Anthropic agrees to a $35 billion cloud computing deal tied to Nvidia-backed Lambda, report says

Anthropic PBC has agreed to a $35 billion cloud computing deal with Lambda, a move aimed at rapidly expanding the amount of computing power the artificial intelligence (AI) developer can use for training and running its models, according to a report citing a person familiar with the matter.

The deal size underscores the scale of compute demand as frontier model makers race to secure more accelerated processing capacity. In the reported arrangement, Lambda is described as a cloud provider backed by Nvidia Corp., linking Anthropic’s infrastructure buildout to Nvidia’s data center ecosystem.

While the announcement was framed as part of Anthropic’s effort to add capacity quickly, the report did not provide additional operational specifics such as the contract length, how quickly capacity would be delivered, or whether the arrangement covers both training and inference workloads (inference being the process of running models to answer requests, as opposed to the process of learning from data).

Nvidia’s involvement, as characterized in the report, centers on Lambda’s backing, which is consistent with the broader industry pattern in which major AI workloads are increasingly tied to Nvidia’s graphics processing units (GPUs) and related software stack. The report did not state whether Nvidia would directly provide equipment under the agreement, or whether backing means something narrower such as investment or supplier relationships.

An agreement of this magnitude would typically be expected to include detailed performance and delivery terms, but the available reporting did not enumerate key commercial elements such as pricing, minimum commit levels, service-level guarantees, or capacity conversion rates from hardware to usable compute.

For Nvidia, deals that channel large AI workloads into its hardware and platforms remain central to its strategy in data center AI. Even without additional detail on the contract mechanics, the news indicates that large model developers are continuing to turn to cloud partners that can provision substantial GPU capacity at speed.

For Anthropic, expanding compute is often a prerequisite for both scaling model training runs and supporting more widespread model deployment. However, as of the report, it was not disclosed what portion of Anthropic’s broader compute requirements the $35 billion figure covers, nor how the company plans to allocate workloads across different providers.

A key uncertainty is transparency. Neither the deal’s full terms nor any company statement was cited in the report available here. Observers will likely watch for follow-up disclosures, including whether Lambda and Anthropic publish additional information on delivery timelines, hardware configuration, and how the compute capacity will be measured and billed.