Anthropic and Microsoft announced general availability of Claude models in Microsoft Foundry, giving Azure customers access to Claude Opus 4.8 and Claude Haiku 4.5 with native authentication, billing, and governance. Usage draws down existing Microsoft Azure Consumption Commitments. Sonnet 5 followed days later at promotional pricing of $2/$10 per million input/output tokens through August 31.
The announcement framing emphasizes enterprise readiness: Entra ID authentication, role-based access controls, Azure billing, and governance within the existing Azure environment. For teams that have been blocked by procurement and vendor onboarding, this removes real friction. The ability to spend budget already committed through MACC without opening a new vendor relationship is the practical unlock.
But the practitioner reaction across LinkedIn and Reddit overwhelmingly focused on what the announcement does not say: where the data goes.
Jiri Forman, Head of Transformation at CETIN, asked the question that got 49 reactions on LinkedIn, more engagement than any other comment in the thread:
Hosted on Azure with all data processing possible within EU?
Murat Yasartas, Global AI Team Lead at Sobi, answered:
Unfortunately it's not, data zone is only US based for now.
Microsoft's own documentation confirms the architecture. Even with the "Hosted on Azure" option, Anthropic remains the "independent data processor" for prompts and outputs. Processing is scoped to "Global" or "DataZone" deployment options, but no European data zone exists for Claude models today. The documentation also notes that automatic safeguards can flag content for Anthropic Trust and Safety review, meaning customer data can leave the Azure boundary on an exceptions basis even under the Azure-hosted option.
On Reddit, a practitioner working with Dutch financial services was direct:
My client, a big Dutch bank, does not allow the use of Anthropic models through Foundry due to this reason.
Radubogdan confirmed the specifics:
Anthropic models 'hosted by Azure' are deployed in Foundry Sweden. Deployment type is 'Global Standard' so inference can happen anywhere.
The Sweden deployment gives a European endpoint address, but "Global Standard" means the actual inference may be routed to US infrastructure.
The contrast with OpenAI models on Azure is the friction point European architects keep hitting. Gregor Beuster, a Member of Technical Staff, noted on LinkedIn:
Still a marketplace model with Anthropic being the operator, which will ruffle some feathers in German enterprise. OpenAI models being Azure native are much easier to get approved.
The distinction matters because OpenAI models on Azure are first-party: Microsoft operates inference, data stays within the Azure trust boundary, and EU data zone deployments are available. Claude models on Foundry are third-party marketplace offerings: Anthropic operates inference as the data processor regardless of hosting option, and the US CLOUD Act applies to Anthropic as a US company.
Anthropic's own documentation makes the gap explicit. The third-party deployment documentation scopes its data residency and compliance guidance specifically to Vertex AI and Bedrock deployments, stating "this section applies when using Vertex AI or Bedrock" and noting that "inference runs in your cloud tenant." Foundry is not included in that scope, meaning the data residency guarantees that European enterprises rely on when using Claude through AWS Bedrock or Google Vertex AI do not extend to Foundry today. Anthropic's regional compliance page lists Microsoft Foundry in Europe as "Coming 2026" but provides no specific date. A Microsoft Q&A question asking for a more specific timeline has been open since April with zero answers.
Thomas Walawgo, an IT architect in ambulatory oncology, captured the gap on LinkedIn:
Today we use Azure OpenAI because guaranteed EU-hosted inference is a key compliance requirement for us. Will Microsoft Foundry enable Claude to run with guaranteed EU-hosted inference?
No answer has been provided.
Alistair Doran, Head of Digital Product Management at BDO, was sharper:
It is very, very disappointing to see it's not available in the European or UK region. And considering what's going on between Anthropic and the US government, you would have thought they might have started prioritising Europe a little bit more.
Even setting aside data residency, capacity is a problem. Jannik Reinhard, a dual Microsoft MVP, pointed out the gap between "generally available" and actually deployable:
For me it is GA if there is capacity in the MS data centers. For now I always have to request this via a form and need luck to get it approved. This is for me not a professional service.
Karl Wirén echoed this:
In practice we still have to request capacity through a form and hope it gets approved, which isn't what we'd expect from a production-grade GA offering.
The Foundry GA is a meaningful step for US-based enterprises and teams without strict data residency requirements. The procurement barrier is real, and removing it matters. But for European enterprises operating under GDPR, financial regulation, or healthcare data requirements, the announcement changes little. The model they need is available on a platform they trust, operated by a company in a jurisdiction they cannot fully control, with inference routing they cannot guarantee stays in Europe.