A research project conducted by Columbia University professors on GPU servers located more than 5,000 miles away in Paraguay has produced results that HIVE Digital Technologies (Nasdaq: HIVE) is seeking as a proof of concept for future distributed AI infrastructure.
HIVE Digital, a publicly traded 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 was presented at the Neural Information Processing Systems Conference (NeurIPS), one of the world’s most influential machine learning conferences along with ICLR and ICML.
What the research actually found
Working from New York City, the Colombian team remotely ran iterative AI training experiments on HIVE’s A40 GPU cluster in Asunción, Paraguay.
The research focused on pre-training neural networks, the fundamental process by which large-scale AI models learn from data before being fine-tuned for a specific task.
Headline discovery is important. After two months of specific code optimizations for HIVE’s A40 hardware, Columbia researchers found that it matched the performance of Nvidia’s H100 GPU, a newer and considerably more expensive generation of chips that is widely considered the gold standard for AI training workloads.
Comparisons were made after normalizing the raw performance of each hardware platform.
The team also ran standard benchmarks on the LLaMA model as part of their research, testing service throughput and latency on the 1.4 billion parameter language model.
“For the use case of pre-training an LLM with up to 1.4B parameters, the results match the H100 results after normalizing the raw performance of each hardware,” said Columbia researchers in the IEOR department.
“This research advances our understanding of modern neural network optimization, including matrix-aware optimizers such as Muon and related scale-invariant methods.”
Trending on TheStreet Roundtable:
- Wall Street Analysts Predict 400% Rise in MicroStrategy Stock
- Billionaire reveals Satoshi’s surprising proposal on Bitcoin
- Mark Zuckerberg eyes new $1 billion market
Why Paraguay and what happens next?
The Asunción results are currently serving as a performance baseline for a large infrastructure project that HIVE is building in Iguazu, Paraguay. The company is building a 100-megawatt substation there, and civil works have already been completed.
The substation is scheduled to be energized in September 2026, and construction of the new Tier-III data center will begin in the fall, with a target service date of late 2027.
HIVE’s bet on Paraguay is rooted in the country’s energy profile. Cheap and abundant hydropower makes it an attractive location for power-intensive computing workloads.
Frank Holmes, HIVE Executive Director, said: “Paraguay has power, it has a strategic location, and now it proves it.”
“High-performance computing doesn’t have to be limited by geography. Our vision is to provide the right power, data center design, software stack, and execution to enable Paraguay to participate directly in the global AI economy through HIVE.”
HIVE’s broader story
When founded in 2017, HIVE was one of the first publicly traded companies to use green energy to mine Bitcoin. We currently operate Tier-I and Tier-III data centers in 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, making it one of Sweden’s largest demand response participants and helping to balance the national power grid.
The collaboration with Columbia represents HIVE’s clearest step yet into AI infrastructure as a commercial business line, moving beyond crypto mining and into the broader market for GPU computing.
”Innovation is in our DNA” said Aydin Kilic, President and CEO of HIVE. ”We believe that the future of AI in Paraguay is bright. ”
Kilic added that good engineering provides great value. He added that HIVE will continue to invest in the community while bringing advancements in its technology stack to data centers around the world.