AI Foundry: The Evolution of Microsoft AI for Enterprise

Introduction 

Building enterprise AI applications used to require choosing between speed and control. You'd either move fast with limited oversight, or build governance into every step and slow down. 

Microsoft Foundry changes that equation. 

The platform unifies agents, models, and tools under one management interface. Your team gets production-grade infrastructure with the flexibility to explore, build, test, and deploy AI applications all within enterprise-ready frameworks that don't slow you down. 

This isn't an incremental improvement. This is how enterprise AI development actually works in 2025. 

What Is Microsoft AI Foundry? 

The Core Definition 

Microsoft Foundry is a unified Azure platform-as-a-service designed specifically for enterprise AI operations. It combines production-grade infrastructure with developer-friendly interfaces, removing the friction between building proof-of-concept AI solutions and deploying them at scale. 

The platform consolidates everything your team needs: 

• Agents: AI agents that understand your business context 
• Models: Access to 11,000+ foundational, open, reasoning, multimodal, and industry-specific models 
• Tools: Built-in tools plus custom integrations for your specific workflows 
• Governance: Unified tracing, monitoring, evaluations, and enterprise setup configurations 

All of this lives under one Azure resource provider namespace with unified role-based access control (RBAC), networking policies, and compliance frameworks. 

The Evolution from Azure AI Foundry 

Microsoft Foundry represents the evolution of Azure AI Foundry, rebranded to better reflect its broader enterprise capabilities beyond traditional AI services. The platform now encompasses agentic AI—applications that can reason, take actions, and operate with significant autonomy within your defined parameters. 

Why Enterprise Teams Choose Foundry 

The 11,000+ Model Advantage 

Access to 11,000+ models sounds impressive until you realize what it actually means operationally. 

Your finance team needs different AI capabilities than your customer service team. Your operations team's requirements differ from your marketing team's. With Foundry, you're not picking one model and forcing every use case into it. 

You get: 

• Foundational models (GPT-class general purpose) 
• Open models (for fine-tuning and customization) 
• Reasoning models (for complex problem-solving) 
• Multimodal models (working with text, image, audio, video) 
• Industry-specific models (pre-trained on domain data) 

The platform includes built-in intelligent model routing. Instead of manually selecting which model to use, the system automatically optimizes model selection in real-time based on performance metrics for your specific use case. 

This isn't just capability. This is operational efficiency. 

Two Portal Versions for Different Workflows 

Microsoft Foundry offers two portals: 

Microsoft Foundry (Classic): Best for teams working with multiple resource types—Azure OpenAI services, Foundry resources, hub-based projects, and Foundry projects in one unified view. This is where teams managing complex enterprise environments live. 

Microsoft Foundry (New): Simplified interface focused on Foundry projects only. For teams building multi-agent applications without managing other Azure resources, this streamlined experience removes unnecessary complexity. 

You switch between them with a toggle in the portal banner, depending on your current task. 

How Enterprise Teams Actually Use Foundry 

Project Types: Foundry Projects vs Hub-Based Projects 

Microsoft Foundry supports two project architectures: 

Foundry Projects (Recommended): 

• Managed under a Microsoft Foundry resource 
• Built for modern agentic AI applications 
• Self-serve environment creation without requiring extra Azure resources 
• Access to new agents and model-centric capabilities in general availability 
• Agents (GA), Foundry API/SDK, evaluations, advanced model management 
• Best for: Teams building new AI applications 

Hub-Based Projects: 

• Hosted by a Microsoft Foundry hub 
• Useful when organizations have administrative teams managing centralized resources 
• Agents available in preview only 
• Certain capabilities limited compared to Foundry projects 
• Best for: Organizations with existing hub infrastructure 

For most new developments, Foundry projects are the right choice. They're where new capabilities land first. 

The Three Stages of Development 

Foundry is structured around how enterprise teams actually develop: 

1. Define and Explore: You define your project goals and explore models and services to find the ones matching your use case. The 11,000+ model catalog and instant benchmarking tools let you test options quickly. 

2. Build and Customize: You actively develop solutions—building agents, fine-tuning models with your own data, grounding models in your knowledge, and integrating with existing systems. You can build through the Foundry portal, through code using SDKs, or both. 

