Google Cloud, part of Alphabet Inc. (NASDAQ: GOOGL), has added an Anthropic-managed option for Claude models that lets enterprise teams choose whether Anthropic or Google operates the model service. The new option is generally available and gives organizations a way to keep Claude workloads on Google Cloud while selecting the data processor and governance model that fit their deployment requirements.
The choice separates access to Anthropic’s newest platform capabilities from Google Cloud’s existing managed controls. Under the Anthropic-managed service, Anthropic operates the stack on Google Cloud infrastructure and acts as the data processor under its own terms. Organizations still authenticate through Google Cloud IAM, retain Cloud Audit Logging and can apply existing Google Cloud billing and eligible commitments, according to the company.
That arrangement gives teams day-zero access to Claude models and Anthropic platform features, including Claude Managed Agents in beta, skills and code execution. For development groups that need new model features as they become available, the option could reduce the lag between an Anthropic feature launch and its availability through a cloud-provider-managed service.
Google Cloud’s existing Google-managed deployment remains available for organizations that place more weight on controls delivered through their existing customer agreements. Google said that option provides VPC Service Controls, customer-managed encryption keys and data residency in the United States, European Union and Asia-Pacific regions. New Claude platform features arrive there after Google integrates them into its managed infrastructure.
The Google-managed service also carries native integrations with Google Cloud services including BigQuery and Model Armor, according to the company. That makes the configuration relevant to teams that want Claude workloads to sit alongside established data and AI-security services. Google did not position one model-hosting option as universally preferable; it framed the choice around an organization’s processor, terms and feature-access requirements.
The two services use the same pricing structure, Google Cloud billing and eligible commitment drawdown, the company said. Both are intended to provide access to the latest Claude models on their release day. Google Cloud listed Claude Fable 5.1, Sonnet 5.5, Haiku 5.5 and Opus 5.5 among the models available through the service. The new capability changes how those models can be operated on Google Cloud, rather than introducing a separate Claude model family.
The operational difference is not a model-performance claim; it is a deployment and accountability choice. Enterprises adopting generative AI often need to decide which party processes their data, how identity and audit records are retained, and whether a service’s newest capabilities can be enabled inside their established controls. The split lets teams make those decisions without moving Claude workloads to a separate cloud billing or identity environment.
For IT and security teams, the Google-managed option preserves an approach centered on cloud-native governance and sovereignty features, while the Anthropic-managed path prioritizes first-party platform access. The practical trade-off is clear: organizations can choose a faster route to Anthropic features or the additional controls that come with Google operating the service, rather than treating model availability and governance as a single deployment decision.