This post is also available in:
עברית (Hebrew)
At CES 2025, Nvidia revealed a groundbreaking innovation: the world’s smallest AI supercomputer, Project Digits. This compact powerhouse packs immense computing capabilities, capable of handling AI models with up to 200 billion parameters. Set to hit the market in May, Project Digits is priced from $3,000, offering cutting-edge performance without requiring the space or power demands of traditional supercomputers.
At its core, Project Digits is powered by Nvidia’s new Grace Blackwell Superchip, designed to deliver a petaflop of AI computing performance. This enables users to prototype, fine-tune, and run sophisticated AI models directly from their desks. Until now, achieving such performance typically meant relying on much larger systems that require much higher energy. But with this revolutionary desktop system, smaller teams and developers can handle advanced AI workloads with no problem.
The system also supports scalable operations—by linking two Project Digits units together, users can process models with up to 405 billion parameters. The larger the model, the more complex and powerful its problem-solving ability, and the Grace Blackwell Superchip ensures that this performance remains accessible to everyday users.
Each unit will feature 128GB of unified memory and up to 4TB of NVMe storage, making it an ideal solution for developers working with large datasets. This is significantly more than traditional computers, highlighting the immense power packed into this compact system. Additionally, Nvidia offers an extensive suite of AI tools with Project Digits, including development kits, orchestration tools, and pre-trained models, which will be available through the Nvidia NGC catalog and Developer portal. These resources enable seamless AI model development and testing directly on the device.
Nvidia’s CEO, Jensen Huang, emphasized the system’s potential to revolutionize AI accessibility: “AI will be mainstream in every application for every industry… Project Digits places an AI supercomputer on every developer’s desk.” While the price point is high, it underscores Nvidia’s confidence in the demand for such powerful AI workstations.
