On October 22, 2026, Anthropic announced the general availability of Claude Models on Google Cloud (Anthropic managed). It is the first Claude service on Google Cloud operated by Anthropic itself: customers keep using their own Google Cloud projects and IAM identities to call Claude, but Anthropic becomes the data processor, and billing runs through the Google Cloud Marketplace, drawing down existing cloud spend commitments. Google's official blog subsequently confirmed the arrangement.
Two options, three differences
The core of this launch is not another entry point, but two parallel "who operates it" options. Google Cloud already offered a Google-managed Claude (Google managed); the new option is the Anthropic-managed one (Anthropic managed). Models, launch cadence, and pricing are the same on both sides — Claude Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5, with new models available on day zero. The differences come down to three things.
First, the data processor is different. Under the Anthropic-managed option, Anthropic is the data processor and Anthropic's Commercial Terms apply; under the Google-managed option, Google is the data processor under the customer's existing Google Cloud agreements. Enterprises with strict data-residency requirements or their own encryption keys are explicitly advised to stay with the Google-managed option.
Second, new features land at different times. The Anthropic-managed option promises new capabilities and betas in step with the main Claude platform, including Claude Managed Agents (long-running agents, still in beta), document uploads, web search, code execution, Agent Skills, and MCP connectors; the Google-managed option only releases them once they are integrated into Google's infrastructure.
Third, the decision logic changes. Anthropic puts it plainly: choose the Anthropic-managed option if you want "the full Claude platform plus Google billing and IAM"; choose the Google-managed option if you need "Google as the data processor plus native Google ecosystem integrations" such as Model Armor or Agent Engine.
What it means for enterprise buyers
For enterprise buyers, this fixes a very practical pain point: AI spend can now legitimately run through the cloud bill and retire commitment spend, instead of going through a separate Anthropic invoice and approval process. With Opus 5.5 fast mode launching at the same time (officially priced at $8 per million input tokens and $40 per million output tokens), the switching cost for teams already running on Google Cloud is low.
One boundary worth noting: day-zero access to new features also means changes to Anthropic's Commercial Terms land on customers faster, so buyers should be clear about whether they care more about "new capabilities first" or "an unchanged data-processing chain." Anthropic also reminds customers with existing discounts to contact their account manager to migrate discounts to the new service, as discounts cannot be applied retroactively.
