NVIDIA said today that RTX Spark laptops are now available from ASUS, Dell, HP, Lenovo, Microsoft and MSI, putting its new class of Windows PCs into the market as a single announcement spreads across multiple makers. The rollout centers on slim, premium machines built for creators, developers and gamers who want local AI on the device itself.
The pitch is simple enough to fit on a spec sheet and broad enough to change how a PC is used: up to 128GB of unified memory, 1 petaflop of FP4 AI performance and a superchip that pairs an NVIDIA Blackwell RTX GPU with an NVIDIA Grace CPU. In practical terms, that means an RTX Spark PC can run large language models and AI agents locally instead of pushing every task to the cloud.
That local-first design is why the launch is drawing attention today. NVIDIA says RTX Spark combines RTX graphics and its AI platform in Windows PCs that also include CUDA, DLSS, OptiX, Reflex, G-SYNC, TensorRT and hardware-accelerated video, giving the machines a familiar desktop feel while adding a new layer of compute underneath. It also ties the product to a broader Windows push, not just a single device debut. A separate report on Microsoft's version of the same platform, Rtx Spark Laptop debuts as Microsoft pairs NVIDIA GPU with 128GB memory, points to how quickly the category is spreading.
The friction is in the promise versus the setup. NVIDIA is selling RTX Spark as a way to make local AI feel practical on a Windows laptop, but the system still asks users to choose which agents to trust and how much access to give them. Guided setup and easy installers are meant to lower that barrier, and NVIDIA lists GitHub Copilot, Hermes, OpenClaw, Perplexity, Pi and WorkBuddy among the agents people can start with.
Dillon Rolnick said Hermes was brought to RTX Spark so people can put open AI models and agents to work locally on a Windows laptop, and he added that personal AI becomes more useful when it can work with users' tools and help get things done. NVIDIA also points to Nemotron 3.5 Bolt 74B-A9B 24GB, a 74-billion-parameter model tuned for tool calling and multistep tasks, as part of the software stack, with the NVIDIA OpenShell secure runtime and Microsoft Execution Containers helping control permissions. The unanswered piece is the one buyers will care about next: which specific RTX Spark models each partner will ship, and when those laptops will actually be in stores.
