from its Prague office rather than remotely. Its hires included Agustin Lebron, a former Jane Street trader and quantitative researcher and the author of *The Laws of Trading*, and Ondřej Bajgar, who returned to Prague after a PhD at Oxford.[21]

Accounts of the deployment timeline differ slightly. Schmid said in 2026 that after a new strategic partnership gave it access to the US, the company deployed its models on "the largest, most complex and most liquid markets in the world" from January 2025 (translated from Czech), and that it had not had a single losing month since.[12] TechCrunch described the agents as having rolled out on crypto markets in 2025 before moving to stock exchanges.[6] Tech.eu reported the company's claim that in 2025 it became the first company to deploy reinforcement learning agents live on key financial markets.[8] The company's own March 2025 funding announcement said EquiLibre was then in the process of moving its technology to the US stock market, where the Tower partnership would let it test the product, and quoted Schmid as saying the company would not want to talk about results for a long time.[21] CzechCrunch wrote at the time that, if all went well, EquiLibre's AI could soon trade stocks, derivatives and other financial instruments.[10]

### Series A (2026)

Schmid told CzechCrunch that the Series A closed at the end of January 2026.[11] The round was made public on 30 June 2026, when TechCrunch reported that EquiLibre was valued at $500 million after an undisclosed Series A led by Creandum.[6] SiliconANGLE and Tech.eu described the valuation as more than $500 million, and The Recursive gave it as €438 million.[7][8][20] Neither EquiLibre nor Creandum disclosed the size of the round, and the company did not give its total funding to date.[6] Creandum vice president Cameron Sellers told TechCrunch that it was the largest single investment the firm "has ever made in one go into a company."[6] Creandum's portfolio page names Staffan Helgesson as the firm's lead for the investment.[9]

Sutton, who received the [Turing Award](https://aiwiki.ai/wiki/turing_award) for 2024,[6] is among EquiLibre's investors, according to SiliconANGLE and Tech.eu; Tech.eu called him one of the company's earliest backers.[7][8] The company said most of the new money would go to computing infrastructure: TechCrunch reported that it planned to bring online what it expects to be one of the largest compute clusters in Central and Eastern Europe. The rest is to be spent on hiring deep learning researchers and engineers.[6][7]

## Technology and approach

EquiLibre's website lists three parts of its approach. The first is reinforcement learning: "self-learning agents that learn to trade just like they would learn to play poker, chess or go." The second is trading itself, where the company says, "Our success is all about the quality of the models and algorithms we develop." The third is [continual learning](https://aiwiki.ai/wiki/continual_learning), since its agents "need to adapt quickly to the ever changing environment and other market participants."[3] According to The Recursive, the agents are trained on historical and live market data and then trade on their own.[20] SiliconANGLE wrote that EquiLibre carried over key technologies from DeepStack, with particular emphasis on self-play, a training method in which a model learns by playing against itself.[7]

Schmid has explained why markets suit reinforcement learning: "The nice thing about trading and markets is that the scoring is super simple: how much money did the agent make?"[6] In the 2026 CzechCrunch podcast, he described the company as a frontier AI research lab rather than a trading firm. He said it trains its own [foundation models](https://aiwiki.ai/wiki/foundation_models) with a large amount of reinforcement learning, and then deploys them for fully automated, high-frequency trading. In his framing, the agent's actions are to buy or sell, its reward is money, and its input is the state of the exchange. He said the models do not read financial statements, news or social media posts and analyze only the sequence of events on the exchange. He added that the team does not know exactly why the network makes a particular decision: while performance keeps improving, it prefers to spend its effort on making it improve faster rather than on explaining it.[12]

Schmid has said EquiLibre wants to "get more from less" out of far fewer GPUs than large trading firms such as Jane Street, which states that it has "tens of thousands of high-end GPUs."[6]

