There is no website behind www.cloud ai.com. The string is a mistyped navigational query — a space where a dot should be — and search engines resolve it to the managed cloud AI platforms of three vendors: Amazon Bedrock, Google’s Gemini Enterprise Agent Platform (the successor surface to Vertex AI) and Microsoft Foundry on Azure. If you landed here trying to reach one of them, the short answer is: pick the console that matches the cloud your workloads already run on, then compare model choice, governance and cost controls inside that console rather than migrating clouds for a single feature. This guide breaks down what each platform actually offers and how engineering teams in Portugal and the EU should choose.

What the Query Actually Resolves To

Typing www.cloud ai.com into a search box tells you something useful about the market: users do not remember vendor URLs, they remember intent. The intent behind this query is “take me to the cloud service where I can build AI applications.” Three services answer that intent, and each is a fully managed, consumption-priced platform rather than a single product.

On the AWS side, Amazon Bedrock powers generative AI for more than 100,000 organizations worldwide, according to AWS’s own product page, spanning startups through regulated enterprises. On the Microsoft side, Azure’s AI surface has consolidated under the Foundry brand. Microsoft positions Foundry as the AI app and agent factory for building, optimizing and governing AI apps and agents at scale, with SDKs, hosted agents and Visual Studio Code integration as first-class entry points.

Google has taken a parallel path. Gemini Enterprise Agent Platform is Google’s open and comprehensive platform for building enterprise-grade agents grounded in enterprise data — a repositioning of the Vertex AI developer surface around agents, model gardens and grounded retrieval rather than raw model endpoints. For a deeper look at how ambiguous “cloud ai” queries behave in practice, see our analysis of the cloud.ia query collision.

Comparing the Three Platforms

All three consoles sell the same core promise — managed access to foundation models plus the surrounding machinery of evaluation, retrieval, guardrails and observability — but their center of gravity differs. The table below summarizes the decision-relevant differences.

The honest engineering summary: for greenfield builds, model access has converged — all three catalogs now span frontier and open-weight models — so the differentiators are identity integration, data gravity and the governance tooling you will actually turn on in production.

How to Choose in the EU

European teams carry an extra decision layer: data residency, the AI Act’s obligations and vendor-specific compliance scopes. AWS states that Bedrock is in scope for ISO, SOC, CSA STAR Level 2, GDPR, FedRAMP High and is HIPAA eligible, and each hyperscaler publishes EU-region availability for its AI services. Bedrock Guardrails can block up to 88% of harmful content and identify correct model responses with up to 99% accuracy, according to AWS — figures worth validating against your own evaluation set before relying on them for a regulated workload. Our engineering guide to cloud AI under the EU AI Act covers the compliance timeline in detail.

Three residency questions should precede any console choice: which EU regions expose the specific models you need, where prompts and fine-tuning data are processed and retained, and whether the vendor commits to not training on your data. AWS, for example, states that Bedrock never stores or uses customer data to train models — get equivalent commitments in writing for every provider in your shortlist.

A Practical Selection Checklist

Work through this ordered checklist before committing a workload to any of the three consoles:

1. Map your data gravity. List where your operational data lives today; the platform co-located with that data usually wins on latency, egress cost and permissioning.
2. Verify model availability per region. Confirm in the provider’s region table that every model in your shortlist is served from an EU region you can commit to contractually.
3. Price the realistic pattern. Model your expected token volumes, caching hit rates and batch share; prompt caching and routing can change monthly spend by double-digit percentages.
4. Test guardrails on your own data. Run adversarial and hallucination checks against your evaluation set rather than accepting benchmark figures.
5. Define the exit path. Document which parts of your pipeline (prompts, evaluation harnesses, retrieval indexes) port to another provider, and keep an abstraction layer where the cost is low.
6. Assign AI Act ownership. Decide who maintains the technical documentation and risk classification before launch, not after.

The pattern that holds across all three vendors in 2026 is consolidation: consoles that used to sell model endpoints now sell agent platforms with governance built in. Choosing a cloud AI platform is therefore less about picking the best model this quarter and more about picking the operating environment — identity, data, observability and compliance — where your team can run AI workloads for the next three years without a forced migration.

Sources

• Amazon Bedrock product page — Amazon Web Services
• Microsoft Foundry documentation — Microsoft Learn
• Gemini Enterprise Agent Platform — Google Cloud Documentation