NVIDIA DGX Spark is a next-generation desktop AI workstation designed specifically for building, fine-tuning, and running large-scale artificial intelligence models locally. Unlike traditional desktops or workstations, DGX Spark represents a new class of personal AI supercomputers, delivering enterprise-grade AI performance in a compact, desk-side form factor.

At the heart of NVIDIA DGX Spark is the NVIDIA GB10 Grace Blackwell Superchip, built on the NVIDIA Grace Blackwell architecture. This advanced platform tightly integrates a high-performance NVIDIA Blackwell GPU with fifth-generation Tensor Cores and a 20-core Arm-based NVIDIA Grace CPU. The CPU and GPU communicate through NVLink-C2C technology, enabling a coherent memory model with significantly higher bandwidth than conventional PCIe-based systems. This architecture ensures faster data movement, lower latency, and optimized AI workflows.

One of the defining advantages of DGX Spark is its 128GB of unified system memory. This shared memory pool allows both CPU and GPU to access the same data without duplication, which is critical for modern AI workloads. With this configuration, developers can work with AI models containing up to 200 billion parameters on a single system. For even larger models, two DGX Spark systems can be connected using NVIDIA ConnectX networking, extending inference capabilities to models up to 405 billion parameters.

Despite its powerful capabilities, NVIDIA DGX Spark is engineered for desktop deployment. Its compact design makes it suitable for offices, research labs, universities, and enterprise innovation teams that need high-performance AI compute without the complexity of rack-mounted servers or constant reliance on cloud infrastructure. The system includes high-speed NVMe storage, modern connectivity options, and enterprise-class networking, making it easy to integrate into existing IT environments.

NVIDIA DGX Spark ships with NVIDIA DGX OS, an Ubuntu Linux–based operating system optimized for AI workloads. It comes preconfigured with NVIDIA’s complete AI software stack, including CUDA, cuDNN, cuBLAS, TensorRT, NCCL, Docker with NVIDIA Container Toolkit, and support for popular AI frameworks such as PyTorch and TensorFlow. Developers also gain access to NVIDIA NIM, AI Blueprints, and NGC containers, enabling rapid development and seamless transition from local experimentation to data center or cloud deployment.

NVIDIA DGX Spark is ideally suited for AI developers, data scientists, enterprises, startups, and educational institutions seeking powerful local AI compute. It accelerates development cycles, reduces dependency on shared cluster resources, and provides a consistent platform aligned with NVIDIA’s enterprise AI ecosystem.