HIVE Merritt GB300 Cluster

HIVE Digital's BUZZ HPC unit announced on 17 August 2026 (the page header says 16 August) a five-year GPU cloud services contract worth about $350M with an unnamed investment-grade enterprise customer, backed by 2,016 NVIDIA Blackwell Ultra GPUs in GB300 NVL72 racks with NVIDIA Quantum-X800 InfiniBand and VAST Data storage on an NVIDIA reference architecture. HIVE says the cluster sits in Bell AI Fabric's Merritt, British Columbia facility, is liquid cooled in a closed loop, runs on hydroelectric power, and is due for delivery and go-live in calendar Q4 2026; HIVE keeps ownership of the hardware and states capex of about $185M including warranties and a customer deposit of about $35M. The release gives no MW figure for this cluster. Bell's 25 March 2026 release separately says BUZZ HPC secured 6.5 MW of gross capacity at the Merritt facility for all of its GPU clusters there, so that figure is a site allocation, not this cluster's draw. A different HIVE financing release from late October 2026 covers a 2,303-GPU GB200 cluster for Cohere that names no site, so it is not recorded here (see coreweave-cambridge-on for another Cohere compute site). Status stays under_construction until HIVE reports the GB300 cluster live; both pages are company or partner statements, so the tier is reported.