On October 23, Google Cloud opened up a new backdoor to enterprise customers – the deployment option for the Claude model hosted by Anthropic is now officially in use. This means that enterprises no longer need to move their data around; they can run Claude directly on Google Cloud’s own infrastructure, and they can access the latest version on the day of model release, without having to wait for new interfaces, schedules, or third-party resale.
The cleverness of this approach lies in “division of labor”. Anthropic plays the role of the data processing party, responsible for delivering the model capabilities to you; while the keys to the enterprise side remain in its own hands—Google Cloud’s Identity and Access Management (IAM), audit logs, and the cloud billing system that is already in use all continue to function as usual, without any additional setup. In other words, the model belongs to Anthropic, but the access control, accounting records, and monitoring sensors are still Google Cloud’s.
Google Cloud has not blocked another path. It still retains the deployment solutions it manages, where Google Cloud is directly responsible for data processing, along with virtual private cloud service control, customer management encryption keys, and data residency capabilities covering the United States, the European Union, and the Asia-Pacific region. Two paths are placed side by side. Enterprises can choose the option that best suits their needs for compliance and control flexibility—if they want Anthropic to host directly and provide lightweight access, they can use the new option; if they require higher standards for data storage locations and encryption control, they can still choose the version with Google providing full hosting. As the computing power and models of the two major AI camps begin to share resources, enterprise customers gain a more flexible choice.