Nvidia has unveiled the RTX Spark, a chip designed for Windows PCs that focuses on running AI agents and large language models locally. This project has been in development with Microsoft for nearly three years.
Create, build, and play: https://t.co/jYkBa10Mq8
— NVIDIA (@nvidia) June 1, 2026
The company positions the RTX Spark as the foundation for personal AI systems that operate directly on the computer without the need for cloud access.
The claimed performance is up to 1 petaflop of computing power and up to 128 GB of unified memory. Nvidia expects users to be able to run agent-based and generative models directly on their PCs.
The chip is built on the Blackwell RTX GPU with 6,144 CUDA cores and a 20-core Grace processor, developed in collaboration with MediaTek. The platform is aimed at Windows 11 and scenarios involving AI agents, content creation, and gaming.
Microsoft Joins the Agent-Driven Windows Initiative
Nvidia and Microsoft have announced their collaboration on infrastructure for agent systems within Windows.
The release mentions new security mechanisms and the NVIDIA OpenShell platform for running agents on user devices. This is about preparing Windows for such scenarios rather than a complete transition of the OS to an agent-based model.
Microsoft is already advancing this area through Copilot and enterprise products, where assistants automate workflows.
Nvidia CEO Jensen Huang described this shift as moving from "PC as a tool" to "PC as an assistant." Essentially, this is about launching a new hardware platform for local AI.
Adobe Optimizes Photoshop and Premiere
Adobe is adapting Photoshop and Premiere for the RTX Spark. Nvidia has reported a twofold increase in AI and graphics performance on the new platform, though independent verification of these claims is lacking.
In May, Adobe added an on-device model for the Remove tool in Photoshop 27.7, allowing users to delete objects without sending data to the cloud. Users can choose between local and cloud processing.
For Premiere, the company introduced AI features, including Object Mask, but has not announced a full transition of the video editor to local AI.
Release Scheduled for Fall 2026
The first PCs based on the RTX Spark are set to launch in the fall of 2026. Partners include ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI, with Acer and GIGABYTE joining later.
Nvidia is doubling down on the AI-PC segment, where AI agents, data analysis, and content generation are performed locally—without a constant reliance on the cloud.
Recall that in March, the company announced plans to launch infrastructure for creating data centers in space.