Microsoft tries to spark new life into Windows

When Microsoft released Windows 11 five years ago, it felt like an operating system that was still being renovated, a work in progress. You’d think by now that those renovations would be complete, with Microsoft turning its attention to Windows 12. Instead, Windows 11 feels like it’s here to stay, forming the foundation for Microsoft’s next big ambition: transforming Windows into an agentic OS.

At a Windows and Surface event this week, Microsoft laid out its plan to bring AI agents to Windows, including hybrid intelligence that lets Windows take advantage of free local models instead of expensive cloud ones. “I think today is the day we’ll start talking and thinking about the frontier ecosystem taking maybe slightly different contours,” said Microsoft CEO Satya Nadella.

After trying to sell businesses on a new Copilot last month, Nadella took a similar approach for his vision of agents and hybrid intelligence for Windows. It was so similar that when he mentioned the security, identity, and management aspects of Microsoft’s platform for securing agents, it almost seemed like he was pitching businesses again instead of AI developers, creatives, and just regular Windows users.

After a somewhat dizzying start, Windows chief Pavan Davuluri appeared onstage to try and explain why agents and hybrid intelligence are so important for Windows users. There were demos of GitHub Copilot coding locally without expensive cloud tokens; a long list of models that can now run locally; examples of agents like OpenClaw, Hermes, and Perplexity; and even deeper integration of Copilot in Windows.

I wasn’t fully convinced by any of these demos, though. I wanted to see what practical benefits local AI models could actually deliver for regular Windows users, instead of just the AI builders or developers Microsoft is trying to win over right now. Those are the people who could help turn its agentic OS vision into reality, though, and many of them are using a MacBook Pro instead of Windows. Microsoft’s answer to tempt them across to Windows? The Surface Laptop Ultra.

The Surface Laptop Ultra is the home for Microsoft’s agentic OS push. Nvidia’s RTX Spark chips power the device, and there’s enough raw GPU power to get AI models exceeding 120 billion parameters running locally. “We built it for developers and creators that push the limit of performance and want portability with exceptional craft,” said Davuluri.

The Surface Laptop Ultra is the flagship RTX Spark device, though there are also similar laptops from Asus, Dell, HP, Lenovo, and MSI. Microsoft’s Surface Laptop Ultra starts at $2,599, and some RTX Spark PCs can run all the way up to an eye-watering $6,999. Much like the very top of the MacBook Pro line, you’d have to really need 128GB of unified memory to spend that kind of money. For many AI builders it might just be worth it, though, especially if you can save thousands in token costs.

For the rest of us, Microsoft had very little to show. A new Windows Search interface looks like it could be one of the best additions to Windows 11 in years, though. Microsoft has totally redesigned search to make it not only more modern, but a heck of a lot faster than before. I’ve gotten to the point where I disable the Bing results in Windows Search, but this new UI now surfaces results in-line and lets you use search more like a launcher.

I wasn’t expecting some grand unveiling of Windows 12 at this event, but I was hoping to hear a lot more about why RTX Spark laptops really do matter. “This is the future we’re building: agents that can work securely on your behalf, intelligence that moves seamlessly between local and cloud, and devices that become more capable than ever before,” said Davuluri.

The problem is Microsoft is still in the early stages of this future it’s betting on for Windows. We had Copilot Plus PCs two years ago, with a big focus on the NPU chips inside that were supposed to revolutionize how consumers use everyday AI on a laptop. They haven’t done that yet, with only a handful of photo, video, and productivity tools making use of the NPU. Microsoft’s Recall feature was supposed to be the star of the NPU show, but the privacy concerns and delayed launch made it totally irrelevant instead.

Nvidia’s presence and GPU dominance may well make RTX Spark a bigger success story than Copilot Plus PCs, but these machines are targeting a niche audience for now. If Microsoft and its partners want to move beyond selling them as MacBook Pro competitors, they’ll eventually need to bring the same capabilities to less expensive, less powerful devices and demonstrate practical reasons for regular Windows users to care. The difference is that RTX Spark PCs are aimed at people who have a much more immediate reason to care about local AI: developers and creators who need to run large models and can potentially save substantial amounts of money by doing so.

For Microsoft, though, it’s the creators and builders that matter most right now. AI agents running in the cloud on Linux virtual machines could spark a wave of desktop Linux app compatibility we’ve never seen before. Microsoft needs to counter that threat by making Windows more agent-friendly.

That might explain why Nadella and Nvidia CEO Jensen Huang spent so much time reminiscing about the early days of the PC and graphics at this event. Microsoft isn’t ready to give up the PC dominance and app compatibility it has spent decades building. But it knows that the way people use computers is changing, and that Windows can’t rely on its historical advantages forever.

Perhaps that gets to the heart of why Microsoft seems to be scrambling to embrace AI agents and hybrid intelligence in Windows. It needs to make sure that, even as computing shifts beyond the traditional PC, Windows remains at the center of it all.

I’m always keen to hear from readers, so please drop a comment here, or you can reach me at notepad@theverge.com if you want to discuss anything else. If you’ve heard about any of Microsoft’s secret projects, you can reach me via email at notepad@theverge.com or speak to me confidentially on the Signal messaging app, where I’m tomwarren.01. I’m also tomwarren on Telegram, if you’d prefer to chat there.

Thanks for subscribing to Notepad.

Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.