About Nebius

Nebius emerged from the restructuring of Yandex, taking the international infrastructure business and the engineers who built it, and refocusing entirely on AI compute. Its flagship data center is in Mäntsälä, Finland, with additional capacity in France and the United States.

The scale is the point. This is a company deploying tens of thousands of GPUs, which puts it in a different bracket from the boutique providers, and it competes for the same large training contracts as CoreWeave.

For a European organisation this is one of the few options offering serious H100 and H200 capacity with the data resident in the EU and a corporate structure to match.

Pricing

Published hourly rates put L40S around $1.20/hr, H100 SXM around $2.00/hr, and H200 around $2.60/hr, with meaningful discounts on committed terms. Reserved capacity for multi-month training is where most of their revenue sits and where the negotiation happens.

Service Details

Beyond raw GPU VMs there is managed Kubernetes, managed PostgreSQL, object storage with an S3-compatible API, and a managed Spark offering, so it functions as a general cloud rather than only a GPU rental desk. That matters when your training pipeline needs somewhere to put the data and something to orchestrate the jobs.

GPU clusters are connected with InfiniBand for multi-node training. There is a Terraform provider and a CLI.

Images are Linux, and the platform assumes a competent infrastructure team. If you want a notebook in a browser, Paperspace is a better fit; if you want a Finnish data center full of H200s, this is the list.

Available in: Finland, France, United States

Learn More

For more information about Nebius, visit their website at https://nebius.com/

Plan	Price/mo	CPU	RAM	Notes
L40S 48GB	—	16 vCPU	200GB	Around $1.20/hr
H100 SXM	—	20 vCPU	200GB	Around $2.00/hr
H200 SXM	—	20 vCPU	250GB	Around $2.60/hr