Argentum AI has signed a $2.5 billion agreement with cloud gaming and GPU infrastructure company Boosteroid and European real estate and digital infrastructure platform DL Invest Group to develop a 300MW AI data center in Bielsko-Biala, Poland. The deal, announced May 15, is one of the largest independent AI compute infrastructure deployments announced in Europe and represents a significant expansion of a facility that DL Invest Group and Boosteroid had previously indicated would reach only 50MW when they formed their joint venture in November 2025.
The agreement establishes a long-term framework for hyperscale AI infrastructure deployments supporting enterprise AI workloads, model training, and next-generation inference systems. Argentum AI will deploy institutional-scale GPU infrastructure within a new high-density platform designed specifically for large-scale compute operations. The project targets tens of thousands of next-generation GPUs, with future deployments of Nvidia GB300 systems included in the build plan.
What Each Party Is Bringing
Argentum AI is an independent, institutional-grade AI cloud platform focused on deploying large-scale GPU infrastructure for enterprise AI applications. Its involvement as the GPU infrastructure and compute layer positions it as the technology operator within a partnership where the other two parties contribute real estate, development, and existing operational GPU experience.
Boosteroid operates GPU infrastructure across 29 data centers in Europe, North America, and South America, with a background in high-performance, low-latency computing developed partly through its cloud gaming service. That operational experience across a distributed global footprint gives the new Bielsko-Biala facility a foundation in running GPU infrastructure at scale rather than commissioning it from scratch.
Why Bielsko-Biala and Why Poland
DL Invest Group, whose founder and CEO Dominik Leszczyński described the deal as one of the key projects supporting independent AI infrastructure development in Central and Eastern Europe, brings the real estate and development infrastructure. The Bielsko-Biala location sits in southern Poland, a market that has attracted growing data center investment given its access to European grid infrastructure, lower land costs than Western European markets, and Poland’s position as one of the fastest-growing data center markets on the continent.
The scale jump from the previously announced 50MW to 300MW is the most significant detail confirmed by Boosteroid directly to Data Center Dynamics. That is a sixfold increase in planned capacity for the same site, reflecting either a significantly expanded partnership scope or a deliberate decision to reframe the project’s ambitions following the Argentum AI agreement.
What the Deal Signals for European AI Infrastructure
The Argentum AI-Boosteroid-DL Invest deal forms part of a broader wave of large independent AI compute deployments across Central and Eastern Europe as operators seek alternatives to the saturated and increasingly expensive primary markets of Frankfurt, Amsterdam, London, and Dublin. Poland has attracted a growing share of this investment thanks to its grid access, land availability, and EU regulatory alignment, which increasingly position the country as a credible location for hyperscale AI deployment.
At $2.5 billion for 300MW, the deal implies a capital cost of approximately $8.3 million per megawatt, consistent with the mid-range of current AI data center construction cost benchmarks and reflecting a facility designed for high-density GPU operations with advanced cooling rather than conventional enterprise compute.
The deal also highlights a structural reality of European AI infrastructure development. Rather than building wholly owned campuses, hyperscalers increasingly rely on partnerships that bring together specialist GPU operators, real estate developers with permitted sites, and compute platforms with established enterprise customer pipelines. This three-party model spreads capital requirements across partners with complementary capabilities and accelerates the path from agreement to operational capacity compared with a greenfield project undertaken by a single operator from scratch.
