NVIDIA DGX Spark

NVIDIA DGX™ Spark belongs to a new class of genAI supercomputer enabling developers to prototype, fine-tune and run inference on large model AI servers locally and seamlessly deploy to the data center or cloud.

Powered by the NVIDIA GB10 Grace Blackwell Superchip, NVIDIA DGX™ Spark delivers 1000 AI TOPS of AI performance in a power-efficient, compact form factor. This yields an HPC AI solution that brings reliable data center AI compute closer to your developers.

This cutting-edge LLM training server, pre-installed with the NVIDIA AI software stack and 128GB of memory, allows developers to run the latest generation of reasoning AI models from DeepSeek, Meta, Google, and others with up to 200 billion parameters locally, and then hand off artifacts to the data center or cloud for scale-out training and serving. The result is faster NVIDIA AI training cycles, simpler handoffs, and a truly scalable AI server footprint. That means DGX for genAI from lab to production.

Whether you’re building new pipelines or standardizing inference, this DGX™ AI platform streamlines everything from experimentation to enterprise deployment.

Watch Keynote Video