There’s lots to like about MSI’s latest flagship gaming laptop, but don't let the marketing fool you. The Stealth 18 AI Studio A1V is so all-in on its artificial intelligence features that MSI put the term in its name. It uses Intel’s 14th-gen Meteor Lake processors with a dedicated neural processing unit (NPU) to handle AI tasks, but this machine excels in more traditional applications.
The NPU is a portion of the Intel Core Ultra 9 185H processor inside the Stealth 18 optimized for tasks that rely on machine learning. This can include image generation or speech-to-text, which are sometimes offloaded to remote servers. Running them locally improves speed, and a dedicated NPU can perform the tasks far more efficiently than a typical CPU.
In practice, however, very few tasks make use of the NPU. If you set that aside, the Meteor Lake Intel processor is still powerful for everyday tasks. Add that to the Nvidia RTX 4090 graphics card, 32 GB of RAM, and a beautiful 4K Mini-LED display, and you’ve got a great gaming laptop and one that'll do a killer job with video editing and visual effects. All without touching the AI stuff.
An Empty Seat for AI
The idea of a dedicated processor for AI tasks has been pretty established in the smartphones world but has only started gaining traction in laptops in recent years. It’s not all that different from how your graphics card works. Graphically intensive tasks like video games require computing power performed in ways that aren't very efficient for CPUs. Graphics processing units (GPUs) are designed to do these tasks better, at the expense of being worse at more generalized tasks like browsing the web. Since your computer has both, it can rely on the GPU when needed but stick with the CPU when it isn’t.
GPUs are also pretty good at doing AI-related tasks, but they’re not ideal for that, either. Hence the need for an NPU. The problem? Unlike graphics-related tasks that can be useful in various scenarioses, NPUs are only good for tasks that rely on machine learning. Despite the generative AI hype, there aren’t lots of features that make use of this power.
In my testing, I'd check the Windows task manager to see how much the NPU was utilized while performing common tasks. Windows’ built-in voice typing function made use of it, but almost nothing else did. This is partly because few applications have added support for NPUs. Even Microsoft’s own Copilot, which the company has announced will run locally at some point in the future, relied on cloud processing when I tried it on the Stealth 18. When I tried running Stable Diffusion locally, it used the GPU rather than the NPU. And while there are ways to get it to use the NPU, you’re going to be diving into some heavy technical modification to make it work. This isn’t a magic bullet that will make every AI task faster.