EquiLibre Technologies: The Ex-DeepMind Poker AI Researchers Now Trading Billions on Wall Street

EquiLibre Technologies, a Prague-based AI lab founded by three ex-DeepMind researchers, has reached a $500 million valuation by applying reinforcement learning algorithms originally designed for poker to financial trading. Its partnership with quantitative trading firm Tower Research Capital now powers billions of dollars in daily trading volume. Here is a fact-checked account of the company's story, its claims, and the significant unknowns that remain.

The Founders and Their Poker AI Pedigree

EquiLibre Technologies was founded by Martin Schmid, Matej Moravčík, and Rudolf Kadlec — three Czech researchers who met during their PhD studies at Charles University and the University of Alberta . They were the lead authors of DeepStack, the first AI to beat professional humans at heads-up no-limit poker (2017), and later co-authored Player of Games (also known as "Student of Games") at DeepMind, a unified algorithm for both perfect and imperfect-information games . Google acquired their DeepStack team in 2017, and all three joined DeepMind in London before leaving in 2022 to start EquiLibre .

From Imperfect-Information Poker to Financial Trading

The core insight behind EquiLibre is that financial markets are structurally analogous to imperfect-information games like poker — not like chess or Go, where all information is visible on the board. In both poker and trading, participants act with hidden information, must manage risk and bluffing dynamics, and need to reason about what other agents believe .

EquiLibre's system uses reinforcement learning and counterfactual regret minimization (CFR) — the same family of algorithms that powered DeepStack — to train AI agents that place short-term directional bets on equities and ETFs . Schmid said the founders realized their poker work was "a perfect foundation for a new kind of trading algorithm" .

The Creandum-Led Series A and $500M Valuation

In mid-2026, EquiLibre announced it had raised a Series A round led by Creandum, the Swedish venture capital firm known for early bets on Spotify and Klarna . The round valued the company at $500 million . Critically, the exact size of the Series A has not been disclosed — neither the company nor Creandum has released the dollar amount . Creandum's vice president Cameron Sellers confirmed to TechCrunch that the investment was the largest single check the firm has ever written in one go into a company . Previous reporting from PitchBook and CB Insights shows EquiLibre had raised roughly $21.3 million across earlier seed and venture rounds prior to the Series A .

The Tower Research Capital "Software Vendor Agreement"

EquiLibre operates under a Software Vendor Agreement (SVA) with Tower Research Capital, a quantitative proprietary trading firm . An SVA is a profit-sharing arrangement in which the external team (EquiLibre) licenses its software and algorithms to Tower, gets access to Tower's global exchange connections and infrastructure, and splits profits with Tower — typically reported as a 50/50 split — while retaining its own intellectual property and brand autonomy .

According to TechCrunch and Tower's own announcement, EquiLibre's algorithms are now trading billions of dollars in daily volume across the S&P 500 and Nasdaq through this partnership . The startup first tested its prototype on crypto markets before transitioning to U.S. equities around 2025 .

The "Perfect Record" Claim

EquiLibre claims its trading agents have delivered "a perfect record of zero negative months since inception" — meaning every calendar month since live trading began has been profitable . This is a striking claim. In a TechCrunch interview, co-founder Martin Schmid acknowledged that while the firm is "still in the early stages," the track record so far is a point of pride . The claim covers both the crypto and equity phases, though the exact time window is not specified.

Skepticism and Open Questions

Several significant caveats remain:

• Undisclosed Series A size — The round size is confidential, so it is impossible to independently verify whether the $500 million valuation is backed by strong fundamentals or was achieved with a relatively small primary investment .
• "Zero negative months" is unaudited — The startup's own reported performance has not been verified by a third-party auditor .
• Very short live track record — EquiLibre only moved to live equities trading after crypto testing, and a few months or even a couple of years of profitable months is statistically thin for a trading strategy .
• The poker-to-markets analogy is unproven at scale — While academically sound, applying CFR algorithms to real-time, high-frequency, multi-asset trading with transaction costs, liquidity constraints, and regime changes is far harder than in a simulated poker game. No independent study has validated EquiLibre's approach over a full market cycle, including a major downturn .
• Transparency — The company does not publicly disclose detailed performance metrics, Sharpe ratios, drawdowns, or the number of strategies running .
• Prague talent market — While the founders are clearly world-class, the company is building its AI research hub in Prague, a smaller talent pool than London or New York, though it has reportedly been hiring senior AI researchers from abroad .

Bottom Line

EquiLibre Technologies has one of the most credible founding teams in AI-driven trading, a real revenue-generating partnership with a major quant firm, and a remarkable self-reported track record. But the undisclosed Series A size, short live history, and lack of audited third-party performance mean the story remains one of high promise with significant unknowns — not yet a proven disruption of quantitative finance.