Nebius
6.7/10Nebius is an AI-focused cloud that rents NVIDIA GPU instances and clusters by the hour, with managed Kubernetes, Slurm, storage and inference services for ML teams.
Data last updated November 2026 · Our methodology
Overview
Nebius is the AI infrastructure cloud run by Nebius Group, the Amsterdam-based, Nasdaq-listed company that took over Yandex N.V.'s international business. Its main product is NVIDIA GPU capacity. You can rent single GPU virtual machines, from L40S and RTX PRO 6000 up to HGX H100, H200, B200 and B300, or large GB200 and GB300 NVL72 clusters connected over InfiniBand. Around this compute it offers managed Kubernetes, Soperator for Slurm on Kubernetes, shared filesystems including WEKA, S3-compatible object storage, managed PostgreSQL, managed MLflow, a container registry and Serverless AI endpoints and jobs. Pricing is pay as you go in USD, charged per GPU-hour or per CPU-hour. On-demand rates run from $1.55 for an L40S to $9.50 for a B300, preemptible capacity starts at $0.79 to $0.99, and multi-month cluster commitments are priced through sales. Ingress, egress and public IPs are free, and the first payment must be at least $25. Teams can manage everything through a CLI, a Terraform provider, gRPC and REST APIs, and SDKs for Go, Python and JavaScript. The compliance list is long and includes SOC 2 Type II, ISO 27001 and ISO 42001, plus alignment with NIS 2 and DORA. Data centers are in Finland, Paris and the US, which matters to EU buyers with data-residency needs. Nebius suits ML researchers, AI startups and enterprises that train or serve models. It is not a general web host: there is no WordPress hosting, CDN or broad platform catalogue. Users also report slower support and gaps in the documentation for advanced setups.
Our Rating
Hosting & Infrastructure6.8/10
AI Software6.6/10
Who Nebius is best for
Good fit
ML engineers get self-serve GPU VMs, a CLI, Terraform and APIs.
AI startups get pay-as-you-go GPUs, cheap preemptible capacity and credit programs.
Large GB200 and GB300 clusters and a wide set of certifications suit big AI programs, though the service catalogue is narrower than a hyperscaler's.
Less suited
It runs AI infrastructure that needs DevOps skills and offers no simple web hosting.
Pros & Cons
Pros
- Newest NVIDIA hardware, including B300 and GB300 NVL72, is listed with published per-hour prices.
- Network egress, ingress and public IPs are free, which removes a common hidden cost on hyperscalers.
- The certification list is broad: SOC 2 Type II, SOC 3, several ISO standards including 27001 and 42001, and CSA STAR.
- EU data centers in Finland and Paris help with GDPR and data residency.
- Infrastructure as code is fully supported through a CLI, a Terraform provider, gRPC and REST APIs, and SDKs.
Cons
- The catalogue is narrow outside AI. The only managed database is PostgreSQL, and there is no CDN or web hosting.
- Some users report slower support responses than AWS or CoreWeave.
- Users note documentation gaps for advanced setups.
- Kubernetes and Slurm workflows assume DevOps expertise.
- Commitment pricing and the largest clusters require contacting sales.
Key Features
Not included
- Managed WordPress
- Shared Hosting
Nebius Pricing
On-demand (pay as you go)
- L40S from $1.55 per GPU-hour
- Minimum first payment of $25
- HGX B200 at $8.50 per GPU-hour
- HGX B300 at $9.50 per GPU-hour
- HGX H100 at $4.50 per GPU-hour
- HGX H200 at $5.40 per GPU-hour
- RTX PRO 6000 at $1.80 per GPU-hour
- Free ingress, egress and public IPs
- CPU instances from $0.06 per vCPU-hour
Preemptible
- Prices are dynamic and can change every 15 minutes
- Spot-style GPU capacity from $0.79 to $0.99 per GPU-hour
Reserved clusters / GB200 and GB300 NVL72
- Contact sales
- Multi-month commitment discounts
Frequently Asked Questions
Network egress and public IPs are free. Object Storage Standard egress costs $0.015 per GiB.
You pay as you go in USD, charged per GPU-hour or per CPU-hour, with storage charged per GiB-month. The first payment must be at least $25. Companies can pay by bank transfer, and individuals pay by card.
Yes. Preemptible GPUs start at $0.79 to $0.99 per GPU-hour, and the price can change every 15 minutes. Multi-month cluster commitments come with discounts negotiated through sales.
Its trust center lists SOC 2 Type II, SOC 3, ISO 27001, 27018, 27701, 27799, 22301 and 42001, and CSA STAR Level 1. It also states alignment with GDPR, NIS 2 and DORA.
Yes. Nebius provides a CLI, a Terraform provider, gRPC and REST APIs, and SDKs for Go, Python and JavaScript.
Wikipedia lists an owned site in Mäntsälä, Finland, a GPU cluster in Paris, and US sites in Kansas City and Vineland, New Jersey, which were under construction at the time. Customers can choose the region where their data is stored.
Similar Tools
See all alternatives
Amazon's public cloud platform, with over 200 pay-as-you-go services for compute, storage, databases, networking and AI, used by startups through large enterprises and governments.
Google's public cloud platform providing pay-as-you-go virtual machines, Kubernetes, serverless, databases, storage, networking, data analytics and AI services for businesses of every size.
Northflank is a Kubernetes-based developer platform for deploying services, jobs, databases and GPU workloads on its managed cloud or on customers' own cloud accounts.