NVIDIA announces 64GB DGX Spark desktop AI system
NVIDIA has announced a 64GB configuration of DGX Spark, a desktop AI system built around the GB10 Grace Blackwell superchip. The configuration will be offered by Acer, ASUS, Dell, Gigabyte, HP and MSI, with availability listed for October 23, 2026, through NVIDIA Marketplace, OEM partners and retail channels.
The system combines a 20-core Arm CPU with a Blackwell GPU and provides up to 1 petaFLOP of FP4 AI computing with sparsity. Its unified 64GB LPDDR5x memory is shared by the CPU and GPU over NVLink-C2C, while storage options range from 1TB to 4TB. NVIDIA says the capacity is suited to current 30–35B-parameter open models, while the maximum local model size is up to 100B parameters, depending on configuration and runtime requirements.
DGX OS includes NVIDIA’s AI software stack, including PyTorch, Jupyter and Ollama. ConnectX-7 networking allows two 64GB systems to combine 128GB of memory and additional compute, with NVIDIA stating that the cluster can reach up to 1.7 times the performance of one 128GB system. The 128GB DGX Spark remains available for larger single-box workloads.
Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.