Everything Microsoft Announced at Build 2026

Microsoft Build 2026 was one of the company’s most announcement-heavy developer conferences. CEO Satya Nadella set the tone, declaring that the industry has entered the age of AI agents.

If you’re a developer, a business, or just someone curious about where technology is headed, here’s everything important that was announced.

1. Seven New MAI Models

For a long time, Microsoft’s AI story was all about OpenAI’s story. At Build, Microsoft changed that and introduced its own model family.

The company launched seven new models under the MAI (Microsoft AI) brand, all trained in-house by the Microsoft AI Superintelligence Team. The headline model is MAI-Thinking-1, Microsoft’s first reasoning model. It has 35 billion active parameters and a 256,000-token context window, which means it can hold a lot of information in its head at once. 

The model was trained entirely from scratch, without distilling knowledge from other AI companies' models, using commercially licensed data. Microsoft says this could appeal to enterprises concerned about legal and compliance risks. The model is now available in private preview through Microsoft Foundry.

The other six models fill out a full toolkit. MAI-Image-2.5 and its Flash variant handle image creation and editing, and already live in PowerPoint and OneDrive. MAI-Transcribe-1.5 is a transcription model that Microsoft says beats both Gemini and OpenAI’s flagship on accuracy across 43 languages. 

MAI-Voice-2 adds support for 15 additional languages and new voice options. Microsoft also introduced MAI Code 1 Flash, a lightweight and fast coding model built specifically for GitHub Copilot workflows. The model is rolling out to all Copilot tiers, including Free, Pro, Pro+, and Max, starting June 2.

2. Surface RTX Spark Dev Box

The biggest hardware news wasn’t a consumer laptop. It was a compact desktop box, called the Surface RTX Spark Dev Box, built specifically for AI developers.

It’s powered by NVIDIA’s RTX Spark platform and delivers up to 1 petaflop of AI compute. It comes with 128 GB of unified memory, which means both the CPU and GPU draw from the same pool, making it much easier to run large AI models locally. Microsoft says it can run models with up to 120 billion parameters right on the machine, without needing to call out to a cloud server at all.

3. Project Solara — A New Platform for Agent-First Devices

One of the more forward-looking announcements at Build was Project Solara, a chip-to-cloud platform Microsoft is building for a world where AI agents live in more places than just a phone or laptop screen.

Microsoft showed two concept reference designs. The first is a wearable badge device powered by next-generation Qualcomm silicon, giving users access to AI agents while on the move, whether they are walking between meetings or working hands-free. No laptop required. 

The second is a desk companion device, a small ambient device powered by MediaTek silicon that sits on your desk and is always available, always aware of context, and always ready to help you plan, think, or manage tasks without interrupting the flow on the main screen.

Neither device is a final product yet. Microsoft is presenting them as reference designs, open concepts that manufacturers and enterprises can build on.

4. Microsoft Scout — Your Always-On Personal Work Agent

Scout is a new personal work agent built on OpenClaw (an open-source autonomous agent project) and Work IQ (the workplace intelligence layer that understands how your organisation actually operates). It connects to Microsoft Teams, Outlook, OneDrive, SharePoint and can take actions directly on your device. It's designed to proactively handle things like meeting prep, scheduling conflicts, and routine tasks without you needing to prompt it every time.

Scout also comes with its own Entra identity, allowing it to operate within the same managed, secure identity system used by employees. This gives IT teams visibility into the actions Scout takes and lets them control what data and systems it can access. Microsoft is making Scout available to Frontier customers immediately, with a wider rollout planned for a later date.

5. Azure Infrastructure for the Agent Era

Microsoft also gave several infrastructure updates.

The Maia 200 AI accelerator chip is already running in production data centres in Iowa and Arizona, with deployments planned for Italy, Australia, and South Korea. Microsoft said the chip delivers the best performance per dollar and per watt in its infrastructure, helping expand AI inference capacity using its own silicon.

The company also revealed that Cobalt 200 virtual machines, powered by Microsoft's custom Arm-based processor, are now in preview and available across more than 10 regions. According to Microsoft, they deliver up to 50% better performance for agentic AI workloads.

Another key announcement was Multipath Reliable Connection (MRC), a new open networking protocol developed with partners including AMD, Broadcom, Intel, NVIDIA, and OpenAI. The technology helps AI training and inference workloads maintain performance by automatically rerouting traffic around network issues, reducing the need for costly restarts.

Microsoft also previewed Azure HorizonDB, a fully managed PostgreSQL database service built for AI agents.

6. Windows 11 Gets a Major Upgrade for Developers

Microsoft announced a series of updates to make Windows 11 a more attractive platform for developers.

The company is adding native Linux-style development capabilities, including support for Linux containers directly within Windows. Microsoft also introduced an experimental Intelligent Terminal and improved command-line tools to streamline development workflows.

The goal is to reduce setup complexity and make Windows more competitive with Linux-based development environments. Microsoft said the new experience offers faster onboarding, integrated container support, and better access to commonly used developer tools.

The company also expanded Windows AI APIs, enabling developers to run AI workloads across CPUs, GPUs, and NPUs. New local AI models, Aion 1.0 Instruct and Aion 1.0 Plan, were also announced for on-device AI scenarios.

7. GitHub Copilot App and GitHub Improvements

Microsoft launched a new GitHub Copilot desktop application, bringing AI-powered coding assistance into a dedicated workspace.

The app gives developers a more integrated experience, reducing the need to switch between browsers, editors, and terminals. Developers can access coding workflows, review pull requests, manage projects, and collaborate with AI agents from a single interface.

Microsoft also announced that Visual Studio will now use the GitHub Copilot SDK as its core AI foundation, helping deliver a more consistent AI experience across the development environment. The company is adding new AI agent capabilities that can assist developers with debugging, performance analysis, and testing.

In addition, Windows Development Configurations are now generally available. This feature lets developers set up a complete development environment using a single WinGet configuration file, automatically installing tools such as WSL, PowerShell 7, Git, GitHub CLI, and VS Code in just a few minutes.

8. Majorana 2 — Microsoft’s Quantum Computing Chip

Microsoft also used Build 2026 to provide an update on its quantum computing efforts, introducing Majorana 2.

The company said the new chip contains qubits that are approximately 1,000 times more reliable than those used in previous designs. Microsoft credits improvements in materials engineering and support from AI-assisted discovery tools for the advancement.

Reliability remains one of the biggest challenges in quantum computing, and Microsoft believes Majorana 2 moves the industry closer to practical, large-scale quantum systems.

The company has previously stated that it aims to achieve commercially useful quantum computing before the end of the decade.

9. Microsoft Discovery — AI for Scientific Research

Microsoft Discovery is now generally available, and it may be one of Microsoft's most ambitious long-term bets.

It’s an enterprise-grade agentic AI platform built for scientific research. Instead of just helping researchers write papers or search the literature, Discovery provides them with a full agentic workflow across the scientific process, including literature review, hypothesis generation, simulation, experimental planning, and analysis. Built on Azure, it uses specialised agents, knowledge graphs, and lab workflow integrations to help researchers move from ideas to insights more efficiently.