Nebius
- Company typePublic
- Founded2024
- HeadquartersSchiphol, Netherlands
- Headcount1,001–5,000
- GTM typeB2B
- OfferingSoftware
Nebius firmographics
Firmographics- Name
- Nebius
- Legal name
- Nebius Group N.V.
- Website
- https://nebius.com
- Company type
- Public
- Founded year
- 2024
- Operating status
- Operating
- Headcount range
- 1,001–5,000 employees
- Ownership category
- akta.pro rank
Nebius industry classification
Industry- Product category
- Cloud AI Infrastructure
- NAICS
- Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518)
- SIC
- Electronic Computers (3571)
- akta.pro primary industry
- GPU/Accelerator Compute Servers (HDABADAC)
- akta.pro secondary industry
- AI Compute Cloud & GPU-as-a-Service (HDAAAAAK)
Keywords
Where Nebius is headquartered
LocationHeadquarters
- HQ city
- Schiphol
- HQ country
- Netherlands
- HQ region
- Europe
Offices8 records
Markets served
Nebius business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Infrastructure, Supply Chain, Technology or R&D, Personnel, Operations, Marketing or Sales
Revenue model
- GPU compute (on-demand and preemptible): Consumption-based revenue from per-GPU-hour billing on NVIDIA H100, H200, B200, B300, GB300, L40S and RTX PRO 6000 instances; priced per-second with no long-term commitment required.
- Reserved/committed capacity: Multi-year reserved capacity with up to ~35% discount versus on-demand pricing, used by enterprises and AI labs needing predictable GPU access.
- Long-term hyperscale capacity contracts: Multi-year dedicated AI infrastructure contracts with prepayment, including a Microsoft contract through 2031 valued at approximately $17.4B and a Meta contract valued up to approximately $27B over five years.
- Managed inference (Token Factory): API-based managed inference billed per token for open-weight and custom LLMs running on Nebius GPU infrastructure.
- Storage and managed cloud services: Recurring revenue from Object Storage ($0.0147–$0.1100/GiB), Shared Filesystem ($0.0800/GiB-month), Managed Kubernetes, KMS, MLflow, and related cloud services.
- Professional services and customer support: Tiered 24/7 support and architecture services bundled with enterprise compute contracts.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Usage-based | Multi-year contract | NVIDIA GB300 NVL72 — quote-based flagship cluster |
| Usage-based | Pay-as-you-go | NVIDIA HGX B300 — $4.30 preemptible / $7.85 on-demand per GPU-hour |
| Usage-based | Pay-as-you-go | NVIDIA HGX B200 — $3.95 preemptible / $7.15 on-demand per GPU-hour |
| Usage-based | Pay-as-you-go | NVIDIA HGX H200 — $2.45 preemptible / $4.50 on-demand per GPU-hour |
| Usage-based | Pay-as-you-go | NVIDIA HGX H100 — $2.15 preemptible / $3.85 on-demand per GPU-hour |
| Usage-based | Pay-as-you-go | NVIDIA RTX PRO 6000 — $0.95 preemptible / $1.80 on-demand per GPU-hour |
| Usage-based | Pay-as-you-go | NVIDIA L40S — from $0.74–$0.90 preemptible / $1.55–$1.82 on-demand per GPU-hour |
| Usage-based | Pay-as-you-go | CPU-only instances — AMD EPYC Genoa from $0.10/hr, Intel Ice Lake from $0.05/hr |
| Usage-based | Pay-as-you-go | Object Storage — $0.0147/GiB (Standard) to $0.1100/GiB (Enhanced) |
| Usage-based | Monthly | Shared Filesystem — $0.0800/GiB-month |
| Hybrid | Multi-year contract | Reserved/committed capacity — up to ~35% discount vs on-demand |
Go-to-market motion6 records
Distribution channels6 records
Marketing channels10 records
Nebius product offering
Product offeringCore offering
Nebius operates a vertically integrated AI cloud platform (Nebius AI Cloud) that provides on-demand and reserved NVIDIA GPU compute (H100, H200, B200, B300, GB300, RTX PRO 6000) on custom hardware with non-virtualized GPUs and InfiniBand, together with managed Kubernetes, Soperator (Slurm on Kubernetes) orchestration, S3-compatible Object Storage, NVMe Shared Filesystem, Token Factory managed inference, ModelOps (managed MLflow), KMS, and Workload Identity Federation. It sells GPU-hours per-second, reserved capacity contracts, multi-year hyperscale deals (Microsoft, Meta), and managed inference APIs to AI developers, startups, enterprises and hyperscalers.
