A research project run by Columbia University professors on GPU servers located more than 5,000 miles away in Paraguay has produced a result that HIVE Digital Technologies (Nasdaq: HIVE) is c
A research project run by Columbia University professors on GPU servers located more than 5,000 miles away in Paraguay has produced a result that HIVE Digital Technologies (Nasdaq: HIVE) is calling a proof of concept for the future of distributed AI infrastructure.
HIVE Digital, a publicly listed data center and digital asset mining company founded in 2017 and traded on both the TSX and Nasdaq, announced on June 22 that it has completed its first AI research collaboration with Columbia University's Department of Industrial Engineering and Operations Research.
The research has been submitted to the Conference on Neural Information Processing Systems (NeurIPS), one of the three most influential machine learning conferences in the world, alongside ICLR and ICML.
Related: HIVE Digital acquires 32 MW Swedish data center after 8 years as tenant
The Columbia team ran iterative AI training experiments remotely on HIVE's A40 GPU cluster located in Asunción, Paraguay, while working from New York City.
The study focused on neural network pretraining, the foundational process by which large AI models learn from data before being fine-tuned for specific tasks.
The headline finding is significant. After two months of code optimisation specifically tailored to HIVE's A40 hardware, the Columbia researchers found that performance matched that of Nvidia's H100 GPUs, a newer and considerably more expensive generation of chip widely regarded as the gold standard for AI training workloads.
The comparison was made after normalising for each hardware platform's raw performance.
The team also tested serving throughput and latency on a 1.4 billion parameter language model, running standard benchmarks on LLaMA models as part of the study.
"In our use case of pretraining LLMs of up to 1.4B parameters, our results match those of H100s after normalizing for each hardware's raw performance," a Columbia researcher from the IEOR department said.
"This work advances our understanding of modern neural network optimization, including matrix-aware optimizers such as Muon and related scale-invariant methods."
The Asunción results now serve as a performance baseline for a larger infrastructure project HIVE is building in Yguazú, Paraguay. The company has a 100 megawatt substation under construction there, with civil works already complete.
The substation is expected to be energised in September 2026, with construction on a new Tier-III data centre beginning in the autumn and a ready-for-service date targeted for the second half of 2027.
HIVE's bet on Paraguay is rooted in the country's energy profile — cheap, abundant hydroelectric power that makes it an attractive location for power-intensive computing workloads.
HIVE executive ehairman Frank Holmes said, "Paraguay has the power, the strategic location, and now the proof point."
"High-performance computing does not need to be limited by geography. Our vision is to bring the right power, data center design, software stack, and execution, so through HIVE, Paraguay can participate directly in the global AI economy."
HIVE was among the first publicly listed companies to mine Bitcoin using green energy when it was founded in 2017. Today it operates Tier-I and Tier-III data centres across Canada, Sweden, and Paraguay, serving both Bitcoin mining clients and high-performance computing customers.
The company has previously collaborated with Intel Corporation on custom mining hardware and became one of Sweden's largest demand-response participants, helping balance the national electrical grid.
The Columbia collaboration represents HIVE's clearest step yet into AI infrastructure as a commercial business line, moving beyond crypto mining into the broader market for GPU compute.
"Innovation is in our DNA," HIVE President and CEO Aydin Kilic said. "We see a bright future for AI in Paraguay."
Kilic added that great engineering can unlock significant value. HIVE will continue to invest in communities while bringing advancements in our tech stack to our data centers globally, he added.
Related: Roundtable 100: Frank Holmes Makes the Case for HIVE Digital Technologies
