Desktop AI Supercomputer
That Actually Fits
Get 1 petaFLOP of AI performance on your desk. No data center required. Just plug it in and start training models up to 200B parameters.
Powered by Industry-Leading Technology
AI Supercomputing, Reimagined
Everything you need to develop, fine-tune, and deploy AI models—right on your desk.
Extreme AI Performance
1 petaFLOP of AI compute with NVIDIA GB10 Grace Blackwell. Train and fine-tune models up to 200B parameters locally.
Unified Memory Architecture
128GB coherent unified system memory. Work with larger datasets and complex models without bottlenecks.
Developer-Optimized
Full NVIDIA AI software stack included. PyTorch, TensorFlow, CUDA—everything pre-configured and ready to go.
Scalable Architecture
Stack two GX10 systems with ConnectX-7. Handle models up to 405B parameters like Llama 3.1.
Advanced Thermal Design
Dual-fan with 7-level control and vapor chamber cooling. 1.6× more efficient than comparable systems.
Compact Design
Ultra-small form factor fits on any desk. Perfect for labs, offices, and edge deployments.
1 PetaFLOP of AI Power
The NVIDIA GB10 Grace Blackwell Superchip delivers desktop AI performance that was impossible just a year ago. Train, fine-tune, and run inference on models you couldn't dream of running locally.
Train the Models That Matter
From DeepSeek to Llama, run the latest AI models without compromise.
Support for 200B Parameter Models
Fine-tune and run inference on the latest reasoning models right on your desk. Need more? Stack two GX10 systems to handle models up to 405B parameters like Llama 3.1.
AI inference optimized up to 70B parameters
Up to 405B parameters with dual-GX10 setup
Broad compatibility with industry frameworks
Enterprise-Grade Networking Built In
NVIDIA ConnectX-7 delivers the speed and security your AI workloads demand.
Ultra-Fast Data Throughput
Lightning-fast bandwidth for distributed AI workloads and rapid model synchronization.
Hardware-Accelerated Security
Built-in TLS, IPsec, and MACsec encryption without CPU overhead.
Precision Time Sync
IEEE 1588v2 PTP for microsecond-level synchronization in edge computing.
Everything You Need, Ready to Go
Complete AI development environment pre-configured and optimized.
Preloaded & Ready
- • NVIDIA DGX OS (Ubuntu)
- • Complete AI software stack
- • Pre-configured frameworks
- • Instant deployment
Optimized Tools
- • PyTorch & TensorFlow
- • CUDA & TensorRT
- • Jupyter notebooks
- • NVIDIA NIMs & Blueprints
Model Support
- • DeepSeek R1 (70B)
- • Llama 3.1 (405B)
- • Meta & Google models
- • Custom frameworks
Built for Real AI Work
From prototyping to production deployment, we've got you covered.
Rapid Prototyping
Test ideas quickly without waiting for cloud resources. Iterate faster, ship better.
Model Fine-Tuning
Customize pre-trained models for your specific use case. Keep your data local and private.
Data Science
Analyze massive datasets with ease. Run complex analytics and ML workflows locally.
Seamless Cloud Integration
Develop locally, deploy anywhere. Move your models to DGX Cloud or any accelerated infrastructure with minimal code changes.
Technical Excellence
Every detail engineered for maximum AI performance.
Computing Power
Connectivity
Physical
Software Stack
Ready to Supercharge Your AI Development?
Join researchers, developers, and data scientists who are pushing AI boundaries with desktop supercomputing.
Available for education, research, and enterprise deployments