HIVE pipes GPU compute 5,000 miles from Paraguay to Columbia University

HIVE Digital Technologies just ran an AI research project across two continents. The GPUs were in Paraguay. The researchers were in New York. Roughly 5,000 miles of fiber connected them. The company, listed on both the TSX and Nasdaq under the ticker HIVE, partnered with Columbia University’s Department of Industrial Engineering and Operations Research. The goal was to test whether serious AI training can happen when the hardware sits on a different continent from the people running it. What actually happened HIVE operates a cluster of Nvidia A40 GPUs in Asunción, Paraguay. Columbia researchers used that cluster remotely from New York to run pre-training workloads for large language models, or LLMs. The collaboration kicked off in March 2026 and wrapped up in June 2026. Over roughly two months, Columbia’s team worked on optimizing their code to squeeze more out of HIVE’s hardware. After those optimizations, HIVE’s A40 GPUs reportedly delivered performance comparable to Nvidia’s H100 chips under specific conditions tied to pre-training workloads of up to 1.4 billion parameters. The project also produced enough data to support a submission to the NeurIPS conference, one of the most p...