Canadian tech giants launch $300M sovereign AI infrastructure deal
Bell Canada will host the operations from its data centre in Merritt, B.C.

Key Takeaways:
- Bell, Cohere, Hypertec, and BUZZ HPC have signed a $220 million sovereign AI infrastructure contract to deploy advanced computing capabilities entirely within Canada.
- The infrastructure will feature 2,304 NVIDIA Grace Blackwell GPUs housed at Bell’s hydro-powered, liquid-cooled data centre facility in Merritt, B.C.
- The unified platform allows Canadian corporate and government entities to run large-scale Cohere language models under strict domestic data sovereignty laws.
The Whole Story:
Bell Canada, Cohere, Hypertec, and BUZZ High Performance Computing (HPC) announced a major infrastructure agreement today to construct and deploy advanced artificial intelligence workloads on sovereign Canadian infrastructure. Anchored by a three-year GPU cloud contract valued at $300 million CAD, the massive collaboration represents a major development in keeping domestic data processing, intellectual property, and enterprise models entirely within Canadian borders.
The partnership integrates four distinct layers of the domestic technology sector into a unified, high-performance computing stack:
- The Facility Layer: Bell Canada will host the operations from its data centre in Merritt, British Columbia. Part of the broader Bell AI Fabric network, this five-acre, seven-megawatt facility is powered entirely by hydro-electric energy and uses closed-loop liquid water cooling to maximize efficiency.
- The Compute Layer: BUZZ HPC, a subsidiary of Vancouver-based HIVE Digital Technologies, will deploy a sovereign cloud utilizing a massive cluster of 2,304 NVIDIA Grace Blackwell GPUs.
- The Hardware Layer: Quebec-based server manufacturer Hypertec will handle the custom system integration, hardware assembly, and on-site engineering support for the complex supercomputing racks.
- The Intelligence Layer: Toronto-headquartered AI firm Cohere will use the secure platform to train and operate its large language models for enterprise and government clients.
By running Cohere’s enterprise-grade solutions on specialized hardware housed inside Bell’s secure network, the initiative solves a critical national gap for public and private sector organizations moving from AI experimentation to full production. Financial institutions and government agencies that require rigorous data isolation and operational oversight can now deploy generative AI tools without routing sensitive information through international server farms.
The high-density NVIDIA hardware deployment is scheduled to become fully operational between late 2026 and early 2027. To fund the upfront hardware procurement, HIVE Digital Technologies utilized a portion of the proceeds from a $115 million convertible note financing arrangement completed in April 2026. Once active, the cluster is projected to add roughly $70 million in annual recurring revenue to HIVE’s growing high-performance computing division.
“Canada has the talent and innovation to lead in AI—what’s been missing is the ambition to bring the right ingredients together. This landmark deal helps close that gap,” said Michel Richer, President of Bell AI Fabric.
Michael Pelosi, Country Manager for Cohere, added that the infrastructure gives enterprise clients complete transparency regarding exactly where their data is handled. The Honourable Evan Solomon, federal Minister of Artificial Intelligence and Digital Innovation, praised the agreement as a vital milestone for retaining critical digital assets, technical talent, and corporate economic growth at home in Canada.