Meta discusses leasing AI computing capacity to Anthropic, aiming to monetize infrastructure

Meta is reportedly in talks with Anthropic about leasing some of its AI infrastructure, a move that could turn the company’s in-house computing buildup into a new revenue line tied to third-party artificial intelligence development.

The report, carried by Yahoo Finance, said the discussions involve renting computing capacity to Anthropic. While the parties’ goals appear aligned around accelerating AI model development, the reporting does not specify the scale of the capacity, the timeline, pricing terms, or whether Anthropic would gain access to specific hardware or managed cloud services.

If the talks result in an agreement, Meta would be expanding beyond its traditional posture as a consumer of cloud and compute services and instead acting as a provider of infrastructure for outside AI workloads. That would put it in the broader orbit of companies that already sell AI-focused cloud and compute capabilities to developers and enterprises, including hyperscalers and cloud platforms.

Anthropic, the AI lab behind the Claude family of models, has sought partnerships and access to compute to train and run large models. Meta’s potential role would reflect an industry pattern in which infrastructure operators, including AI builders and cloud providers, look to monetize expensive data-center assets rather than treat them purely as internal costs.

Meta’s public communications around AI infrastructure have often emphasized its role in supporting products and research, but it has not, in the information available here, confirmed any commercial arrangement with Anthropic. The latest reported development therefore remains at the discussion stage, not a finalized contract.

In sector terms, the possible shift matters because AI compute has become a bottleneck and a budget line for many organizations. Renting capacity can shorten deployment timelines for model builders and reduce the need to procure and operate dedicated infrastructure from scratch, but it also requires providers to meet performance, reliability, and supply expectations that typical internal systems may not be built to support.

Still, significant questions remain. The report does not disclose whether Meta would provide raw compute, a platform-layer service, or something closer to an enterprise offering with support and service-level commitments. It also does not say which specific systems or regions would be involved, whether the arrangement would be exclusive, or how Meta would handle security and data-governance requirements for external customers.

What to watch next is whether Meta or Anthropic confirms the discussions and whether any details emerge on scope and terms. Industry observers will likely focus on whether this becomes a broader strategy for Meta, or a one-off deal tied to the current wave of AI commercialization.