What Is an RTX Spark PC and Why It Matters
RTX Spark PCs are a new class of Windows systems built to run local AI agents directly on the machine, combining a Blackwell RTX GPU, an Nvidia Grace CPU, and large unified memory so that everyday apps can host assistants that plan, create, and automate without depending on the cloud. Instead of treating the computer as a grid of icons, RTX Spark PCs are designed so the system can carry out tasks on your behalf, from searching local files to generating content. Nvidia and Microsoft announced the platform at Computex, positioning it as the next step for creators, developers, and gamers who want on-device AI computing. These Windows AI acceleration features aim to make local AI agents feel like teammates rather than tools, while still keeping data on your desktop where you control it.
Hardware and Performance: 1 Petaflop Power and 2x Faster Inference
Under the hood, an RTX Spark PC pairs an Arm-based 20-core Nvidia Grace CPU, co-developed with MediaTek, with a Blackwell RTX GPU and up to 128GB of unified memory. Nvidia says this configuration can deliver up to 1 petaflop of AI performance, enough to support demanding models for local AI agents, media generation, and Windows AI acceleration workloads. The company has worked with the llama.cpp community to add multi-token prediction and other GPU optimizations, achieving 2x inference performance on popular agentic models running locally. These gains matter in practice: faster token generation means smoother conversational agents, quicker code completion, and more responsive productivity flows, all without sending each prompt to the cloud. Multi-GPU enhancements for llama.cpp and ComfyUI further extend this on-device AI computing performance for enthusiasts with dual RTX setups.
Secure, On-Device AI Agents with Microsoft and OpenShell
Running local AI agents safely at scale requires more than raw speed, so Nvidia and Microsoft are aligning RTX Spark PCs with new Windows security primitives. These features provide identity, containment, policy, and end-to-end protections designed for agent workloads. On top of this foundation, Nvidia’s OpenShell runtime lets users and developers define what local AI agents can and cannot do, set when to favor local models over cloud services, and disguise personal information in any request that must leave the device. According to Nvidia, this approach is meant to keep agents “under full user control” while they operate across Windows applications. Projects like Hermes Agent and OpenClaw are already integrating OpenShell and Microsoft’s security tools, bringing personal, private, on-device AI computing to standard RTX Spark PC desktops and laptops.
From Enterprise Deskside to Consumer Laptops
RTX Spark is part of a broader effort to spread local AI agents across both professional and consumer hardware. Nvidia and Microsoft highlighted RTX Spark laptops and desktops from major PC brands including Microsoft, Dell, HP, Lenovo, ASUS, and MSI, all expected to arrive this fall. At the high end, Nvidia DGX Station for Windows brings data-center-class GPUs and CPUs into a deskside system for advanced inference and enterprise workflows. For individual users, RTX Spark PCs aim to make on-device AI computing as common as discrete graphics, with all-day battery life for slim laptops and efficient desktop towers. Nvidia’s collaboration with H Company adds another layer: computer-use tools and desktop agent harnesses that let agents see the screen and operate mouse and keyboard input, even in apps without APIs, while still running models locally on RTX and DGX platforms.
Creative and Productivity Apps Get Local AI Upgrades
RTX Spark PCs are not only about infrastructure; they are also about the apps people already use for creative and professional work. Adobe is rearchitecting Photoshop and Premiere to take advantage of the performance and memory profile of RTX Spark systems, improving AI-accelerated editing and generative features on-device. Blender is adding Nvidia DLSS 4.5 Ray Reconstruction, and ComfyUI will gain RTX Video Frame Generation, giving artists and 3D creators faster previews and higher-quality renders with Windows AI acceleration. Nvidia Broadcast 2.2 adds Studio Voice optimizations, while Stream Deck integration arrives for both Broadcast and Project G-Assist, tying AI features into streaming and gaming workflows. Together, these software updates position the RTX Spark PC as a capable hub for local AI agents that support content creation, automation, and real-time media without constant cloud round trips.