Personal AI Supercomputer on the Desk with NVIDIA DGX Spark

For organizations and institutions looking to start their AI journey—from development to experimentation—NVIDIA DGX Spark is a new-generation AI Supercomputer designed to deliver enterprise-class performance in a compact, desktop-ready form factor, without the need to rely on cloud GPUs at all times.

The system comes with the NVIDIA AI Software Stack pre-installed from the factory. Simply connect a monitor, keyboard, and mouse, and it is ready for immediate use.

The size of your hand and a weight of only 1.2 kg, DGX Spark brings data-center-level

AI power into a form factor that is accessible to everyone.

Who DGX Spark Is For

Educational Institutions & Universities
Ideal for AI, Machine Learning, Robotics, and Data Science programs
- Enables students and researchers to work hands-on with large language models (LLMs) at the 200–405B parameter scale
- Reduces budget constraints and dependency on limited cloud GPU queues
- Supports enterprise-grade research and industry collaboration

Organizations Starting with AI
- Develop AI capabilities in-house without investing in large data center infrastructure
- Deploy as an AI workstation or a personal AI cloud
- Run AI workloads offline or locally, ensuring sensitive data remains within the organization and aligned with GDPR and compliance requirements
- Supports clustering of up to two systems to handle larger models and workloads
Key Parameters & Capabilities
DGX Spark Platform
Powered by the NVIDIA Grace Blackwell GB200 architecture, DGX Spark delivers AI performance that you fully control:

- Memory: 128GB unified memory
- Memory Bandwidth: 273 GB/s
- Storage: 4TB SSD
- Power & Deployment: Standard power (240W); no data center required
- Operating Environment: Pre-installed NVIDIA AI Software Stack
- Compute Performance: Up to 1 PetaFLOP (FP4) using NVFP4 technology
- Large Model Support:
- Up to 200B parameters on a single system
- Expandable to 405B+ parameters by linking two systems
- Designed to handle large models and multiple concurrent workloads efficiently

Available DGX Spark-Based Systems
- MSI EdgeXpert MS-C931
- Lenovo ThinkStation PGX
- ASUS Ascent GX10
- Dell Pro Max with GB10
- Acer Veriton GN100 AI Mini Workstation