The next phase of the AI revolution begins at SpaceX, with NVIDIA CEO Jensen Huang handing over the first DGX Spark to Elon Musk. This compact powerhouse brings unprecedented AI capabilities directly to where innovation happens.

Huang's visit to the SpaceX facility, filled with rocket engines, marked the start of DGX Spark's rollout. He recounted delivering the first DGX system to OpenAI, emphasizing how Spark expands that mission.

DGX Spark: A Pocket-Sized Supercomputer

The DGX Spark is engineered to deliver a petaflop of AI performance with 128GB of unified memory. This allows it to locally run models with up to 200 billion parameters. It's tailored for developers, researchers, and creators seeking supercomputer-level performance on the go.

Huang highlighted the significance of the delivery, stating, “Imagine delivering the smallest supercomputer next to the biggest rocket.” This emphasizes the shift of AI development from centralized data centers to edge environments.

Inside the DGX Spark

This compact system, about the size of a hardcover book, is packed with cutting-edge technology:

NVIDIA GB10 Grace Blackwell Superchip: Delivers up to 1 petaflop of AI performance.
128GB of unified CPU-GPU memory: Enables local prototyping, fine-tuning, and inference.
NVIDIA ConnectX networking: Facilitates clustering, with NVIDIA NVLink-C2C enabling 5x PCIe bandwidth.
NVMe storage: Ensures speed, with HDMI output for visuals.

Software and Applications

DGX Spark includes the complete NVIDIA AI software stack. This includes frameworks, libraries, pre-trained models, and NVIDIA NIM microservices. This enables a range of workflows, such as:

Customizing image-generation models like FLUX.
Building vision search and summarization agents with NVIDIA Cosmos.
Deploying optimized chatbots using Qwen.

Partnerships and Availability

The DGX Spark is already empowering innovators across various fields. PC giants such as Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, and MSI are rolling out systems that leverage its capabilities.

Additionally, organizations like Ollama, NYU Global Frontier Lab, Zipline, Arizona State University, and Refik Anadol’s studio are already utilizing DGX Spark.

DGX Spark will be generally available starting Wednesday, Oct. 15, on NVIDIA.com and through partners worldwide. This availability marks a significant step in democratizing access to high-performance AI computing.