## Funding

| Round | Date | Amount | Investors | Valuation |
|---|---|---|---|---|
| Pre-seed | 2022 | $6 million[21] | Credo Ventures, Miton, RockawayX[21][10] | Not disclosed |
| Seed | Announced March 2025 | $10 million[21][10] | Blossom Capital (lead, according to Dealroom data cited by TechCrunch)[6][22] | $140 million according to Dealroom data cited by TechCrunch[6]; The Recursive gives €122.8 million[20] |
| Series A | Closed January 2026; announced 30 June 2026[11][6] | Not disclosed[6] | Creandum (lead)[6][9] | About $500 million ("more than $500 million" in some reports; €438 million per The Recursive)[6][7][20] |

Schmid said FTX also took a stake in 2022, which EquiLibre later bought back. The size of that stake has not been reported.[12] SiliconANGLE and Tech.eu name Richard Sutton as an investor. Tech.eu calls him one of the earliest backers, but neither outlet says which round he joined.[7][8]

## People

### Founders

| Name | Role | Background |
|---|---|---|
| Martin Schmid | CEO and co-founder; managing director of the Czech entity[3][2] | PhD in algorithmic game theory; formerly project lead and senior research scientist at DeepMind and project lead and research scientist at IBM; co-author of DeepStack and Player of Games[3][4][5] |
| Matej Moravčík | Chief scientific officer and co-founder[3] | PhD in algorithmic game theory; formerly research scientist at DeepMind and IBM; co-author of DeepStack and Player of Games[3][4][5] |
| Rudolf Kadlec | CTO and co-founder[3] | PhD in episodic memory modeling (Charles University, 2015); formerly tech lead and senior research engineer at DeepMind and senior research scientist at IBM; co-author of Player of Games[3][14][5] |

### Advisory board

The company website lists the following advisers:[3]

| Adviser | Description given by the company |
|---|---|
| [Richard Sutton](https://aiwiki.ai/wiki/richard_sutton) | Turing Award laureate, former distinguished research scientist at DeepMind, professor of computing science at the University of Alberta |
| Michael Bowling | Former site lead and research team lead of DeepMind's Edmonton office; professor at the University of Alberta |
| Csaba Szepesvári | Lead of DeepMind's reinforcement learning theory team; professor at the University of Alberta; principal investigator at the Alberta Machine Intelligence Institute |
| Michal Pěchouček | Former chief technology officer of Avast; president of the Czech Technical University in Prague |
| Murray Campbell | Distinguished research scientist at IBM Research; co-author of [Deep Blue](https://aiwiki.ai/wiki/deep_blue) |

## Other activities

Schmid gave a public talk at the Czech Technical University in April 2025 titled "From Poker AI to Algorithmic Trading: Game Theory and Reinforcement Learning in Action."[17] EquiLibre's careers page promotes a bot-programming competition for generals.io, which it organizes with ÚFAL, the Institute of Formal and Applied Linguistics at Charles University. The competition had a sprint stage and a final "marathon" stage, with a $10,000 first prize in the final.[23][24][25] TechCrunch noted that EquiLibre shares a building in Prague with another AI startup, BottleCap AI.[6]

## Reception and analysis

Several outlets noted that EquiLibre's performance claims could not be checked independently. TechCrunch reported the company's statement that its agents had "a perfect record of zero negative months since inception."[6] The Recursive and Tech Funding News both pointed out that EquiLibre had not published performance data or methodology to support that statement; Tech Funding News added that the period covered and drawdown figures were also missing.[20][19] Tech Funding News also called the company's claim to be the first AI lab to use reinforcement learning live on US markets "hard to confirm."[19] Descriptions of its trading volume vary as well. The company and most outlets say billions of dollars a day, while SiliconANGLE wrote "billions of dollars worth of trade per month."[3][6][7]

TechCrunch wrote that the startup could be overtaken by better-resourced competitors, citing Jane Street's statements about using reinforcement learning and its GPU fleet. Schmid's response: "This is not a winner-takes-all market."[6] Sellers argued that the addressable market is "one of the biggest on earth" and said EquiLibre defines itself as "a lab first, not a finance firm." Schmid told TechCrunch he was not motivated by making markets efficient: "I'm doing this because we are all excited about building new things that have never been built before."[6] TechCrunch also placed EquiLibre among a group of AI startups founded by DeepMind alumni and attracting large venture rounds, alongside [Ineffable Intelligence](https://aiwiki.ai/wiki/ineffable_intelligence).[6]

