NVIDIA unveiled its latest technological breakthrough at this year's CES: NVIDIA Project DIGITS, a personal AI supercomputer that brings extraordinary computing power to AI researchers, data scientists, and students.
The heart of the system NVIDIA Project DIGITS is brand new GB10 Superchip, based on architecture Grace Blackwell and provides 1 petaflop of performance in FP4 precision, which is enough to power AI models with up to 200 billion parametersThe entire system runs on a regular electrical outlet, meaning a supercomputer can now sit right on your desk.
A supercomputer on your desk? Yes, that's now possible.
While we're used to NVIDIA setting new AI milestones every year, Project DIGITS is something special. It's the smallest AI supercomputer ever, allowing researchers and developers to prototype, train, and optimize AI models on their desktops. Once the model is ready, they can easily transfer it to the cloud or data centers for larger-scale implementation.
Jensen Huang, founder and CEO of NVIDIA, emphasized at the presentation that artificial intelligence will be an integral part of every industry and application in the future. With Project DIGITS, they want to make the Grace Blackwell platform accessible to millions of developers, accelerating innovation and expanding the boundaries of what is possible in the world of AI.
GB10 Superchip: Small but Brutally Powerful – The Heart – NVIDIA Project DIGITS
The heart of Project DIGITS is GB10 Superchip, which combines NVIDIA Blackwell GPU and Grace CPU in ultra-efficient system-on-a-chip (SoC) design. With 20 Arm cores, it ensures high energy efficiency, and the NVLink-C2C connection enables extremely fast communication between the processor and the graphics unit. The system includes 128 GB of unified storage, which enables faster processing of large models, and up to 4TB NVMe storage, which provides enough space for large data sets.
Another unique feature of the GB10 Superchip is the ability to connect two Project DIGITS devices via NVIDIA ConnectX networking technology. This means that developers can run even larger models – up to 405 billion parameters, which is comparable to high-end AI data centers. NVIDIA developed the GB10 in collaboration with MediaTek, the leading manufacturer of Arm SoC chips, which further improves the ratio between performance and energy efficiency.
Why is this groundbreaking?
Currently, most AI development is done in data centers or the cloud, which means high costs for renting processing capacity. With Project DIGITS, NVIDIA is changing this by enabling AI models to be developed and tested directly on the desktop, without the need for constant access to the cloud.
At the same time, the system is fully compatible with existing NVIDIA solutions, meaning developers can seamlessly move from prototyping on a local device to larger scale implementations in the cloud or data centers. All of this is enabled by a unified Grace Blackwell architecture and support for the entire NVIDIA software ecosystem, including PyTorch, TensorFlow, Jupyter notebooks, NeMo and RAPIDS.
With this, NVIDIA is democratizing access to artificial intelligence and enabling every developer to create breakthrough AI models – regardless of whether they work in an academic lab, startup, or corporation.
When and for how much?
Project DIGITS will be available in May 2025, with a starting price of $3,000Considering the capabilities it offers, this is a surprisingly affordable price for an AI supercomputer that can sit right on your desk.
Conclusion: The AI Revolution in Your Workspace
With Project DIGITS, NVIDIA is not just delivering another powerful device – it is creating an entirely new paradigm for AI development. AI, once the domain of massive data centers, is now coming directly to the desks of researchers, developers, and innovators around the world.
Will DIGITS truly become an indispensable AI tool or will it remain a niche product for enthusiasts? Only time will tell. But judging by NVIDIA's past successes, we are on the verge of another technological revolution that will reshape the future of artificial intelligence.
