1 hour ago
19
HIVE Digital Technologies just wrapped up something that sounds like it shouldn’t work: running AI training workloads on GPUs sitting in Paraguay, controlled by researchers in New York, over 5,000 miles away.
The project, a collaboration with Columbia University’s Department of Industrial Engineering and Operations Research, used HIVE’s A40 GPUs located in Asunción to handle remote large language model pre-training and optimization workloads. It ran for approximately two months, and the company says the results were good enough to submit a research paper to NeurIPS, one of the most prestigious machine learning conferences in the world.
Older GPUs, newer tricks
HIVE claims that after optimization, its A40 GPUs delivered performance metrics that matched or exceeded those of Nvidia’s H100 GPUs when normalized for raw hardware capabilities. The A40 is a workstation-class GPU that’s considerably cheaper and older than the H100, which has been the gold standard for AI training since its launch.
That’s a meaningful finding. While 1.4 billion parameters is small compared to frontier models from OpenAI or Anthropic, it covers a huge swath of practical AI applications. Fine-tuning, domain-specific models, and research workloads all frequently operate at this scale.
Paraguay’s unlikely AI play
HIVE’s GPU cluster in Paraguay went live in March 2026 as the company’s first AI deployment in the country, running on its BUZZ AI Cloud platform.
Paraguay generates massive amounts of renewable hydroelectric power, much of it from the Itaipú Dam, one of the largest hydroelectric facilities on the planet. HIVE’s initial capacity there sits at 300 megawatts. A new substation in Yguazú is targeting an additional 100 megawatts of capacity, with energization planned for September 2026. Civil works for that expansion have already been completed, with construction scheduled for fall 2026. Beyond that, HIVE has a new Tier-III data center slated to begin operations in the second half of 2027.
What this means for investors
HIVE trades on both the TSX Venture Exchange and NASDAQ under the ticker HIVE.
The bear case is equally clear. Model sizes are growing rapidly, and 1.4 billion parameters is relatively modest. The optimization gains that let A40s match H100s at that scale may not translate to larger models where the H100’s architectural advantages become more pronounced.
The NeurIPS submission timeline is worth tracking. Conference decisions typically come several months after submission, and acceptance would provide significant third-party validation of HIVE’s technical claims.
Investors should also watch the September 2026 substation energization deadline and the 2027 Tier-III data center timeline closely.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·