NVIDIA has unveiled two new personal AI supercomputers—the DGX Spark and DGX Station—designed to empower developers, researchers, and data scientists with advanced AI capabilities directly at their desktops.
 |
| NVIDIA has unveiled two new personal AI supercomputers—the DGX Spark and DGX Station—designed to empower developers, researchers, and data scientists |
DGX Spark
- Overview: Formerly known as Project DIGITS, the DGX Spark is a compact AI supercomputer tailored for individuals and small teams aiming to develop and test large AI models without relying on traditional data centers.
- Specifications: It is powered by the NVIDIA GB10 Blackwell Superchip, featuring advanced functionalities such as FP4 support and delivering up to 1,000 trillion operations per second (TOPS) of AI compute.
- Availability: Preorders are currently open on NVIDIA's official website, with deliveries expected to commence in the summer of 2025.
DGX Station
- Overview: The DGX Station is a high-performance desktop supercomputer equipped to handle more demanding AI workloads, making it suitable for complex model development and extensive data processing tasks.
- Specifications: This system incorporates the NVIDIA GB300 Blackwell Ultra chip, offering 20 petaflops of AI performance and 784GB of unified memory, facilitating the management of large-scale AI models.
- Availability: The DGX Station is scheduled for release later in 2025, with specific dates to be announced.
Collaborations and Market Impact
NVIDIA has partnered with leading computer manufacturers—including ASUS, Dell Technologies, HP Inc., and Lenovo—to produce these systems, ensuring widespread availability and support. This initiative underscores NVIDIA's commitment to making advanced AI tools more accessible, potentially accelerating innovation across various sectors.
By bringing the power of the Grace Blackwell architecture to individual workspaces, NVIDIA aims to facilitate seamless AI development and deployment, bridging the gap between conceptualization and real-world application.