---
type: analysis
slug: on-google-cloud-google-s-lock-in-tools-now-come-only-with-the-slower-claude
title: "On Google Cloud, Google's lock-in tools now come only with the slower Claude route"
description: "Anthropic's new Google Cloud route gives customers Google IAM, a Google invoice, commitment drawdown and same-day Claude features. Model Armor, Agent Engine, regional residency and Google as data processor stay on the Google-managed route. Every team that wants the newest agent features now has to give up Google's own integrations to get them."
published_at: 2026-10-26T18:00:00.000Z
topics: ["google","anthropic","claude"]
source_urls: ["https://claude.com/blog/claude-models-on-google-cloud-anthropic-managed","https://claude.com/resources/articles/claude-models-on-google-cloud-anthropic-managed","https://platform.claude.com/docs/en/about-claude/models/overview","https://code.claude.com/docs/en/mcp"]
author: "Andrey Reshetnikov"
author_profile: https://newruntime.com/owner-profile.md
---
# On Google Cloud, Google's lock-in tools now come only with the slower Claude route

Anthropic's new Google Cloud route gives customers Google IAM, a Google invoice, commitment drawdown and same-day Claude features. Model Armor, Agent Engine, regional residency and Google as data processor stay on the Google-managed route. Every team that wants the newest agent features now has to give up Google's own integrations to get them.

## Summary

Anthropic has made Claude Models on Google Cloud (Anthropic managed) generally available. Google Cloud customers get the Claude Platform feature set using Google Cloud IAM, one Google Cloud invoice and drawdown of existing Google commitments. Anthropic runs the service, acts as data processor, and says new features and betas will arrive the same day as on its native platform. The Google-managed Claude route continues for customers who need Google as processor, regional data residency, or Model Armor and Agent Engine.

## What happened

Google Cloud now has two Claude routes, and the announcement spells out exactly how they differ.

The Anthropic-managed route is, according to Anthropic's technical documentation (d2), the Claude API running on Anthropic infrastructure and reached through a Google Cloud project. Google provides access and billing. According to the announcement, customers get Google Cloud IAM credentials, a single Google Cloud invoice and commitment retirement, and Anthropic promises features and betas on the same day as the native platform. Google's own documentation (d1) confirms immediate access to Anthropic's latest models, features, research previews and betas, and says usage can draw down Google Cloud commitments. The listed capabilities are Managed Agents (beta), code execution, computer use, agent skills, the MCP connector (beta), Files API, citations, web search, web fetch, prompt caching and batch processing. The announcement says the route is served through US and Global endpoints and names Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 as available models. Anthropic is the data processor.

The Google-managed route stays. Anthropic points customers there if they need Google Cloud as data processor, strict regional data residency, or native Google Cloud integrations such as Model Armor or Agent Engine.

On pricing, the announcement says Fast Mode for Opus 5.5 is GA on this offering, claims up to 2.5x faster output-token generation, and gives Anthropic's list price of $8 per million input tokens and $40 per million output tokens. Anthropic's models overview lists Haiku 5.5 from $0.10/$0.50 per million tokens. These are Anthropic's Claude Platform list prices. The sources give no general pricing for the Google route and do not say which existing discounts apply.

## What the archive adds

The archive shows Anthropic has done this before. Its AWS partner page (d3) lists "Claude Platform on AWS," described as access to the complete Claude Platform inside the AWS environment. Google Cloud now follows the same pattern: a cloud-managed Claude listing next to an Anthropic-operated one that the cloud bills for. Anthropic's models overview also lists matching model IDs for Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS. So Anthropic already keeps one model naming scheme across several cloud channels, and the routes differ in features and contract terms, not in model names. The archive has no data on how customers divided between native and cloud-managed routes on AWS. It also has no data on whether the cloud-managed routes got features later than the native platform. That leaves "same-day" as a promise with no track record behind it.

