Three former DeepMind researchers who built an artificial intelligence system capable of defeating professional poker players have quietly adapted the same reinforcement learning technology to trade stocks — and early results suggest the gambit is succeeding. Their Prague-based laboratory, EquiLibre Technologies, has reached a valuation of $500 million after closing a Series A financing round led by Creandum, which the venture firm describes as the single largest investment it has ever made in a company. The common thread between the poker table and the trading floor is reinforcement learning, a training technique in which self-learning models are driven by rewards. And as EquiLibre CEO Martin Schmid puts it, the scoring system in financial markets is brutally simple: how much money did the agent make?
From DeepStack to Wall Street
The three founders — Schmid, chief technology officer Rudolf Kadlec, and chief scientific officer Matej Moravcik — first made their mark in the AI research world as visiting PhD students at DeepMind’s inaugural international AI research office in Edmonton, Alberta. While there, they built DeepStack, the first AI program to defeat human professionals at no-limit Texas hold ’em, a milestone that demonstrated how reinforcement learning could handle imperfect information scenarios where bluffing and incomplete knowledge are central. After Alphabet shut down the Edmonton office in 2023, the trio decided to return to their home country, Czechia, to build EquiLibre. They drew on a deep pool of talent from the Czech diaspora at Google and other tech companies, telling friends and former colleagues: “We are moving back to Prague, do you want to join us?” That strategy helped the startup assemble an initial team in 2022 and reach its current headcount of 25 people.
How Reinforcement Learning Bridges Poker and Trading
The question most observers ask is straightforward: what does beating humans at poker have to do with beating the stock market? The answer lies in the structure of both domains. Poker is a game of incomplete information — players cannot see each other’s cards, must infer hidden states, and make sequential decisions under uncertainty. Financial markets share the same essential character: traders operate with imperfect information, prices reflect the collective actions of many participants, and every decision carries a probabilistic outcome. Reinforcement learning excels in precisely these conditions. The AI agent explores an environment, takes actions, receives feedback in the form of rewards or penalties, and iteratively refines its strategy. In poker, the reward is chips won. In trading, the reward is return on capital. The mathematical framework is nearly identical. EquiLibre’s algorithms were first deployed on cryptocurrency markets in 2025, where they began demonstrating consistent profitability. The startup then expanded into equity markets in partnership with Tower Research Capital, a quantitative trading firm, where its systems now handle billions of dollars in daily trading volume across the S&P 500 and Nasdaq.
A Track Record of Zero Negative Months
EquiLibre claims that its trading agents have delivered a perfect record of zero negative months since inception, meaning the portfolio has finished each calendar month with net gains. This is the kind of performance metric that grabs the attention of quantitative hedge funds and venture investors alike. Cameron Sellers, vice president at Creandum, noted that the total addressable market for algorithmic trading in global financial markets is one of the largest on earth, and that many funds over the years have generated profits that make most venture-backed successes appear modest by comparison. The size of the Series A round was not disclosed, but the $500 million valuation represents a significant leap from EquiLibre’s previous seed round, which Dealroom data indicates was a $10 million raise led by Blossom Capital at a $140 million valuation. Pre-seed backers included Credo, a CEE-focused venture firm that has also backed ElevenLabs and UiPath.
A Lab First, Not a Finance Firm
Despite operating in one of the most profit-driven sectors of the global economy, EquiLibre’s founders are explicit about their identity. Sellers confirmed that the startup defines itself as a lab first, not a finance firm. Schmid himself is candid about his motivations. He does not come from a finance background, and he is not driven by a desire to make markets more efficient. “I’m doing this because we are all excited about building new things that have never been built before, and this is a lot of fun to build,” he said. This distinction matters because it shapes how the company allocates resources and prioritizes research. Unlike a pure trading firm that might optimize exclusively for short-term returns, EquiLibre is structured to pursue longer-term AI research questions, with the trading application serving as both a testbed and a revenue engine. The founders’ academic roots run deep. They worked closely with professors who are now part of the startup’s advisory board, including Rich Sutton, who received the Turing Award in 2024 for his foundational contributions to reinforcement learning.
Prague as a Talent Advantage
The decision to base EquiLibre in Prague rather than in a traditional AI hub like San Francisco or London might seem counterintuitive, but Schmid argues it has become a strategic asset. The Czech capital offers a stable talent pool in an industry where poaching is rampant. “It’s much easier to keep the good people here, because there’s not a new sexy AI thing happening every two months,” he said. The company is not alone in the city. BottleCap AI, another notable startup, operates out of the same building. But EquiLibre is emerging as one of the region’s most significant AI employers by research caliber. The next phase of growth involves scaling compute infrastructure: the company plans to bring online what it expects will be one of the largest computing clusters in Central and Eastern Europe.
The Competitive Landscape and the Jane Street Factor
EquiLibre is entering a field where the incumbents are formidable. Jane Street, one of the world’s most profitable quantitative trading firms, has stated publicly that it already uses reinforcement learning in combination with large language models and any other technique required to train effective models. The firm also claims to operate tens of thousands of high-end GPUs, a compute advantage that most startups cannot match. EquiLibre is pursuing a different approach: squeezing more performance out of fewer chips and doing more with less. “Because we started four years back, we believe we are ahead,” Schmid said, referencing the skepticism that reinforcement learning faced in trading applications when EquiLibre was founded. That skepticism has faded. RL has become the standard approach in many quantitative strategies. Still, the risk of being leapfrogged by better-resourced competitors is real. Jane Street recently reported a sharp jump in trading revenue, pulling in $16.1 billion in the first quarter of 2026, a figure that underscores the scale of the players EquiLibre is up against. Schmid, however, sees the market differently than the poker table. “This is not a winner-takes-all market,” he said. And he may be right. The global equity and derivatives markets are vast, fragmented, and deep enough to accommodate multiple successful algorithmic strategies operating simultaneously.
The Broader Frontier AI Race
EquiLibre is part of a broader wave of startups founded by DeepMind alumni that are attracting significant venture capital interest. Another recent example is Ineffable Intelligence, which raised $1.1 billion to build AI systems that learn without human data. Most of these ventures are based in the United Kingdom, but EquiLibre is a notable exception, anchoring its operations in Central Europe. The startup’s trajectory also reflects a larger shift in how reinforcement learning is perceived. When the founders began their work, RL was viewed by many in finance as an exotic academic curiosity. Today it is a standard tool in the quant arsenal. The evolution of DeepStack from a poker-playing novelty to the engine behind billions of dollars in daily trading volume is one of the more compelling narratives in applied AI. EquiLibre’s next challenge is to scale its research and infrastructure without losing the research-first culture that has defined its early years. If the company can maintain its edge in reinforcement learning while expanding its compute capacity and fending off competition from much larger players, it may well achieve its stated ambition of becoming the AI lab in trading. And even in a market that is not winner-takes-all, being recognized as the definitive research leader would be a victory that rivals anything the founders accomplished at the card table.
