NVIDIA unveils AI supercomputers for desks with DGX Spark, Station

NVIDIA has revealed that leading system manufacturers from Taiwan will produce the DGX Spark and DGX Station personal computing systems focused on artificial intelligence workloads.

The company confirmed ongoing collaborations with Acer, GIGABYTE and MSI, expanding the reach of DGX Spark and DGX Station devices. The systems are intended for a global community of developers, data scientists and researchers requiring high levels of performance and efficiency from personal AI supercomputers.

The company stated that enterprises, software businesses, government agencies, startups and research bodies now demand workstations that can provide server-level AI capability in desktop form. According to NVIDIA, the rise of advanced agentic AI systems, which support autonomous task execution and decision-making, is intensifying this need.

The DGX Spark and DGX Station platforms are powered by the NVIDIA Grace Blackwell platform and are designed to enable model prototyping, fine-tuning and inference at desktop and data centre scales.

Jensen Huang, Founder and Chief Executive Officer of NVIDIA, said: "AI has revolutionised every layer of the computing stack — from silicon to software. Direct descendants of the DGX-1 system that ignited the AI revolution, DGX Spark and DGX Station are created from the ground up to power the next generation of AI research and development."

The DGX Spark system incorporates the NVIDIA GB10 Grace Blackwell Superchip and fifth-generation Tensor Cores, delivering up to 1 petaflop of AI compute and 128GB of unified memory. The system is capable of seamless model export to NVIDIA DGX Cloud or any compatible accelerated cloud or data centre infrastructure.

The company indicated that DGX Spark's compact format allows developers, researchers, data scientists and students to advance generative AI projects and accelerate workloads across multiple industries.

The DGX Station is designed for intensive AI workloads and features the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. This enables up to 20 petaflops of AI performance and provides 784GB of unified system memory. The system also includes an NVIDIA ConnectX-8 SuperNIC, supporting up to 800Gb/s networking speeds for connectivity and scale across multiple stations.

DGX Station can be used as a dedicated desktop for a single user working with local data on advanced AI models, or as a centralised compute resource for multiple users. The model supports partitioning using NVIDIA Multi-Instance GPU (MIG) technology, allowing up to seven distinct instances, each with its own resources. This feature positions the system as a personal cloud for data science and AI development teams.

Both DGX Spark and DGX Station mirror the software environments found in industrial-scale AI facilities, operating with the NVIDIA DGX operating system and preconfigured with the most recent NVIDIA AI software stack. They also come with access to NVIDIA NIM microservices and NVIDIA Blueprints.

With support for common tools, including PyTorch, Jupyter and Ollama, users of DGX Spark and DGX Station can prototype, fine-tune and run inference on local hardware, with the capability to deploy workloads to DGX Cloud or compatible cloud and data centre infrastructure.

Dell Technologies is one of the first companies to develop DGX Spark and DGX Station solutions to meet the increasing enterprise demand for robust local AI computing technology.

Michael Dell, Chairman and Chief Executive Officer of Dell Technologies, said: "There's a clear shift among consumers and enterprises to prioritise systems that can handle the next generation of intelligent workloads. The interest in NVIDIA DGX Spark and NVIDIA DGX Station signals a new era of desktop computing, unlocking the full potential of local AI performance. Our portfolio is designed to meet these needs. Dell Pro Max with GB10 and Dell Pro Max with NVIDIA GB300 give organisations the infrastructure to integrate and tackle large AI workloads."

HP is also supporting the developments for AI-centric computing, and will provide offerings based on the new hardware.

Enrique Lores, President and Chief Executive Officer of HP, said: "Through our collaboration with NVIDIA, we are delivering a new set of AI-powered devices and experiences to further advance HP's future-of-work ambitions to enable business growth and professional fulfillment. With the HP ZGX, we are redefining desktop computing — bringing data-center-class AI performance to developers and researchers to iterate and simulate faster, unlocking new opportunities."

DGX Spark systems will be available for purchase from Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo and MSI as well as from global channel partners starting in July. Availability for DGX Station is expected from ASUS, Dell Technologies, GIGABYTE, HP and MSI later in the year.