NVIDIA RTX Spark is a chip for Windows laptops and small desktops that joins an NVIDIA graphics processor and main processor, so AI models and agents can run on the PC itself.
In plain words
RTX Spark puts an NVIDIA Blackwell graphics processor with up to 6,144 cores beside a Grace main processor with up to 20 cores, linked at 600 GB/s. They share up to 128 GB of memory, and NVIDIA quotes one petaflop of AI computing in FP4, a compact number format. NVIDIA says it can run models of up to 120 billion parameters with a 1-million-token context. It runs NVIDIA's CUDA software, so code written on it runs unchanged on other NVIDIA GPUs. NVIDIA and Microsoft announced it on 7 October 2026.
A quick example
Imagine a developer on a long train ride with no internet. On an RTX Spark laptop, the coding agent and its model both run on the laptop itself, so the work carries on offline and the code never leaves the machine.
Where you'll meet it
- Running and fine-tuning AI models on a Windows PC instead of in the cloud.
- Always-on AI agents on compact desktops.
- 3D work, video editing and games on slim laptops.
Related
Sources
- DocsNVIDIA, Microsoft Kick Off a New Beginning for Windows PCs With RTX Spark and AI Agents, NVIDIA · read 12 Oct 2026
- DocsNVIDIA RTX Spark, NVIDIA · read 7 Oct 2026
Checked by WhatIsAI against the sources above · 12 Oct 2026. Spotted an error? Report it.