Last Updated on June 19, 2025 by Arnav Sharma
The Azure updates at Build 2025 were loud and clear — Microsoft is going all-in on giving developers superpowers in the cloud. Whether you’re building AI apps, orchestrating agents, managing data lakes, or just looking for a simpler way to get from idea to impact, Azure brought the goods this year.
Here’s what you really need to know from the Azure announcements, without the fluff.
Azure AI Foundry becomes the agent factory
Microsoft launched Azure AI Foundry last year, but now it’s grown up. It’s no longer just about hosting models. It’s about designing, customizing, deploying, and running fully-fledged AI agents at scale.
Foundry now includes
- Access to over 10,000 open-source models from Hugging Face
- Full fine-tuning support with a new developer tier (no hosting fees)
- Integration with OpenAI models like GPT-4o and Elon Musk’s Grok 3
- Model playgrounds to test, tweak, and deploy your own AI flows
And if you’re working with sensitive data or in healthcare, there’s even a reference architecture for building multi-agent workflows in areas like tumor board decisions. Yes, this is real.
Agents that think, remember, and get things done
Azure AI Foundry Agent Service is now generally available. It lets developers create agents that can
- Work together in multi-agent workflows
- Connect securely to data from Bing, SharePoint, Databricks, and more
- Be monitored, traced, and evaluated with observability tools
These aren’t basic bots. These are smart, task-aware, scalable agents that can run with or without human intervention and even handle regulated workloads.
Observability for AI apps is no longer optional
When you deploy AI agents, you want to know what they’re doing and why. Foundry now includes real-time observability dashboards that track
- Performance
- Cost
- Safety
- Prompt traces
You can evaluate agents at every stage, from playground to GitHub commit to production. This makes debugging and governance less of a headache and a whole lot faster.
Smart model routing and templates for real-world speed
Choosing the right model for the job has always been tricky. Now Foundry’s Model Router can auto-select the best model for your prompt. If the model underperforms, Foundry can roll back to a better one.
They also introduced pre-built templates for common use cases, so you can go from zero to working prototype in minutes. Think of these like blueprints for GenAI applications that actually work.
AI search gets an upgrade with Agentic Retrieval
Azure AI Search now includes Agentic Retrieval, a smarter way to find answers using natural language and context history. This is retrieval augmented generation done right, with
- Step-by-step planning and execution
- Transparent query logs
- Context memory for better agent results
This is ideal if you’re building chatbots, custom copilots, or any AI that needs to search knowledge efficiently.
Bring the cloud experience to your local device
One of the biggest surprises was Foundry Local, which brings the power of Azure AI to Windows and macOS. It lets you
- Run open-source models locally
- Avoid cloud costs
- Build AI apps with full privacy and offline support
For developers working in finance, healthcare, or regulated environments, this is a game changer.
Enterprise-grade security for AI apps
Foundry is also baking in serious security features like
- Prompt shielding to prevent jailbreaks
- PII redaction using Azure AI Language
- Integration with Microsoft Defender for Cloud for runtime threat detection
- Support for new compliance standards like the EU AI Act
This means you can build and ship AI workloads with the confidence that they won’t expose your business or your users.
Databases get a generative boost
It wasn’t just AI making waves. Microsoft also introduced
- Cosmos DB integration with Fabric and AI Foundry
- PostgreSQL support in GitHub Copilot for VS Code
- SQL Server 2025 preview with built-in AI capabilities
- Global Secondary Indexes for faster queries in Cosmos DB
These changes make it easier to bring your data into the agentic future without reinventing the wheel.
Wrap up
The Azure news this year felt mature, ambitious, and finally cohesive. Microsoft isn’t just saying “AI is the future”, it’s giving developers the tools to build that future now.
From low-code agent builders to pro-code model routing, from on-device inferencing to enterprise-grade observability, Azure is setting the bar high for building intelligent, secure, and scalable AI apps.
I help organisations secure their cloud infrastructure and stay ahead of evolving cyber threats. Microsoft MVP and Certified Trainer, author of Mastering Azure Security, and founder of arnav.au — a platform for practical Cloud, Cybersecurity, DevOps and AI content.