Meta Platforms is in early discussions to lease artificial intelligence computing capacity to Anthropic in a potential transaction worth up to $10 billion. The proposed agreement would give the developer of Claude access to Meta’s extensive data center infrastructure while creating a new revenue stream for the social media company, according to the NYT.
Anthropic reportedly proposed the arrangement in June 2026. Under the terms being considered, the company would make monthly payments to Meta over a two-year period, with both parties retaining the ability to exit the agreement early.
The discussions remain preliminary, and there is no assurance that the companies will reach a final agreement. The terms could also change as Meta evaluates how to structure and deliver computing services to an outside customer.
A completed transaction would represent a significant expansion beyond Meta’s established advertising-driven business. The company could begin generating revenue from its investments in AI chips, data centers, networking equipment and power infrastructure.
Meta has traditionally built computing infrastructure to support its own products, recommendation systems and AI models. Unlike Amazon, Microsoft and Google, it does not currently operate a mature public cloud business that sells computing capacity to companies at scale.
The potential Anthropic agreement could provide Meta with an initial major customer as it explores a broader cloud infrastructure operation. The company has reportedly considered selling computing capacity and hosting AI models that outside developers could access through Meta’s data centers.
This strategy would place Meta in more direct competition with established cloud providers such as Amazon Web Services, Microsoft Azure and Google Cloud. It could also challenge specialized AI infrastructure companies that provide high-performance computing capacity to model developers.
For Anthropic, the proposed agreement would add another source of the computing power needed to train and operate Claude. Advanced AI systems require large numbers of specialized chips for model development and for processing requests from customers using chatbots, coding tools and application programming interfaces.
Diversifying infrastructure providers can help an AI company increase capacity while reducing its dependence on any single cloud or hardware partner. Anthropic already uses multiple chip architectures and infrastructure providers to support its AI development and commercial services.
The proposed Meta deal demonstrates how access to chips, electricity and data center capacity has become a central competitive factor in the AI industry. Model developers increasingly need long-term infrastructure commitments to support growing customer demand and the development of more computationally intensive systems.
For Meta, selling capacity could help produce financial returns from infrastructure that is not immediately required for internal projects. The company could also reallocate computing resources between outside customers and its own AI programs as its requirements change.
The arrangement would be notable because Meta and Anthropic also compete in the broader AI market. Meta develops its own models and consumer AI products, while Anthropic offers the Claude family of models to individuals, developers and enterprises.
Any final agreement would therefore combine commercial cooperation with competitive considerations. Meta would need to protect its internal infrastructure priorities while providing Anthropic with sufficient reliability, security and computing performance.
The talks highlight a possible evolution in Meta’s role within the AI economy. Rather than using its computing infrastructure exclusively to support advertising, social media and internal model development, the company could emerge as a large-scale supplier of AI computing services to outside organizations.
