NVIDIA has announced two new desktop AI supercomputers – the DGX Spark and DGX Station – at GTC 2025. These personal AI supercomputers are aimed at making advanced AI development tools more accessible, empowering individuals and smaller organizations to conduct significant AI research and development without the need for large data centers.
Formerly known as Project DIGITS, the DGX Spark is presented as the world’s smallest AI supercomputer (approximately the size of a small book). It is powered by the NVIDIA GB10 Grace Blackwell Superchip, which features a Blackwell GPU with fifth-generation Tensor Cores and FP4 support. This configuration enables users to train and run large AI models with up to 200 billion parameters directly from their home or office.
Also, the configuration delivers up to 1,000 trillion operations per second (TOPS) of AI computing, making it suitable for fine-tuning and inference tasks with the latest AI reasoning models. The GB10 Superchip utilizes NVIDIA’s NVLink-C2C interconnect technology, providing a CPU-GPU coherent memory model with five times the bandwidth of fifth-generation PCIe, optimizing performance for memory-intensive AI workloads.
Now coming to the DGX Station, this computer brings data center-level AI performance to the desktop. It is equipped with the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, offering a substantial 784GB of coherent memory. This system is designed to accelerate large-scale training and inference workloads.
The GB300 Superchip includes a Blackwell Ultra GPU connected to a high-performance Grace CPU via NVLink-C2C, ensuring efficient system communication. Additionally, the DGX Station features the NVIDIA ConnectX-8 SuperNIC, supporting networking speeds of up to 800Gb/s, which facilitates high-speed connectivity between multiple DGX Stations.
Both systems are integrated with NVIDIA’s full-stack AI platform, enabling users to seamlessly transition their models from desktop environments to the DGX Cloud or other accelerated infrastructures with minimal code changes.
“AI has transformed every layer of the computing stack. It stands to reason a new class of computers would emerge — designed for AI-native developers and to run AI-native applications. With these new DGX personal AI computers, AI can span from cloud services to desktop and edge applications,” Jensen Huang (founder and CEO of NVIDIA) said in his statement.
Speaking of availability, reservations for the DGX Spark are currently open on NVIDIA’s website, while the DGX Station is expected to be available from manufacturing partners later this year. Meanwhile, global system builders including ASUS, Dell, HP, and Lenovo are collaborating with NVIDIA to develop these systems.
Earlier, some reports suggested that the company will launch a compact ‘Jetson Thor’ computer for AI-powered humanoid robots in 2025. The development comes at a time when the AI supercomputer market is expected to expand significantly over the next decade, reaching an estimated value of $38.79 billion by 2034.
