NVIDIA ships RTX Spark laptops with up to 128GB unified memory

NVIDIA said its RTX Spark laptops are available globally today through ASUS, Dell, HP, Lenovo, Microsoft and MSI, marking the company’s latest push to bring AI-focused hardware into slim Windows PCs. The launch matters because the systems combine NVIDIA Blackwell RTX graphics, CUDA software and up to 128GB of unified memory in a portable form factor aimed at creation, gaming and local AI workloads, according to NVIDIA’s Oct. 16, 2026 announcement.

Evidence and context

NVIDIA said RTX Spark systems deliver up to 1 petaflop of FP4 AI performance and are designed to run large language models and AI agents locally. The company said the laptops can support demanding workflows such as editing 8K 4:2:2 video, rendering 3D scenes of more than 90GB and running supported games at 1440p at over 100 frames per second, although those performance figures were presented by NVIDIA and depend on the specific configuration and software being used.

The company described RTX Spark as a single superchip pairing an NVIDIA Blackwell RTX GPU with an NVIDIA Grace CPU. NVIDIA said MediaTek contributed to the custom CPU design. The platform also includes NVIDIA CUDA, DLSS, OptiX, Reflex, G-SYNC, TensorRT and hardware-accelerated video, with the company positioning the devices as portable systems that can handle both creative and AI workloads on device.

NVIDIA and Microsoft both framed the launch around local AI computing on Windows. Jensen Huang, NVIDIA’s founder and chief executive, said RTX Spark brings NVIDIA AI, RTX graphics and CUDA together with Microsoft Windows for creating, gaming and running personal AI agents. Yusuf Mehdi, Microsoft’s executive vice president, said Windows PCs powered by RTX Spark expand local compute for creators and developers.

Developer, creator and gaming support

NVIDIA said the laptops ship with day-one support for local AI agents and creative apps, including GitHub Copilot, Hermes, OpenClaw, Perplexity, Pi and WorkBuddy, alongside NVIDIA’s own tools such as OpenShell and agent skills. The company said supported agents can work with tools including NVIDIA Broadcast, RTX Remix and Omniverse libraries. It also said RTX Spark can serve AI inference to compatible Windows, Linux and macOS computers on a home network through NVIDIA PAIR.

For creative users, NVIDIA said Adobe is adding TensorRT acceleration in AI-editing features across Photoshop, Premiere and Substance 3D. The company said Photoshop and Premiere have native RTX Spark support through beta releases at launch, while Substance support is due later. NVIDIA also highlighted Broadcast 2.3, beta RTX Video features and new RTX Video nodes in ComfyUI.

On the software development side, NVIDIA said RTX Spark brings coding, testing and prototyping onto one machine with native CUDA support, Windows Developer Config and Windows Subsystem for Linux. The company said it and Microsoft prepared tools including the first PyTorch-CUDA build for Windows on Arm, the first Python build native by default, the first native Conda installer and support for 97% of the top 15,000 Python packages, as cited in the announcement.

NVIDIA said launch game support includes Cyberpunk 2077, Fortnite, Gears of War: E-Day, Marvel Rivals, World of Warcraft and NBA 2K27. The company said NBA 2K27 uses DLSS 5 with 3D-Guided Neural Rendering, and said a new 2x faster AI model available today can boost DLSS 5 frame rates by up to 1.4x.

Availability and pricing

NVIDIA said RTX Spark laptops are available now through the websites of ASUS, Dell, HP, Lenovo, Microsoft and MSI, as well as major retailers. Pricing starts at $2,599, with configurations and pricing varying by region, according to the company. The initial lineup includes ASUS ProArt P14 and P16, Dell XPS 16, HP OmniBook Ultra and OmniBook X 14, Lenovo Yoga Pro 9n, Microsoft Surface Laptop Ultra, and MSI Prestige N16 Flip AI+ and Prestige N16.

The announcement did not provide a separate shipment target, sales forecast or follow-up catalyst beyond the stated global availability date.