Skip to content

Claude direct or through AWS, Google or Microsoft?

[PUBLISHED_DATE] 2026.11.20
[READ_TIME] 6 MIN
[AUTHOR_ID] MARKETING OPS

It depends on who you want processing the data, where, and which features you need. On Amazon Bedrock and Google Cloud, the cloud provider is the data processor; on Anthropic's own API, Claude Platform on AWS and Microsoft Foundry, Anthropic is. The clouds add residency options and reuse contracts you already have, but they don't offer every feature.

One clarification first: this decision applies to applications your team builds on the API and to tools like Claude Code. The Claude chat app for staff, on the Team and Enterprise plans, is bought from Anthropic.

What changes with each option?

As of October 2026, according to Anthropic's documentation:

Option Who operates it Data processor Residency What to know
Anthropic's API Anthropic Anthropic Global or US only Every feature. US-only inference costs 1.1 times the standard rate for recent models
Claude Platform on AWS Anthropic, with AWS billing and access Anthropic Same options as Anthropic's API Same data policies as Anthropic's API; billed through AWS Marketplace
Claude in Amazon Bedrock AWS AWS Global or regional, with a 10% premium for regional No Anthropic personnel access to the inference infrastructure. Doesn't include, among others, Agent Skills, the MCP connector, code execution or the Compliance API
Claude on Google Cloud (Agent Platform, formerly Vertex AI) Google Google Global, multi-region (US or EU) or regional, with a 10% premium for the last two Doesn't include, among others, Agent Skills, the MCP connector, code execution or the Compliance API
Claude in Microsoft Foundry Anthropic, on Azure or Anthropic infrastructure Anthropic, as an independent processor for Microsoft Global or US Data Zone When hosted on Azure, prompts and completions stay in Azure; only usage metadata and content flagged by safety systems go to Anthropic. Doesn't include the Compliance API

What matters most at a bank, insurer or health system?

Four questions frame the decision:

  1. Who do you already have a contract and due diligence with? If your organization has already vetted and contracted AWS, Google Cloud or Microsoft, using Claude there can reuse that file, the billing and the identity controls. Confirm the contract covers the specific AI service.
  2. Who do you want as the data processor? If your policy requires data to stay with the cloud provider you already audit, Bedrock or Google Cloud handles that. If you want the direct relationship with Anthropic and the full feature set, use Anthropic's API.
  3. Where must data be processed? Every option offers a US choice, and each one adds cost.
  4. Which features do you need? If your use case relies on skills, MCP connectors, code execution or the Compliance API for audit, confirm the platform has them before you decide.

For banks, this is a third-party decision like any other. The interagency guidance on third-party relationships (OCC Bulletin 2023-17) applies whether you contract Anthropic directly or reach Claude through a cloud provider; what changes is which contract and which controls you're relying on. We cover the questions to ask in AI vendor due diligence.

What's the HIPAA path on each platform?

The path determines which business associate agreement (BAA) covers you:

  • Anthropic's API: Anthropic's BAA, with HIPAA readiness enabled on the organization.
  • Amazon Bedrock: AWS's BAA. Bedrock is on AWS's list of HIPAA eligible services, and you have to configure it consistent with HIPAA. Anthropic's own documentation points regulated organizations that need HIPAA, FedRAMP High, IL4 or IL5 to Bedrock rather than Claude Platform on AWS.
  • Claude Platform on AWS: Anthropic's documentation says HIPAA readiness isn't available.
  • Microsoft Foundry and Google Cloud: confirm in writing with Anthropic and the cloud provider what your agreement covers before sending PHI, because the documentation isn't consistent.

We go deeper in is Claude HIPAA compliant?

What do we recommend?

There's no single answer. As a starting point:

  • If you already have a contract and governance on one cloud, start with Claude on that cloud and confirm it has the features you need.
  • If you need the full feature set, such as skills, connectors or the Compliance API, use Anthropic's API.
  • Either way, use enterprise accounts managed by the organization, never personal accounts, and document the decision in the vendor file.

An honest caveat: the features available on each cloud change every few weeks. Review the decision every quarter.

How do we handle it at Andes?

We work with Claude every day and help IT and compliance choose the path, deploy in your own cloud when required, and document the decision for compliance to review (how we work with Claude). If you're still choosing how to manage model risk, read model risk management for generative AI.

Do you know which path your AI takes?

If every project picks its own cloud, you need one organizational decision. See our AI for regulated institutions service and let's make it with your team.


Sources

  • Claude Docs, API and data retention: https://platform.claude.com/docs/en/manage-claude/api-and-data-retention
  • Claude Docs, Claude in Amazon Bedrock: https://platform.claude.com/docs/en/build-with-claude/claude-in-amazon-bedrock
  • Claude Docs, Claude Platform on AWS: https://platform.claude.com/docs/en/build-with-claude/claude-platform-on-aws
  • Claude Docs, Claude on Google Cloud: https://platform.claude.com/docs/en/build-with-claude/claude-on-vertex-ai
  • Claude Docs, Claude in Microsoft Foundry: https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry
  • Claude Docs, Data residency: https://platform.claude.com/docs/en/manage-claude/data-residency
  • AWS, HIPAA Eligible Services Reference: https://aws.amazon.com/compliance/hipaa-eligible-services-reference/
  • OCC Bulletin 2023-17, Third-Party Relationships: Interagency Guidance on Risk Management: https://www.occ.gov/news-issuances/bulletins/2023/bulletin-2023-17.html
Direct_Answers

Frequently asked questions

The cloud provider. According to Anthropic's documentation, on Amazon Bedrock and Google Cloud's Agent Platform, AWS or Google is the data processor and their terms and controls apply. On Anthropic's own API, Claude Platform on AWS and Microsoft Foundry, Anthropic is the processor.

Amazon Bedrock is on AWS's list of HIPAA eligible services, so it can be covered under your AWS BAA, as long as you configure it consistent with HIPAA. Anthropic's own BAA does not apply on Bedrock, because AWS is the processor there.

Yes. According to Anthropic's documentation, US-only inference on its API costs 1.1 times the standard rate for recent models, and regional endpoints on Amazon Bedrock and Google Cloud carry a 10% premium over global ones.

No. On Amazon Bedrock and Google Cloud, features such as Agent Skills, the MCP connector, code execution and the Compliance API are not available. Check that the features you need are on the platform before you choose.

READY FOR THE NEXT DEPLOYMENT?

Start a Technical Consultation