**Meta in Early Discussions to Lease Computing Capacity to Anthropic in $10 Billion Deal**
Meta, the parent company of Facebook, is reportedly in early negotiations to lease computing capacity to Anthropic, the AI research company, under a two-year agreement potentially worth up to $10 billion. According to *The New York Times*, the proposed arrangement would allow Anthropic to access Meta’s computing infrastructure while creating a new external revenue stream for Meta.
The details of the deal remain preliminary, with no specifics provided regarding the facilities, chip models, or the amount of capacity involved. Anthropic would reportedly pay in monthly instalments, and both companies would retain the option to terminate the agreement early. Neither Meta nor Anthropic has commented on the reports.
—
### Anthropic Broadens Its Infrastructure Mix
Anthropic has been actively expanding its computing infrastructure to support its AI models and services. The potential Meta agreement would add to its existing relationships with Amazon Web Services (AWS), Google, SpaceX, and TeraWulf.
Anthropic already operates its Claude AI across multiple accelerator platforms, including AWS Trainium, Google TPUs, and Nvidia GPUs. The company has also secured significant TPU capacity from Google and Broadcom, scheduled to come online in 2027.
One of its largest agreements is with SpaceX, reportedly costing $1.25 billion per month through May 2029 for access to the Colossus computing infrastructure. Additionally, Anthropic has entered a 20-year data centre lease with TeraWulf, expected to generate around $19 billion in contracted revenue over the initial term.
These agreements vary in structure and technical scope. For example, Anthropic’s work with AWS includes long-term Trainium capacity and engineering collaboration on processor optimization and the Neuron software stack. However, it remains unclear whether Meta would provide similar software integration or dedicated computing clusters under the proposed arrangement.
—
### Meta Looks Beyond Internal Workloads
A deal with Anthropic would represent a shift in how Meta utilizes its data centre estate. Historically, the company has built infrastructure primarily to support its own AI development, advertising systems, and consumer services.
Meta anticipates capital expenditures of between $125 billion and $145 billion in 2026, a significant increase from the approximately $72 billion spent in 2025. Much of this spending is directed toward data centres and AI infrastructure.
The company operates a global data centre network using CPUs, GPUs, and its internally developed MTIA accelerators. It is also expanding its Richland Parish campus in Louisiana to 5 gigawatts of computing capacity. While Meta has not specified which facilities might be used for the Anthropic agreement, it continues to source computing capacity from external providers such as CoreWeave and Nebius.
Meta has also been hiring senior executives from major tech firms to support its infrastructure expansion, including Dave Brown from AWS, to oversee data centre operations.
—
### Compute Lease or Cloud Platform?
The reported negotiations focus on access to computing capacity rather than the creation of a full public cloud platform. It is unclear whether Meta would offer dedicated clusters, customer-facing management tools, operational support, or broader managed services.
Established cloud providers bundle processor capacity with storage, databases, networking, security tools, billing, and technical support. Meta has not announced a comparable external service offering. Additionally, there is no indication that Anthropic’s own models would be part of the agreement.
If the deal proceeds, Meta would be both a buyer and supplier of computing infrastructure, gaining experience in serving external AI developers without positioning itself as a full hyperscale cloud provider. However, analysts note that such an arrangement would place Meta in direct competition with specialist infrastructure providers like CoreWeave and Nebius.
The lack of disclosed details—such as processor types, utilisation levels, service terms, or pricing—makes it difficult to assess the agreement’s impact on enterprise AI infrastructure markets.
—
### FAQ
**What is the reported agreement between Meta and Anthropic?**
Meta is in early discussions to lease computing capacity to Anthropic under a two-year agreement potentially worth up to $10 billion. The deal would allow Anthropic to use Meta’s data centre infrastructure while generating revenue for Meta.
**What details have been disclosed so far?**
No information has been released about the specific facilities, chip models, or capacity involved. Payments are expected to be made monthly, and both parties would retain the option to terminate the agreement early.
**How does this fit into Anthropic’s infrastructure strategy?**
The agreement would expand Anthropic’s infrastructure mix, complementing its existing partnerships with AWS, Google, SpaceX, and TeraWulf.
**What does this mean for Meta’s business model?**
The deal would mark a shift for Meta, moving from using infrastructure solely for internal purposes to leasing capacity to external AI companies. It would also provide experience in offering large-scale computing resources without becoming a full public cloud provider.
**Is Meta planning to launch a public cloud service?**
No. The reported arrangement involves compute capacity leasing rather than a comprehensive cloud platform with managed services, storage, or security tools.
**When will the deal be finalized?**
The negotiations are still preliminary, and no timeline for completion has been disclosed. The $10 billion figure represents the maximum potential value, not a confirmed contract.
—
### Conclusion
The potential lease agreement between Meta and Anthropic highlights the growing intersection of AI development and infrastructure supply. For Anthropic, the deal would provide additional compute resources as it scales its AI models. For Meta, it represents an opportunity to monetize its expanding data centre footprint while avoiding the full commitments of becoming a public cloud provider.
While the discussions are still in early stages, they reflect a broader trend of hyperscalers and tech giants exploring flexible infrastructure partnerships to meet the rising demands of AI. As the AI infrastructure market continues to evolve, such agreements may become increasingly common—balancing proprietary control with the need for scalable, specialised compute resources.
