Microsoft announced ten major innovations for Azure AI Foundry at Microsoft Build 2025 on May 19, revealing the platform now supports more than 70,000 customers, processes 100 trillion tokens per quarter, and powers 2 billion daily enterprise search queries. Corporate Vice President Asha Sharma positioned the platform as transforming software development from months-long processes to production solutions deliverable in days.
The expanded model catalogue includes Grok 3 from xAI available immediately, with Flux Pro 1.1 from Black Forest Labs and Sora via Azure OpenAI coming soon in preview. Microsoft now offers over 10,000 open-source models from Hugging Face in Foundry Models, supporting full fine-tuning including LoRA/QLoRA and DPO methods. A new developer tier for fine-tuning eliminates hosting fees for experimentation.
Azure AI Foundry Agent Service achieved general availability, with more than 10,000 organisations including Heineken, Carvana, and Fujitsu already using the platform to automate complex business processes. The fully managed service provides ready-to-use templates, actions, and connectors to more than 1,400 enterprise data sources including SharePoint, Microsoft Fabric, and third-party systems.
Multi-agent orchestration enables connected agents to call each other as tools, with new workflows providing stateful layers for managing context and handling long-running processes like financial approvals and supply chain operations.
Agentic retrieval in Azure AI Search introduces a multi-turn query engine that improves answer relevance by up to approximately 40% on complex, multi-part questions through parallel searches and comprehensive answer compilation with citations.
Microsoft Entra Agent ID provides first-class identity management for AI agents, appearing in Microsoft Entra directory with access controls and permissions. Security administrators can apply Conditional Access policies, multi-factor authentication, and least-privilege roles to agents. Always-on observability features provide end-to-end monitoring with built-in metrics on latency, throughput, usage, and quality plus detailed trace logs of agent reasoning steps.
The platform positions Microsoft as the unified AI-native development experience connecting Visual Studio Code, GitHub, and Azure for full-stack AI development from cloud to edge. Foundry Local runtime for Windows and Mac enables offline AI applications for manufacturing floors and remote field service scenarios. Integration with Azure Arc allows centralised management of on-device AI deployments, establishing Azure AI Foundry as an comprehensive AI factory delivering generative AI where enterprise data resides.