NVIDIA took the stage at GTC Taipei 2026 to announce RTX Spark, a new chip designed to run personal AI agents directly on Windows PCs. No cloud required.
The announcement reflects how both NVIDIA and Microsoft want to reframe the PC as less of a tool you operate, and more of an AI collaborator that works alongside you.
New Chip, Big Claims
RTX Spark delivers 1 petaflop of AI performance and up to 128 GB of unified memory for its laptop models. NVIDIA is calling it the first Windows chip built specifically for running personal AI agents on-device. Slim laptops powered by RTX Spark are promised to last all day on a single charge.
On the performance side, NVIDIA says the chip can handle 3D scenes over 90 GB, 12K video editing, 4K AI video generation, and 120-billion-parameter AI models with up to one million tokens of context, all running locally.
For gaming, the RTX Spark targets 1440p at over 100 FPS. We saw demos of 007: First Light and Pragmata running on RTX Spark notebooks, and both looked promising, at least based on the preview.
These demos are most likely running on the top tier RTX Spark chips so we have to wait and see how the actual numbers stacks up with various brand models and or SKUs.
Windows Around Agents
NVIDIA and Microsoft are collaborating on a native Windows experience for AI agents, including new security features and something called NVIDIA OpenShell.
It’s basically a way to let agents run securely on your main device without opening the door to obvious security risks.
Adobe for RTX Spark
Adobe is rebuilding Photoshop and Premiere specifically for RTX Spark, promising twice the AI and graphics performance compared to what’s available today. Expect other software vendors to follow suit.
Availability
RTX Spark devices (laptops and compact desktops) are coming this fall from ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI. Acer and GIGABYTE will follow.
The AI PC space is getting crowded, with Qualcomm and Apple already competing for the same buyers. NVIDIA is betting its AI pedigree gives RTX Spark an edge, particularly with developers and businesses that want more AI capability running locally.
