Summary
Alex Finn tests the Microsoft Surface Ultra with the new Nvidia RTX Spark chip, a laptop pitched as the first built specifically for local AI. The RTX Spark is the same silicon found in the DGX Spark, and it supports up to 128 GB of unified memory. Finn’s unit has 64 GB, which he uses to see how far local models can go. Microsoft sent the laptop for free, and Finn says the video itself is not sponsored by them.
Rather than covering ports and display specs, the review is aimed at AI enthusiasts. Finn explains how local AI differs from cloud AI: instead of prompts traveling to a data center full of GPUs and back, all the computation happens on the device, which means no internet requirement, no per-use subscription cost, and better privacy.
He then shows how to load local models using Hermes Agent, a free open-source agent that runs on any computer and makes it easy to connect local models. The walkthrough applies to any hardware, even inexpensive machines, so viewers do not need this laptop to follow along.
The video closes by weighing whether a powerful local setup can realistically replace paid AI subscriptions, making it useful for anyone considering running models on their own hardware.
📺 Source: Alex Finn · Published November 02, 2026
🏷️ Format: Review







