Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Sovereign AI Reference Architectures for Google Cloud Dedicated

What is Google Cloud Dedicated?

Google Cloud Dedicated (GCD) creates a separate, standalone 'universe' of Google Cloud services, based on dedicated hardware, with sales, support, and operations by a local partner. It is engineered specifically for regulated industries like the public sector and healthcare with strict sovereignty requirements. While it is built on the same industry-leading technologies that power Google's public cloud, the Google Cloud Dedicated instances in France and Germany are operated exclusively by fully independent local providers (such as S3NS in France).

This hybrid approach offers the best of both worlds by wrapping hyperscale innovation in local control. Google Cloud Dedicated translates this sovereignty from a policy objective into a reality. A key benefit of this architecture is that customers can receive the latest in cloud features, while knowing that all updates are strictly monitored, reviewed, inspected, and approved by the local partner. Further, given the exclusively local and independent operations, Google does not have any access to customer data nor can it disrupt service availability to customers.

Sovereign AI in Practice: A Reference Implementation

Google Cloud Dedicated brings Google’s advanced, optimized AI stack directly into the dedicated sovereign infrastructure. Because your data lifecycle is entirely contained within the local region, model training and inference happen locally. Your proprietary datasets and strategic insights never leave the platform boundary, ensuring you can lead with cutting-edge AI agents while remaining perfectly compliant with regional mandates.

This repository is intended to get you started with Google Cloud Dedicated by providing practical examples of AI solutions and more broadly, acquainting you with the product. The demonstration contained herein is the first of many that show how Google Cloud Dedicated can meet stringent sovereign needs, answering key questions such as: how can I innovate with my data?

Please note that these examples are provided as educational demos; they are not audited, optimized, or intended for production deployment.

We are actively working to expand this repository with additional examples over the next couple of months. Stay tuned.

Sovereign Tax Anomaly Detection with BigQuery ML & Gemma

This Tax Anomaly Detection sample application leverages BigQuery ML and a RAG pipeline with Google’s open-weight Gemma LLM to optimize tax enforcement while ensuring data sovereignty. Built for regulated industries and the public financial sector, the architecture guarantees local data residency by running all operations within Google Cloud Dedicated infrastructure.

As the first in a series of reference architectures for regulated public sectors, this application provides a blueprint for secure, localized AI innovation.

Follow the Tax Anomaly Detection - User Guide for a step-by-step walkthrough to deploy the infrastructure and workflow.

Sovereign Health Insurance Risk Analysis with BigQuery ML & Gemma

This Health Insurance Risk Analysis sample application demonstrates how to optimize health insurance risk analysis and claims verification while ensuring data sovereignty. By leveraging BigQuery ML for health risk modeling and a regional open-weight Gemma LLM for claims document verification, the architecture guarantees local data residency within Google Cloud Dedicated infrastructure.

As another key reference architecture for highly regulated sectors, this application provides a blueprint for secure, localized data modeling and automated claim checks.

Follow the Health Insurance Risk Analysis - User Guide for a step-by-step walkthrough to deploy the infrastructure and verification workflow.

Sovereign Multi-Universe Federation with Bank of Anthos

This Multi-Universe Federation sample application demonstrates how highly regulated organizations, such as retail banks and financial institutions, can achieve multi-universe digital resilience and data sovereignty by federating workloads across standard Google Cloud Platform (GCP) and Google Cloud Dedicated (GCD) environments. By leveraging continuous database replication, Workforce Identity Federation (WIF), and Storage Transfer Service (STS) object sync, the architecture maintains a near-real-time "Sovereign Standby" mirror of a core retail banking stack (Bank of Anthos).

As a reference architecture for disaster recovery and operational continuity, this application provides a blueprint for deterministic emergency failover, ensuring core operations can run completely isolated on GCD when needed.

Follow the Multi-Universe Federation - User Guide for a step-by-step walkthrough to deploy the cross-universe infrastructure, database replication, and failover workflow.

General Prerequisites & Setup

Before starting, you must be onboarded on Google Cloud Dedicated (GCD) by your partner of choice and have your identity provider and permissions configured. Note that this is a prerequisite for all other steps. You can refer to the S3NS documentation as an example of the organisation setup process.

Configuration Reference

The following table provides a detailed Configuration Reference across various GCD environments, as compared to GCP. Please replace the placeholders in the commands and configurations given in each demo README with the values corresponding to your target universe.

Variable GCP GCD France (GA) GCD Germany (Preview)
UNIVERSE_WEB_DOMAIN cloud.google.com cloud.s3nscloud.fr cloud.berlin-build0.goog
UNIVERSE_API_DOMAIN googleapis.com s3nsapis.fr apis-berlin-build0.goog
UNIVERSE_NAME google s3ns berlin
UNIVERSE_PREFIX n/a s3ns eu0
UNIVERSE_REGION n/a u-france-east1 u-germany-northeast1

Google Cloud CLI

Initialize the gcloud CLI for your GCD universe. Use the values from the Configuration Reference table to replace the <PLACEHOLDERS> in the commands below.

First, create a Workforce Identity Federation (WIF) login configuration:

AUDIENCE=locations/global/workforcePools/mypool/providers/myprovider
# Replace with values from the Configuration Reference table
UNIVERSE_WEB_DOMAIN="<UNIVERSE_WEB_DOMAIN>"
UNIVERSE_API_DOMAIN="<UNIVERSE_API_DOMAIN>"
UNIVERSE_NAME="<UNIVERSE_NAME>"
WF_POOL_FILE_PATH="/tmp"

gcloud config configurations create $UNIVERSE_NAME
gcloud config configurations activate $UNIVERSE_NAME
gcloud config set universe_domain $UNIVERSE_API_DOMAIN

gcloud iam workforce-pools create-login-config $AUDIENCE \
  --universe-cloud-web-domain="$UNIVERSE_WEB_DOMAIN" \
  --universe-domain="$UNIVERSE_API_DOMAIN" \
  --output-file="$WF_POOL_FILE_PATH/wif-login-config-$UNIVERSE_NAME.json" \
  --activate

Once the above file has been created and the gcloud profile configured, run the following command to login to the organization with gcloud. This will prompt a web browser that will allow login to the organization with the configured IdP.

gcloud auth login \
  --login-config=$WF_POOL_FILE_PATH/wif-login-config-$UNIVERSE_NAME.json \
  --no-launch-browser

gcloud auth application-default login \
  --login-config=$WF_POOL_FILE_PATH/wif-login-config-$UNIVERSE_NAME.json

Copyright 2026 Google LLC

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    https://www.apache.org/licenses/LICENSE-2.0


Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.

About

Sovereign AI Reference Architectures for Google Cloud Dedicated

Resources

Contributing

Stars

3 stars

Watchers

0 watching

Forks

Used by

Contributors