Qualcomm took us to Maui for Snapdragon Summit, but the Microsoft Surface showcase on the show floor quickly became less about specifications and more about what these new Snapdragon-powered machines can actually do.
The demos ranged from tiny interface touches to surprisingly demanding creative workloads, offering a glimpse at how Microsoft sees local AI evolving on Windows PCs. And some of it was much more impressive in person than it sounds on paper.
Haptics make Surface feel more connected
The first demo was also the easiest to dismiss: haptic feedback. Microsoft showed how tactile feedback is built into the Surface touchpad and pen, with demonstrations across apps including Affinity, DaVinci Resolve and PowerPoint. Align an object, snap something into place, scrub through a timeline or make certain mistakes, and the device responds with a tiny physical pulse.
It sounds like a small feature. It doesn’t feel like one.
The closest comparison is smartphone haptics. Vibration feedback was once something users barely noticed, but good haptics eventually became part of what made a premium smartphone feel polished. Microsoft’s implementation on Surface creates a similar physical bridge between an action and its result. The interesting part is that the value becomes clearer after going back to a conventional trackpad or pen.
Of course, it is not going to make a laptop faster, but it makes interacting with one feel more deliberate. And for creative applications, where users are constantly moving, aligning, selecting and manipulating objects, that extra layer of feedback could become surprisingly useful.
Your phone just became part of the PC
Microsoft also showed a more understated feature that could end up being one of the most useful: deeper integration between Phone Link, an Android phone and Copilot CLI.
In the demo, a Nothing Phone was connected to the Surface laptop, after which the user could interact with files on the phone from the PC using natural language. Instead of manually opening the Gallery, hunting through hundreds of images and copying them across, the user simply asked Copilot CLI to find suitable images from the phone’s gallery for a social media post.
The AI searched through the available files, analysed the images, identified the relevant ones and copied them into a new folder on the Surface. Microsoft already supports accessing and transferring Android photos and files through Phone Link, but this demo showed what happens when an AI agent becomes the interface for that workflow.
What made the demo particularly interesting was how little the user actually had to do. There was no elaborate prompt engineering and no digging through folders. The instruction was essentially a normal sentence, with Copilot handling the steps underneath. Microsoft is clearly pushing toward a model where the PC understands the task rather than simply waiting for the user to operate individual applications.
The best part for me was that the demo neatly tied into Microsoft’s broader push for on-device AI. The phone provided the data, the Snapdragon-powered Surface did the heavy lifting locally, and the AI quietly took care of all the tedious steps in between.
Snapdragon X2 Elite gets a proper creative workout
The bigger spectacle was a creative workflow that would traditionally have demanded considerably more powerful hardware.
A Microsoft 3D artist started with a football sketched manually in Photoshop. That 2D concept was then taken into Blender, where AI helped turn it into a 3D asset. The artist could manipulate the model, scale it, add elements, and then push it back to Photoshop to work with lighting and shadows, while also introducing animation and camera movements.
The important bit wasn’t simply that Blender was running. The entire workflow remained responsive while the Surface convertible was running on battery. The demo then moved into generative AI. The final output was fed into an image-generation workflow to create a more realistic football-stadium scene, complete with lighting and environmental effects. From there, Microsoft showed a short video generated from the concept, turning what started as a rough sketch into a moving cinematic-style scene.
Now obviously, none of these individual AI tricks are particularly new anymore. What was interesting was seeing them chained together on a portable 2-in-1. And that’s exactly where the NPU story becomes much easier to understand. Microsoft’s new Snapdragon X2 Surface machines are designed to handle AI workloads locally where possible, with the NPU taking on workloads that would otherwise fall to the CPU or GPU. The new X2 Plus models announced at Snapdragon Summit offer an NPU capable of up to 80 TOPS, while Microsoft’s broader X2 lineup also includes more powerful Elite configurations.
The demo also made an important point about the difference between theoretical AI performance and practical use. Nobody watching it needed to know how many TOPS the processor had. The more useful question was simply: Can it actually do the work? In this case, the answer appeared to be yes.
Ink Canvas makes AI feel less like homework
Microsoft’s Ink Canvas demo was perhaps the most interesting from a usability perspective.
Instead of opening a chatbot and figuring out exactly how to describe an idea, Ink Canvas starts with something far more natural: a blank visual canvas. Users can add images, scribble notes with a pen, highlight things, and arrange elements into a mood board.
The clever part is what happens next. A user can use the lasso tool to select specific parts of the canvas and ask the AI to work with that context. The same approach can be used to generate images or create text summaries from the material on the board.
That removes one of the biggest annoyances with generative AI: having to translate a visual idea into a very specific text prompt. Want a particular object moved? Draw where it should go. Want to explain what matters in an image? Circle it. Want to build a visual concept? Put the pieces together first and let the AI figure out the next step.
Microsoft says Ink Canvas will be pre-installed on supported touch-enabled Surface devices. The canvas experience can work with local device capabilities, while the generative functions require connectivity and a Microsoft 365 subscription.
Can Snapdragon really take on Nvidia, though?
I also asked one of Microsoft’s designers how Snapdragon X2 Elite compares with Nvidia’s RTX Spark platform, particularly given Microsoft’s own push toward much more powerful local AI hardware.
The response was fairly pragmatic. Nvidia has the clear advantage in raw computing power and memory, which becomes increasingly important when running larger models or juggling several demanding workloads simultaneously. RTX Spark is effectively aimed at a different class of machine, with configurations offering enormous memory capacity and substantially more AI compute.
Snapdragon is trying to deliver enough local AI and graphics capability in a machine that can still be thin, convertible, and battery-powered.
But the conversation didn’t frame Snapdragon as simply “AI for mainstream users.” The designer’s view was that a professional creator can absolutely get meaningful creative work done on an X2 Elite Surface. The demo itself made that argument fairly convincingly. The difference is really about the ceiling. Nvidia’s hardware gives creators substantially more headroom when workloads become extreme, while Snapdragon is trying to deliver enough local AI and graphics capability in a machine that can still be thin, convertible, and battery-powered.
That distinction is perhaps the most interesting takeaway from Microsoft’s demos. The future of AI PCs may not be about finding one chip that does everything. It could instead be about making increasingly capable AI available at every level, from workstation-class machines built around huge amounts of memory and compute to Surface devices that can quietly run surprisingly sophisticated creative workflows while sitting in a café.
