NVIDIA has unveiled DGX Spark, the world’s smallest AI supercomputer, priced at $3,999, marking a major step in bringing petascale computing power to individual developers and enterprises alike. The company said the compact yet powerful system can deliver a petaflop of performance with 128 GB of unified memory, enabling it to handle advanced AI workloads from large model inference to agentic AI development.
In a symbolic moment, CEO Jensen Huang personally delivered the first DGX Spark units to Elon Musk, continuing a partnership that began in 2016 when Musk received NVIDIA’s original DGX-1 system. The gesture underscores the ongoing collaboration between two of the most influential figures in AI and computing innovation.
At the heart of DGX Spark is the GB10 Blackwell Superchip, NVIDIA’s latest breakthrough in AI hardware. The system can perform inferences on AI models with up to 200 billion parameters and fine-tune models containing up to 70 billion parameters, making it ideal for both enterprise-scale research and local AI experimentation.
NVIDIA said that DGX Spark is built to empower developers, researchers, and enterprises to design and deploy AI agents and software stacks directly on-site, without relying solely on cloud infrastructure. This local compute capability opens new possibilities for privacy-sensitive and high-performance applications, especially in fields like healthcare, finance, robotics, and edge AI.
The company confirmed that early adopters of DGX Spark include Microsoft, Google, Hugging Face, Meta, and SpaceX, among other global AI leaders. These organizations are expected to leverage the system for developing next-generation AI models and applications that demand real-time, high-efficiency computation.
KyungHyun Cho, professor at NYU Global Frontier Lab, noted the transformative potential of the device, stating, “DGX Spark allows access to peta-scale computing on a desktop and enables rapid prototyping and experimentation for advanced AI algorithms, even in privacy- and security-sensitive applications like healthcare.”
With DGX Spark, NVIDIA continues to push the boundaries of accessible AI infrastructure — compressing supercomputing power into a desktop form factor and redefining how innovation, experimentation, and AI development can scale across industries.
