IA en un minutoNewsBusiness and products
Google Cloud now offers Claude managed directly by Anthropic
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
Anthropic can now serve Claude on its own inside Google Cloud, which until now only offered it under Google's management. Under that route, data falls under Anthropic's terms and its new features, such as managed agents in beta, arrive from launch day. Both cost the same and are paid on the Google Cloud invoice, though the announcement only lists customer-managed keys and data residency for Google's option.
Google Cloud now offers Claude, Anthropic's AI models, in two ways: managed by Google, as before, or managed by Anthropic itself, an option that became generally available today.
Michael Gerstenhaber, Google Cloud's vice president of product management for Gemini, announced it on the company's blog. Under the new arrangement, Anthropic runs the service on Google Cloud infrastructure and acts as the data processor, under its commercial terms rather than the contracts customers already hold with Google. In return, it opens up its platform features from launch day, such as Claude Managed Agents (managed agents, still in beta), skills and code execution. Companies keep signing in through Google Cloud identity and permissions (IAM) and retain their audit logs.
In the Google-run option, which already existed, Google Cloud remains the data processor under each customer's current agreements. That is the option that provides VPC Service Controls perimeter controls, customer-managed encryption keys (CMEK), data residency in the U.S., the EU and Asia-Pacific, and integration with services such as BigQuery and Model Armor. New Claude platform features reach it later, as Google integrates them.
Both arrangements get new models on their launch day, cost the same and are charged on the same Google Cloud invoice, drawing down any spending commitments that qualify. The available models include Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. The announcement does not mention data residency or customer-managed keys for the Anthropic option.
Why it matters · analysis and opinion
Google letting Anthropic run Claude inside its own cloud says a lot about how the market is settling: the big providers no longer compete only on their models, but on being the place where companies concentrate their AI spending, even when the model belongs to someone else. For customers, the decision stops being technical and becomes a legal one. Whoever picks the Anthropic route gains speed with new features, but changes contract and the party responsible for processing their data. Anyone who needs to keep data in Europe or control their own keys has, for now, only one clear option. Before switching anything on, the sensible step is for whoever handles data protection to compare the two frameworks and decide which workloads can go each way.
Official source: Google Cloud · Written with the help of AI: how we make the news
What does it mean for your business? · analysis and opinion
This affects companies and freelancers in Spain and Latin America that already use Claude through Google Cloud or are considering it for their own projects. For anyone using Claude through its chat app or directly through the Anthropic API, without going through Google Cloud, this news changes nothing.
In practice, it is worth checking carefully which option gets switched on. With the Anthropic-managed option, the data is processed by Anthropic under its own Commercial Terms, not under the agreement the company already has with Google Cloud, so it makes sense to read those terms and record the change in the record of processing activities if personal data from customers or employees is being sent. If the company needs its data to stay in the EU or wants to use its own encryption keys, the news only mentions those safeguards for the Google-managed option, and the prudent choice is to stay with it until this is confirmed for the other one.
There is no difference in cost: pricing and the Google Cloud bill are the same, and both options draw down any eligible spending commitments the company has with Google. The risk lies in the new features: Claude Managed Agents is still in beta, and our advice is to test it on internal tasks without sensitive data before using it in customer-facing processes.
