NVIDIA Opens RTX Spark Preorders for Windows AI PCs

NVIDIA RTX Spark preorders opened on October 7 as NVIDIA and Microsoft presented hardware and operating-system work for local AI agents on Windows. NVIDIA says RTX Spark laptops will be available October 16, with compact desktops going on sale in November. The same event previewed DGX Station for Windows.

Those are three different availability statements. A laptop preorder, a forthcoming compact desktop and a preview of a larger workstation should not be treated as products all shipping today.

RTX Spark brings the CUDA stack to Windows PCs

According to NVIDIA’s announcement, RTX Spark combines a Blackwell RTX GPU with a Grace CPU and offers configurations with up to 128GB of unified memory. NVIDIA describes the platform as running its full CUDA software stack.

Microsoft presented Surface Laptop Ultra around RTX Spark. NVIDIA also names Acer, ASUS, Dell, HP, Lenovo, MSI and Gigabyte among system manufacturers. The announcement describes both laptops and compact desktops; a buyer still needs the configuration and availability of the specific manufacturer’s system.

For developers, the relevant claim is software continuity between a local Windows machine and larger NVIDIA systems. That is a platform claim, not a benchmark of a particular model or an independently tested laptop.

DGX Station for Windows is a separate preview

The event also previewed DGX Station for Windows using the GB300 Grace Blackwell Ultra Desktop Superchip. NVIDIA lists 748GB of coherent memory and up to 20 petaFLOPS of FP4 AI compute for that platform.

These specifications concern DGX Station, not an RTX Spark laptop. Likewise, FP4 compute is a stated low-precision performance measure; it is not a prediction of every model’s throughput.

The announcement directs readers to register for availability information. It does not supply a general shipping date for DGX Station for Windows.

Windows agent infrastructure was part of the event

NVIDIA reports that Microsoft announced general availability of Microsoft Execution Containers, or MXC, as operating-system infrastructure for agents running persistently in the background. Microsoft described OS control and governance as part of that design.

That software announcement sits alongside the hardware roadmap. It does not establish that every application can immediately run an autonomous agent or that each announced PC has passed the same application tests.

For context on the hardware distinctions, browse AI chip explainers. The next practical checks are the manufacturer’s actual configuration, shipping date and supported workflow.

Source

NVIDIA announcement, October 7, 2026, reviewed October 8, 2026. Hardware and performance specifications are NVIDIA’s published claims; this report does not present them as independent measurements.