EquiLibre Technologies is trying to turn the poker AI that beat professional players into a trading engine, and investors have now valued the Prague startup at $500 million.
The clean version of this story is almost too neat: three former DeepMind researchers beat no-limit Texas hold'em, then took the same kind of reasoning into stock and crypto markets. You should be careful with stories that tidy. Markets are not poker tables. But the connection here is real enough to matter.
Martin Schmid, Matej Moravcik and Rudolf Kadlec were visiting PhD students at Google DeepMind's Edmonton research office when they helped build DeepStack, the 2017 system described as the first AI to beat professional players at no-limit Texas hold'em. That game mattered because the machine couldn't see the whole board of information. It had to act through bluffing, hidden hands and decisions that changed after every bet. It didn't solve a toy problem. It solved a messy one.
That is why EquiLibre is worth watching. According to TechCrunch's reporting on the funding, the startup's algorithms have been trading billions in daily volume across the S&P 500 and Nasdaq, after first trading in crypto markets from 2025. The company also claims it has had zero negative months since inception. That's the kind of claim you don't simply applaud. You look for the audit trail, the counterparty, the drawdown history and the capital actually at risk.
The funding says investors are willing to look. Blossom Capital led EquiLibre's seed round at a $140 million valuation. Creandum has now led the Series A at a $500 million valuation, and partner Cameron Sellers told TechCrunch it was the largest single check the firm has written. The exact round size wasn't disclosed. That missing number matters, but the valuation jump still tells you something plain: European venture capital is no longer treating AI trading systems as a clever research side project.
Frankly, the interesting part isn't that EquiLibre uses AI. Everyone says that now. The interesting part is the kind of AI. Most machine learning in trading still begins with old market data: feed in price sequences, search for patterns, predict the next move, then hope the pattern survives contact with live money. Reinforcement learning works differently. The agent learns by acting and by getting punished for bad decisions.
That difference is exactly what made DeepStack useful in poker. A system trained only on old hands would learn what players used to do. A system trained through repeated play has to deal with opponents who adjust. Markets are full of those opponents, even when they don't sit across from you with chips in front of them. Other funds see the same prices, chase the same signals and erase easy edges as soon as they become visible.
The Tower Research partnership announced in 2024 was the first public sign that EquiLibre had moved beyond an academic story. Tower is one of the serious names in high-frequency trading, and it doesn't need charity projects from former DeepMind researchers. If an outside team is contributing signals there, the only question that matters is whether the trades work after costs, slippage and competition from everyone else trying to do the same thing.
That doesn't make EquiLibre proven in the way a long-running audited fund is proven. It means the company has cleared a harder test than a demo day chart. There is a difference. A startup can show a beautiful backtest and still fail when spreads widen, liquidity disappears or the strategy starts moving the market against itself. A live trading relationship with Tower is more meaningful than a slide deck because somebody has to settle the trades.
Creandum's check now turns that proof into pressure. EquiLibre has to decide whether it remains an exclusive technology partner, adds more hedge fund clients, or eventually builds a fund structure around its own models. Each route changes the company. Selling signals is not the same business as managing capital, and managing capital is not the same business as running a research lab with venture funding behind it.
The strongest version of EquiLibre's argument is simple: if poker taught machines how to reason through hidden information, markets are the bigger table. The weaker version is also simple: finance has humbled plenty of beautiful models before. You don't need to choose between those views yet. You only need to notice that three researchers who once trained an AI to survive bluffs are now being valued as if that skill can survive live markets.
Also read: Etched bets $800 million that transformer silicon will outlast the GPU era • Bitcoin ETFs just posted their worst month on record and the buyers stepping in are not who you think • Bending Spoons prices its Nasdaq IPO above range as Wall Street bets on AI-powered software roll-ups
