$3,000 for data center power on your desk? This is the crazy, but realistic bet that Nvidia is making with its Project DIGITS in collaboration with MediaTek.
After the announcement of the RTX 50, Nvidia mainly mentioned artificial intelligence… and a one more thing with Project DIGITS, a personal supercomputer that could revolutionize the way we develop and use AI.
To go further
Nvidia announces the GeForce RTX 5090, RTX 5080, RTX 5070 Ti and RTX 5070 at CES 2025: follow the new features live
Far from traditional PCs or servers, this newcomer creates its own category by offering computing power previously reserved for data centers, all in a surprisingly compact format.
Project DIGITS is based on an innovation: the chip GB10 Grace Blackwell. Developed in collaboration with MediaTekthis chip combines a 20-core ARM CPU with the latest Nvidia innovations and Blackwell architecture in terms of AI calculation. The result? Computing power reaching 1 petaflop, the equivalent of 1000 operations per second in FP4 precision. To put that number into perspective, that’s about 1000 times more powerful than a current high-end laptop.
The device does not skimp on technical specs: 128 GB of LPDDR5 memory, up to 4 TB of NVMe storage, and NVIDIA ConnectX network connectivity. But most impressive remains its ability to run AI models with up to 200 billion parameters – a performance that rises to 405 billion when two units are connected together.
3000 dollars
While the price of $3,000 may seem high for a computer, it must be put into perspective with the current costs of AI development. Today, most developers and researchers must rent cloud resources at often very high prices, or invest in even more expensive infrastructure. Project DIGITS could therefore represent a significant saving in the long term.
Nvidia has clearly designed its product for a specific audience: AI developers, researchers and data science students. The device comes with the Linux-based Nvidia DGX OS and the entire Nvidia AI software suite pre-installed, allowing immediate use for model prototyping, fine-tuning, and edge application development.
Rather than embarking on the race for ARM Windows PCs as some anticipated, Nvidia chose to create a new market segment. This positioning avoids direct confrontation with the PC giants while strengthening the company’s position in its preferred field: AI and high-performance computing.
This strategy could well pay off. By offering a personal supercomputer at a relatively affordable price, Nvidia really wants to imbue the market with its position as a leader in AI.
