Google Cloud customers can now use the full Claude developer platform while paying through Google Cloud and drawing down eligible cloud commitments. Anthropic operates the service and acts as the data processor, even though the charges appear on a Google Cloud invoice.
The Anthropic-managed offering is generally available alongside the existing Google-managed route. Both use Google Cloud identity and billing, but they differ in platform-feature availability, operating terms and documented governance controls.
Anthropic’s October 22, 2026, release notes record the launch, with Google publishing its explanation on October 23. Both paths promise new Claude models on day zero. Only the Anthropic-managed option promises the full Claude Platform feature set and same-day betas.
Google Billing Does Not Make Google the Processor
According to Google’s deployment comparison, the Anthropic-managed service runs on Google Cloud infrastructure under Anthropic’s commercial terms. Organizations authenticate through Google Cloud IAM, retain Cloud Audit Logging and consolidate charges through Google Cloud billing. Eligible usage also counts toward existing Google Cloud commitments.
The Google-managed path keeps Google Cloud as the data processor under existing Google Cloud customer agreements. It also supports Google billing and eligible commitment drawdown.
IAM determines who can access a service; a cloud invoice determines how usage is paid for. Neither, by itself, establishes which company processes the data.
Enterprise buyers still need to address procurement and data governance separately. A team may be able to use its existing Google project, permissions and spending commitment while needing approval for Anthropic to act as processor under different terms. The expanded access lets customers keep Google Cloud’s purchasing arrangements while choosing between two operating arrangements for Claude.
Day-Zero Models Are Not Day-Zero Platform Features
Google says both options receive the latest frontier Claude models on day zero, at the same pricing. The launch roster includes Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5.
Developers will see the difference in the services available around those models. Anthropic says its managed route provides the complete native Claude Platform feature set, with new features and betas arriving the same day they launch on the native platform. On the Google-managed path, those features become available as they are integrated into Google’s infrastructure.
Anthropic groups the available capabilities around several kinds of work:
- Running agents and tools: Claude Managed Agents in beta, code execution, computer use, agent skills and the beta advisor strategy.
- Connecting information and systems: the Files API, citations, web search, web fetch and the beta MCP connector.
- Managing cost and throughput: prompt caching and asynchronous batch processing.
The offering also includes access to the Claude Console for configuring agents, monitoring usage and costs, and managing workspace access.
Access to the same model does not necessarily provide the same hosted execution environment, file handling or tool connections. Matching model availability may be sufficient for a team using basic model inference. For an application built around Managed Agents or a newly released platform API, feature timing can determine whether a deployment path supports the application at all.
Same-day beta access is still beta access. The announcement promises availability; it does not say every newly available capability has reached general availability or satisfies a particular enterprise’s production requirements.
Fast Mode Has a Specific Price and a Narrow Speed Claim
Anthropic says Opus 5.5 Fast Mode is generally available on the Anthropic-managed Google Cloud offering. Its published rates are:
- Input: $8 per million tokens.
- Output: $40 per million tokens.
The company claims up to 2.5 times faster output-token generation, positioning the mode for interactive agents, live debugging and rapid iteration.
That vendor claim concerns generation speed. It is not an independently measured reduction in total application latency. An agent can also spend time retrieving documents, executing code or waiting for external tools, so faster output generation does not establish a corresponding speedup across the whole workflow.
Google’s statement that the two deployment paths have the same pricing should be read separately from Anthropic’s Fast Mode announcement. The supplied announcements establish Fast Mode availability and its rates on the Anthropic-managed path; they do not establish equivalent Fast Mode availability on the Google-managed path.
Customers with existing discounts have a check to make before starting: Anthropic tells them to contact their Anthropic or Google Cloud account executive. Discounts cannot be applied retroactively to usage. Commitment drawdown and negotiated discounts remain separate checks, even when charges arrive on the same cloud invoice.
Residency Requirements Can Decide the Deployment
The Anthropic-managed service currently offers US and Global endpoints. Those labels should not be treated as interchangeable with every contractual residency guarantee available through the Google-managed service.
Google documents its managed option with data residency in the US, EU and APAC, alongside VPC Service Controls and customer-managed encryption keys. It also highlights native integrations with services including BigQuery and Model Armor. Anthropic separately identifies Agent Engine integration as a reason to choose the Google-managed route.
