HIVE Digital Technologies Ltd. (NASDAQ: HIVE) has launched its first operational AI cloud deployment in Paraguay, marking a step in its strategy to expand beyond cryptocurrency mining into high-performance computing infrastructure. The company said its BUZZ AI Cloud platform in Asunción is now running live GPU workloads, with an academic research team from Columbia University using the system remotely from New York.
The deployment represents the first active GPU cluster under HIVE’s broader plan to integrate AI and high-performance computing capabilities into its renewable energy footprint in South America. The infrastructure is housed in a Tier III data center operated by a local telecommunications provider and is designed to support intensive AI model training and inference workloads.
First live deployment under Paraguay expansion strategy
HIVE’s move into AI computing builds on its existing presence in Paraguay, where it already operates facilities powered by hydroelectric energy. The company said the new GPU cluster is intended to serve as a proof of concept for delivering large-scale AI compute across long distances, linking North American users with infrastructure based in South America.
The Asunción site is supported by Paraguay’s abundant renewable energy capacity and a nationwide fiber network, which together provide the power and connectivity required for data-heavy computing tasks. The company indicated that future expansion will depend on demand for AI cloud services and available capital, with additional Tier III facilities planned in Yguazú.
Management views the deployment as an initial step toward building a scalable AI infrastructure platform in the region, leveraging cost advantages in energy and growing connectivity to attract international workloads.
Columbia University research team tests AI workloads
The BUZZ Cloud platform is currently being used by a research team from Columbia University to conduct large language model (LLM) training experiments. The collaboration is non-commercial and aimed at generating performance data that will help guide HIVE’s future infrastructure decisions.
The team is working on pre-training foundation models, starting with smaller architectures ranging from 200 million to 2 billion parameters, and scaling toward larger models exceeding 8 billion parameters. Their work includes optimizing algorithms to improve efficiency and reduce computational and memory requirements.
“This deployment marks our first live GPU compute workload in Asuncion and provides the real-world performance data we need to guide our Tier-III expansion roadmap,” said Frank Holmes, Executive Chairman of HIVE.
Early testing has shown efficiency improvements compared to standard training methods, with researchers evaluating results using metrics such as loss, perplexity, and downstream task performance. The team’s research on LLM reasoning has also been accepted for publication in a peer-reviewed journal.
According to HIVE, running these workloads from New York to Paraguay allows the company to measure latency, throughput, and system performance under real-world conditions. These insights are expected to inform the company’s roadmap for scaling commercial AI infrastructure through 2027.
Paraguay positioned as emerging AI infrastructure hub
HIVE’s expansion reflects a broader view that Paraguay could become a competitive location for AI and high-performance computing infrastructure. The country offers a combination of low-cost renewable electricity and expanding telecommunications networks, which are critical for supporting large-scale data processing.
The company currently has access to approximately 300 megawatts of hydroelectric power in Paraguay, with additional capacity under development. This energy base, combined with enterprise-grade connectivity, provides a foundation for hosting compute-intensive workloads originating from outside the region.
Paraguay’s recent economic growth and relatively stable policy environment have also supported foreign investment in infrastructure projects. HIVE indicated that these factors could contribute to increased demand for AI computing services across South America.
“Having a research team from Columbia University running LLM training workloads on HIVE’s infrastructure in Asunción is a powerful validation of what we are building,” said Aydin Kilic, President and CEO of HIVE.
The company said it plans to use insights from the current deployment to refine its approach and determine how quickly to scale additional capacity, depending on customer demand and market conditions.
Strategy shifts from mining to AI and HPC
HIVE was originally established as a digital asset mining company, focused on using renewable energy to support cryptocurrency operations. The company is now expanding its business model to include GPU-based computing services, reflecting growing demand for AI infrastructure globally.
The transition highlights a broader industry trend in which companies with access to large-scale energy resources are repurposing infrastructure to support AI workloads. For HIVE, the integration of AI computing into its existing operations represents a shift toward more diversified revenue streams.
The company’s leadership has emphasized a phased approach, using initial deployments like the Asunción cluster to validate performance before committing to larger-scale investments.
About HIVE Digital Technologies
Founded in 2017, HIVE Digital Technologies builds and operates data centers powered by renewable energy across Canada, Sweden, and Paraguay. The company focuses on both digital asset mining and high-performance computing, including GPU-based AI infrastructure, with an emphasis on environmentally sustainable operations.
More details about the announcement are available via the source.
