NVIDIA DGX Spark Founders Edition ANZ 4TB - Personal AI Supercomputer
The NVIDIA DGX Spark Founders Edition is a revolutionary compact AI supercomputer powered by the NVIDIA GB10 Grace Blackwell Superchip. Delivering up to 1 petaFLOP of AI performance in a desktop form factor, this powerhouse enables developers, researchers, and data scientists to build, fine-tune, and run large AI models locally with up to 200 billion parameters. With 128GB of unified system memory, 4TB NVMe storage, and the NVIDIA AI software stack preinstalled, DGX Spark brings enterprise-grade AI computing to your desk in a compact 150mm × 150mm × 50.5mm chassis weighing just 1.2kg.
Why Choose the NVIDIA DGX Spark?
DGX Spark belongs to a new class of personal AI computers designed from the ground up for autonomous agents, large language models, and advanced machine learning workflows. Perfect for AI developers, machine learning engineers, data scientists, and researchers who need powerful local compute for prototyping, fine-tuning, and inference without relying on cloud resources.
Key Features
-
GB10 Grace Blackwell Superchip: Integrated 20-core Arm CPU with Blackwell GPU featuring 5th-generation Tensor Cores and 4th-generation RT Cores
-
1 PetaFLOP AI Performance: Up to 1,000 TOPS of AI computing power at FP4 precision for rapid inference and training
-
128GB Unified Memory: LPDDR5x coherent system memory with 273GB/s bandwidth for handling massive AI models
-
4TB NVMe Storage: Self-encrypting M.2 storage for secure local AI workflows and large datasets
-
Advanced Connectivity: 10GbE RJ45, ConnectX-7 NIC (200Gbps), Wi-Fi 7, Bluetooth 5.4 LE, 4x USB-C, and HDMI 2.1a
-
Dual-System Clustering: Connect two DGX Spark units to work with models up to 405B parameters
-
Preloaded AI Stack: NVIDIA AI software stack ready out of the box for immediate deployment
-
Compact & Efficient: Ultra-compact desktop form factor with 240W max power consumption
Unmatched AI Performance for Local Workflows
The DGX Spark's GB10 Grace Blackwell Superchip integrates a high-performance 20-core Arm processor (10 Cortex-X925 + 10 Cortex-A725 cores) with an NVIDIA Blackwell architecture GPU. This system-on-chip delivers breakthrough AI performance with fifth-generation Tensor Cores supporting FP4 precision, enabling up to 1 petaFLOP of parallel throughput. Combined with 128GB of unified system memory, DGX Spark can prototype, fine-tune models with up to 70 billion parameters, and run inference on AI models up to 200 billion parameters entirely locally.
Purpose-Built for AI Development
Whether you're developing autonomous AI agents, fine-tuning large language models from DeepSeek, Meta, Google, Qwen, or NVIDIA, or accelerating data science workflows with NVIDIA RAPIDS, DGX Spark provides the performance and memory you need. The preinstalled NVIDIA AI software stack includes frameworks for edge AI application development with NVIDIA Isaac, Metropolis, and more. Build and run secure, local AI agents with NVIDIA OpenShell without relying on cloud-based token generation.
Enterprise Connectivity & Expandability
DGX Spark features robust connectivity options designed for professional AI workflows:
-
Networking: 1x RJ-45 10GbE Ethernet, ConnectX-7 Smart NIC with 200Gbps throughput, Wi-Fi 7, Bluetooth 5.4 LE
-
I/O Ports: 4x USB Type-C (including power delivery), 1x HDMI 2.1a with multichannel audio
-
High-Speed Interconnect: 2x QSFP network connectors (ConnectX-7) for clustering two DGX Spark systems
-
Video Processing: 1x NVENC encoder and 1x NVDEC decoder for media workflows
Compact Desktop Form Factor
Despite its massive computing power, DGX Spark measures just 150mm × 150mm × 50.5mm and weighs only 1.2kg (2.6 lbs). The small form factor chassis fits easily on any desk, making enterprise-grade AI accessible without dedicated server rooms or complex cooling infrastructure. Operating temperature range of 5°C to 30°C ensures reliable performance in typical office environments.
What's in the Box
- NVIDIA DGX Spark Founders Edition (ANZ) with 4TB storage
- Power adapter and cable
- Quick Start Guide
- Safety and regulatory documentation
Full Specifications
Compute
| GPU |
NVIDIA Blackwell Architecture with 5th Generation Tensor Cores, 4th Generation RT Cores |
| CPU |
20-core Arm processor (10 Cortex-X925 + 10 Cortex-A725) |
| AI Performance |
Up to 1 petaFLOP (1,000 TOPS) at FP4 precision |
| System Memory |
128GB LPDDR5x unified system memory, 256-bit interface, 4266 MHz |
| Memory Bandwidth |
273 GB/s |
| Storage |
4TB NVMe M.2 with self-encryption |
Connectivity
| Ethernet |
1x RJ-45 10 GbE |
| High-Speed Networking |
ConnectX-7 Smart NIC (200 Gbps), 2x QSFP connectors |
| Wireless |
Wi-Fi 7, Bluetooth 5.4 LE |
| USB |
4x USB Type-C (one for power delivery) |
| Display |
1x HDMI 2.1a with multichannel audio |
| Video Processing |
1x NVENC, 1x NVDEC |
Power & Environment
| TDP (GB10 SoC) |
140W |
| Max System Power |
240W |
| Operating Temperature |
5°C to 30°C (41°F to 86°F) |
| Operating Humidity |
10% to 90% (non-condensing) |
| Operating Altitude |
Up to 3,000 meters (9,843 feet) |
Physical
| Chassis Type |
Small form factor (SFF) |
| Dimensions (L × W × H) |
150mm × 150mm × 50.5mm |
| Weight |
1.2 kg (2.6 lbs) |
Software & AI Capabilities
| Preinstalled Software |
NVIDIA AI software stack |
| Supported Model Size (Single Unit) |
Up to 200 billion parameters |
| Fine-Tuning Capacity |
Up to 70 billion parameters |
| Dual-System Clustering |
Up to 405 billion parameters (256GB combined memory, 8TB total storage) |
| AI Frameworks |
NVIDIA Isaac, Metropolis, RAPIDS, OpenShell, and more |
Warranty & Support
| Warranty (New Units) |
1 year limited warranty |
| Part Number |
940-54242-0002-000 (Founders Edition, ANZ, 4TB) |
Specifications are sourced from the official NVIDIA product page and technical documentation and may be updated. This product is designed for AI development, research, and data science workloads. Not designed for medical, military, aircraft, space, or life support applications.
