NVIDIA DGX Spark AI Desktop — 128GB Unified Memory, 4TB

NVIDIA DGX Spark at a glance

The NVIDIA DGX Spark AI desktop brings the GB10 Grace Blackwell platform into a compact personal AI supercomputer for developers, researchers, data scientists and advanced creators who want to explore demanding AI workloads locally. Its 128GB unified memory architecture gives the CPU and GPU access to a shared memory pool, while the 4TB storage capacity provides room for models, datasets and development environments.

Why it stands out

• GB10 Grace Blackwell platform: built for local inference, model development and AI experimentation.

• 128GB unified memory: designed for working with larger AI models without a separate discrete GPU memory pool.

• 4TB storage: practical space for software, model files and project data.

• DGX OS and NVIDIA software: a purpose-built environment for the CUDA and AI development ecosystem.

• Desktop-scale footprint: brings serious AI computing to a desk rather than requiring a conventional data-center setup.

Who it is for

This NVIDIA DGX Spark AI desktop is best suited to technical users who want a dedicated environment for prototyping, inference, fine-tuning and research. NVIDIA documents support for AI models up to 200 billion parameters and up to 1 PFLOP FP4 performance with sparsity, making the platform particularly relevant to large-model experimentation and local development.

Points to consider

DGX Spark is a specialized AI computer, not a general-purpose desktop replacement for every workflow. Its ARM-based platform means buyers should confirm that their preferred applications, containers and development tools support the required architecture. For best performance, NVIDIA recommends using the supplied power adapter.

Our verdict

The NVIDIA DGX Spark AI desktop is a focused option for builders who want substantial unified memory, NVIDIA’s AI software stack and a compact local platform. It makes the most sense when local model development and experimentation matter more than a traditional consumer PC specification sheet.