The full Claude Platform, billed through Google Cloud.

What is new?

Anthropic made Claude Models on Google Cloud (Anthropic managed) generally available on 22 October 2026. Google announced the new deployment option on 23 October 2026 on the Google Cloud Blog. There are now two ways to use Claude through Google Cloud: the existing option operated by Google and the new one operated by Anthropic.

According to Google, both deliver the same models from day zero, at the same prices and with Google Cloud billing. The difference is who acts as data processor and whether new Claude Platform features are available immediately. Available models include Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5.

Which features are included?

With the Anthropic-managed option, Google Cloud customers get the full feature set of the Claude API, according to Anthropic. New features and betas ship on the same day as on the native Claude Platform. Customers also get access to the Claude Console, where teams write prompts, configure agents, monitor usage and costs and manage access per workspace.

Fast mode for Claude Opus 5.5 is also generally available on this option. According to Anthropic, it generates output tokens up to 2.5 times faster and costs 8 US dollars per million input tokens and 40 US dollars per million output tokens.

• Agents: Claude Managed Agents (beta), code execution, computer use, agent skills, advisor strategy (beta)
• Knowledge and systems: Files API, citations, web search and web fetch, MCP connector (beta)
• Cost and performance: prompt caching and batch processing

How do access and billing work?

Access runs through existing Google Cloud IAM credentials; according to the documentation, API keys are not supported on this option. Setup creates a new Anthropic organization for the Google Cloud project, separate from any existing one. You need permission to purchase in Google Cloud Marketplace, to enable services on the project and to open the Claude Console.

Anthropic meters usage and reports it to Google Cloud in Claude Consumption Units, with one unit equal to 0.01 US dollars. Charges appear on the Google Cloud bill, Anthropic sends no separate invoice, and usage counts toward existing Google Cloud commitments like other Marketplace purchases. If you already have a discount, Anthropic advises contacting your account executive before you start, because discounts are not applied retroactively.

What does this mean for data protection and logging?

On the Anthropic-managed option, Anthropic is the data processor and runs the service under Anthropic's commercial terms. According to the documentation, some supporting services, such as Anthropic's trust and safety systems, process data outside Google Cloud. The Claude API retention policy applies; Zero Data Retention is available on request, and an existing arrangement on another Anthropic organization does not carry over automatically.

Cloud Audit Logs records API activity in your own project. According to the documentation, Data Access logs, which cover routes such as messages, batches, files and skills, are off by default. If you need traceability, you have to turn them on deliberately.

Anthropic-managed or Google-managed?

The Anthropic-managed option is currently available with US and Global endpoints. According to Google, the Google-managed option offers data residency in the US, the EU and APAC, VPC Service Controls, customer-managed encryption keys (CMEK) and integrations such as BigQuery and Model Armor. New platform features arrive there only once Google has integrated them. According to Anthropic's documentation, customer-managed keys via Google Cloud KMS are also available on the Anthropic-managed option.

Anthropic explicitly recommends the Google-managed option for customers that need Google as data processor, have strict regional data residency requirements or want native integrations such as Model Armor or Agent Engine. For businesses in Europe with EU requirements, this is usually the deciding point.

• Not supported on the Anthropic-managed option: API keys, MCP tunnels, memory stores on self-hosted environments and the pre-built self-hosted sandbox workers.

What should businesses do now?

First clarify whether your use cases require EU data residency or Google as data processor. If so, stay with the Google-managed option. If not, and you need new features such as Managed Agents without delay, test the Anthropic-managed option in a separate project, review the contract terms with data protection and procurement, and turn on Data Access logs. According to the documentation, both organizations can run in parallel and workloads can move gradually; in code, the AnthropicGoogleCloud client replaces the previous AnthropicVertex client.

Is the Anthropic-managed option more expensive? According to Google, the same prices apply as on the Google-managed option. Is there an EU region? Anthropic currently lists only US and Global endpoints. Do existing API keys work? No, authentication runs only through Google Cloud credentials. The free AI System Check helps sort out platform and data protection requirements. These notes are suggestions, not customer results.