Hive Digital Technologies has launched its AI cloud platform from a data center in Asunción, Paraguay.
It is already supporting research projects on large language models (LLMs) at Columbia University in the US.
The announcement marks a significant step in the company’s evolution toward advanced computing services, demonstrating its infrastructure’s ability to connect high-performance workloads between North and South America with low latency and operational efficiency.
The deployment builds on the infrastructure Hive has been developing in Paraguay in recent years, based on access to large-scale renewable hydroelectric power and the progressive construction of data centers.
Building on this foundation, the company has evolved from less complex facilities toward more advanced environments, including Tier III data centers equipped for artificial intelligence and HPC workloads, which house GPU clusters designed for model training and inference.
This approach responds to a clear strategy: to use energy-intensive infrastructure – traditionally linked to the digital mining business – as a starting point for building scalable computing capacity geared toward the AI economy.
The use case with Columbia University demonstrates Paraguay’s potential as a digital infrastructure platform for international workloads. The combination of abundant energy, competitive costs, and fiber connectivity is positioning the country as an emerging destination for high-performance computing deployments.
Furthermore, the HIVE model envisions progressive capacity growth based on demand, with the potential to significantly expand installed capacity and consolidate the country as a regional hub for AI and HPC.
This development comes alongside the global growth of the company’s high-performance computing business. Hive has already secured contracts worth up to $30 million for its GPU cloud platform, reinforcing the commercial viability of its commitment to AI.
The company plans to continue expanding its GPU deployment and international presence, relying on a hybrid model that combines its own data centers with third-party facilities to accelerate time-to-market.
This piece was automatically translated from DCD's Spanish site and edited by a member of DCD's editorial staff.
Comments