Updated October 25, 2026
· ~2 min read
A year ago the model you wanted often decided the cloud. That is mostly over. Claude is on Bedrock, Vertex AI and, since June 2026, generally available in Microsoft Foundry. OpenAI's models reached general availability on Bedrock in June 2026, alongside Azure. Gemini remains exclusive to Google Cloud, and open-weight families (Llama, Mistral, Qwen, DeepSeek, gpt-oss, Gemma) are on all three in some form.
So the choice now rests on the things around the model: where your data and identity already live, how you buy capacity, what the agent runtime costs and what your procurement team has already signed.
All three state that prompts and outputs are not used to train the providers' models, and all three offer private connectivity. The differences are in routing. Bedrock runs in-Region by default, with cross-Region inference profiles that route within a geography (US, EU, APAC) or globally for more capacity. Foundry offers Global, Data Zone (US or EU) and Regional deployments; Claude on Foundry is available as global standard with a US data zone option. Vertex AI offers regional endpoints and a global endpoint. If a contract says data must stay in the EU, check the specific model and deployment type on each platform. Availability varies by model, and the newest models often launch on global routing first.
Question 1 of 3
Which cloud holds the data the model needs to read?
**Bedrock:** your data, IAM and networking are on AWS. The widest choice of capacity tiers, AgentCore for running agents, and inference that stays inside your VPC boundary through PrivateLink. **Foundry:** your company runs on Microsoft. Entra ID, Microsoft 365 data, Azure consumption commitments that Claude and OpenAI usage can draw down, and the most mature OpenAI deployment options. **Vertex AI:** you want Gemini, or your data is in BigQuery. Strong data and search integration and a free tier for Agent Engine. Multi-cloud for models alone rarely pays. A gateway in front of one platform, with a second as fallback for one or two models, is usually enough.
Production GenAI Platform on AWS
Deploy secure, scalable AI systems using AWS-native GenAI services. RAG, agent orchestration, and guardrails: production-ready.
Amazon Bedrock Enterprise AI Platform
A production AI platform on Amazon Bedrock: RAG pipelines, agent orchestration, guardrails, multi-tenant cost tracking, and the observability needed to run AI in production.
Secure LLM Gateway on AWS
One governed way for every team to call AI models: keys held centrally, budgets and rate limits per team, fallbacks between providers, and hardening that assumes the gateway is a target.
We work with startups and growing companies on AWS architecture decisions. Tell us about your situation and we will recommend the right approach.
As of October 25, 2026. Provided as is, for general information, and accurate to the best of our knowledge on the date shown. Not legal, compliance, tax, financial or other professional advice. Prices, limits and vendor services change; check current vendor documentation before relying on anything here. Terms of use.