NVIDIA Project DIGITS: All You Need To Know About the Blackwell AI Supercomputer

"The amount of computation we need is incredible and we truly envision a society that can scale computation to produce ever more novel and advanced intelligence. Intelligence is our most valuable asset which can be applied to solve some of the most challenging problems. This 'scaling law' is driving enormous demand for NVIDIA computing.", said NVIDIA CEO Jensen Huang at the much-awaited CES 2025 Keynote on 6 January 2025. 

As a result of the growing demands for AI scaling, NVIDIA released yet another breakthrough in AI “Project DIGITS” which is now causing waves across the AI space. It is a $3,000 personal AI supercomputer powered by the state-of-the-art GB10 Grace Blackwell Superchip. It is designed to change the way we prototype, fine-tune and deploy AI models in 2025. Continue reading as we explore more about the NVIDIA Project DIGITS. 

Get Grace Blackwell at Your Fingertips

NVIDIA Project DIGITS (Deep Learning GPU Training System) aims to bring AI supercomputing to every desk. It is built with the new GB10 Grace Blackwell Superchip, delivering up to 1 petaflop of AI computing performance. Designed explicitly for prototyping, fine-tuning, and inference tasks, PROJECT DIGITS offers capabilities once reserved for massive data centres. 

“AI will be mainstream in every application for every industry. With Project DIGITS, the Grace Blackwell Superchip comes to millions of developers. Placing an AI supercomputer on the desks of every data scientist, AI researcher, and student empowers them to engage and shape the age of AI.”

-Jensen Huang, founder and CEO of NVIDIA, during the CES 2025 keynote. 

The GB10 Superchip is a system-on-a-chip (SoC) powered by the NVIDIA Grace Blackwell architecture, offering up to 1 petaflop of AI performance at FP4 precision. It features an NVIDIA Blackwell GPU with the latest CUDA cores and fifth-generation Tensor Cores, connected via NVLink-C2C chip-to-chip interconnect to a high-performance NVIDIA Grace CPU. The CPU includes 20 power-efficient Arm-based cores.

What Sets Project DIGITS Apart?

The NVIDIA GB10 Superchip is not a regular “AI Supercomputer" would therefore be the next big thing in AI with innovative features:

1. 1 Petaflop Performance at FP4 Precision: Project DIGITS is powered by the Grace Blackwell architecture, the system ensures seamless AI operations.
2. Unified Memory and High Storage: With 128GB of unified memory and up to 4TB of NVMe storage, Project DIGITS can effortlessly run large-scale AI models with up to 200 billion parameters. To improve performance, two DIGITS systems can interlink to handle models up to 405 billion parameters.
3. NVIDIA ConnectX Networking: It enables high-speed networking for seamless data transfers, essential for demanding AI applications.
4. Compact Design: Project DIGITS operates on a standard electrical outlet and fits comfortably on a desk, democratising access to advanced AI computing.

Grace Blackwell Architecture for Scalable AI Prototyping and Deployment

The Grace Blackwell architecture empowers enterprises and researchers to efficiently prototype, fine-tune, and test AI models on local Project DIGITS systems running Linux-based NVIDIA DGX OS. These models can then be seamlessly deployed to NVIDIA DGX Cloud VMs or on-premises data centre infrastructure for production. This end-to-end workflow allows developers to leverage the same Grace Blackwell architecture and NVIDIA AI Enterprise software platform throughout their local prototyping and cloud scaling phases.

Project DIGITS offers users access to a comprehensive range of NVIDIA AI software, providing ample tools for experimentation and prototyping. This includes development kits, orchestration tools, frameworks, and models found in the NVIDIA NGC catalogue and available on the NVIDIA Developer portal. Developers can fine-tune models using the NVIDIA NeMo framework, accelerate data science workflows with NVIDIA RAPIDS libraries, and run frameworks like PyTorch, Python and Jupyter notebooks.

Prototyping, Fine-Tuning and Inference

Project DIGITS allows you to prototype, fine-tune and run inference on large AI models of up to 200 billion parameters locally while offering seamless deployment capabilities to the data centre or cloud environments.

For developers requiring even greater computational capabilities for fine-tuning and Inference, Hyperstack’s GPU-as-a-Service solution provides instant access to powerful NVIDIA GPUs like the NVIDIA H100 SXM and the NVIDIA A100.  These systems are built with exceptional capabilities for fine-tuning and inference for large-scale AI applications. 

• 1-Click Deployment: You can quickly start your AI workloads without setting up complexity.
• Instant Access to High-End GPUs: Get access in minutes to powerful cloud GPUs for AI to get superior performance for all your AI needs.
• High-Speed Networking: Get faster networking speeds of up to 350Gbps for seamless data transfer for fine-tuning and inference workloads. Learn how to Improve LLM Fine-Tuning and Inference with High-Speed Networking
• Cost-Effective Solutions: With flexible GPU configurations, Hyperstack ensures you only pay for what you need.

Start Your AI Journey with Hyperstack's Cloud Infrastructure.

FAQs

What is Project DIGITS?

Project DIGITS is a compact AI supercomputer powered by NVIDIA's GB10 Grace Blackwell Superchip. It delivers up to 1 petaflop of AI performance, allowing users to prototype, fine-tune and deploy large AI models locally.

What are the key features of Project DIGITS?

The key features of NVIDIA Project DIGITS include:

• Processor: NVIDIA GB10 Grace Blackwell Superchip, featuring an NVIDIA Blackwell GPU with latest-generation CUDA cores and fifth-generation Tensor Cores, connected via NVLink-C2C chip-to-chip interconnect to a high-performance NVIDIA Grace CPU.
• Memory: 128GB of unified and coherent memory.
• Storage: Up to 4TB of NVMe storage.
• Networking: NVIDIA ConnectX networking for high-speed data transfers.
• Form Factor: Compact design suitable for desktop placement.

What AI models can Project DIGITS handle?

Project DIGITS can run large AI models with up to 200 billion parameters locally. For even larger models, two systems can be linked to handle models up to 405 billion parameters. 

What software and frameworks does Project DIGITS support?

Project DIGITS supports NVIDIA's AI software stack, including frameworks like PyTorch, Python, and Jupyter notebooks. It also integrates with NVIDIA's NGC catalogue, providing access to a wide range of AI tools and libraries. 

When will Project DIGITS be available?

Project DIGITS is expected to be available in May 2025.