Google Cloud customers can now buy the full Claude experience without leaving their Google bill. Google and Anthropic have made an Anthropic-managed deployment option for Claude generally available on Google Cloud, announced today on the Google Cloud blog — meaning new Claude models and platform features arrive on Google Cloud the day they ship, rather than whenever Google gets around to integrating them.
The move completes a one-two punch for Google Cloud this week. Days after it started selling OpenAI’s frontier models through its Model Garden, Google Cloud is now the only hyperscaler offering the complete Claude platform alongside its own Gemini models on a single invoice — a direct pitch to enterprises that want to consolidate AI spending instead of juggling separate vendor contracts.
Two ways to run Claude on Google Cloud
The headline change is choice. Google Cloud now sells Claude in two flavours, and the difference comes down to who operates the stack and how fast new features land.
The new Anthropic-managed option puts Anthropic in charge of operating the service as the data processor, running on Google Cloud infrastructure under Anthropic’s own commercial terms. Customers sign in with their existing Google Cloud IAM credentials, keep Cloud Audit Logging, and pay through their normal Google Cloud invoice — with usage counting down their existing committed-use contracts. In exchange, they get day-zero access to new Claude models and every new Claude platform feature the moment it launches natively.
The existing Google-managed option remains the pick for customers with strict governance requirements. There, Google Cloud stays the data processor under each customer’s current agreements, and the offering includes VPC Service Controls, customer-managed encryption keys, and data residency across the US, EU and Asia-Pacific, plus native hooks into BigQuery and Model Armor. New Claude platform features reach this route only after Google integrates them into its own infrastructure — the trade-off for tighter control.
What the full platform includes
Anthropic’s own Claude platform feature set is the real substance of the announcement. On the Anthropic-managed option, customers get the complete toolkit: Claude Managed Agents (currently in beta) for running long-lived agents without managing infrastructure, Python code execution inside API calls, computer use for driving browsers and desktops, agent skills for teaching Claude repeatable workflows, and an advisor strategy mode for boosting agent quality.
The data side is equally complete: a Files API for uploads, citation grounding in source documents, live web search and web fetch, plus an MCP connector in beta. Cost controls come through prompt caching and batch processing for high-volume asynchronous work, and customers also get access to Claude Console, Anthropic’s development environment for configuring agents, monitoring usage and managing workspace security.
One detail will matter to speed-sensitive teams: Fast Mode on Claude Opus 5.5 is generally available through the new option, offering up to 2.5x faster output token generation at 8 and 40 dollars per million input and output tokens respectively. The service launches with US and Global endpoints and ships Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, with new models promised on the Anthropic-managed route as they launch.
Why It Matters
For enterprise buyers, this removes the feature lag that has quietly annoyed Google Cloud’s Claude customers. Under the Google-managed route, capabilities like managed agents and skills arrived only after Google wired them into its own stack — while competitors buying Claude directly from Anthropic, or from AWS and Azure, got them on day one. That lag is now optional rather than mandatory, and the choice maps neatly onto the industry’s split between compliance-first buyers and velocity-first builders.
The billing mechanics are the other lever. Drawing Claude spend down against existing Google Cloud commitments turns AI from a separate procurement headache into an extension of the cloud bill — exactly the convenience play Google Cloud made with OpenAI’s models days earlier. The catch, which Anthropic itself flags, is discounts: customers with negotiated pricing need to sort that with their account executive before switching, because discounts cannot be applied retroactively to usage.
The timing is also strategic for Anthropic. With a public listing expected as soon as November, every enterprise distribution channel that shortens the path from evaluation to production spend counts. And the skills story got a standards boost this week too: Anthropic donated its Agent Skills specification to the Agentic AI Foundation, the same mechanism now shipping day-zero on Google Cloud.
Background
Google and Anthropic’s partnership is one of the industry’s oldest: Google has invested billions in Anthropic since 2023 and Google Cloud has resold Claude models through Vertex AI for years, with Anthropic gaining access to vast fleets of Google TPUs for training. The new Anthropic-managed option formalises a division of labour the two companies have been building toward — Google provides the infrastructure, billing and identity; Anthropic operates the model platform itself. It lands in the same week Google Cloud also became OpenAI’s third cloud home, cementing the platform’s pitch as the neutral ground where enterprises can run every frontier model family on one contract.
Source: Google Cloud Blog, Claude.com
0 Comments