Qualcomm Discrete NPU Spotted in Dell Pro Max Plus laptop at DTW

Dell Technologies World recently wrapped up in Las Vegas, and a big theme for the event was distributing AI workloads out of the datacenter, down to the edge and ultimately to the individual endpoint devices. For lots of devices this means using laptops based on a CPU with a built-in NPU like the Ryzen AI 300 series or Intel Core Ultra 200 series. However, Dell and Qualcomm have come up with another option; an add-in NPU that takes the place of a discrete GPU in a laptop to enable higher-performance AI processing than any built-in NPU could hope to offer.

What you are looking at is the Dell Pro Max Plus laptop, fitted with a pair of Qualcomm’s Cloud AI 100 processors on a card that sits where one might normally install an NVIDIA GPU. This configuration is essentially a prototype. There are only a handful of these systems so far, each hand-assembled. Each of the Cloud AI 100 processors comes paired with 32GB of memory, but they are presented to the system in a unified manner, and they can be loaded up with models requiring up to 64GB of accelertator memory. Dell said that they have tested running several models, including Llama 4 Scout which is a 109 billion parameter AI model.

This AI accelerator is intended to be combined with Dell’s Pro AI Studio, a software solution designed to help make AI more accessible to developers with pre-validated AI models and essentially push-button deployment. On the show floor they were demonstrating using this accelerator and the Pro AI Studio for AI-powered development tasks, such as creating and modifying a game engine.

That 64GB of NPU memory in a mobile form factor is a key advantage. 100+ billion parameter models like Llama 4 scout simply won’t fit into the memory that comes on the NVIDIA RTX PRO 5000 Blackwell Generation, which is equipped with 24GB of RAM. Additionally, by comparison Dell claims this new Qualcomm discrete NPU consumes less power than NVIDIA’s latest GPU, though the exact specifics are still forthcoming.

Of course you give up the other capabilities of a GPU. The Qualcomm accelerator cannot render graphics like the Blackwell card can, but Qualcomm and Dell are betting there is a market for high-performance AI acceleration in a mobile form factor.

Exact specifications are light on the ground at the moment since these are effectively prototypes. The Cloud AI 100 processor itself is not new itself; you can read about it here  and here. The form-factor is the new part, and both Dell and Qualcomm have yet to iron out all of the exact details for the pending retail release. Minor things like pricing, and which systems such a card might find itself available in are yet to be worked out, but Dell was adamant that they are marching towards release.

Final Words

This one is neat mostly because of the form factor. Apparently it was a demonstration that came together just for Dell Tech World 2025. The real question is whether people are willing to give up the rendering and media pipelines of the NVIDIA GPU for the additional memory capacity and performance per watt for the Qualcomm NPU. Perhaps there is also a bit of whether folks are willing to give up NVIDIA and the CUDA solutions for Qualcomm as well. Another competitor in this space is likely to be the NVIDIA GB10-based solutions when they launch for client devices and solutions like the AMD Ryzen AI Max+ 395. Larger pools of LPDDR5X may end up being the go-to in this form factor for AI workloads as LLMs are often memory capacity bound.