Nvidia has introduced RTX Spark, a new superchip designed to put personal AI agents at the center of Windows computers. Announced at GTC Taipei, the company says the platform is meant to shift PCs from tools users operate manually to systems that can carry out tasks more proactively.
The chip pairs Nvidia’s Blackwell RTX GPU with a Grace CPU and is billed as delivering up to 1 petaflop of AI performance. Nvidia says the package is aimed at slim laptops and compact desktops, with up to 128GB of unified memory and a focus on efficiency as well as raw performance. The first systems are expected from partners including ASUS, Dell, HP, Lenovo, Microsoft Surface and MSI, with availability slated for this fall. Acer and Gigabyte are also expected to follow.
A central pitch for RTX Spark is that it gives users a way to run AI agents on-device rather than relying only on the cloud. Nvidia and Microsoft said they are working together on a Windows environment that includes new security primitives and Nvidia OpenShell, a runtime intended to help keep agents under user control.
The companies said the security tools are designed to support identity, containment and policy enforcement for native agent experiences. Nvidia OpenShell adds controls for what agents are allowed to do, can route some prompts to local models based on privacy settings, and can obscure personal information before queries are sent to cloud models.
That approach is already drawing support from agent developers such as Hermes Agent and OpenClaw, according to the companies. The goal is to let users run local assistants that can handle Windows applications, manage cross-app workflows, generate images and video, write code for plug-ins and apps, and search files on a local device.
Nvidia founder and CEO Jensen Huang framed the launch as a reinvention of the personal computer, saying users will increasingly ask their PC to complete work rather than launching apps one by one. Microsoft CEO Satya Nadella said the collaboration is intended to bring more intelligent computing to everyday Windows devices.
Nvidia says RTX Spark is not just for agents. The company is positioning it as a full-stack platform for creative work, development and gaming. It says users will be able to handle large 3D scenes, edit 12K video, run large language models locally and play games at 1440p with more than 100 frames per second.
The platform also adds support for new RTX features, including DLSS 4.5 Ray Reconstruction and RTX Video with 4x Frame Generation. Nvidia said more than 100 software vendors are adopting or supporting the platform, among them Adobe, Blackmagic Design, Blender, CapCut, ComfyUI and OTOY.
Adobe is among the most prominent partners. The company is reworking Photoshop and Premiere for RTX Spark, with an emphasis on faster AI tools, better performance and tighter integration with Windows agents. Adobe said its upcoming versions will use the chip’s unified memory, GPU and software stack to improve editing, color work and rendering.
The first RTX Spark systems are expected to be thin Windows laptops with all-day battery life and premium displays, along with small desktop PCs. Nvidia said the laptops will measure as little as 14 millimeters thick and weigh around three pounds, depending on the design.
The company is pitching the hardware as a new foundation for personal computing, one built around local intelligence rather than cloud-only services. Whether that model gains traction will likely depend on how well developers, PC makers and Microsoft can translate the promise of on-device agents into products users trust and actually use.