Microsoft Build 2025 | Satya Nadella Opening Keynote

Key Concepts

Agentic Web: An emerging platform shift towards open, scalable, and agent-centric applications.
Copilot: AI-powered assistant integrated into various Microsoft products (Visual Studio, VS Code, Microsoft 365) to enhance productivity and automate tasks.
MCP (Model Context Protocol): A protocol for secure and open communication between agents and services, enabling access to content and actions.
Foundry: A comprehensive app platform for the AI age, providing tools for model selection, orchestration, RAG, and deployment.
Copilot Studio: A low-code platform for building custom agents that connect to Copilot and Teams.
Copilot Tuning: A feature that allows enterprises to fine-tune Copilot models with their own data, workflows, and style.
NLWeb: A way for anyone with a website or API to easily make it an agentic application.
Microsoft Discovery: A platform for science that brings together the entire stack to accelerate the scientific process.
RAG (Retrieval-Augmented Generation): A technique for improving the accuracy and relevance of AI-generated content by grounding it in external knowledge sources.

Tools and Infrastructure Updates

Visual Studio: Enhanced with .NET stand support, live preview, improved Git tooling, and a new cross-platform debugger. Moving to a monthly release cadence.
VS Code: 100th release includes improved multi-window support and easier staging. Copilot is being open-sourced within VS Code.
GitHub: Doubling down for developers building any applications. Trust, security, compliance, auditability, data residency are even more critical today.
GitHub Copilot: Evolving from code completions to chat, multi-file edits, and now agents. Can upgrade frameworks (Java 8 to Java 21, .NET 6 to .NET 9) and migrate on-premise apps to the cloud.
GitHub Coding Agent: A full coding agent built into GitHub, capable of autonomously completing tasks like bug fixes, new features, and code maintenance.
Azure AI Foundry: A production line for intelligence, offering model choice, orchestration, RAG, and enterprise-wide deployments. Supports 1,900 models, including those from OpenAI, Grok from xAI, Mistral, and Llama.
Windows AI Foundry: A platform for building on-device AI capabilities on Windows, supporting CPUs, GPUs, and NPUs. Includes Foundry Local and allows customization of Phi Silica SLM using LoRA adapters.
Azure Local: Brings the power of the app server and app building capability to the edge and clients.

Agentic Capabilities and Applications

Copilot in Microsoft 365: Integrates Chat, Search, Notebooks, Create, and Agents into a unified interface. Includes Researcher and Analyst agents for expertise at your fingertips.
Multi-Agent Orchestration in Copilot Studio: Enables agents to collaborate on complex tasks, such as onboarding a new hire.
Copilot Tuning: Allows enterprises to customize Copilot models with their own data, workflows, and style.
Autonomous SRE Agent: Automatically triages, root causes, and mitigates live site issues, logging incident reports as GitHub issues.
NLWeb: Allows anyone with a website or API to easily make it an agentic application.
Healthcare Agent Orchestrator: Used by Stanford Medicine to orchestrate tumor board meetings, bringing together patient history, radiology, and clinical trial data. Now available to everyone in Foundry.
VibeTravel App: A chat-based app that uses Azure AI Foundry to provide an AI travel agent.
Microsoft Discovery: A platform for science that brings together the entire stack to accelerate the scientific process.

Model Updates and Choices

OpenAI Models: Azure offers same-day access to new OpenAI models, including Sora coming next week.
Model Router: Automatically chooses the best OpenAI model for the job.
Grok from xAI: Coming to Azure, offering reasoning, deep search, and response models in one. Aims to reason from first principles and minimize error.
Mistral: Available with sovereign deployment in the EU region.
Llama: A full herd of Llama models coming to Azure.
Hugging Face: Access to over 11,000 frontier and open-source models in Foundry.
Phi Silica: A model that will revolutionize inference compute on the PC.

Data and Infrastructure Enhancements

SQL Server 2025: Launching with closer integration between the data layer and intelligence.
Cosmos DB Integration: Integrating Cosmos DB directly into Foundry for storing conversational history and RAG application needs.
Azure Databricks: Connecting data in Genie spaces or in AI functions to Foundry.
LLM Integration in Postgres SQL: Allows natural language and SQL to mix together in queries.
Cosmos DB in Fabric: Unifies structured and semi-structured data for AI.
Digital Twin Builder in Fabric: Allows easy creation of digital twins with no-code/low-code.
Shortcut Transformations in OneLake: AI-driven ETL with prebuilt AI-powered transformations.
Copilot in Power BI: Allows chatting with data, visually exploring, and analyzing across multiple reports and semantic models.
Nvidia GB200 on Azure: Azure is the first cloud to bring Nvidia GB200 online at scale, generating 865,000 tokens per second.
Azure Data Center Expansion: 70+ data center regions, with 10 new data centers opened in the past three months.
AI WAN: 400-terabyte backbone connecting data centers.
Cobalt VMs: Arm-based VMs powering Teams, Defender, and partner workloads.
Azure Local: Offers the ability to run things disconnected.

Security and Compliance

GitHub: Trust, security, compliance, auditability, data residency are even more critical today.
Entra ID: Agents now get their own identity, permissions, policies, and access controls.
Purview: Integrates with Foundry to ensure end-to-end data protection.
Defender: Integrates with Foundry to protect agents from threats.
Compliance Capabilities: Most comprehensive set of compliance capabilities in the cloud.
Sovereign Controls: Data residency, confidential computing for cloud infrastructure.

Key Examples and Case Studies

NFL Scouting Combine: Used Azure AI Foundry to provide a full prospect view with AI-powered filtering and analysis.
Stanford Medicine Tumor Board: Used a healthcare agent orchestrator to bring together patient history, radiology, and clinical trial data.
Aisin and Kato San: Built an app with Foundry to help people with auditory processing disorder.
Met Office: Using Azure to better predict the weather, ensuring timely and accurate forecasting.
World Bank in Nigeria and Peru: Using Copilot to enhance lesson plans, improve student outcomes, and grow in their own careers.
Microsoft Discovery Immersion Coolants: Used Microsoft Discovery to find a promising candidate for an immersion coolant that does not rely on forever chemicals.

Notable Quotes

Satya Nadella: "We're just about getting into these middle innings of another platform shift. And these middle innings are where all things happen, all things scale."
Satya Nadella: "It's not about any one tool, any one agent or any one form factor. It's about the coming together of all of it for us as individuals and as Teams of software engineers to be able to express our craft."
Sam Altman: "This is like true software engineering task delegation, and I think it'll only get better from here."
Elon Musk: "With Grok, especially with Grok 3.5 that is about to be released, it's trying to reason from first principles."
Kevin Scott: "An agent is a thing that a human being is able to delegate tasks to."
Jensen Huang: "The alignment between our two organizations to build the most advanced infrastructure in the world, the most advanced AI factories in the world, and the best of times are yet to come."

Synthesis/Conclusion

The Microsoft Build conference highlights a significant platform shift towards an "Agentic Web," characterized by open, scalable, and AI-driven applications. Microsoft is investing heavily in tools, infrastructure, and partnerships to empower developers to build these new agentic experiences. Key announcements include the open-sourcing of Copilot in VS Code, the introduction of GitHub Coding Agent, the expansion of Azure AI Foundry, and the unveiling of Microsoft Discovery for scientific research. The focus is on creating a comprehensive ecosystem where developers can leverage AI to automate tasks, enhance productivity, and solve complex problems across various domains, from software engineering to healthcare and scientific discovery. The emphasis on open standards like MCP and NLWeb aims to foster interoperability and innovation within the agentic web.