--- title: "Anthropic's Claude Opus 4.8 Hits General Availability Inside Microsoft Foundry" subtitle: "Claude Opus 4.8 and Haiku 4.5 graduate from preview to GA in Microsoft Foundry, unlocking Azure-native billing, MACC drawdown, and a new 'hosted on Azure' mode with US data residency" url: "https://alphasignal.ai/news/anthropic-s-claude-opus-4-8-hits-general-availability-inside-microsoft-foundry" author: "AlphaSignal Newsroom" author_url: "https://alphasignal.ai/editorial-team" company: "Claude" company_url: "https://alphasignal.ai/company/claudeai" published: "2026-06-29T17:56:07.000Z" updated: "2026-06-29T18:12:57.277Z" type: "News" topics: - "Api" - "Llms" subtopics: - "Long Context" upvotes: 5967 read_time: "7 min" source_url: "https://claude.com/blog/claude-in-microsoft-foundry" comments: 0 --- # Anthropic's Claude Opus 4.8 Hits General Availability Inside Microsoft Foundry > Claude Opus 4.8 and Haiku 4.5 graduate from preview to GA in Microsoft Foundry, unlocking Azure-native billing, MACC drawdown, and a new 'hosted on Azure' mode with US data residency Anthropic's Claude models have officially graduated from preview to general availability inside [Microsoft Foundry](https://azure.microsoft.com/en-us/products/ai-foundry), the company's unified AI development platform on Azure. The move is more than a label change: it introduces a new hosting mode that runs inference on Azure infrastructure operated by Anthropic, adds a US data zone for teams with residency requirements, and lets enterprise customers pay for Claude directly out of their existing Microsoft Azure Consumption Commitment budgets. ## What just shipped Two models are available at launch in the GA tier: **Claude Opus 4.8**, Anthropic's most capable generally available model for coding and agentic workflows, and **Claude Haiku 4.5**, the lightweight, cost-efficient option. Both are accessible through the Messages API and ship with prompt caching and extended thinking (Anthropic's term for giving the model extra compute time to reason through hard problems before responding) enabled from day one. The [official announcement](https://claude.com/blog/claude-in-microsoft-foundry) introduces a two-mode architecture worth understanding: - **Hosted on Azure** (new): Inference runs on Azure infrastructure, operated by Anthropic as the data processor. You get Azure authentication via Microsoft Entra ID, consolidated Azure billing, and access to a US data zone for data residency requirements. - **Hosted on Anthropic** (formerly the Foundry Preview): Routes to Anthropic's own infrastructure. Gives you access to the full API feature set and models not yet available on the Azure-hosted path. Anthropic says the goal is eventual feature and model parity between the two modes. For now, if you need a model or feature that's only on the Anthropic-hosted path, you can switch without leaving the Foundry environment. ## The billing detail that matters most to enterprises The headline technical feature is the hosting mode, but the headline procurement feature is MACC eligibility. MACC, or Microsoft Azure Consumption Commitment, is the pre-negotiated cloud spend budget that large enterprises lock in through Enterprise Agreements with Microsoft, typically spanning one to three years. Claude usage can now draw down that pre-committed Azure budget, which for a Fortune 500 buyer is the difference between a six-week vendor approval process and a one-click add to an existing contract. Pricing flows through the customer's Azure agreement with no separate Anthropic billing relationship to set up, no second set of API keys to manage, and no separate vendor security review to complete. For teams already deep in the Azure ecosystem, that removes a meaningful amount of procurement friction. ## The backstory: a $30B bet on Azure This GA launch is the commercial payoff of a partnership that has been building for months. Anthropic committed to purchase $30 billion of Azure compute capacity and to contract additional capacity up to one gigawatt, scaling its rapidly growing Claude AI model on Microsoft Azure powered by NVIDIA. That is not a typical vendor relationship. It is a structural commitment that aligns Anthropic's own training and inference infrastructure with Azure's buildout. As part of the broader partnership, NVIDIA and Microsoft committed to invest up to $10 billion and up to $5 billion respectively in Anthropic. The financial entanglement makes the GA launch feel less like a product decision and more like the operational consequence of a strategic alliance that was already locked in. The partnership also has a competitive logic for Microsoft. Microsoft's defining 2026 move is a three-way model hedge: it holds a stake in OpenAI valued at roughly $135 billion, has embedded Anthropic's Claude inside Copilot, and is shipping its own in-house MAI models. No other company sells frontier AI from three rival sources at once. Adding Claude to Foundry as a GA, MACC-eligible option strengthens that position, especially as OpenAI formally ended its exclusive cloud partnership with