Product overview
Nebius operates a single unified AI platform called Nebius AI Cloud, organized as a platform-plus-modules architecture rather than a single monolithic product. The core platform provides Nebius Compute (NVIDIA Blackwell/Blackwell Ultra/Hopper GPUs), Networking (virtual networks, subnets, NAT), and AI Storage (object, shared filesystem, block, plus a WEKA-integrated high-performance tier). On top of this foundation, Nebius sells a portfolio of integrated modules and products: Token Factory for managed inference of foundation, reasoning, and vision-language models via an OpenAI-compatible API; Tavily by Nebius for agentic web search; Tendem (Toloka x Nebius) for human-in-the-loop validation exposed as an MCP server; AI Orchestration (Soperator managed Slurm, SkyPilot, Ray, Anyscale) for workload scheduling; Serverless AI (Jobs, Endpoints, DevPods) for serverless GPU workloads; DataOps (managed PostgreSQL) and ModelOps (managed MLflow) for data and ML lifecycle; and the Agents Blueprint reference architecture plus the Physical AI Workbench and NVIDIA OSMO Managed by Nebius for AI agents and Physical AI respectively. Avride (autonomous vehicles subsidiary), the Builder Program, and Nebius Academy are part of the broader Nebius offering but are customer-facing programs rather than cloud modules.
Differentiator
Problem solved
Functional benefit
Brands
- Nebius AI Cloud: Full-stack AI cloud platform offering GPU compute, training, inference, storage, and managed services for AI developers and enterprises.
- Nebius Token Factory
- Avride
- Tavily by Nebius
- Nebius Academy
Products and services
- Nebius AI Cloud End-to-end AI cloud platform for training, fine-tuning, and inference, purpose-built for AI workloads on NVIDIA Blackwell, Blackwell Ultra and Hopper GPUs, integrating Compute, Networking, AI Storage, AI Orchestration, Serverless AI, Token Factory, Tavily, Tendem, DataOps and ModelOps for AI developers, startups, enterprises and hyperscalers.
- Nebius Compute On-demand and reserved GPU and CPU compute on Nebius AI Cloud with NVIDIA Blackwell, Blackwell Ultra and Hopper GPUs (H100, H200, B200, GB200, RTX PRO 6000 Blackwell Server Edition), shared and dedicated instances, priced per GPU-hour with per-second billing and reservations at up to ~35% discount.
- Nebius AI Cloud 3.6 Latest major version release of the Nebius AI Cloud platform, expanding GPU capacity and platform capabilities across compute, storage, orchestration and managed services. Generally available.
- Nebius Token Factory Managed inference service providing on-demand access to open-source, fine-tuned and proprietary foundation, reasoning, embedding and vision-language models via an OpenAI-compatible API, with dedicated deployments, autoscaling and 99.9% uptime SLA; SOC 2 Type II, HIPAA and ISO 27001 certified.
- Tavily by Nebius Search and crawling API purpose-built for AI agents and LLM applications, providing real-time, optimized and context-rich web content retrieval, used as the default agentic search layer in the Nebius Agents Blueprint.
- Tendem (Toloka x Nebius) Human-in-the-loop validation service for AI agents and pipelines, exposed as an MCP server that combines Toloka's global crowd-expert workforce with Nebius AI Cloud infrastructure to deliver labeling, validation and feedback at production scale.
- Nebius AI Orchestration Workload orchestration and scheduling layer for AI/ML jobs on Nebius AI Cloud, providing Soperator (managed Slurm on Kubernetes), SkyPilot job scheduling, Ray/Anyscale distributed compute and a scheduler-agnostic control plane for hybrid and multi-region GPU workloads.
- Nebius Serverless AI Three serverless AI offerings on Nebius AI Cloud: Jobs for batch GPU inference, Endpoints for scaled-to-zero GPU inference APIs, and DevPods for managed notebook and IDE environments, with usage-based pricing on B200 and GB200 tiers. Generally available.
- Nebius DataOps (Managed PostgreSQL) Fully managed PostgreSQL database service on Nebius AI Cloud, available in single-instance, replicated high-availability, and sharded configurations for AI/ML data pipelines, vector store and operational workloads.
- Nebius ModelOps (Managed MLflow) Fully managed MLflow service on Nebius AI Cloud for ML lifecycle management, providing experiment tracking, model registry and model deployment based on the open-source MLflow standard.
- Nebius Networking Networking primitives for Nebius AI Cloud, including Virtual Networks (VPCs), subnets, IP management, NAT gateways and high-throughput interconnects designed for distributed AI/ML training and multi-node GPU workloads.
