Anthropic seeks more control over AI compute as it explores data center leasing with Google support

Anthropic, the artificial intelligence company backed by Amazon and others, is reportedly moving to take greater control of the computing infrastructure behind its models, and is pursuing data center leases that would place it in direct control of operating its own facilities. The effort centers on securing more ownership over the physical layer of AI, including the servers and the data center operations required to run large-scale training and inference workloads.

According to a report carried by Yahoo Finance, the company is exploring arrangements that would shift Anthropic from being solely a customer of cloud or third-party capacity to a position where it can influence or oversee how the facilities are run. In practical terms, that means data center leases where Anthropic would play a larger role in operating the facilities, rather than simply relying on a provider to manage everything end to end.

The same report characterizes Google as a key partner in the plan, suggesting Anthropic is engaging with Google to support the data center push. While details of the exact deal structure were not provided in the portion of information available for this story, the framing points to a relationship that goes beyond standard cloud usage, potentially involving deeper coordination on capacity planning and how compute is provisioned for Anthropic’s models.

The timing of Anthropic’s infrastructure move aligns with a broader industry pattern, in which leading AI builders increasingly treat compute as a strategic input rather than a back-office cost. As model sizes grow and as companies compete on faster deployment and steadier throughput, controlling the supply chain for servers, networking, and data center capacity becomes a competitive lever.

A data center lease model can offer benefits for an AI lab that expects sustained and rising demand. More control over facilities can help reduce variability in performance and availability, and it can also make it easier to tailor infrastructure to specific hardware and workload patterns. It can also shift some cost risk from a pure consumption model, where spending tracks usage, to a structure where longer-term commitments are reflected in operating plans.

For Google, a partnership that supports Anthropic’s infrastructure plans would fit with the company’s continuing investment in data center expansion and AI-related capacity. Even when cloud services are involved, deep commitments around physical infrastructure can create longer-duration demand for compute resources, and they can reinforce a supplier role for critical components such as data center hardware, networking, and power infrastructure.

Still, key specifics are not disclosed in the available reporting. The exact locations of the facilities, the expected timeline, the scale of the leases, whether the arrangements involve dedicated servers or fully managed operations, and how much of the stack Anthropic would control were not spelled out. It is also unclear how the plan would interact with Anthropic’s existing compute and deployment approaches, including any reliance on public cloud or other suppliers.

What to watch next is whether Anthropic or Google provides more concrete terms, including facility scope, duration of leases, and operational responsibility. Confirmation of deal size, geography, and hardware configuration would help investors and customers gauge how quickly Anthropic’s approach could scale and what it might announcement for the direction of AI infrastructure spending across the sector.