Elon Musk Receives NVIDIA DGX Spark: Petaflop AI Supercomputer at SpaceX

The next AI revolution has officially begun at SpaceX's Starbase, Texas. NVIDIA CEO Jensen Huang personally delivered the inaugural DGX Spark system to Elon Musk, marking the start of the rollout for what NVIDIA calls the world’s smallest AI supercomputer.
Huang's visit to the SpaceX facility involved handing over the DGX Spark to Musk amidst the backdrop of towering rocket engines. This delivery signals NVIDIA's commitment to pushing AI development beyond traditional data centers.
DGX Spark: Supercomputing Power in a Compact Package
The DGX Spark is designed to bring supercomputer-class performance to a wider range of users. This includes developers, researchers, and creators who need powerful AI capabilities in a portable format.
Inside, the DGX Spark boasts impressive specifications:
- NVIDIA GB10 Grace Blackwell Superchip: Delivers up to 1 petaflop of AI performance at FP4 precision.
- 128GB of unified CPU-GPU memory: Enables local prototyping, fine-tuning, and inference without relying on cloud instances.
- NVIDIA ConnectX networking and NVLink-C2C: Provides high-speed connectivity for clustering and data transfer.
- NVMe storage and HDMI output: Ensures fast storage access and visual display capabilities.
Beyond the hardware, the DGX Spark includes the full NVIDIA AI software stack, including frameworks, libraries, pretrained models, and NVIDIA NIM microservices. This allows for immediate deployment of AI workflows.
Partnerships and Broader Availability
NVIDIA is collaborating with partners like Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, and MSI to integrate DGX Spark into their systems. These partnerships aim to transform desktops into AI development platforms.
The DGX Spark is already being used by various organizations, including:
- Ollama: Rewriting how developers run large language models locally.
- NYU Global Frontier Lab: Prototyping algorithms for privacy-sensitive applications.
- Zipline: Pushing the boundaries of autonomous delivery.
- Arizona State University: Running robotics simulations and vision models at the edge.
- Refik Anadol’s studio: Blending art and AI with petaflop performance.
DGX Spark will be generally available starting Wednesday, Oct. 15, on NVIDIA.com and through partners worldwide.