How CoreWeave and Meta are Scaling AI Data Centres

CoreWeave has confirmed an expanded long-term agreement with Meta Platforms to supply AI cloud capacity through to December 2032, in a deal valued at around US$21bn.

The agreement builds on an existing relationship between the two companies and centres on scaling infrastructure to support Meta’s AI operations.

Meta is gaining access to infrastructure that supports ongoing AI development and deployment by securing capacity through to 2032.

Long-term agreements such as this can help stabilise supply in a market where demand often outpaces available resources.

Their partnership with CoreWeave builds on data centre capacity at the very core of AI development, with the latter hyperscaler delivering dedicated infrastructure designed for high-performance workloads.

As AI models grow in size and complexity, demand is rising for environments capable of handling intensive compute requirements and networking at scale.

Distributed data centre infrastructure

CoreWeave states that the capacity will be deployed across multiple locations, forming a distributed data centre footprint rather than relying on a single facility. This will improve resilience and maintain steady performance if one location is disrupted.

A distributed model allows Meta to run workloads closer to where data is generated or needed, in turn reducing latency and improving responsiveness.

It also supports scalability, enabling additional capacity to be brought online without disrupting current operations.

The deployment will integrate the NVIDIA Vera Rubin platform, a next-generation computing architecture built for advanced AI workloads.

By implementing this technology into its data centres, CoreWeave will provide higher processing efficiency and faster training times for large-scale models.

That said, the integration proves both the cost and complexity of building AI-ready data centres, as well as the competitive pressure to deploy advanced models quickly.

Scaling for AI workloads

The agreement reflects broader trends across the data centre sector, where operators are adapting facilities to meet the needs of AI-driven workloads.

Traditional enterprise applications place different demands on infrastructure compared with AI training and inference, which require specialised hardware like GPUs and high-bandwidth interconnects. 

With Meta's AI expertise, CoreWeave will be able to deploy AI effectively.

CoreWeave also works with AI labs and large enterprises, providing access to infrastructure that might otherwise require substantial upfront investment.

“This is another example that leading companies are choosing CoreWeave’s AI cloud to run their most demanding workloads,” said Michael Intrator, Co-founder, CEO and Chairman of CoreWeave.

Long-term capacity planning

The scale of the agreement, extending over several years and valued in the billions, shows how organisations are committing long-term resources to secure access to capacity.

Long-term deals offer visibility into future utilisation, allowing data centre providers to plan expansions and invest in new technologies across their facilities.

The inclusion of emerging platforms such as NVIDIA Vera Rubin also shows how CoreWeave and Meta align their infrastructure roadmaps with advances in hardware.

CoreWeave has noted that further details of the agreement are outlined in its filing with the US Securities and Exchange Commission, which provides additional context on the structure and scope of the deal.

As AI continues to evolve, the role of data centres shifts from general-purpose infrastructure to specialised environments designed for performance and scalability.

This agreement between CoreWeave and Meta demonstrates how partnerships are shaping that transition, with long-term commitments underpinning the next phase of AI growth.