NVIDIA RTX Spark laptops go on sale at $2,399 with up to 128GB of unified memory

NVIDIA announced on its blog that the first Windows laptops built around RTX Spark are now on sale. The platform pairs a Blackwell RTX GPU with a Grace CPU on a single superchip, targets slim premium Windows PCs, and starts at $2,399.

According to NVIDIA's specifications, RTX Spark delivers up to 1 petaflop of AI compute and up to 128GB of unified memory. Because the CPU and GPU share one memory pool, model weights, context and intermediate results do not have to be shuttled between VRAM and system RAM — one of the practical bottlenecks for running large models locally.

Initial systems come from ASUS, Dell, HP, Lenovo, Microsoft and MSI, and NVIDIA says additional machines are on the way. The company frames the pitch around three workloads: local AI, content creation and gaming, with the full NVIDIA AI platform packed into a portable device.

To mark the launch, NVIDIA held retail events in six cities. In Berlin, timed to the Festival of Lights, the TV Tower was lit with RTX Spark-themed lighting and the main event ran at XPERION, the gaming venue inside Saturn at Alexanderplatz, where attendees queued for hours to try the machines.

NVIDIA founder and CEO Jensen Huang made a surprise appearance in Berlin, greeting the crowd, handing out popcorn, posing for selfies and giving away five personally signed RTX Spark laptops. The other events were held at the JD MALL Shuangjing store in Beijing, Vatan Bilgisayar Topkapı in Istanbul, the COEX East Lobby in Seoul, a Best Buy in San Francisco and Muse Garden in Taipei.

For developers, NVIDIA says it has published tested recipes, playbooks and resources for running local AI on RTX Spark. That suggests the company is selling more than silicon: it wants the laptop to be an entry point into its local AI software stack.

The price is the part to watch. At $2,399, RTX Spark sits in the premium thin-and-light bracket, and NVIDIA will need real local inference performance to justify the premium. The 128GB unified memory option is what moves running a large model on a laptop from demo to deliverable workload.

Next comes supply and independent testing: configuration differences across the six OEMs, power draw and battery life, and large-model throughput at 128GB will determine whether this is a marketing category or a durable product line. NVIDIA says more systems are coming, so the pace of OEM follow-up is worth tracking.