Buzz HPC’s £220 Million Deal Puts Canada’s Sovereign AI Push on Nvidia Racks
Buzz HPC signed a three-year sovereign AI contract with Bell Canada and Cohere, with 2,304 Nvidia Grace Blackwell GPUs planned for Bell Canada’s Merritt data center.
Canada’s AI Factory Deal Gets a Compute Layer
Hive Digital Technologies subsidiary Buzz HPC has signed a three-year sovereign AI contract with Bell Canada and Cohere, moving Canada’s AI infrastructure debate from policy language to a named data-center deployment.
The deal is valued at £220 million and centers on GPUs that Buzz HPC will provide to Cohere inside Bell Canada’s facility in Merritt, British Columbia.
The customer use case is specific.
The Merritt cluster is intended to give Cohere Canadian compute for foundation-model work and AI products aimed at public-sector and enterprise buyers.
The announcement is more concrete than a general cloud-capacity pledge: it names the infrastructure operator, the telecom and data-center partner, the model company, the site, and the intended customer base.
The Rack Details Are Specific
Buzz HPC’s hardware order covers 2,304 Nvidia Grace Blackwell GPUs configured around GB200 NVL72 rack-scale systems.
Nvidia Quantum InfiniBand networking links the systems, giving the project a clear high-performance computing profile rather than a vague “AI cloud” label.
The financing trail is also visible.
The GPUs were bought using proceeds from Hive’s April 2026 $115m convertible note financing round.
Aydin Kilic, Hive’s CEO, expects the cluster to go live in late 2026 or early 2027 and add approximately $70m of annualized revenue.
Plans for the 6.5MW Merritt deployment were disclosed in March 2026.
Those figures give the project measurable operating markers.
Buzz HPC has to turn purchased GPUs, Bell Canada data-center capacity, and Cohere demand into a running cluster on the disclosed timeline.
The annualized revenue figure is meaningful only if the deployment enters service, maintains enough utilization, and supports Cohere workloads without slipping into another infrastructure announcement cycle.
Sovereignty Claim Still Has Operating Questions
Frank Holmes, executive chairman of Hive Digital Technologies, framed the partnership as a way to commercialize Canadian AI talent with domestic industrial infrastructure.
Craig Tavares, Buzz HPC’s president and COO, tied the stack to Bell’s national platform, Cohere’s enterprise AI solutions, Hypertec’s Canadian-built GPU servers, and Nvidia’s full-stack AI infrastructure.
The operating case remains unfinished.
Buzz HPC has not named Cohere workload volumes, government customers, corporate customers, utilization commitments, or service-level terms.
Hive/Buzz is already working with Bell Canada on several deployments, including earlier Nvidia AI infrastructure work in Manitoba, and Hive is also planning a 320MW data center in the Greater Toronto area.
For now, the Merritt project gives Canada’s sovereign AI push a concrete cluster to judge.
The key test is whether the late 2026 or early 2027 go-live happens, whether Cohere runs production workloads for government and corporate customers in Canada, and whether Hive can turn the cluster into the approximately $70m annualized revenue Kilic described.
