The Future of AI Supercomputing in a Compact Form

Today, NVIDIA and its partners unveiled a revolutionary new product in the world of artificial intelligence: DGX Spark, the smallest AI-powered supercomputer ever built. This device combines enormous computing power with a compact size, making it ideal for office spaces or laboratories.

Power Meets Portability

The DGX Spark is built for developers working with large AI models. Packed with the NVIDIA GB10 Grace Blackwell Superchip, it offers up to 1 petaflop of performance, 128 GB of unified memory, and a bandwidth of 273 GB/s. This enables local training and inference of models with up to 200 billion parameters without the need for cloud services. It’s a portable powerhouse, weighing just over 1 kg and consuming only 240 watts—significantly lower than previous systems.

The Revolution is Here

The innovation is about performance and accessibility. DGX Spark runs on NVIDIA's full stack of artificial intelligence software. Developers can get started on projects right away without lengthy configuration. it is optimized for a wide range of tasks and is already being tested by major companies such as Google, Meta, and Microsoft. Even research institutions, like NYU, are exploring its potential for sensitive fields such as healthcare.

Partners Take the Lead

Several of NVIDIA’s partners are launching their own compact AI supercomputers. MSI, ASUS, Dell, and Gigabyte are rolling out products based on the GB10 chip, each offering impressive specifications and tailored for different sectors, from fintech to medicine. These devices are proof that the future of AI supercomputing is no longer tied to massive server rooms—it can now fit right on your desk.

The DGX Spark is redefining the future of AI computing. Whether you’re a developer, researcher, or innovator, this tiny titan is ready to power your next big idea.