- Nebius AI Storage High-performance, scalable storage for AI workloads on Nebius AI Cloud, including S3-compatible Object Storage, parallel high-throughput Shared Filesystem on NVMe for AI training, Block Storage, and a WEKA-integrated high-performance tier.
- Nebius Agents Blueprint Reference architecture and blueprint for building enterprise AI agents on Nebius AI Cloud, integrating Nebius Token Factory, LangChain Deep Agents, LangSmith, Pinecone (including Pinecone Nexus), Snowglobe by Guardrails AI, NVIDIA NIMs and Tendem. Generally available.
- Physical AI Workbench by Nebius Open-source orchestration framework for physical AI pipelines, integrating NVIDIA Cosmos for synthetic data generation, NVIDIA OSMO for orchestration, NVIDIA Isaac Sim/Lab for robot simulation and Nebius Compute for robotics and autonomous systems training.
- NVIDIA OSMO Managed by Nebius Nebius-managed deployment of NVIDIA OSMO for orchestrating multi-stage, multi-container Physical AI and robotics workflows on Nebius AI Cloud infrastructure.
Quantifiable outcome
- 43% better TCO for fine-tuning vs AWS
- +5 more outcomes
Companies that use Nebius
Customer profileNamed customers14 records
Segments7 records
Ideal customer profiles4 records
Nebius technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration18 records
AI capability13 records
Feature12 records
Nebius partnerships and signals
Strategic signalPartnerships
Ten partnerships are on record, tiered flagship, core and minor.
- MicrosoftflagshipMulti-year dedicated AI infrastructure agreement valued at approximately $17.4B through 2031, anchored by a significant prepayment. The deal is among the largest disclosed AI infrastructure contracts of 2025 and underpins Nebius's contracted backlog.
- MetaflagshipFive-year agreement under which Nebius will provide $12B of dedicated capacity, with options to expand total contract value to up to approximately $27B over the term.
- Bloom EnergycoreStrategic partnership for up to 328 MW of fuel-cell power capacity to provide behind-the-grid, around-the-clock clean electricity for Nebius data centers.
- Kao Datacore10-year, 22 MW agreement giving Nebius access to Kao Data's Harlow, UK data center campus, supporting Nebius's UK AI infrastructure initiative (~$1.7B investment, 65 MW by 2027).
- NVIDIA (Reference Platform / Physical AI Living Lab)flagshipBeyond capital, Nebius is an NVIDIA Reference Platform Cloud Partner and Exemplar Cloud and operates an NVIDIA Physical AI Living Lab, deploying NVIDIA reference architectures including HGX B200/B300 and the upcoming Vera Rubin platform.
- Eigen AI (acquired)coreAcquired for approximately $643M to bring inference and model-optimization capabilities in-house, expanding Nebius's Bay Area engineering presence and Token Factory roadmap.
- Clarifai (core team + IP license)coreAcquired Clarifai's core engineering team and licensed its IP portfolio covering AI inference and compute-orchestration methods.
- TD SYNNEXcoreGlobal channel partner for distribution of Nebius AI Cloud to enterprises and resellers; part of Nebius's broader Partner Program.
- Rowan UniversityminorEducation partnership supporting academic research and AI talent pipeline, including access to Nebius AI Cloud for university workloads.
- UK Government / AI MinistercorePublic partnership and policy engagement supporting Nebius's UK AI infrastructure build-out (~65 MW by 2027, ~$1.7B investment), including sovereign AI initiatives.
Scale indicators14 records
Recent moves8 records
Expansion highlights7 records
Nebius competitors and assessment
Company assessmentDirect peers
- CoreWeave: Pure-play GPU cloud provider offering NVIDIA H100/H200/Blackwell instances, storage, and managed services to AI labs and enterprises — the closest direct competitor to Nebius AI Cloud with similar hyperscale capacity contracts and MLPerf benchmark focus.
- Lambda Labs: GPU cloud and AI workstation company offering on-demand NVIDIA H100/B200 clusters for training and inference, targeting AI startups, researchers, and enterprises with a similar self-serve plus reserved-capacity model.
- Crusoe Energy: AI cloud operator building large-scale, energy-optimized GPU data centers, with a focus on stranded/clean power — competes directly for hyperscale AI training contracts and large NVIDIA cluster deals.
- Together AI: AI cloud platform offering GPU compute and managed inference for open-source and custom foundation models with a token-based API — directly comparable to Nebius's Token Factory and self-serve GPU model.
