# Landing zones: Azure AI landing zone, AWS AI landing zone, Google Cloud AI landing zone, Multi-cloud / hybrid compared

> A landing zone is the secure foundation your AI workloads “land” on: accounts or subscriptions, private networking to model endpoints, identity, policies, logging, quotas and budgets, all as code. Wit

Source: https://aibyos.com/platforms/landing

⛶ Platforms & Cloud

# Landing zones

Which to choose, and when: Azure AI landing zone, AWS AI landing zone, Google Cloud AI landing zone, Multi-cloud / hybrid.

[Talk to an expert](https://aibyos.com/contact)

A landing zone is the secure foundation your AI workloads “land” on: accounts or subscriptions, private networking to model endpoints, identity, policies, logging, quotas and budgets, all as code. Without one, every AI project reinvents security, and shadow AI appears with public endpoints and keys in code.

## Why you’d want one

-   New AI projects start in days with approved defaults
-   No public model endpoints or leaked API keys
-   Central cost, quota and audit visibility

[Find your best fit interactively →](https://aibyos.com/platforms#guide-landing)

### Azure AI landing zone

##### Choose it when

-   Built on Microsoft’s Cloud Adoption Framework landing zones
-   Azure OpenAI / AI Foundry behind private endpoints, with API Management as the AI gateway
-   Entra ID identity, Defender and Purview integration for regulated organisations

##### Watch out for

-   Model quotas are regional; plan capacity early
-   Many moving parts; policy-as-code is essential

### AWS AI landing zone

##### Choose it when

-   Control Tower / Landing Zone Accelerator with strong account isolation
-   Bedrock gives Claude, Llama, Mistral, Amazon Nova and others via VPC endpoints
-   Service control policies and guardrails to enforce region and model use

##### Watch out for

-   Model availability varies by region
-   Multi-account design needs careful planning up front

### Google Cloud AI landing zone

##### Choose it when

-   Organisation hierarchy and shared VPC from Google’s foundation blueprints
-   Vertex AI with Gemini and partner models (including Claude)
-   VPC Service Controls create a perimeter against data exfiltration

##### Watch out for

-   Best when data and teams are already on Google Cloud
-   Perimeter design can block legitimate flows if rushed

### Multi-cloud / hybrid

##### Choose it when

-   Regulatory, resilience or commercial reasons to use several providers
-   On-prem or sovereign-cloud GPUs combined with public models
-   A central AI gateway gives one policy layer across clouds

##### Watch out for

-   Highest complexity: only when there is a clear reason
-   Needs strong platform team and automation