Nvidia Launches 64GB DGX Spark to Lower AI Entry Cost

Nvidia is launching a 64GB version of its DGX Spark platform and GB10 SoC on October 23 to provide a lower-cost entry point for local AI users. The move comes as a global silicon and memory supply crunch has driven the market price of the existing 128GB model up to between $7,000 and $9,000. The new configuration starts at $4,999 and maintains the same 20-core Arm CPU and 273 GB/s shared memory bandwidth as the higher-end model, though it features half the RAM and storage.

To facilitate scaling, Nvidia is releasing the Sync Cluster Assistant, a software tool that automates the connection of multiple Spark systems. The accompanying Sync app will include a Model Launcher for the automatic deployment of models such as Qwen 3.8 27B. The hardware will be available through several partners, including Acer, Asus, Dell, Gigabyte, HP, and MSI.

Gigabyte has already announced its specific implementation, the 64GB unified memory version of the AI TOP ATOM desktop AI development system. This system integrates the CUDA software ecosystem and the Gigabyte AI TOP Utility to support inference and retrieval-augmented generation. Using ConnectX-7 networking and Nvidia Sync, users can cluster up to four units to expand compute and memory pools. Gigabyte is currently testing the platform for agentic AI applications, recently demonstrating a scientific computing workflow using Nvidia Nemotron open models and the NemoClaw open agent blueprint.