Skip to content
Rush Commerce
AI & Automation3 min read

Claude on Google Cloud, Anthropic managed: pick your processor

Anthropic-managed Claude on Google Cloud is GA: full Claude Platform, GCP IAM and billing, commit drawdown. The catch is who processes your data.

Claude on Google Cloud now comes in two flavors. On October 22, Anthropic made Claude Models on Google Cloud (Anthropic managed) generally available: the full Claude Platform, signed in with Google Cloud IAM and billed on your Google Cloud invoice. The older Google-managed option stays. Same models, same prices. The real choice is who processes your data, and whose terms you sign.

What actually happened

Anthropic's announcement and Google's companion post lay it out:

  • Full feature set, day zero. Managed Agents (beta), code execution, computer use, agent skills, Files API, citations, web search and fetch, the MCP connector (beta), prompt caching, batch, and the Claude Console. New features and betas land here the day they ship.
  • Google plumbing. You authenticate with existing Google Cloud IAM, keep Cloud Audit Logging, and Claude usage draws down your existing Google Cloud commitment.
  • Anthropic is the data processor, under Anthropic's Commercial Terms, running on Google Cloud infrastructure.
  • US and Global endpoints only for now. Models: Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5. Fast Mode on Opus 5.5 is GA at $8 per million input tokens and $40 per million output.
  • Discounts are not retroactive. If you have a negotiated rate, Anthropic says to call your account executive before you start.

The Google-managed option keeps Google as processor under your existing Google Cloud agreement, with VPC Service Controls, customer-managed encryption keys, US/EU/APAC residency, and native BigQuery and Model Armor hooks. New platform features reach it later.

Why Anthropic-managed Claude on Google Cloud matters for your business

This is a compliance decision dressed up as a feature list. If you need EU data residency or CMEK, the Anthropic-managed lane isn't for you yet. If you want Managed Agents and skills now, the Google-managed lane will make you wait. Pick per workload, not per company.

Update your subprocessor list. Moving from the Google-managed endpoint to the Anthropic-managed one changes who touches your customer data and which contract governs it. If you promised clients a processor list, that list just changed.

The commit drawdown is real money. A shop sitting on unused Google credits can now spend them on Claude without a separate Anthropic bill. Do the math on your actual token volume first.

Keep the endpoint a config value. Two Google lanes, plus the Anthropic API direct, plus Bedrock. If switching between them means rewriting code, you've built the lock-in yourself.

Key takeaways

  • Anthropic-managed Claude on Google Cloud went GA on October 22, 2026, with full Claude Platform features on day zero
  • It uses Google Cloud IAM, Cloud Audit Logging and Google billing, and draws down existing commitments
  • Anthropic is the data processor under its Commercial Terms; the Google-managed option keeps Google as processor
  • Only US and Global endpoints today; EU residency, CMEK and VPC Service Controls stay on the Google-managed side
  • Confirm negotiated discounts before your first call; they don't apply retroactively

Your data map should decide the endpoint, not the other way around. We build AI workflows where provider, region and processor are config you own, so a compliance change is a switch, not a rebuild. See how we build, or send us the workloads you need to place.

Sources: Anthropic, Google Cloud blog.

  • #claude
  • #google-cloud
  • #anthropic
  • #data-processor
  • #vendor-lock-in
TR

Tommy Rush — Founder, Rush Commerce

Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More

Get The Rush Report weekly — one email, zero fluff.

Keep reading

AI & Automation3 min read

Cortex XDR One: inventory the AI agents on your laptops

Palo Alto's Cortex XDR 6.0 and XDR One find AI agents, MCP servers and plugins on endpoints and block bad actions. Here's the inventory to start today.

Read it
AI & Automation3 min read

Tipalti AP Agent and MCP: It Routes Invoices, It Doesn't Pay

Tipalti's new AP Agent codes and routes non-PO invoices, and its MCP server puts AP data in Claude Enterprise. Payments still need a human. Here is what to set up.

Read it