xFusion FusionXpark GB10 Personal AI Supercomputer - 128GB / 2TB with Technical Support
PRODUCT DESCRIPTION
xFusion FusionXpark GB10 Personal AI Supercomputer 2 TB (FusionXpark GB10_2T) — 128GB Unified Memory — xFusion Technical Support Service Included
FusionXpark GB10 — Master AI in Your Hands
The xFusion FusionXpark GB10 is a portable AI agent development platform built on the NVIDIA GB10 Grace Blackwell Superchip. It delivers up to 1 PFLOP of FP4 AI performance and 128 GB of coherent unified memory in a 1.2 kg desktop box, with NVIDIA DGX OS and the NVIDIA AI software stack pre-installed — the same NVIDIA GB10 platform as the DGX Spark, in a design-award-winning xFusion chassis available in gold or silver finish. This configuration ships with a 2 TB self-encrypting NVMe SSD.
Prototype, Fine-Tune, Infer and Deploy Agents — Locally
Run inference on models up to 200 billion parameters and fine-tune models up to 70 billion parameters without a cloud dependency. FusionXpark covers the full workflow from model prototyping to AI agent deployment, and natively supports the NVIDIA ecosystem (CUDA, TensorRT, NIM, NGC containers) with a clear path from desktop to production. It keeps sensitive data on-premise for predictive maintenance, medical imaging, financial forecasting and other regulated workloads.
Built for Developers, Researchers, Educators and the Edge
An easy-to-set-up personal supercomputing platform for AI teaching and research, and an end-to-end desktop environment for edge AI application development. A built-in NVIDIA ConnectX-7 SmartNIC lets you cluster units for larger models.
Technical Support Service Bundled — Deployment-Ready in Singapore
Every FusionXpark GB10 from SourceIT includes the xFusion Technical Support Service for the unit, a GST invoice and local support in Singapore. 2 TB gives room for multiple model checkpoints and datasets side by side.
xFusion FusionXpark GB10 2TB Datasheet — Technical Specifications
General
| Model | xFusion FusionXpark GB10 — 2 TB |
| Configuration Code | FusionXpark GB10_2T |
| Category | Desktop Personal AI Supercomputer / Portable AI Agent Development Platform |
| Finish | Gold or Silver; custom signature engraving supported |
| Support | xFusion Technical Support Service included; local support via SourceIT (Singapore) |
Processor — NVIDIA GB10 Grace Blackwell Superchip
| Architecture | NVIDIA Grace Blackwell (CPU + GPU Unified, NVLink-C2C) |
| CPU | 20-Core Arm — 10 × Cortex-X925 + 10 × Cortex-A725 |
| GPU | NVIDIA Blackwell Architecture, 6,144 CUDA Cores |
| Tensor Cores | 5th Generation |
| RT Cores | 4th Generation |
| Supported Data Formats | TF32, FP16, BF16, INT8, FP8, FP6, FP4 |
AI Performance
| Tensor Performance | Up to 1 PFLOP FP4 |
| Inference | Models Up to 200 Billion Parameters (Single Unit) |
| Fine-Tuning | Models Up to 70 Billion Parameters |
Memory & Storage
| System Memory | 128 GB LPDDR5x Coherent Unified Memory |
| Memory Interface | 256-bit, 273 GB/s Bandwidth |
| Storage | 2 TB Self-Encrypting NVMe M.2 SSD |
Networking & I/O
| SmartNIC | NVIDIA ConnectX-7 |
| Ethernet | 1 × 10 GbE RJ-45 |
| Wireless | Wi-Fi 7, Bluetooth 5.4 |
| USB | 4 × USB Type-C |
| Display | 1 × HDMI 2.1a |
| Audio | HDMI Multi-Channel Audio Output |
| Video Engines | 1 × NVENC, 1 × NVDEC |
Software
| Operating System | NVIDIA DGX OS (Pre-Installed) |
| AI Stack | NVIDIA AI Software Stack; NVIDIA AI Enterprise supported (additional licence required) |
Physical
| Dimensions | 150 mm × 150 mm × 50.5 mm |
| Weight | 1.2 kg |
| Power | External Power Adapter |
Workloads & Use Cases
Developers: model prototyping, fine-tuning, inference and AI agent deployment in one local workflow, with broad NVIDIA ecosystem support.
Local enterprise deployment: keeps data private while accelerating predictive maintenance, medical imaging, financial forecasting and similar workloads.
Research & education: an easy-to-set-up AI teaching platform that gives each student or lab a personal-supercomputing experience.
Edge: end-to-end edge AI application development on the desktop, from robotics and vision to smart-city pilots.
Ordering Information — GB10 AI Supercomputers at SourceIT
| xFusion FusionXpark GB10 1TB | FusionXpark GB10_1T + Technical Support Service |
| xFusion FusionXpark GB10 2TB | FusionXpark GB10_2T + Technical Support Service (This Product) |
| xFusion FusionXpark GB10 4TB | FusionXpark GB10_4T + Technical Support Service |
| NVIDIA DGX Spark 4TB | 940-54242-0007-000 — In Stock |
| ASUS Ascent GX10 1TB / 2TB / 4TB | GX10-GG0007BN / GX10-GG0030BN / GX10-GG0031BN |
| NVIDIA QSFP112 400G DAC Cable (Links 2 Units) | X0101G00400A — In Stock |
Datasheet, Downloads & Resources
Frequently Asked Questions — xFusion FusionXpark GB10 2TB
Is the FusionXpark GB10 the same as an NVIDIA DGX Spark?
It is built on the same NVIDIA GB10 Grace Blackwell Superchip with 128 GB unified memory, ConnectX-7 networking and NVIDIA DGX OS pre-installed, so it runs the same NVIDIA AI software stack. xFusion adds its own chassis design (gold or silver, custom signature), 1 TB / 2 TB / 4 TB storage options and xFusion technical support.
What is included in the price?
The FusionXpark GB10 2 TB unit together with the xFusion Technical Support Service for that unit, supplied with a GST invoice by SourceIT.
What AI models can it run?
A single unit handles inference on models up to 200 billion parameters and fine-tuning up to 70 billion parameters — Llama, Qwen, DeepSeek, Mistral and similar open models locally.
Can I link two units?
Yes — the built-in NVIDIA ConnectX-7 SmartNIC supports multi-node clustering; pair with the NVIDIA QSFP112 400G DAC cable.
Does it need NVIDIA AI Enterprise?
No. The NVIDIA AI software stack on DGX OS is ready to use; NVIDIA AI Enterprise is supported as an optional add-on with a separate licence.
Which storage size should I choose?
1 TB suits prototyping and single-model work; 2 TB gives room for multiple model checkpoints; 4 TB is recommended for fine-tuning datasets, large model libraries and video workloads.
