NVIDIA’s latest leap in artificial intelligence began in an unlikely place – beside the world’s largest rocket. Founder and CEO Jensen Huang personally delivered the first DGX Spark, billed as the world’s smallest AI supercomputer, to Elon Musk at SpaceX’s Starbase facility in Texas.
Huang’s arrival marked the start of DGX Spark’s global rollout. Walking through rows of SpaceX engineers, he joined Musk for an informal meeting at the company cafeteria as preparations were underway for the eleventh test of Starship, the world’s most powerful launch vehicle.
Reflecting on NVIDIA’s history, Huang recalled his delivery of the first DGX system to OpenAI in 2016. Speaking of the new launch, he remarked: “Imagine delivering the smallest supercomputer next to the biggest rocket.”
The new DGX Spark system represents NVIDIA’s effort to take AI performance out of the data centre and place it directly into the hands of developers, researchers, and creators. Roughly the size of a hardcover book and weighing just 1.2kg, the device delivers a petaflop of AI performance and features 128GB of unified CPU-GPU memory. It is powered by NVIDIA’s GB10 Grace Blackwell Superchip and supported by NVIDIA’s full AI software stack, including frameworks, libraries, pretrained models, and NIM microservices.
This configuration allows users to fine-tune and run AI models locally without relying on Cloud infrastructure. NVIDIA said the system could handle models with up to 200 billion parameters, making it suitable for workloads ranging from generative image tools and summarisation agents to optimised chatbots.
The compact system also incorporates NVIDIA NVLink-C2C connectivity – providing five times the bandwidth of PCIe – along with ConnectX networking, NVMe storage, and HDMI output.
The DGX Spark’s debut at SpaceX symbolised more than a product launch. It marked a new phase for NVIDIA’s AI ecosystem, which began with the introduction of the DGX-1 nine years ago. Now, the company is extending that legacy to the desktop and Edge environments.
OEMs including Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, and MSI are collaborating with NVIDIA to bring DGX Spark systems to market, combining compact design with access to NVIDIA’s full AI platform.
Initial units are already being used by early adopters across diverse sectors:
- Ollama in Palo Alto, enabling developers to run large language models locally
- NYU Global Frontier Lab, advancing privacy-sensitive algorithm research
- Zipline, supporting autonomous delivery systems
- Arizona State University, testing robotics and computer vision models
- Refik Anadol Studio, merging AI with creative design
Each delivery forms part of NVIDIA’s broader effort to “put petaflop AI within arm’s reach”, enabling innovation in spaces ranging from robotics labs to creative studios.
DGX Spark will be generally available from 15th October on NVIDIA’s website and through global partners.
