Microsoft entices developers to build more Windows AI apps

Microsoft’s annual developer conference is rather like a litmus test, showing the general direction of travel of the software industry. Given the massive footprint of Windows in desktop computing, the conference showcases new technology that Windows developers can use to build applications.

As has been the case over the last few years, artificial intelligence (AI) is a big focus for Microsoft, with AI functionality not only being deeply embedded into the Windows operating system but Microsoft is also opening up this functionality.

The Copilot+ PC, which embeds a neural processing unit (NPU),  offers what Microsoft sees as a standard platform for Windows AI developers.

There is a very real case for deferring PC hardware upgrades, as there appears to be diminishing returns on the supposed step improvements the new devices claim to offer. But an NPU gives the whole industry a chance to showcase a piece of technology that is entirely new to the PC experience – built-in AI. However, there has been little reason to go out and get this new functionality, due to the lack of genuinely useful applications.

So Build 2025 was a chance for Microsoft to convince developers that creating AI functionality in their applications is the way ahead.

Looking at the announcements Microsoft has made during Build 2025, there is an AI inference engine, Windows ML, built into Windows 11, which can be accessed via an application programming interface (API), and, significantly, can be trained on custom data.

It may be just terminology, but it reminds us very much of the era of client-server computing. While at that time, data resided on servers, which eventually moved from private datacentres to public cloud infrastructure, an AI-based PC offers local AI processing. Data can be processed locally, as well as in the cloud.

For some applications, the amount of data processing makes cloud-based AI impractical. For instance, Volker Rölke, a senior machine learning computer scientist at Adobe, sees an opportunity in using the new machine learning APIs in Windows to process the terabytes of data in its video and motion editing suites like Premiere Pro and After Effects.

Beyond the built-in AI, Microsoft CEO, Satya Nadella, sees a future of multiple AI agents all working together to automate workflows and business business processes.

From a Computer Weekly perspective, we think Windows ML is interesting as it shows a genuine push from the software giant away from cloud computing towards edge-based computing. Microsoft will argue that it is giving developers choice as to where they choose to run AI processing. But with the level of untapped NPU power that is likely to grow as more and more AI PCs ship, there is a concerted industry effort to make these things useful.