Debarghya Das's "The Ultimate Neolab List" includes EquiLibre among its frontier-lab entries but marks it as a borderline case: "Pure research team calling itself 'a lab first', but end use is proprietary trading and compute needs are modest."[26]

## See also

- [List of neolabs](https://aiwiki.ai/wiki/list_of_neolabs)
- [Neolab](https://aiwiki.ai/wiki/neolab)
- [AI in finance](https://aiwiki.ai/wiki/ai_in_finance)
- [Pluribus (poker AI)](https://aiwiki.ai/wiki/pluribus)
- [Google DeepMind](https://aiwiki.ai/wiki/google_deepmind)
- [Reinforcement learning](https://aiwiki.ai/wiki/reinforcement_learning)

## References

1. Shead, Sam. ["Three former DeepMinders are developing A.I. to pick stocks and crypto"](https://www.cnbc.com/2022/04/05/deepmind-alum-want-to-make-an-ai-that-can-pick-stocks-and-crypto.html). CNBC, 5 April 2022.
2. Ministry of Justice of the Czech Republic, Public Register. ["EquiLibre Technologies s.r.o., C 359182, full extract (úplný výpis)"](https://or.justice.cz/ias/ui/rejstrik-firma.vysledky?subjektId=1142137&typ=UPLNY) (in Czech). Data valid as of 26 October 2026.
3. EquiLibre Technologies. ["EquiLibre Technologies" (official website)](https://equilibre.ai/). Accessed 27 October 2026.
4. Moravčík, Matej; Schmid, Martin; Burch, Neil; Lisý, Viliam; Morrill, Dustin; Bard, Nolan; Davis, Trevor; Waugh, Kevin; Johanson, Michael; Bowling, Michael. ["DeepStack: Expert-Level Artificial Intelligence in Heads-Up No-Limit Poker"](https://arxiv.org/abs/1701.01724). *Science* 356 (6337): 508-513, 2017. doi:10.1126/science.aam6960.
5. Schmid, Martin; Moravčík, Matej; Burch, Neil; Kadlec, Rudolf; et al. ["Student of Games: A unified learning algorithm for both perfect and imperfect information games"](https://arxiv.org/abs/2112.03178). *Science Advances* 9, eadg3256, 2023. First posted to arXiv as "Player of Games," 6 December 2021.
6. Heim, Anna. ["The DeepMind trio who built a poker AI are now making money for quant hedge funds"](https://techcrunch.com/2026/06/30/the-deepmind-trio-who-built-a-poker-ai-are-now-making-money-for-quant-hedge-funds/). TechCrunch, 30 June 2026.
7. Deutscher, Maria. ["AI stock trading startup EquiLibre raises funding at a $500M+ valuation"](https://siliconangle.com/2026/07/01/ai-stock-trading-startup-equilibre-raises-funding-500m-valuation/). SiliconANGLE, 1 July 2026.
8. Reynolds, John. ["Ex-DeepMind researchers land record Creandum funding to scale AI agents for Nasdaq"](https://tech.eu/2026/07/01/ex-deepmind-researchers-land-record-creandum-funding-to-scale-ai-agents-for-nasdaq/). Tech.eu, 1 July 2026.
9. Creandum. ["EquiLibre" (portfolio company page)](https://creandum.com/commitments/equilibre). Accessed 27 October 2026.
10. Magalhães, Filip. ["Jejich umělá inteligence vítězila v pokeru. Se stamiliony v kapse jdou Češi změnit obchodování na burze"](https://cc.cz/jejich-umela-inteligence-vitezila-v-pokeru-se-stamiliony-v-kapse-jdou-cesi-zmenit-obchodovani-na-burze/) (in Czech). CzechCrunch, 6 March 2025.
11. Magalhães, Filip. ["Naučili AI vítězit v pokeru, teď mění obchodování na burze. Hodnota české firmy přesáhla 10 miliard"](https://cc.cz/naucili-ai-vitezit-v-pokeru-ted-meni-obchodovani-na-burze-hodnota-ceske-firmy-presahla-10-miliard/) (in Czech). CzechCrunch, 2 July 2026.
