• Anthropic reportedly agreed to spend US$35 billion on computing from Nvidia-backed Lambda, with Hut 8 developing the Texas data-centre capacity behind the service
  • The arrangement puts the cloud provider, hardware supplier and data-centre developer in separate parts of the delivery chain

The fact

Anthropic has reportedly agreed to spend US$35 billion on cloud computing from Nvidia-backed Lambda, Bloomberg reported, citing a person familiar with the matter. Hut 8 is developing the Texas data centre associated with the capacity. Anthropic declined to comment, while Lambda, Nvidia and Hut 8 had not publicly confirmed the agreement when the report was published.

The reported deal adds another large compute commitment to Anthropic's infrastructure programme. Lambda supplies GPU-based cloud computing, while Hut 8 is developing large-scale data-centre capacity in Texas. Nvidia supplies the accelerator hardware and has also backed Lambda. The reporting does not establish how much physical capacity is allocated specifically to Anthropic or when the full contracted compute will become available.

The assessment

Anthropic’s reported commitment is with Lambda, but the compute ultimately depends on infrastructure being delivered elsewhere. Nvidia provides the accelerator hardware, while Hut 8 is developing the Texas data-centre capacity needed to host it. The contract therefore commits demand before all of the underlying capacity is necessarily in service. Hut 8 still has to deliver the facility, the site has to be powered and equipped, and Lambda has to make the resulting compute available to Anthropic.

For BTW readers, the US$35 billion figure is best read as a commercial commitment rather than a measure of capacity already online. The more useful test is how quickly the facilities, power and hardware behind that contract move into service.

What to watch

Watch for formal confirmation of the Anthropic-Lambda agreement, disclosure of the capacity attached to it and construction milestones at Hut 8's Texas facilities. Power delivery, GPU installation and service start dates would show when the reported commercial commitment begins turning into usable computing capacity.