The new laptops combine Blackwell graphics and Grace processing with up to 128GB of unified memory, targeting local AI agents, creative workloads and gaming from $2,599.
NVIDIA and Microsoft have unveiled RTX Spark, a new class of Windows PCs aimed at gaming, content creation and on-device AI workloads. The company says the first systems pair an NVIDIA Blackwell RTX GPU with a Grace CPU in a single superchip, and can ship with as much as 128GB of unified memory and up to 1 petaflop of FP4 AI performance.
The pitch is straightforward: let users run large language models and personal AI agents locally rather than depending on cloud services. NVIDIA says RTX Spark supports agents such as GitHub Copilot, Hermes, OpenClaw, Perplexity, Pi and WorkBuddy, with easier installation and configuration than more manual workstation setups. It has also tuned its Nemotron 3.5 Bolt 74B-A9B model for tool use and multi-step agent tasks, a sign that the company is pushing beyond raw model execution towards more structured automation.
Security controls are part of that strategy. NVIDIA OpenShell and Microsoft Execution Containers are intended to limit what agents can access and help manage permissions over system resources. NVIDIA also introduced PAIR, which allows RTX Spark systems to share AI inference capacity with other machines on a local network, including Windows, Linux and macOS devices. In practice, that turns the laptop into both a personal workstation and a small local AI hub.
The company is also using the platform to court creators. According to NVIDIA, RTX Spark machines are designed for workloads such as 8K 4:2:2 video editing and 3D rendering of scenes as large as 90GB, including when running on battery power. Adobe is adding TensorRT acceleration for AI features in Photoshop, Premiere Pro and Substance 3D, with beta support for Photoshop and Premiere available at launch and Substance support to follow later. NVIDIA Broadcast 2.3 brings improved noise suppression and Studio Voice Mic Profiles, while RTX Video adds up to 4K super resolution and 4x frame generation in beta.
For developers, the platform includes native CUDA support, compatibility with Windows Subsystem for Linux, PyTorch-CUDA for Windows on Arm, and native Python and Conda tooling. NVIDIA says it expects compatibility with 97% of the top 15,000 Python packages. The company also highlights MediaTek’s role in the custom CPU design, which it says was shaped around efficiency, performance and connectivity.
RTX Spark is not just being positioned as a workstation platform, but as a gaming one too. NVIDIA says the systems support DLSS 5, ray tracing, G-SYNC, Reflex and other RTX features, with launch titles including Cyberpunk 2077, Fortnite, Gears of War: E-Day, Marvel Rivals, World of Warcraft and NBA 2K27. The first laptops are on sale globally from $2,599, with models from ASUS, Dell, HP, Lenovo, Microsoft and MSI, although configuration and regional pricing vary.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
