Two ways to buy Claude on Google Cloud. Only one ships Anthropic’s newest features on day zero. Since October 22, 2026, Claude on Google Cloud comes in an Anthropic-managed flavor — generally available, billed on your Google Cloud invoice, under your existing IAM credentials. Anthropic operates the service and acts as your data processor. Best of all, usage draws down your existing Google Cloud commitments.
What Anthropic actually launched
Anthropic’s official announcement (Oct 22, 2026) makes Claude Models on Google Cloud (Anthropic managed) generally available — confirmed by Google the next day.
The headline feature is day-zero access. New models and new Claude Platform features now ship the same day they go live on Anthropic’s native platform — no more waiting for Google to wire each feature in.
The commercial plumbing stays Google:
- Authentication: your existing Google Cloud IAM credentials — no separate Anthropic accounts
- Billing: a single Google Cloud invoice
- Commitments: Claude usage draws down your existing Google Cloud commitments
- Audit: you keep Cloud Audit Logging
A data processor is the party legally responsible for handling your data. Under this route it’s Anthropic (under Anthropic’s Commercial Terms), not Google — the legal distinction the whole two-route design hangs on.
The two routes, side by side
Both routes deliver the same models on day zero, at the same pricing, on the same bill. Google’s Michael Gerstenhaber, VP of product management for Gemini at Google Cloud: “The difference comes down to who acts as your data processor and whether you need access to the new Claude platform features when they launch.”
| Feature | Anthropic-managed (new) | Google-managed |
|---|---|---|
| New models | Day zero | Day zero |
| Platform features & betas | Day zero | Arrive as Google integrates them |
| Data processor | Anthropic (Commercial Terms) | Google Cloud (existing agreements) |
| Governance extras | IAM + Audit Logging | VPC Service Controls, CMEK, US/EU/APAC data residency |
| Native integrations | Claude Console | BigQuery, Model Armor |
| Billing | Single invoice, commitment drawdown | Single invoice, commitment drawdown |
The lineup: Claude Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5, with new models shipping on the Anthropic-managed route as they launch. Endpoints are currently US and Global.
The money angle: spend commitments, not new budgets
Here’s what matters for freelancers, agencies, and small teams. A cloud commitment is money a company has already promised Google Cloud — use-it-or-lose-it credit. Claude usage on the managed route draws down that existing commitment.
So instead of asking a client to approve a brand-new AI vendor contract, you route the work through spend they’ve already committed. The invoice stays the Google Cloud invoice they already pay.
Fast mode on Claude Opus 5.5 is also GA on this route: $8 per million input tokens, $40 per million output tokens, with up to 2.5x faster output for interactive agents and live debugging. Source: Anthropic announcement, Oct 22 2026.
Those are fast-mode prices, not flat rates across models. Prompt caching and batch processing cut costs further on repeated context and bulk workloads.
One catch from the fine print: if you hold a discount, talk to your Anthropic or Google Cloud account executive before you start — discounts can’t be applied retroactively.
How it works: from zero to billed agents
How to profit from this launch
The opportunity isn’t the models — it’s the procurement. Every Google Cloud customer just got a simpler way to buy serious AI:
- Sell “agents on your existing bill.” No new vendor, no new contract, no new budget line. Build the agent, route inference through the client’s commitments, and charge a setup fee plus a monthly retainer for monitoring and tuning.
- Productize day-zero features. Managed Agents run long tasks without infrastructure babysitting; agent skills package repeatable workflows; code execution analyzes data inside an API call. Sellable services: weekly report agents, inbox triage, data cleanup.
- Price per outcome, pay per token. Opus 5.5 fast mode — 2.5x faster output at $8/$40 per million tokens — is built for interactive use: live debugging assistants, customer-facing chat, rapid prototyping.
- Offer the route audit. Regulated clients need Google-managed (VPC controls, CMEK, residency); product teams want Anthropic-managed (day-zero features). Charge for the one-hour assessment that picks correctly.
Anthropic’s own walkthrough of building on Google Cloud:
Frequently asked questions
What is Claude Models on Google Cloud (Anthropic managed)?
The full Claude Platform — models plus Managed Agents, skills, code execution, computer use — on Google Cloud, operated by Anthropic and billed on your Google Cloud invoice.
How is it different from the Google-managed version?
Who processes your data (Anthropic vs Google), when features arrive (day zero vs later), and the extras: Claude Console vs VPC controls, CMEK, residency, and BigQuery/Model Armor.
Which models are available?
Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5 — with new models shipping on the Anthropic-managed route as they launch.
What does Opus 5.5 fast mode cost?
$8 per million input tokens, $40 per million output, up to 2.5x faster output. Fast-mode prices — not flat rates for all models.
Does usage count toward my Google Cloud commitments?
Yes — it draws down your existing commitment. Holders of discounts should confirm with their account executive first: no retroactive application.
Which route should my business choose?
Anthropic-managed for day-zero features; Google-managed for strict governance — Google as data processor, VPC controls, CMEK, US/EU/APAC residency.
Keep Reading
Sources: Anthropic (Oct 22, 2026); Google Cloud Blog (Oct 23, 2026).