Google Cloud adds Anthropic-managed option for Claude

Google Cloud made an Anthropic-managed deployment option for Claude generally available, opening a second route to the models alongside its existing Google-managed service.

The choice preserves Google Cloud IAM authentication, API routing, Cloud Audit Logs and Marketplace billing, while changing who processes customer data and which contracts govern the service. Google’s product documentation says Google and Anthropic act as separate, independent data processors for their respective components. Google terms cover the Google Cloud services; Anthropic’s Commercial Terms, Data Processing Addendum and Usage Policy cover the Anthropic-managed Claude API and Console. Customers authenticate through Google Cloud rather than with an Anthropic API key.

The Google-managed option works differently: Google says it is the sole data processor. Google positions that route for workloads requiring inference within its security boundary, as well as controls such as VPC Service Controls, customer-managed encryption keys, regional data residency and specified compliance coverage. It also supports native integrations including BigQuery, Model Armor and Agent Engine.

One part of the infrastructure boundary remains unresolved in Google’s published descriptions. The launch post says Anthropic operates the stack as data processor on Google Cloud infrastructure. The product documentation says storage and inference occur on Anthropic-managed infrastructure and that requests may be routed outside Google Cloud. Neither source explains how those statements fit together.

Feature timing is another dividing line. Google says Anthropic-managed customers are intended to receive new models, Claude platform features, research previews and betas when Anthropic releases them. The Google-managed service gets platform features after Google completes its integration or according to the Model Garden release schedule; Google has not provided a fixed timetable for individual features. The launch post also says both options have the same pricing and provide frontier models on day zero. Those claims were not independently corroborated in the supplied evidence.

Activating the service creates a separate Anthropic organization for the customer’s Google Cloud project. Settings from an existing first-party Anthropic organization—including API keys, workspaces, HIPAA controls and zero-data-retention settings—do not carry over. Anthropic’s Commercial Terms say customers retain rights to their inputs and own outputs, and that Anthropic does not train models on customer content from covered services.

Google says Anthropic-managed usage can count against eligible Google Cloud commitments. That use remains subject to the Marketplace Commitment Drawdown Policy, including the applicable service drawdown rate, exclusions and a cumulative cap of 25% of the corresponding minimum commitment. The supplied sources do not establish the product-specific rate or exclusions.