NVIDIA RTX Spark Compact Desktops Arrive with Up to 128 GB Unified Memory for Local AI

On November 18, 2026, NVIDIA announced that NVIDIA RTX Spark compact desktops are now available for purchase. The first models come from ASUS, Dell and MSI, with more from Acer, GIGABYTE, HP and Microsoft on the way. These mini PCs focus on high-performance local AI.

What happened?

The new PCs use the RTX Spark N1X superchip: a 20-core NVIDIA Grace CPU paired with a Blackwell RTX GPU featuring 6,144 CUDA cores and up to 128 GB of unified memory. NVIDIA says this delivers up to 1 petaflop of AI performance (FP4). Systems run Windows and include 10 Gb Ethernet. More details on the official NVIDIA blog.

They can run language models of up to 120 billion parameters locally, render 90 GB+ 3D scenes in real time and edit 4:2:2 video smoothly. Dedicated desktops can keep AI agents running 24/7 while a laptop is elsewhere.

Why does this matter?

The focus is local AI: data and code stay on the PC, avoiding cloud API costs for supported tasks. Tools such as Hermes Agent, NemoClaw and Switchyard (which routes between local and cloud) help creators and developers. Adobe, Blender (with DLSS Ray Reconstruction) and ComfyUI have native or accelerated support.

Beyond AI, the platform is gaining game support from studios including EA, Embark and Ubisoft, combining productivity, creation and entertainment in small machines.

What changes in practice?

Users with multiple PCs can use PAIR (Personal AI Router) software to distribute inference across idle machines on the local network. Prices start around US$4,400–4,500 for 64 GB configurations, according to retail reports from The Verge and VideoCardz.

Official source: NVIDIA blog and coverage from The Verge, VideoCardz and others.

Image credit: NVIDIA | Source: NVIDIA Blog

By Grok