Nvidia has officially renamed the compact AI supercomputer formerly known as Digits the DGX Spark and dissected it to launch this summer alongside its larger counterpart, the DGX Station.
Announced at the GTC 2025, these "personal AI supercomputers" use Nvidia's cutting-edge Grace Blackwell platform to bring powerful AI processing to desktop users without relying on constant data center connectivity.
The DHX Spark, priced at $3,000, is a Mac Mini-sized device that packs Nvidia's GB10 Blackwell Superchip. With fifth-generation Tensor Cores and FP4 support, it can perform up to 1000 trillion operations per second (TOPS) of AI computing.
With 128GB of unified memory and up to 4TB of NVMe SSD storage, the Spark is designed to execute AI workloads locally, making it suitable for developers working with huge AI models. It supports inference and fine-tuning of models like Nvidia's Cosmos Reason world foundation model and GR00T N1 robot foundation model.
Meanwhile, the newly announced DGX Station is a more powerful device designed for AI developers, academics, and data scientists. This desktop supercomputer has Nvidia's GB300 Blackwell Ultra Superchip, which delivers a remarkable 20 petaflops of AI performance and 784GB of unified memory.
While its pricing is unknown, its features make it an excellent choice for users who need to prototype, fine-tune, and run large-scale AI models on personal systems.
To increase accessibility, Nvidia has collaborated with leading OEMs like Asus, Dell, HP, and Lenovo to develop versions of the DGX Spark. Pre-orders are available now, and delivery will be this summer. Later this year, Asus, Dell, HP, Boxx, Lambda, and Supermicro will release their own unique versions of the DGX Station.
Nvidia's latest products highlight the increasing competitiveness in the local AI hardware industry, where competitors like AMD are also pushing the limits of AI computing.
With personal AI supercomputers growing more powerful and accessible, the DGX Spark and DGX Station represent a big step toward enterprise-grade AI performance on desktops.