Skip to main content
This article provides pricing information for applications in Nebius AI Cloud. Pricing depends on the deployment option you choose.

​
How charges and prices work

Each group of chargeable items in this article has two time units associated with it:
  • Billing unit: The minimum unit of usage for which you can be charged.
  • Pricing unit: The unit of usage for which the prices are shown.
Charges for units smaller than the pricing unit are calculated proportionally.
For example, for GPUs on running VMs, the billing unit is 1 second, and the pricing unit is 1 hour (3600 seconds). For 30 minutes of usage, you will be charged half the hourly price.
Prices in US dollars (USD, $) apply to all customers except for companies from Israel, where prices in Israeli shekels (ILS, ₪) apply instead. All prices are shown without any applicable taxes, including VAT. Due to rounding errors, usage costs shown in the web console and final charges may slightly differ from calculations based on the prices in this article.

​
GPU prices from June 1, 2026

Starting June 1, 2026, prices for standalone applications with NVIDIA H200 and H100 GPUs are updated. The article lists prices before and from this date.

​
Standalone Applications

Applications deployed using the Standalone deployment option are charged for computing resources and storage.

​
Prices

Nebius AI Cloud charges you for the computing resources allocated to running applications and the storage size. Public access to applications is free of charge.

​
Computing resources

You are charged for computing resources in your running applications. You are not charged for computing resources in stopped applications.
  • Billing unit: 1 second
  • Pricing unit: 1 hour (3600 seconds)
The platform is available in the eu-north1 and us-central1 regions.
Charges are based on the resource preset that you are using. For example, using a 8gpu-128vcpu-1600gb application for 1 hour costs 8 × $4.50 = $36.00. For 730 hours (~ 1 month), this amounts to 730 × $36.00 = $26,280.00.
The platform is only available in the eu-north1 region.
Charges are based on the resource preset that you are using. For example, using a 8gpu-128vcpu-1600gb application for 1 hour costs 8 × $3.85 = $30.80. For 730 hours (~ 1 month), this amounts to 730 × $30.80 = $22,484.00.

​
NVIDIA® L40S PCIe with Intel Ice Lake

The platform is only available in the eu-north1 region.
Charges are based on the resource preset that you are using. For example, using a 1gpu-16vcpu-64gb application for 1 hour costs $1.35 + 16 × $0.012 + 64 × $0.0032 = $1.7468. For 730 hours (~ 1 month), this amounts to 730 × $1.7468 = $1275.164.

​
NVIDIA® L40S PCIe with AMD Epyc Genoa

The platform is only available in the eu-north1 region.
Charges are based on the resource preset that you are using. For example, using a 1gpu-16vcpu-96gb application for 1 hour costs $1.35 + 16 × $0.01 + 96 × $0.0032 = $1.8172. For 730 hours (~ 1 month), this amounts to 730 × $1.8172 = $1326.556.

​
Non-GPU AMD EPYC Genoa

The platform is available in the eu-north1, us-central1 and uk-south1 regions.
Charges are based on the resource preset that you are using. For example, using a 4vcpu-16gb application for 1 hour costs 4 × $0.025 + 16 × $0.005 = $0.18. For 730 hours (~ 1 month), this amounts to 730 × $0.18 = $131.40.

​
Non-GPU Intel Ice Lake

The platform is only available in the eu-north1 region.
Charges are based on the resource preset that you are using. For example, using a 4vcpu-16gb application for 1 hour costs 4 × $0.025 + 16 × $0.005 = $0.18. For 730 hours (~ 1 month), this amounts to 730 × $0.18 = $131.40.

​
Storage

You are charged for the disks used by deployed applications. Charges are based on disk sizes, regardless of the amount of used space.
  • Billing unit: 1 byte per 1 second
  • Pricing unit: 1 GiB per 730 hours (230 bytes per 2,628,000 seconds ~ 1 month)

​
Kubernetes

Applications deployed by using the Kubernetes deployment option run on Managed Kubernetes clusters. Applications themselves are not subject to their own charges. Only Managed Kubernetes charges apply to clusters and nodes that host the applications.

​
Virtual machine

Applications deployed by using the Virtual machine deployment option run on Compute virtual machines. Applications themselves are not subject to their own charges. Only Compute charges apply to the virtual machines that host the applications.