ComputeLabs Research
HIVE’s three-year Bell AI Fabric agreement covers 2,304 NVIDIA GB200 NVL72 GPUs, adding $75 million in contracted annual recurring revenue.
· ComputeLabs Research · from the August 17, 2026 edition
HIVE described its Bell AI Fabric arrangement as an approximately $225 million, three-year sovereign-AI agreement supporting Cohere. The agreement covers deployment of 2,304 NVIDIA GB200 NVL72 GPUs.
HIVE said the agreement adds $75 million in contracted annual recurring revenue, or ARR. Combined with approximately $35 million of GPU-cloud ARR already live, HIVE reported approximately $110 million of active and contracted GPU-cloud ARR.
The company is targeting approximately $200 million of GPU-cloud ARR by the fourth quarter of calendar 2026. HIVE explicitly made that target subject to market conditions and successful deployment, so it is a management target rather than contracted revenue already in service.
Separately, HIVE signed a letter of intent for long-term colocation at its 32-megawatt Big Boden facility in Sweden, representing approximately $45 million of potential annual colocation revenue. Because that item is an LOI rather than a definitive contract, HIVE’s cited approximately $155 million HPC pipeline includes potential, not entirely contracted, revenue.

