NVIDIA DGX Spark – Personal AI Supercomputer with Blackwell GPU and Arm CPU for Developers and Researchers

The NVIDIA DGX Spark is a compact, power-efficient AI supercomputer powered by the NVIDIA GB10 Grace Blackwell Superchip, delivering 1000 AI TOPS of performance at FP4 precision. Designed for AI developers, researchers, and data scientists, it enables prototyping, fine-tuning, and inferencing AI models with up to 200 billion parameters locally. It supports AI workloads, data science, edge applications, 128GB of unified system memory, high-speed ConnectX networking, and NVIDIA’s AI software stack.

With Blackwell architecture GPU, 20-core Arm CPU, Tensor and RT cores, 1TB or 4TB NVMe storage, WiFi 7, 4x USB4 ports and 10 GbE connectivity, it offers high efficiency in a desktop-friendly 1.2 kg form factor. It is ideal for developing AI models, validating applications, and deploying edge solutions, seamlessly integrating with data centers and cloud environments.

NVIDIA previously launched Project DIGITS, a personal AI supercomputer powered by the new GB10 Grace Blackwell Superchip, designed for AI researchers, data scientists, and students. We have also covered the UDOO BOLT, a supercomputer with twice the power of a MacBook Pro 13.

NVIDIA DGX Spark AI Supercomputer Specifications:

• SoC: NVIDIA GB10 Grace Blackwell Superchip
  • CPU: 20-core Arm (10x Cortex-X925 + 10x Cortex-A725)
  • GPU: Blackwell architecture with 5th-gen Tensor Cores and 4th-gen RT Cores
  • VPU:
    • 1x NVENC
    • 1x NVDEC
• Memory: 128GB LPDDR5x unified system memory (256-bit, 273GB/s bandwidth)
• Storage: 1TB or 4TB NVMe M.2 SSD (self-encrypting)
• AI Performance: Up to 1000 AI TOPS at FP4 precision
• Display: 1x HDMI 2.1a (multichannel audio output)
• Audio: HDMI multichannel output
• Networking:
  • 1x RJ-45 10GbE Ethernet
  • ConnectX-7 Smart NIC for high-speed networking
  • WiFi 7
  • Bluetooth 5.3
• USB: 4x USB 4 Type-C (up to 40Gbps)
• Power Consumption: 170W
• Physical Dimensions: 150 x 150 x 50.5 mm
• Weight: 1.2 kg

According to the company, the NVIDIA DGX Spark operates on NVIDIA DGX OS, which is built on Ubuntu 22.04 and supports the NVIDIA Cosmos Reason world and GR00T N1 robot foundation models. Its AI platform provides tools, frameworks, libraries, and pre-trained models, enabling developers to prototype, fine-tune, and run reasoning AI models from DeepSeek, Meta, Google, and more—handling up to 200 billion parameters locally.

Developers in the US, UK, France, Italy, Germany, and Spain can pre-order the NVIDIA DGX Spark through partners like ASUS, Dell, and HP. Pricing is expected to range between $3,000 and $4,000, with availability starting in May because, at the time of writing, the exact price is not given. For more details, visit the product page or the official announcement, which also introduces the larger DGX Station, designed like a desktop PC.

Please follow and like us: