Claude on Google Cloud, run by Anthropic: the full platform, but Anthropic processes the data

Anthropic is making the full Claude Platform available through Google Cloud. "Claude Models on Google Cloud (Anthropic managed)" was announced as generally available on Thursday 22 October at 21:21 UTC, and Google Cloud followed with its own blog post dated 23 October.

For organisations that already have a Google Cloud agreement and spending commitments, this means they can adopt new Claude features without waiting for Google to integrate them, and without setting up a new procurement and access path. The difference lies in who processes the data: in this option, Anthropic, not Google, operates the service.

What Anthropic is launching

According to Anthropic, Google Cloud customers on the new option get the full set of Claude Platform features. New features and betas are meant to ship the same day they launch on Anthropic's own platform. The company lists Claude Managed Agents (beta), code execution, computer use, agent skills, the Files API, citations, web search, the MCP connector (beta), prompt caching and batch processing, among others.

Access is controlled through existing Google Cloud IAM. Usage is billed on a single Google Cloud invoice and, according to Anthropic, draws down the customer's existing Google Cloud commitment. Google's documentation specifies that drawdown follows Google's Marketplace commitment policy. Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 are available.

Anthropic also says Fast mode for Opus 5.5 is generally available on this option. It promises up to 2.5 times faster output generation and costs $8 per million input tokens and $40 per million output tokens.

Who processes the data

Claude was already available on Google Cloud in an option where Google runs the models. That option remains. There, Google is the data processor under the customer's existing Google agreements, with VPC Service Controls, customer-managed encryption keys and data residency in the US, the EU and Asia-Pacific. New Claude features arrive there as Google integrates them.

In the new option, Anthropic operates the service and acts as data processor under Anthropic's Commercial Terms. Google's documentation describes Google and Anthropic as separate, independent data processors for their respective parts: Google for access control, API routing, audit logs and billing; Anthropic for the Claude API and Console themselves.

The sources do not fully agree on where data ends up. Google's blog post says Anthropic runs the stack on Google Cloud infrastructure. Google's documentation says storage and inference take place on Anthropic-managed infrastructure that may route outside Google Cloud. Inference can be pinned to a specific geography per request with the inference_geo parameter. Anthropic says the service is currently available with US and Global endpoints.

What does not carry over

Setup creates a new, separate Anthropic organisation for the Google Cloud project. According to Google's documentation, API keys, workspaces, HIPAA controls, Zero Data Retention and other settings from an existing Anthropic organisation do not carry over. Zero Data Retention may be available on request.

Discounts also need attention up front. Anthropic asks customers with an existing discount to contact their Anthropic or Google Cloud account executive before starting, because discounts cannot be applied retroactively. Rate limits and quotas are managed by Anthropic, while Google Cloud Support handles frontline cases with specialised backline support from Anthropic.

What leaders should settle

For an organisation on Google Cloud, this is mainly a procurement and compliance choice rather than a technical one. The new option gives faster access to agent features and can use money already committed to Google. It also moves data processing to Anthropic, under Anthropic's terms.

Three things should be settled before anyone switches it on. Legally, it should be clear which data processing agreement applies and how it covers international transfers of personal data. Technically, where inference and storage happen for the organisation's data should be confirmed in writing. Commercially, existing discounts and any Zero Data Retention arrangement should be in place before traffic starts.

Organisations that need EU residency, Google's security perimeter or Google's compliance certifications still have the Google-managed option. The features and terms above are as stated by Anthropic and Google. None of the sources says when a European endpoint will be offered for the Anthropic-managed option.

Sources and media

• Claude on X: general availability announcement, 22 October 2026
• Anthropic: Claude Platform is now available on Google Cloud
• Google Cloud: Expanding Claude availability and deployment choice on Google Cloud
• Google Cloud documentation: Use Claude models on Google Cloud (Anthropic managed)

Source: Anthropic.

Illustration: AI-generated with Grok Bot / hogby.ai