Prague-based AI lab EquiLibre Technologies valued at $500M after applying poker AI to quant trading

EquiLibre Technologies, a Prague-based AI lab founded by three former DeepMind researchers, has been valued at $500 million after raising an undisclosed Series A. The round was led by Creandum, which described the investment as the largest single commitment the firm has ever made to a single company.

The startup’s technology applies reinforcement learning—an AI training technique where self-learning models are incentivized by rewards—to trading. EquiLibre’s algorithms have been trading billions in daily volume across the S&P 500 and Nasdaq in partnership with quant firm Tower Research Capital, the company said. EquiLibre claims its systems have maintained a perfect record of zero negative months since inception, including after expanding from crypto markets in 2025 to stock exchanges.

EquiLibre’s founders—CEO Martin Schmid, CTO Rudolf Kadlec, and CSO Matej Moravcik—previously built DeepStack, the first AI program to defeat professional players at no-limit poker. Their work originated during their time as visiting PhD students at DeepMind’s Edmonton office, which Alphabet closed in 2023.

The company positions itself as “a lab first, not a finance firm,” emphasizing the engineering challenge over financial outcomes. Schmid stated the motivation is building new things rather than market efficiency, noting the founders lack finance backgrounds.

EquiLibre’s Series A valuation of $500 million represents a significant increase from its prior $140 million valuation in a $10 million seed round led by Blossom Capital, according to Dealroom data. The startup also previously raised two additional funding rounds with pre-seed backing from Credo, a CEE-focused VC that backed ElevenLabs and UiPath.

The company plans to scale its compute infrastructure, aiming to deploy one of the largest compute clusters in Central and Eastern Europe. Despite competition from firms like trading giant Jane Street, which claims to use reinforcement learning with large language models and tens of thousands of high-end GPUs, EquiLibre says it aims to achieve more with fewer chips.