Azure OpenAI Service

Azure OpenAI, now sold as Azure OpenAI in Foundry Models, gives a company access to OpenAI's GPT models inside its own Azure subscription. Developers use it to add chat, coding, and document features to internal and customer-facing apps while keeping data inside Azure's network. Admins log in through Microsoft Entra ID or an API key and can restrict traffic to private endpoints and virtual networks, avoiding the open internet. Microsoft states the service carries over 100 compliance certifications, including SOC 2 Type 2, ISO 27001, and HIPAA support, and publishes a 99.9 percent uptime commitment. Pricing follows either pay-as-you-go per token or reserved provisioned throughput for steady, high-volume workloads.

This fits a company that already runs on Azure and needs OpenAI's models under its own compliance and networking controls rather than calling OpenAI directly. A common starting point is a pilot inside one business unit, connecting the model to internal documents through Azure AI Search, then expanding to other teams. As a planning figure, pay-as-you-go pricing runs close to OpenAI's own published rates, generally a few dollars per million tokens, while provisioned throughput charges a flat hourly rate for guaranteed capacity once usage is steady. Access to the newest models sometimes needs an approval request and can lag OpenAI's own API by weeks, and regional availability varies by model, so confirm your Azure region offers the specific model and version you need.