Why did HIVE Digital Technologies work with Columbia University?

HIVE Digital Technologies collaborated with Columbia University to evaluate the performance of its GPU infrastructure in Paraguay for artificial intelligence workloads. The research focused on training large language models using HIVE Digital Technologies' A40 GPU cluster.

What did the Columbia University study find about HIVE Digital Technologies?

Researchers found that HIVE Digital Technologies' optimised A40 GPU systems matched Nvidia H100 performance after normalising for each platform's raw hardware capabilities. The study was based on large language model pretraining and benchmark testing.

How was the HIVE Digital Technologies research conducted?

Columbia University researchers remotely operated HIVE Digital Technologies' GPU infrastructure in Asunción, Paraguay while working from New York. The team conducted artificial intelligence training experiments and tested a 1.4 billion parameter language model.

What does the Paraguay expansion mean for HIVE Digital Technologies?

HIVE Digital Technologies said the study provides a benchmark for its planned 100-megawatt infrastructure expansion in Yguazú, Paraguay. The company expects its new substation to be energised in September 2026, with a Tier-III data centre targeted for service in the second half of 2027.

How is HIVE Digital Technologies expanding beyond Bitcoin mining?

HIVE Digital Technologies has been developing artificial intelligence and high-performance computing services alongside its Bitcoin (CRYPTO) mining operations. The company said the Columbia University collaboration supports its strategy of growing GPU computing infrastructure as an additional business segment.

Image for illustrative purposes only. Not a real photo.

HIVE cites Columbia study on Paraguay GPUs

Written by Mahathir BayenaPublished Jun 25, 2026, 1:10 AM

HIVE Digital Technologies said a Columbia University research project validated the performance of its GPU infrastructure in Paraguay.
Researchers reported that optimised A40 GPU systems matched the performance of Nvidia H100 chips after adjusting for hardware capabilities.
The findings support HIVE's plans to expand its artificial intelligence infrastructure business in Paraguay.

HIVE Digital Technologies said a research collaboration with Columbia University's Department of Industrial Engineering and Operations Research demonstrated the effectiveness of its GPU infrastructure located in Asunción, Paraguay.

The study involved researchers in New York remotely running artificial intelligence training workloads on HIVE's A40 GPU cluster in Paraguay and focused on neural network pretraining techniques used in large language models.

“In our use case of pretraining LLMs of up to 1.4B parameters, our results match those of H100s after normalising for each hardware's raw performance,” said a Columbia University researcher from the Department of Industrial Engineering and Operations Research.

The research, which has been submitted to the NeurIPS machine learning conference, also evaluated throughput and latency using a 1.4 billion parameter language model and standard LLaMA benchmarks.

HIVE said the results provide a performance benchmark for its planned 100-megawatt infrastructure expansion in Yguazú, Paraguay, and following the announcement there was no immediate share price reaction reported.

The company currently operates data centres in Canada, Sweden and Paraguay and was originally established as a cryptocurrency mining business focused on using renewable energy sources for Bitcoin (CRYPTO:BTC) mining.

HIVE said a new substation in Paraguay is expected to be energised in September 2026, with construction of a Tier-III data centre planned to begin later in the year and commercial operations targeted for the second half of 2027.

At the time of reporting, Bitcoin price was $60,743.34.

HIVE cites Columbia study on Paraguay GPUs

Frequently asked questions

Assets related to this article