Skip to main content
Starting October 1, 2025, GPUs, vCPUs and RAM of each virtual machine and standalone application with NVIDIA B200, H200 or H100 GPUs are charged together rather than separately. This change becomes visible in your usage details and prices on October 22, 2025, with all usage from October 1 onwards recalculated retroactively. This does not affect your costs. What you pay per hour stays the same.

​
What changes

Before October 1, 2025, usage of virtual machines (VMs) and standalone applications is recorded under separate billing items for GPUs, vCPUs and RAM. Starting October 22, VMs and applications that use NVIDIA B200, H200 and H100 GPUs appear under new, unified billing items. Each of these new items covers GPU, vCPU and RAM usage. The usage of VMs and applications with the listed GPUs from October 1–21 is also recalculated and shown under the new billing items. The prices of the new items are calculated to ensure that the hourly rate of using a VM or an application from October 22 stays the same as before.
For example, the cost of using a VM or an application with one NVIDIA H200 GPU (preset 1gpu-16vcpu-200gb) for one hour before October 22 is calculated like this: 1 (GPU) × $2.668 + 16 (vCPUs) × $0.012 + 200 (GiB RAM) × $0.0032 = $3.50 Starting October 22, the unified hourly price of using the same VM or application is $3.50. This price is also applied retroactively to usage from October 1–21.
For pricing details, see Compute pricing in Nebius AI Cloud and Standalone Applications pricing in Nebius AI Cloud. In the Billing section of the web console, the Usage, Commitment discounts (under Discounts) and List prices pages show three lines (GPU, vCPU, RAM) for usage and prices before October 1 and one unified line for usage and prices starting October 1. When you view usage or prices for a period that includes October 1, all four lines appear.
For example, if your VM with eight NVIDIA H200 GPUs (preset 8gpu-128vcpu-1600gb) ran for 100 hours in September 2025 and then for another 100 hours in October 2025, this is how the web console shows this usage, starting October 22:
  • September 2025
    ProductUsageUnitTotal
    NVIDIA® H200 NVLink with Intel Sapphire Rapids. GPUGPU hour$
    NVIDIA® H200 NVLink with Intel Sapphire Rapids. CPUCPU hour$
    NVIDIA® H200 NVLink with Intel Sapphire Rapids. RAMGiB hour$
    Grand total$2800.00
  • October 2025
    ProductUsageUnitTotal
    NVIDIA® H200 NVLink with Intel Sapphire RapidsGPU hour$
    Grand total$2800.00
When you view usage across September and October, the web console shows all four lines. If you view October usage before October 22, the web console shows three lines: for GPU, vCPU and RAM billing items.

​
What does not change

The changes do not affect how much you pay for using VMs and applications. Your effective hourly rate stays unchanged. The changes do not apply to the following resources and usage in Compute and Standalone Applications:
  • VMs and applications with NVIDIA L40S GPUs
  • VMs and applications without GPUs
  • Storage resources (volumes)
  • Usage before October 1, 2025

​
See also