HIVE Digital Technologies has signed a three-year GPU cloud contract with Bell AI Fabric, a subsidiary of Bell Canada. The deal is valued at about $225 million. Once the hardware is running, HIVE expects it to bring in roughly $75 million in annual recurring revenue.
The contract runs through HIVE’s BUZZ High Performance Computing arm. It covers 2,304 NVIDIA GB200 GPUs at a site in Merritt, British Columbia. The facility uses renewable hydroelectric power. Deliveries are expected late 2026 or early 2027. That timing matters because the revenue target depends on the cluster being live.
The structure matters
HIVE owns the chips when the contract ends. That is the detail worth noting. To pay for the roughly $165 million cluster, HIVE secured a $129 million fixed-rate loan. The loan carries 9% interest over three years and was finalized in late October 2026. The loan term lines up with the contract term. Three years of customer payments against three years of debt service. It is a deliberate match.
The loan covers most of the equipment cost, not all of it. HIVE is using debt for the bulk of the purchase and letting the customer payments carry the load. If the plan works, HIVE finishes with the GPUs on its books and the debt repaid.
What to watch
Delivery delays are the near-term risk. A late cluster would push back revenue while the loan clock keeps running. The late 2026 to early 2027 window is the first checkpoint.
The bigger scorecard is HIVE’s $200 million GPU cloud annual recurring revenue target for the end of 2026. Contracted ARR at BUZZ HPC was about $185 million as of September 30, 2026. The gap will be closed by bringing new clusters online, I think, not by signing more paper.
Bigger GPU push
This is not HIVE’s first large GPU deal this year. In August 2026, it signed a five-year contract worth about $350 million, expected to add roughly $70 million in annual recurring revenue. Across 2026, HIVE has signed GPU agreements with about $600 million in total contract value.
HIVE still mines Bitcoin and runs data centers in Canada, Sweden, and Paraguay. The Bell deal shows how the company is trying to fund GPU growth with customer contracts. Perhaps the main test is simple: can it turn signed capacity into working capacity on time?