3. Observe and Improve: You monitor performance through tracing and evaluation tools. You identify where the application isn't meeting requirements and refine it. You integrate safety and security systems before moving to production. 

SDKs and APIs: Building with Confidence 

Consistent Development Experience 

Microsoft Foundry provides a unified API and SDK framework designed specifically for agentic applications. This consistency matters because it means: 

• Your Python developers, C# developers, and JavaScript developers can use the same logical API structure 
• You're not translating between different syntaxes and paradigms 
• Code written in one language is easier for teams to review and maintain 

Available SDKs: 

• Python (stable) 
• C# (stable) 
• JavaScript/TypeScript (preview) 
• Java (preview) 

The Microsoft Foundry for VS Code Extension lets developers explore models and develop agents directly in their preferred development environment. 

This isn't just convenient. This is developer productivity at scale. 

Governance and Security at Enterprise Scale 

Built-In for Compliance, Not Bolted On 

Foundry projects act as secure units of isolation. Your finance team's AI agents don't see your customer service team's data. Projects share file storage, conversation history, and search indexes within teams, but maintain isolation between teams. 

The platform includes a unified RBAC, letting you control who can access what at granular levels. You can bring your own Azure Key Vault for encryption key management. You can bring your own storage for agent services. 

For regulated industries (finance, healthcare, government), this matters enormously. You're not retrofitting compliance. It's built in. 

Availability, Pricing, and Getting Started 

Where You Can Build 

Foundry is available in most regions where Foundry Tools are available. If your organization operates globally, you can deploy in regions meeting your data residency requirements. 

Cost Structure 

The platform is free to explore and test. You pay only at deployment level for models you use, services you consume, and compute you consume. 

The actual costs depend on your architecture: 

• Which models you're using (prices vary by model) 
• How many API calls your agents make 
• Compute resources for running agents 
• Data storage for conversation history and files 

You're only charged for what you actually use. 

FAQs 

What is Microsoft AI Foundry exactly?

Microsoft Foundry is a unified Azure platform where enterprise teams build, deploy, and manage AI applications and autonomous agents. It combines 11,000+ models, enterprise governance, and developer-friendly tools in one integrated environment. Unlike point solutions, Foundry gives you everything you need without switching between platforms. 

When was Azure AI Foundry launched?

Azure AI Foundry evolved into Microsoft Foundry. The rebranding reflects the platform's expansion from traditional AI services into agentic AI capabilities—autonomous systems that can reason and act. The current Microsoft Foundry platform continues this evolution with enhanced governance, more models, and deeper Azure integration.

How do I choose between Foundry projects and hub-based projects?

Use a Foundry project if you're building new applications with agents or models. Use a hub-based project only if your organization has an existing hub infrastructure and you need to continue within that structure. New capabilities launch on Foundry projects first.

Can I use Microsoft Foundry with my existing Azure infrastructure?

Yes. Foundry integrates with your existing Azure environment. You can bring your own Key Vault for secrets management, your own storage accounts, your own networking configurations. You control how Foundry connects to your broader Azure ecosystem.

What models can I access?

You have access to 11,000+ models across multiple categories: foundational models for general purpose tasks, open models you can fine-tune, specialized reasoning models, multimodal models, and industry-specific pre-trained models. You can compare and benchmark models for your specific use case.

How do I get my team started with Foundry?

Start building with Microsoft Foundry Our team can help you evaluate which project type makes sense for your use case, set up your initial environment, and plan your first AI application. We work with organizations building everything from customer service agents to internal knowledge systems. 

For deeper implementation planning and custom architecture: 

Explore our AI and ML solutions  See how organizations are building secure, scalable AI applications within existing compliance frameworks and integrating them with systems like Microsoft Fabric, GitHub, and Visual Studio. 

The Bottom Line 

Microsoft Foundry isn't asking you to choose between speed and control, between innovation and governance, between exploration and production. 

It gives you all of them. 

With 11,000+ models, unified project management, enterprise-grade security, and developer SDKs that actually reduce friction, Foundry is how enterprise AI development works when it works well. 

The question isn't whether you need AI in your organization. That's settled. The question is whether you'll build it on infrastructure designed for enterprise scale or on infrastructure that requires constant workarounds. 

Foundry is the former.