12. Holzman, Ondřej. ["Jejich první AI vyhrávala v pokeru a koupil ji Google. Teď chtějí Češi porazit giganty z Wall Street"](https://cc.cz/jejich-prvni-ai-vyhravala-v-pokeru-a-koupil-ji-google-ted-chteji-cesi-porazit-giganty-z-wall-street/) (in Czech; excerpts from the Money Maker podcast). CzechCrunch, 2 August 2026.
13. Schmid, Martin. ["Search in Imperfect Information Games"](https://arxiv.org/abs/2111.05884) (doctoral thesis, Charles University, Prague). arXiv:2111.05884, 10 November 2021.
14. Kadlec, Rudolf. ["DyBaNeM: Bayesian Model of Episodic Memory"](https://artemis.ms.mff.cuni.cz/main/papers/DyBaNeM-thesis.pdf) (doctoral thesis). Charles University in Prague, Faculty of Mathematics and Physics, 2015.
15. Brown, Noam; Sandholm, Tuomas. ["Superhuman AI for heads-up no-limit poker: Libratus beats top professionals"](https://doi.org/10.1126/science.aao1733). *Science* 359 (6374): 418-424, 2018.
16. Kadlec, Rudolf; Schmid, Martin; Bajgar, Ondrej; Kleindienst, Jan. ["Text Understanding with the Attention Sum Reader Network"](https://arxiv.org/abs/1603.01547). ACL 2016. arXiv:1603.01547.
17. Artificial Intelligence Center, Czech Technical University in Prague. ["Guest talk by Martin Schmid"](https://www.aic.fel.cvut.cz/events/guest-talk-by-martin-schmid). Event of 24 April 2025.
18. Schmid, Martin; Burch, Neil; Lanctot, Marc; Moravcik, Matej; Kadlec, Rudolf; Bowling, Michael. ["Variance Reduction in Monte Carlo Counterfactual Regret Minimization (VR-MCCFR) for Extensive Form Games using Baselines"](https://arxiv.org/abs/1809.03057). arXiv:1809.03057, 9 September 2018.
19. Chesnokova, Sofia. ["Prague's EquiLibre raises Series A at $500M valuation to scale AI trading agents"](https://techfundingnews.com/equilibre-series-a-500m-ai-trading-agents-creandum/). Tech Funding News, 1 July 2026.
20. Stanislavova, Yoana. ["EquiLibre Technologies Raises Series A at €438M Valuation to Scale AI Trading Agents"](https://www.therecursive.com/equilibre-technologies-raises-series-a-at-eur438m-valuation-to-scale-ai-trading-agents/). The Recursive, 1 July 2026.
21. EquiLibre Technologies (press release via FeedIT.cz). ["Český AI startup Equilibre získal od investorů stovky miliónů korun a jde na americkou burzu (IPO to ale ještě není)"](https://feedit.cz/2025/03/06/cesky-ai-startup-equilibre-ziskal-od-investoru-stovky-milionu-korun-a-jde-na-americkou-burzu-ipo-to-ale-jeste-neni/) (in Czech). 6 March 2025.
22. Blossom Capital. ["EquiLibre Technologies" (portfolio page)](https://www.blossomcap.com/portfolio/equilibre-technologies). Accessed 27 October 2026.
23. EquiLibre Technologies. ["Careers"](https://equilibre.ai/hiring). Accessed 27 October 2026.
24. Generals Competition. ["Generals Competition: generals.io bot competition"](https://www.generals.bot/) (presented by Equilibre and ÚFAL). Accessed 27 October 2026.
25. ÚFAL. ["Institute of Formal and Applied Linguistics"](https://ufal.mff.cuni.cz/). Charles University. Accessed 27 October 2026.
26. Das, Debarghya. ["The Ultimate Neolab List"](https://debarghyadas.com/neolabs/). Snapshot of 27 October 2026.