At CES 2025, Nvidia introduced Project Digits, its first personal AI supercomputer, aimed at addressing the growing demand for generative AI. Designed as an accessible yet powerful platform for developers and data scientists, Digits provides a comprehensive solution for managing the entire lifecycle of AI models, from inference to fine-tuning and full-scale development.
With a starting price of $3,000, this compact device is set to democratize AI innovation, making advanced capabilities available to individuals and smaller organizations.
At the core of Project Digits is the Nvidia GB10 Grace Blackwell Superchip, which combines a Blackwell GPU and a 20-core Grace CPU connected by the NVLink-C2C interconnect. This integration ensures rapid data transfer between components, enabling efficient handling of complex AI workloads. With 128GB of unified memory shared between the CPU and GPU, the system eliminates delays caused by data transfers, while high-speed NVMe storage provides quick access to datasets during training and execution. Despite delivering one petaflop of AI performance, Digits operates efficiently on a standard wall outlet, bypassing the need for specialized power and cooling systems.
Preloaded with Nvidia’s robust AI software stack, Project Digits integrates seamlessly into the Nvidia AI ecosystem. Its Linux-based DGX OS is optimized for high-performance computing, while the Nvidia AI Enterprise platform and NeMo framework provide essential tools for developing and fine-tuning large language models. RAPIDS libraries accelerate data workflows, and compatibility with popular frameworks like PyTorch, Python, and Jupyter Notebooks ensures that developers can work in familiar environments. The device also supports cloud and data center integration, allowing for scalable AI model deployment.
Project Digits supports AI models with up to 200 billion parameters, a capability previously reserved for large-scale supercomputing environments. By connecting multiple Digits units, users can scale to handle models as large as 405 billion parameters. Additionally, the system bridges Linux and Windows environments through the Windows Subsystem for Linux, making it versatile for various user preferences.
Nvidia collaborated with MediaTek to design an energy-efficient CPU, enhancing Digits’ power efficiency. The device’s ability to operate locally ensures data privacy and compliance, reducing reliance on cloud infrastructure. Targeted at AI researchers, developers, students, and smaller teams, Digits opens new possibilities for prototyping, fine-tuning large language models, researching new algorithms, and deploying AI-powered solutions.
The Project Digits is expected to be available in May 2025. By combining affordability, power, and flexibility, it empowers a broader range of users to explore and innovate in the rapidly evolving field of AI.
