Nvidia Launches DGX Spark AI Supercomputer for $3,999 Starting October 15

Nvidia has officially launched the DGX Spark, a compact, desktop-sized AI supercomputer designed to bring high-end AI computing power directly to developers, researchers, and innovators. Available starting October 15, the device is priced at $3,999—up from its initial $3,000 announcement—and is being sold through Nvidia’s website and select partners, including Micro Center in the U.S. The DGX Spark is not a typical PC; it runs a customized Ubuntu-based Linux system preloaded with Nvidia’s AI tools for training and running large language models (LLMs) and image generation models. Despite its small size, it packs a powerful punch, featuring a Grace Blackwell GPU and a 20-core ARM-based CPU, delivering up to 1 petaflop of AI performance and 1,000 TOPS (trillions of operations per second). This level of compute is far beyond what standard consumer PCs can handle, enabling local execution of models with up to 200 billion parameters—something previously only possible in massive cloud data centers. The device comes with 128GB of unified memory and up to 4TB of NVMe SSD storage, making it ideal for training and experimenting with complex AI models without relying on cloud infrastructure. While its 240W power draw is modest compared to older data center systems like the DGX-1, which consumed 3,200W, the Spark’s efficiency and performance make it a significant leap for edge AI computing. Nvidia has positioned it as a “new class of computer,” not just a PC, emphasizing its transformative potential for AI development. Nvidia’s CEO Jensen Huang personally delivered the first units to key figures in the AI world, including Elon Musk at SpaceX’s Starbase facility in Texas—coinciding with the 11th test flight of Starship. The gesture underscored the device’s role in advancing AI at the frontier of innovation. Other early adopters include OpenAI, Microsoft, and creative studios like Refik Anadol’s, highlighting its use in both research and artistic applications. The DGX Spark is not intended for general consumers or everyday tasks. It’s built for developers and researchers who need serious AI compute power on their desks. While it outperforms any standard desktop in AI workloads, it still falls short of the raw TOPS of high-end consumer GPUs like the RTX 5090. However, its combination of portability, performance, and specialized software makes it a unique tool for AI experimentation. Multiple OEMs, including Acer, Asus, Dell, and MSI, are releasing their own versions of the Spark, with Acer’s Veriton GN100 priced at $3,999. These models will expand access while maintaining the core capabilities of the original. Despite its high cost, Nvidia sees the Spark as a strategic move to democratize access to AI supercomputing. By placing petaflop-level AI power on desks, Nvidia aims to accelerate innovation, foster a new generation of AI developers, and drive the next wave of breakthroughs—beyond simple chatbots or surface-level AI tools. The company also hinted at a larger, more powerful version called DGX Station, though its release date remains unannounced. As AI continues to evolve, the DGX Spark represents a pivotal shift: moving the most advanced AI work from remote data centers to the very hands of creators, where ideas begin.