The Company
EquiLibre Technologies is a Prague-based AI lab founded by Martin Schmid (CEO), Rudolf Kadlec (CTO), and Matej Moravcik (CSO) — three former DeepMind researchers without traditional finance backgrounds. While at DeepMind, they built DeepStack, published in Science in 2017, the first AI program to defeat professional players at heads-up no-limit Texas hold'em. On June 30, TechCrunch and EU-Startups reported that EquiLibre had closed a Series A at a €438 million (~$500M) valuation led by Creandum, which confirmed this was the largest single investment the firm has ever made in one company.
Why Poker and Markets Are the Same Problem
EquiLibre's core architectural claim is that poker and market trading share the same underlying mathematical structure — imperfect-information sequential games where optimal strategy requires modeling what counterparties believe, not just reacting to observed data. Reinforcement learning techniques like counterfactual regret minimization and self-play, which powered their poker breakthrough, are naturally suited to this structure. The model doesn't predict the market; it finds strategies that are robust across all the opponent behavior it might encounter. This is a meaningful departure from the standard quant stack, which combines factor models, statistical arbitrage, and more recently LLM-assisted signal generation. EquiLibre's system is trained entirely through self-play — competing against itself across simulated conditions rather than fitting historical patterns that may not persist.
What They're Claiming
The headline number: zero negative months since going live on crypto markets in 2025, and subsequently on equities. That's an extraordinary track record at face value. The details that matter — AUM size, instrument liquidity, intra-month drawdown profiles — shape how quant allocators interpret it. Creandum's conviction, committing to the largest single check in the firm's history, suggests the due diligence found those answers satisfying. EquiLibre's advisory board adds credibility: Rich Sutton, who received the Turing Award in 2024 for foundational RL research, is an active advisor — a signal that the architectural claims are grounded in current research, not retrofitted for the pitch deck.
What This Means for Builders in Quantitative Finance
For fintech and quant infrastructure builders, the EquiLibre funding maps a concrete market shift. RL-based trading systems are now raising at serious valuations, which means institutional buyers are treating them as a credible category rather than an academic curiosity. The team-without-finance-background pattern is also worth noting: three researchers who understood the mathematical structure of the problem outperformed the conventional assumption that market expertise is the prerequisite. For teams building data infrastructure, backtesting platforms, or execution tooling for quant strategies, RL-based systems require different support architectures than factor models. They generate evaluation data through self-play rather than historical replay, and their performance attribution doesn't map to traditional alpha decomposition. Builders who understand that difference will be better positioned as this category scales.
Where This Goes
EquiLibre will not be the last team to route game-theoretic AI into financial markets. The techniques that solved poker are being explored for options pricing, market microstructure, and multi-player auction design across multiple research groups. The common thread: wherever imperfect information and strategic interaction define the problem, RL approaches trained on adversarial simulation can find strategies that pure statistical modeling misses. The €438 million bet on EquiLibre is an early signal that institutional capital has started taking those claims seriously at production scale.