## Trend signal

Model labs are separating the commercial relationship from the runtime. Enterprises like to buy AI through committed cloud spend, because procurement has already approved that contract and the money must be used. Labs, meanwhile, release agent features (managed agents, MCP connectors, computer use, skills) at a pace a re-hosted listing has trouble matching. The "Anthropic runs it, the cloud bills it" model meets both needs: procurement keeps its contract and IAM, and the lab keeps the runtime. AWS (d3) and now Google Cloud show it repeating. Our read is that a cloud's own AI integrations (guardrails, agent runtimes, residency controls) are becoming the main way it holds customers once the model runtime belongs to someone else.

## Why now

1) Enterprise AI budgets sit inside cloud commitments. Anthropic's and Google's documentation (d1, d2) both lead with Google billing, IAM and commitment drawdown, which removes the need for a separate procurement contract. 2) Day-one features are the selling point. The announcement and Google's documentation (d1) both stress immediate access to betas and research previews, which only works if Anthropic runs the service. 3) Precedent: Anthropic's AWS partner page (d3) already lists Claude Platform on AWS, so offering the same route on Google Cloud fits an existing pattern. 4) Customer segmentation: the announcement itself assigns residency, Google-as-processor and Model Armor/Agent Engine needs to the Google-managed route, which lets both companies offer two products instead of one compromise.

## Connections to other events

The new part of this launch is the explicit split, not the availability. Everything Google uses to tie a Claude workload to its platform sits on one route: Model Armor for guardrails, Agent Engine for agent hosting, regional residency, and Google as processor. Everything that makes Claude's agent features current sits on the other: same-day betas, Managed Agents, the MCP connector, agent skills. The announcement itself sorts customers this way.

Our inference about Anthropic's motive is that it gets to control the release schedule on an important cloud. On the Google-managed route, how fast Claude features arrive depends on Google. On this route, Anthropic sets the schedule and holds the data-processor role. We infer that the processor role also gives Anthropic the operating relationship with the customer: support, quotas, and visibility into which workloads grow. Neither company says this.

What does this mean for Google's lock-in? Google keeps the invoice, IAM and commitment drawdown, so Claude spend still counts toward Google contracts. That is likely why Google agreed, though this is inference. It also likely keeps that spend from moving to a direct Anthropic contract or to AWS. What Google loses is the integrations it uses to keep customers. A team that picks the Anthropic route for agent features is choosing Anthropic's agent stack (Managed Agents, the MCP connector) over Agent Engine, and some other guardrail layer over Model Armor. Google's lock-in shrinks to billing and identity, which are the easiest layers to move later. The cost is unequal: the customers who can least give up Google's integrations, regulated ones that need residency or Google as processor, are the ones held on the route without the same-day feature promise.

## Why it matters

For builders on Google Cloud, the choice of route is now an architecture decision. If your agents depend on the newest Claude features, the Anthropic-managed route is the only one with a same-day promise. In exchange you give up Model Armor, Agent Engine and residency beyond the US and Global endpoints, so you will need guardrails and agent hosting that don't come from Google. If you need Google as processor or in-region data, you stay on the Google-managed route and should plan for features that may arrive later. The sources don't measure that lag.

For buyers, the immediate gain is that Claude spend on the Anthropic route can retire Google commitments. Before moving anything, ask your account team which existing discounts apply, because the sources don't say. For platform teams, Anthropic's shared model IDs across clouds make routing simple. The real work is tracking which features each route supports.

## What happens next

- **By 2027-04-30, at least one Claude Platform feature or beta will appear in Anthropic's Anthropic-managed Google Cloud documentation (d2) while Google's partner-model docs for the Google-managed Claude route do not list it, and the gap will last 30 days or more.** (6 months, 60%). Watch for: Compare the feature list on platform.claude.com's Anthropic-managed page with Google Cloud's Google-managed Claude docs every month. Reason for the probability: the same-day promise applies only to the Anthropic route, and Managed Agents and the MCP connector are still in beta.
- **By 2027-10-31, Anthropic's documentation for the Anthropic-managed Google Cloud route will list an endpoint other than US and Global (for example an EU region).** (12 months, 30%). Watch for: A new regional endpoint or data-residency wording on the d2 documentation page. Reason the probability is low: the announcement explicitly sends strict-residency customers to the Google-managed route, which suggests no regional endpoint is planned soon.
- **By 2027-04-30, Google Cloud documentation will state that Model Armor or Agent Engine works with the Anthropic-managed Claude route.** (6 months, 25%). Watch for: Google's Anthropic-managed page (d1) or the Model Armor/Agent Engine docs naming the Anthropic-managed route. Reason the probability is low: the announcement presents these integrations as reasons to choose the Google-managed route.

