Limited Time Offer: Save $350 on your Ai Cluster → Shop Now

--

Days

--

Hours

--

Mins

--

Secs

Sale Ended

TECHNOLOGY AT A GLANCE

Explore the capabilities

NVIDIA GB10 performance

Up to 1 petaFLOP of theoretical FP4 AI performance using sparsity, combining a Blackwell GPU with a 20-core Arm CPU.

128GB unified memory

Supports inference with models up to 200 billion parameters, depending on model precision and configuration

4TB NVMe SSD

Fast local storage for AI models, datasets, and project files, with self-encryption support.

High-speed connectivity

Includes 10Gb Ethernet and NVIDIA ConnectX-7 networking at 200Gbps for connecting systems and expanding your AI setup.

Your workspace. Supercharged for AI.

Bring dedicated AI computing into your everyday workspace with NVIDIA DGX Spark. Powered by the GB10 Grace Blackwell Superchip and 128GB of unified memory, this compact system lets you prototype, fine-tune, and run AI models locally. Connect from your existing computer and turn ideas into working AI applications with the preinstalled NVIDIA AI software stack.

Connectivity for Scalable AI Development

NVIDIA DGX Spark combines everyday workstation connectivity with high-speed networking for local AI development, shared storage, and scalable compute workflows.

  • High-speed networking: 10Gb Ethernet and NVIDIA ConnectX-7 networking support fast connections to compatible storage, network infrastructure, and AI systems.
  • USB-C connectivity: Four USB-C ports connect peripherals, external storage, and accessories for a complete local AI workspace.
  • Display connectivity: HDMI 2.1a connects a monitor for development, visualization, and day-to-day workstation use.
  • Dedicated power input: The 240W external power supply supports dependable performance in a compact desktop footprint.

SEE IT IN ACTION

Meet the product

TECHNICAL DETAILS

Built into the system

NVIDIA GB10 Grace Blackwell Architecture

NVIDIA DGX Spark is powered by the NVIDIA GB10 Grace Blackwell Superchip, which combines a 20-core Arm CPU with NVIDIA Blackwell GPU architecture. The CPU includes 10 Cortex-X925 cores and 10 Cortex-A725 cores, while the GPU includes fifth-generation Tensor Cores and fourth-generation RT Cores for modern AI workloads.

Up to 1,000 AI TOPS of Compute

The GB10 platform delivers up to 1,000 AI TOPS at FP4 precision with sparsity, based on theoretical performance. This level of local compute is designed for prototyping, fine-tuning, and running large AI models without relying exclusively on remote infrastructure.

128GB Coherent Unified Memory

128GB of LPDDR5X coherent unified memory gives the CPU and GPU access to a shared memory pool for AI workloads. A 256-bit memory interface and up to 273GB/s of memory bandwidth support moving data efficiently between system components during model development and inference.

4TB Encrypted NVMe Storage

The 4TB M.2 PCIe NVMe SSD provides local capacity for AI models, datasets, tools, and project files. Self-encryption support adds another layer of protection for data stored on the system.

High-Speed Networking and Wireless Connectivity

A 10GbE RJ-45 port supports high-speed wired network access, while NVIDIA ConnectX-7 provides 200Gbps networking for compatible high-bandwidth infrastructure. Wi-Fi 7 and Bluetooth 5.4 provide additional wireless connectivity options for a flexible desktop setup.

Compact AI Desktop with Modern I/O

DGX Spark runs NVIDIA DGX OS, a Linux-based operating system built for AI development. Its compact 150 × 150 × 50.5mm chassis weighs approximately 2.65 lb and includes four USB-C ports, including power delivery, HDMI 2.1a with multichannel audio, and dedicated NVENC and NVDEC hardware for video encoding and decoding. A 240W external power supply is included.