Are you a client?
Sign in to view the full news archive.
At its Build 2025 conference, Microsoft laid out its vision for the future of AI-powered development, introducing a host of tools and partnerships that aim to position the company as a central and increasingly supplier-agnostic platform for building enterprise AI. Announcements ranged from autonomous coding agents, to new support for third-party AI models, with Microsoft signalling a strategic shift away from its exclusive alignment with OpenAI.
One of the big announcements was Microsoft expanding its Azure AI Foundry to include new AI models including xAI’s Grok 3, Meta’s Llama, and models from European startups Mistral and Black Forest Labs, bringing the total number of models it offers to Azure customers to more than 1,900. Notably, these models will run directly within Microsoft’s own data centres, allowing it to provide high availability guarantees amid growing concerns over service reliability in AI infrastructure.
Another major development is the evolution of GitHub Copilot into a fully-fledged “coding agent”. Unlike its predecessor, which simply assisted with code snippets, the new agent can complete complex development tasks with minimal human input, from diagnosing bugs to executing full fixes. The move arrives just days after OpenAI previewed its own competing ‘Codex agent’.
Microsoft is also deepening its enterprise focus with services like ‘Copilot Tuning’, which allows businesses to create domain-specific agents using their own workflows and data. It also introduced ‘Windows AI Foundry’, which offers a unified platform supporting the AI developer lifecycle across training and inference, with simple model APIs for vision and language tasks. Another new platform, ‘Discovery’ has been built to empower researchers to transform the R&D discovery process with agentic AI, accelerating the time to market for new products and expanding the discovery process for scientists.
Lastly, Microsoft is betting on a future defined by what it terms the “open agentic web.” Through initiatives like Model Context Protocol (MCP) and the new NLWeb project (a standard likened to HTML for AI), the company envisions a decentralised, conversational internet where AI agents perform tasks on behalf of users across services and platforms.
The rapid adoption of AI tools and services has already been a massive boon for Microsoft, with its AI run rate now $13bn as of Q3, up 175% yoy. (See - Microsoft reports solid Q3 and strengthens European digital commitments). 15 million developers are already using GitHub Copilot, whilst enterprise adoption is surging. Over 230k organisations, including 90% of Fortune 500 companies are using Copilot Studio to build customised AI agents and automations, with both Fujitsu and NTT DATA highlighted as examples of organisations leveraging Azure AI Foundry to build apps that prioritise leads and surface client insights.
Posted by: Simon Baxter at 10:05