Companies using Google Cloud can now choose between two ways of accessing Claude models: one managed by Google, with data residency in the European Union and customer-managed encryption keys, and a new one managed by Anthropic on the same infrastructure, which offers agent features sooner. The Anthropic-managed option is generally available, and both share the same pricing and billing, according to Noticias.ai, reporting on Google’s announcement.
Claude’s two options on Google Cloud in 30 seconds
- Google Cloud keeps the Google-managed version of Claude and adds another option operated by Anthropic.
- Google’s option offers CMEK, VPC Service Controls, and data residency in the EU.
- Anthropic’s option brings Claude Managed Agents (beta), skills, and code execution from day one, with Anthropic acting as data processor.
- Both have the same price, are billed through Google Cloud, and count against signed spending commitments.
The change was explained on October 23, 2026, by Michael Gerstenhaber, vice president of Gemini Product Management at Google Cloud, in the company’s blog. The available models are Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, and Claude Haiku 5.5. Cloud News covered one of them when Anthropic launched Claude Opus 5.5, with more agentic power at a lower cost.
Two ways to operate on the same cloud
The core difference lies in who operates the service and who is accountable for the data. In the Google-managed option, Google Cloud is the data processor under the contracts the customer already has with the cloud provider, and the service integrates with the platform’s usual control tools: VPC Service Controls perimeters, customer-managed encryption keys (CMEK), data residency in the United States, the EU, and Asia-Pacific, and native connections to BigQuery and to Model Armor, Google’s service for filtering model inputs and outputs.
In the Anthropic-managed option, Claude’s maker operates the platform on Google’s infrastructure and acts as data processor under its own commercial terms. In exchange, the customer gets, from day one, Claude platform features that Google hasn’t yet integrated into its own version, such as Claude Managed Agents, still in beta, skills, and code execution. In both cases, authentication runs through Google Cloud IAM, Cloud Audit Logging records are kept, and spending shows up on the Google Cloud bill.
Sovereignty versus speed
For a European IT leader, the choice resembles other cloud sovereignty decisions: either contractual and technical control over where and how data is processed takes priority, or immediate access to the model provider’s latest features does. With the GDPR in play, switching data processors means reviewing contracts, records of processing activities, and, in many cases, impact assessments. Google’s announcement doesn’t specify which regions the Anthropic-managed option runs in, a detail worth confirming before putting it into production with personal data.
The move fits Google Cloud’s strategy of offering models from multiple providers under a single bill, which the same blog post extended to OpenAI’s models as well. For companies with signed spending commitments with Google, being able to count Claude usage against them is one of the most practical arguments in the announcement.
Frequently asked questions
How can Claude be used on Google Cloud?
In two ways: a Google-managed option, with data residency and CMEK, or an Anthropic-managed option, which offers Claude’s platform features sooner.
Who processes the data in each option?
Google Cloud in the Google-managed option and Anthropic in the Anthropic-managed option, each under its own contracts.
Is there EU data residency for Claude on Google Cloud?
Yes, in the Google-managed option, which offers data residency in the United States, the EU, and Asia-Pacific.
Which Claude models are available?
Claude Fable 5.1, Claude Opus 5.5, Claude Sonnet 5.5, and Claude Haiku 5.5.
Sources:
