Galaxy Digital reported a net loss of $295 million for the first quarter of 2025. The losses were cited due to digital asset depreciation and a $57 million charge related to winding down Helios’s data center mining operations. The company also confirmed that its reorganization and domestication to the United States was completed on May 13, ahead of a planned Nasdaq listing on May 16. The first-quarter loss equates to $0.86 per diluted share. As of March 31, Galaxy’s equity capital stood at $1.9 billion, with cash and net stablecoins totaling $1.1 billion. Despite the Q1 loss, the firm estimated operating income of $160–$170 million so far in Q2, with equity capital increasing to $2.2 billion as of May 12. You might also like: Robinhood acquires WonderFi to drive Canadian expansion CoreWeave expanded capacity In a key infrastructure update, AI infrastructure firm CoreWeave exercised its option to expand capacity at Galaxy’s Helios data center campus. The expansion—Phase II of the existing partnership—adds 260 megawatts of critical IT load for AI and high-performance computing workloads. Galaxy Digital’s 15-year AI hosting agreement with CoreWeave was announced during its Q4 2024 earnings call, marking a potentially transformative move for the company. The deal involved repurposing 200 MW of its Helios mining facility in West Texas to support 133 MW of critical IT load for CoreWeave’s GPUs. It was expected to generate $4.5 billion in revenue over its term, averaging around $300 million annually, which represented 70% of Galaxy’s total 2024 revenues. The terms were considered “very favorable” compared to other high-performance computing deals, with estimated annual revenue of $2.26 million per MW at approximately 90% EBITDA margins. Deliveries under the new agreement are expected to begin in 2027. Combined with an earlier 133 MW lease agreement signed in March, CoreWeave’s total committed capacity at Helios will reach approximately 393 MW. You might also like: Dow dips after Monday’s rally as CPI slows and S&P 500 outlook improves