## Limits and uncertainty

The sources don't give general pricing for the Google route, rate limits or quotas on either route, which discounts apply, or plans for regional endpoints. "Same-day parity" is a vendor promise, and nothing yet shows how it holds up on this route. The sources also say nothing about whether the Google-managed route actually trails on features, so the size of the gap is unknown. Our claims about Anthropic gaining the customer relationship and usage visibility, and about Google agreeing in order to keep spend from leaving, are inferences; neither company says either. Google may also add Model Armor or Agent Engine support to the Anthropic route later, which would close the split this take is built on. We have no data on how many customers will move between routes.

## Skill to build now

Build a capability test harness for each route. First exercise: write a script that, for each route you can reach (for example Claude Platform direct, Anthropic-managed Google Cloud and Google-managed Google Cloud), sends one minimal request per feature: prompt caching, batch processing, Files API upload, web search, the MCP connector and a Managed Agents call. For each, record pass or fail, the error text and the date in a table. Run it weekly on a schedule so the table shows how features differ between routes over time, and when features arrive on each.

## Reader takeaway

If you run Claude on Google Cloud, write down your hard constraints before your next commitment renewal: processor identity, residency region, and any dependence on Model Armor or Agent Engine. If none of them applies, test the Anthropic-managed route on one non-regulated agent workload and compare which features actually work. If any applies, stay on the Google-managed route and track feature arrival dates so you have numbers when you negotiate.

## Open question

Once Anthropic runs Claude inside Google's billing, will Google respond by adding Model Armor and Agent Engine to the Anthropic-managed route, or by keeping those integrations for the Google-managed Claude route and its own Gemini models? In other words, will Google compete on integrations or use them to keep customers in place?

## Sources

- [Claude Models on Google Cloud (Anthropic managed) is generally available](https://claude.com/blog/claude-models-on-google-cloud-anthropic-managed)

- [Claude Platform is now available on Google Cloud | Claude by Anthropic](https://claude.com/resources/articles/claude-models-on-google-cloud-anthropic-managed)

- [Models overview](https://platform.claude.com/docs/en/about-claude/models/overview)

- [Claude Code CLI gains MCP Tasks for durable long-running jobs and Elicitations for mid-task user input.](https://code.claude.com/docs/en/mcp)

- [Use Claude models on Google Cloud (Anthropic managed)](https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/partner-models/claude/anthropic-managed)

- [Claude Models on Google Cloud (Anthropic managed)](https://platform.claude.com/docs/en/build-with-claude/claude-models-on-google-cloud/anthropic-managed)

- [Azure OpenAI frequently asked questions](https://learn.microsoft.com/en-US/azure/cognitive-services/openai/faq)

- [Agent Platform Pricing | Google Cloud](https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing)

- [Introducing Claude Opus 5.5 | Anthropic](https://www.anthropic.com/claude-opus-5-5)

- [The 2026-07-28 Specification | Model Context Protocol Blog](https://blog.modelcontextprotocol.io/posts/2026-07-28/)

- [The AWS MCP Server is now available in six additional AWS Regions](https://aws.amazon.com/about-aws/whats-new/2026/10/aws-mcp-server-six-additional-regions/)

- [Google Cloud MCP servers stability commitment](https://docs.cloud.google.com/mcp/mcp-gcp-stability-commitment)

Author: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Google](https://newruntime.com/topics/google.md): Explore the google topic hub.
- topic hub: [Anthropic](https://newruntime.com/topics/anthropic.md): Explore the anthropic topic hub.
- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- collection: [Analysis](https://newruntime.com/analysis.json): Open the latest authorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)