Rental costs for Nvidia H100, H200, B200 and B300 GPUs will increase on Oct 1 across both on-demand and dedicated inference endpoints. On-demand H100 rates rise from $3.85 to $4.50 per hour, while H200, B200 and B300 prices increase by approximately 20%. Separate price hikes for Token Factory’s dedicated inference endpoints range from 16% to 20%, with the B300 endpoint moving from $8.10 to $9.70 per hour.

Nebius will transition its preemptible virtual machine capacity to a dynamic spot pricing model on Oct 8, replacing previous fixed discounts with real-time rates based on regional demand and available capacity. This shift allows users to set a maximum price limit or follow the market rate for spare capacity. The rate increases for on-demand and managed services mark the second price hike for the provider since May, though existing contracted prices for long-term customers remain unaffected.