Microsoft in April 2026, and within days Amazon invested $50 billion to bring OpenAI's models to AWS Bedrock. ## Who wins, who watches carefully The clearest winners are Azure-native enterprise teams that have been evaluating Claude but couldn't get it past procurement. Anthropic's enterprise market share jumped from 12% to 32% in 2025 as organizations began optimizing for price-performance rather than single-vendor lock-in. GA status and MACC eligibility removes the last major procurement barrier for the Azure-heavy slice of that market. Early adopters are already vocal. Momentic, an AI-powered testing platform, reports serving millions of tokens per minute with the reliability their customers depend on after moving to Azure Foundry. Everstar, a nuclear safety analysis company, compressed a safety analysis that would have taken 200 human days into a single day using the combination of Claude and Azure infrastructure. The more complicated picture is for teams with strict EU data residency requirements. Claude models currently run on Anthropic-managed infrastructure, not on Azure's regional compute. That is a different operational profile from OpenAI models, which run natively on Azure. EU-native inference support in Foundry is listed as "Coming 2026" with no more specific date. Until that lands, EU-regulated workloads may still need to route through AWS Bedrock or GCP Vertex AI, where Anthropic models run inside the cloud provider's own regional infrastructure. ## What developers need to know to get started The integration uses Anthropic's standard API surface, so existing code requires minimal changes. Foundry is supported by the C#, Java, PHP, Python, and TypeScript SDKs. The Go and Ruby SDKs do not currently support Microsoft Foundry. Authentication is handled through Microsoft Entra ID rather than Anthropic API keys. A starter kit uses a single `azd up` command to provision a Foundry account, project, and Claude model deployments using either Bicep or Terraform, then wires the Anthropic SDK and the Claude Code CLI to call your deployment over Microsoft Entra ID with no API keys to manage. The endpoint shape is straightforward: ``` import anthropic from azure.identity import DefaultAzureCredential # Authenticate via Azure credentials (no Anthropic API key needed) credential = DefaultAzureCredential() token = credential.get_token("https://cognitiveservices.azure.com/.default") client = anthropic.Anthropic( base_url="https://.services.ai.azure.com/anthropic", api_key=token.token, ) message = client.messages.create( model="claude-opus-4-8", max_tokens=1024, messages=[{"role": "user", "content": "Explain prompt caching."}] ) ``` A few things worth pinning before you roll out to a team: model aliases like `sonnet` and `opus` resolve to Claude Code's built-in default for Foundry, which can lag the newest release and may not yet be available in your account. Foundry has no startup model check, so requests fail silently when the default is unavailable. Pin specific model version strings in your deployment config. Foundry uses a two-level hierarchy: resources contain your security and billing configuration, while deployments are the model instances you call via API. You'll first create a Foundry resource, then create one or more Claude deployments within it. ## What comes next Anthropic has signaled that more capabilities and models are on the way for the Azure-hosted path, with the explicit goal of closing the gap with the Anthropic-hosted option. The two immediate things to watch: EU data residency support (currently undated but publicly committed to for 2026), and whether additional Claude models, including the newer Opus and Sonnet variants, make it into the Azure-hosted tier. Claude Fable 5, Claude Opus 4.7, Claude Opus 4.6, and Claude Sonnet 4.6 already have a 1M-token context window on Microsoft Foundry in the Anthropic-hosted path, suggesting the pipeline for Azure parity is already in motion. For teams building production agents on Azure, the practical question is no longer whether Claude is available, but which hosting mode fits your compliance posture and whether your workload can tolerate the current data routing model while EU support catches up. ## Sources - [X post by @claudeai](https://x.com/claudeai/status/2071653958905467027) - [Claude in Microsoft Foundry is now generally available | Claude by Anthropic](https://claude.com/blog/claude-in-microsoft-foundry) - [Claude in Microsoft Foundry - Claude Platform Docs](https://platform.claude.com/docs/en/build-with-claude/claude-in-microsoft-foundry) - [Microsoft, NVIDIA and Anthropic announced new strategic partnerships | Anthropic](https://www.anthropic.com/news/microsoft-nvidia-anthropic-announce-strategic-partnerships) - [Anthropic's Claude Becomes First-Party Option in Azure AI Foundry | Krasa.ai](https://www.krasa.ai/news/anthropic-claude-first-party-azure-ai-foundry-microsoft-build-2026) - [Deploy and use Claude models in Microsoft Foundry | Microsoft Learn](https://learn.microsoft.com/en-us/azure/foundry/foundry-models/how-to/use-foundry-models-claude)