Nvidia just announced a new $3,000 AI supercomputer during the CES 2025 keynote. Currently, it’s called Project Digits. This will be available for purchase starting in May 2025, targeting developers, researchers, and even students. So what exactly is this AI supercomputer, and how does it differ from a normal one? Here’s everything you need to know about Nvidia’s Project Digits.
What Is Project Digits?
Project Digits is a Mac-mini-like device designed to run AI models. It is powered by Nvidia’s GB10 Grace Blackwell Superchip. That’s a bit of a mouthful, but it combines a powerful Blackwell GPU with fifth-generation Tensor Cores and a high-performance Grace CPU featuring 20 Arm-based cores. Interestingly, you can run it using your wall socket power outlet which was previously not possible.
In simpler terms, it comes with a super chip specifically designed for AI calculations and tasks that delivers a petaflop of AI performance at FP4 precision (make the calculations faster by making approximations).
Also Read:
- Samsung’s Bixby Makes a Comeback with Powerful AI Upgrade
- Google Maps Wrapped: How to See Your 2024 Travel Recap
- What Are AI Agents and Why Everyone Is Talking About Them?
Users can leverage this device for AI tasks like prototyping, fine-tuning AI models, and even running an entire large AI model directly on the device. It supports AI models with up to 200 billion parameters, and if you need more, you can link two units together using Nvidia ConnectX to handle 405-billion-parameter models. For context, with this setup, you can run Meta’s largest Llama 3.1 405B model locally.
In addition, it includes 128GB of unified memory and up to 4TB of NVMe storage, enabling it to handle massive datasets and complex models required for AI.
Coming to software, the device runs on the NVIDIA DGX OS, which is built on top of open-source OS Ubuntu Linux. Developers can use popular AI frameworks like PyTorch, Python, and Jupyter notebooks. Nvidia’s AI software library, including tools like NeMo for model fine-tuning and RAPIDS for data science workflows, is also supported. Once models are built locally, they can be deployed to Nvidia DGX Cloud or data centers without compatibility issues.
Pricing and Availability
Project Digits will start at $3,000 and is set to launch in May 2025 through Nvidia and its partners. For AI researchers, data scientists, developers and students, who are looking to experiment with AI or scale up their projects, this could be a game-changing tool to bridge the gap between local development and large-scale deployment.
