Advancing Windows for AI development

At Build 2025, Microsoft reaffirmed its commitment to empowering the global developer community by showcasing new tools and capabilities designed to make Windows the ultimate development platform in the age of artificial intelligence. With a focus on flexibility, performance, and security, Windows continues to evolve into a powerful foundation for AI innovation, both on-device and in the cloud.

Advancing Windows for AI Development: New Platform Capabilities and Tools Introduced at Build 2025

The company emphasized its mission to make Windows the best platform for developers, with intelligent integrations across software, silicon, and hardware. From Windows 11 to Windows 365, Microsoft is supporting a range of developer needs including AI workflows, productivity, and core IT operations.

Introducing Windows AI Foundry

One of the most significant announcements was the introduction of Windows AI Foundry, a comprehensive platform that builds upon the Windows Copilot Runtime. It supports the AI development lifecycle from model selection and optimization to deployment across client and cloud environments.

Key components of Windows AI Foundry include:

• Windows ML, the native AI inferencing runtime that supports efficient model deployment across AMD, Intel, NVIDIA, and Qualcomm hardware.
• Foundry Local and integrations with model catalogs such as Ollama and NVIDIA NIMs to provide ready-to-use open-source models optimized for Windows silicon.
• Inbox AI APIs for language and vision tasks, like text summarization and object removal, powered by on-device models for Copilot+ PCs.
• Support for LoRA (Low-Rank Adaptation) to fine-tune the Phi Silica small language model using custom data.
• Semantic Search and Knowledge Retrieval APIs for building RAG scenarios and natural language search functions using local data.
   

Evolution of Windows ML for AI Inference

Windows ML now serves as the backbone of AI model execution on Windows. Designed with direct feedback from developers and silicon partners, it offers:

• Simplified deployment without bundling separate ML runtimes or drivers.
• Future-proof adaptability to new AI hardware generations.
• A robust toolset via the AI Toolkit for VS Code, including quantization and optimization utilities.
   

Windows ML is already being used by developers at Adobe, McAfee, Topaz Labs, Wondershare, and others to deliver high-performance AI applications across different hardware.

Built-In AI APIs for Rapid Integration

Microsoft is providing developers with out-of-the-box AI APIs for language and vision tasks, available in the Windows App SDK 1.7.2. These APIs run locally and are optimized for privacy, security, and performance on NPUs.

Notably, these include:

• Text summarization and rewrite tools
• Image description and recognition
• Super-resolution and segmentation features
   

These APIs are being leveraged by developers from Dot Vista, Filmora, Pieces, and iQIYI to rapidly embed AI features into their applications.

Windows AI Foundry

Semantic Search and Retrieval-Augmented Generation (RAG)

New semantic search APIs provide developers with capabilities to perform intuitive, context-aware searches on app data. These APIs support both semantic and lexical search, and they are optimized for privacy and performance on all device types, including Copilot+ PCs.

They also enable retrieval-augmented generation (RAG), allowing developers to enhance LLM outputs with grounded data from their apps.

Model Context Protocol: Enabling the Agentic Future

With the rise of AI agents, Microsoft is introducing native support for Model Context Protocol (MCP) in Windows 11. MCP enables AI agents to interface directly with Windows apps and expose specific app functionalities.

Key MCP components include:

• MCP Registry: A secure registry of MCP servers accessible to agents.
• MCP Servers: System features like File System and Windowing, as well as developer-defined functionalities, exposed to agents through secure APIs.
   

Leading AI developers including Anthropic, Perplexity, OpenAI, and Figma are already collaborating with Microsoft to integrate MCP capabilities.

App Actions: Unlocking Discoverability

Microsoft introduced App Actions on Windows, allowing developers to define discrete actions within their apps that can be surfaced contextually by the system. These actions enhance user engagement and discoverability.

Early adopters of App Actions include Zoom, Filmora, Todoist, Raycast, Goodnotes, and Spark Mail.

LoRA for Fine-Tuning Phi Silica Models

To facilitate easy fine-tuning of language models, Microsoft introduced LoRA support for Phi Silica, allowing developers to tailor model performance for specific tasks by modifying a limited set of parameters. This feature is now available in public preview for Snapdragon X NPUs and will roll out to Intel and AMD platforms later.

Developers can initiate fine-tuning via the AI Toolkit in VS Code, and deploy the resulting adapters using the Phi Silica API.

Empowering Local Development with AI Workstations

To support local inferencing and compute-heavy workloads, Microsoft is partnering with OEMs to offer AI-capable developer workstations. Highlights include:

• Dell Pro Max Tower for high-performance desktop development
• HP Z2 Mini G1a for compact, powerful workstations
• Copilot+ laptops like the Dell Pro Max 16 Premium and HP Zbook Ultra G1a for mobile development
   

VBS Enclave SDK for Secure Computing

Security continues to be a priority. The new VBS Enclave SDK, now in public preview, offers a Trusted Execution Environment for secure data handling, including cryptographic operations. It supports:

• Interface definition and parameter validation
• Enclave creation and thread pool management
• Encryption and telemetry support
   

This technology forms the foundation of Microsoft’s Recall experience and is now available for developers to use in their own apps.

Post-Quantum Cryptography for Windows and Linux

Microsoft is also preparing for the quantum era by integrating post-quantum cryptography (PQC) into Windows Insiders builds and Linux via SymCrypt-OpenSSL. This early access allows developers to test compatibility and prepare for future cryptographic standards.

Open Sourcing Windows Subsystem for Linux (WSL)

Microsoft announced the open sourcing of WSL, empowering developers to contribute to and customize the Linux virtualization layer on Windows. This move fulfills a long-standing community request and will enable greater flexibility and performance improvements.

Enhancements to Developer Tools

Microsoft continues to refine its development tools:

• WinGet Configuration now supports DSC V3 for capturing application states and replicating environments.
• Advanced Windows Settings centralizes powerful options for developers, such as GitHub integration in File Explorer.
• Command Palette in PowerToys offers quick access to commands and applications, reducing context switching.
• Edit, a new command-line text editor, allows developers to make quick file changes directly in the terminal.
   

Microsoft Store: Growing the App Ecosystem

The Microsoft Store now boasts over 250 million monthly users and is expanding support for developers with features such as:

• Free account registration
• FastTrack program for Win32 app submissions
• Beta App Campaigns to drive user acquisition
• Enhanced analytics and discoverability features
   

The Store’s new AI Hub experience, including AI badges and featured experiences, highlights Windows as a leader in consumer-facing AI tools.

Looking Ahead

Build 2025 underscores Microsoft’s vision for an open, secure, and intelligent Windows platform built for AI innovation. With continued partnerships, new APIs, enhanced security, and developer-centric tooling, Windows is positioned to empower the next generation of app creators and AI experiences. Microsoft invites developers everywhere to explore these capabilities and continue building the future of AI on Windows.