- Vultr: Cloud provider offering GPU instances (including H100/L40S), bare metal, and object storage with per-hour pricing — competes in the same self-serve, per-second-billed AI cloud segment.
Broad incumbents
- Amazon Web Services (AWS): Largest hyperscale cloud with EC2 GPU instances (P5/H100, P5e/H200, Blackwell-based), SageMaker, Bedrock, and Bedrock Marketplace — broad incumbent that bundles AI compute with networking, identity, and enterprise procurement that Nebius challenges on TCO.
- Microsoft Azure: Hyperscale cloud with Azure ND H100/H200/B200 VMs, Azure AI Foundry, and global enterprise relationships — simultaneously a customer of Nebius and a competitor for the same AI workload budgets.
- Google Cloud Platform (GCP): Hyperscale cloud offering A3 (H100) and A4 (Blackwell) GPU instances, Vertex AI, and TPU alternatives — competes for both AI training capacity and managed inference workloads with a deep, integrated software stack.
- Oracle Cloud Infrastructure (OCI): Hyperscale cloud with OCI Cluster Networking and H100/Blackwell GPU bare-metal instances, often used for large AI training clusters — competes directly for multi-gigawatt AI training contracts.
Emerging players
- Paperspace (DigitalOcean): GPU cloud (now part of DigitalOcean) offering H100/A100 instances and notebooks for AI developers and startups — overlaps with Nebius's self-serve and AI startup segment.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat7 records
Key risks6 records
Key highlights7 records
Customer concentration
Nebius social profiles
Digital presenceNebius compliance and trust
Trust signalCompliance3 records
Nebius financial estimates
Financial estimateRevenue estimate
Valuation estimate
Nebius leadership team
Management profileNumber of profiles
Profiles10 records
Nebius subsidiaries and ownership
Company hierarchySubsidiaries2 records
Nebius funding detail
Funding detailFunding overview
Funding rounds10 records
Investors13 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Nebius M&A and investment
M&A and investmentM&A3 records
Investments2 records
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Nebius
What does Nebius do?
Nebius operates a vertically integrated AI cloud platform (Nebius AI Cloud) that provides on-demand and reserved NVIDIA GPU compute (H100, H200, B200, B300, GB300, RTX PRO 6000) on custom hardware with non-virtualized GPUs and InfiniBand, together with managed Kubernetes, Soperator (Slurm on Kubernetes) orchestration, S3-compatible Object Storage, NVMe Shared Filesystem, Token Factory managed inference, ModelOps (managed MLflow), KMS, and Workload Identity Federation. It sells GPU-hours per-second, reserved capacity contracts, multi-year hyperscale deals (Microsoft, Meta), and managed inference APIs to AI developers, startups, enterprises and hyperscalers.
Is Nebius a public or private company?
Nebius is a public company. It is classified as public and is currently operating.
When was Nebius founded?
Nebius was founded in 2024. It employs 1,001 to 5,000 people.
Where is Nebius based?
Nebius is headquartered in Schiphol, Netherlands, in the Europe region.
How does Nebius make money?
Six revenue lines are on record. GPU compute (on-demand and preemptible) is the primary driver. The others are reserved/committed capacity, long-term hyperscale capacity contracts, managed inference (Token Factory), storage and managed cloud services and professional services and customer support.
Who are Nebius's main competitors?
Direct peers on record are CoreWeave, Lambda Labs, Crusoe Energy, Together AI and Vultr. Broad incumbents are Amazon Web Services (AWS), Microsoft Azure, Google Cloud Platform (GCP) and Oracle Cloud Infrastructure (OCI). Paperspace (DigitalOcean) is listed as an emerging player.
Does Nebius have an API?
Yes. Nebius offers a public, OpenAI-compatible REST API for Token Factory managed inference. Developers can call foundation, fine-tuned, reasoning, vision-language, and embedding models (including NVIDIA NIMs and Llama-based models) via a uniform endpoint and migrate OpenAI workloads with minimal code changes. The API supports serverless endpoints and dedicated deployments, with a separate Inference API documented at https://docs.tokenfactory.nebius.com/. MCP (Model Context Protocol) is available: Tendem (Toloka x Nebius) is exposed as an MCP server for human validation, and the Nebius Agents Blueprint architecture is MCP-compatible. Developer documentation is at docs.tokenfactory.nebius.com.
What industry is Nebius in?
Nebius's product category is Cloud AI Infrastructure. Its primary akta.pro industry code is HDABADAC, GPU/Accelerator Compute Servers, with a secondary code of HDAAAAAK, AI Compute Cloud & GPU-as-a-Service. Its NAICS code is 518 and its SIC code is 3571.