Claude Models Now Available in Microsoft Foundry

Claude Opus 4.8 and Claude Haiku 4.5 now run inside Microsoft Foundry on Azure infrastructure, as confirmed in the official Microsoft Azure announcement and Anthropic documentation.

The models execute on NVIDIA GB300 GPUs within Azure Foundry’s multi-node clusters. Inference leverages NVIDIA’s full CUDA stack and NVLink interconnects for low-latency tensor parallelism, while Anthropic’s inference stack manages dynamic batching and KV-cache offloading across GPU memory. Users select US data regions for processing. Access occurs through the Messages API, which already supports prompt caching and extended thinking.

Azure customers keep their current login, billing, and policy controls. A single invoice covers both Azure and Claude usage. Enterprise Agreement customers can count Claude consumption against their Azure spend commitments.

This setup removes the need for separate Anthropic accounts or new payment flows. It also places Claude workloads next to other Azure AI services already used by the same organizations.

Microsoft gains a high-profile model family without building it.

Anthropic gains distribution inside an existing enterprise environment that many large buyers already trust for data residency and compliance. The arrangement sidesteps the usual friction of adding a new vendor.

The move arrives as several frontier labs test direct cloud partnerships rather than relying solely on their own hosted endpoints. OpenAI maintains a similar Azure relationship, while Google keeps most Gemini traffic on its own network. Microsoft now hosts two of the three largest model providers.

Claude users inside Azure do not lose features. The Messages API remains the entry point, so existing prompt patterns, caching configurations, and extended reasoning calls transfer without change. Anthropic continues to manage model updates and safety layers.

Enterprise buyers gain one more option when they already face pressure to consolidate vendors and show spend against committed Azure contracts. The inclusion of Claude usage in those commitments creates an immediate budget lever for teams that previously treated Anthropic as a separate line item.

For example, a global bank running fraud-detection pipelines on Azure OpenAI can now route its most complex compliance-review prompts to Claude Opus 4.8 via the same Messages API endpoint, satisfying US data-residency rules while applying the spend against its existing Azure commitment - eliminating a second procurement cycle.

Teams that need US-region inference for regulatory reasons can now satisfy that requirement inside the same Azure tenant they use for storage, identity, and other workloads. The shift reduces the number of network perimeters and audit boundaries they must maintain.

Questions remain about capacity and pricing behavior over the next quarter. Foundry customers will watch whether Microsoft adjusts reserved-capacity discounts for Claude workloads or whether Anthropic changes rate limits once traffic scales inside Azure.

Another signal to monitor is whether other Anthropic models follow the same route. If Opus 4.8 and Haiku 4.5 see strong adoption, the next logical step would be wider model availability under the same integration.

Developers already inside Azure gain a practical test bed. They can route a portion of traffic to Claude through the Messages API, compare latency and output quality against existing Azure OpenAI deployments, and decide whether to shift more prompts without creating new vendor relationships.

The integration therefore functions as both a product launch and a procurement simplification. Organizations that have delayed testing Claude because of separate billing or compliance reviews now face fewer obstacles.

The core pressure lands on teams that still manage multiple AI vendors through separate contracts and approval processes. Those teams now have a concrete path to fold one more frontier model into an existing Azure footprint.

The next measurable indicators will appear in quarterly earnings and customer case studies. Microsoft will likely highlight any new Foundry wins that include Claude usage. Anthropic will report whether enterprise adoption accelerates through the Azure channel. Both signals will show how quickly the integration moves from announcement to material workload share.

Readers who already run Azure workloads can test the new endpoints today through the Messages API. Those who manage separate Anthropic contracts can begin evaluating the consolidated billing path in the coming weeks.