Nebius Signs Massive AI Infrastructure Deal with Meta Built on NVIDIA’s Vera Rubin Platform

Amsterdam-based AI cloud company Nebius Group has signed a long-term infrastructure agreement with Meta that could reach $27 billion in total contract value, marking one of the largest AI compute supply deals announced to date.

The agreement centers on building large-scale AI infrastructure powered by NVIDIA’s next-generation Vera Rubin platform, positioning Nebius as a key provider of high-performance compute capacity for Meta’s expanding AI ambitions.

Under the five-year deal, Nebius will deliver $12 billion in dedicated AI compute capacity across multiple data center locations starting in early 2027. In addition, Meta has committed to purchasing up to $15 billion in additional compute capacity from future Nebius clusters, bringing the potential total value of the partnership to approximately $27 billion.

The AI Cloud Infrastructure Race

The agreement reflects a broader shift in the AI industry: the race is no longer just about building better models, but about owning the infrastructure required to train and run them at scale.

Training frontier AI systems now requires enormous clusters of GPUs interconnected with high-speed networking and optimized software stacks. Companies developing large models—from Meta to OpenAI and Anthropic—are increasingly dependent on specialized AI cloud providers capable of delivering this infrastructure reliably.

Nebius is positioning itself as one of those providers.

The company is building a full-stack AI cloud platform, designed to support the entire lifecycle of AI development—from data ingestion and training to inference and production deployment. This type of vertically integrated infrastructure is becoming increasingly important as organizations move from experimentation with AI to deploying production systems that require stable, high-throughput compute resources.

Why the NVIDIA Vera Rubin Platform Matters

A central technical component of the deal is Nebius’s deployment of the NVIDIA Vera Rubin platform, which represents the next generation of GPU-accelerated AI infrastructure.

While NVIDIA’s current generation architectures—such as Hopper—power many of today’s AI workloads, Vera Rubin is designed specifically for the next wave of large-scale AI systems. The architecture is expected to deliver significant improvements in:

• Training throughput for large models
• Memory bandwidth and interconnect speeds
• Cluster-scale GPU efficiency
• Performance for multimodal and agentic AI systems

Large-scale deployments of the platform will allow AI companies to train increasingly complex models while reducing the operational overhead associated with distributed training across massive clusters.

Nebius plans to build some of the first large-scale Vera Rubin deployments, which will form the backbone of the compute capacity supplied under the Meta agreement.

Flexible Capacity for AI Customers

An interesting aspect of the arrangement is how Nebius plans to allocate the infrastructure.

The company intends to sell a portion of the cluster capacity to third-party AI cloud customers, including startups and enterprises building AI products. Any remaining capacity will be purchased by Meta under its commitment agreement.

This structure allows Nebius to scale its infrastructure for a broader customer base while still guaranteeing utilization through Meta’s demand.

For AI cloud providers, this hybrid model helps mitigate a major risk associated with building multi-billion-dollar compute clusters: ensuring that the infrastructure remains fully utilized once deployed.

Building the Backbone of AI Development

Nebius has been positioning itself as a specialized infrastructure provider for the AI era. Headquartered in Amsterdam and listed on NASDAQ under the ticker NBIS, the company is focused on delivering the underlying compute systems required for modern AI development.

Its platform targets a growing ecosystem of developers and enterprises building:

• generative AI models
• autonomous agents
• multimodal AI systems
• large-scale AI applications

By offering scalable compute clusters optimized for training and inference, Nebius aims to compete with hyperscale cloud providers while focusing specifically on AI workloads.

The Bigger Picture: AI Infrastructure as a Strategic Asset

The Meta agreement highlights how AI infrastructure is becoming one of the most valuable assets in the technology sector.

Large AI labs are increasingly securing long-term supply contracts for compute capacity, similar to how energy or semiconductor companies secure long-term production agreements.

This shift reflects a simple reality: the future of AI development depends not just on algorithms, but on access to enormous amounts of compute.

With multi-billion-dollar clusters powered by next-generation GPU architectures, the infrastructure layer is rapidly becoming the foundation upon which the next generation of AI systems will be built.