The compact yet powerful Spark is capable of managing advanced AI models while still being desk-friendly.
The compact yet powerful Spark is capable of managing advanced AI models while still being desk-friendly.
Nvidia will begin selling its DGX Spark “personal AI supercomputer” this week. The device is robust enough for users to develop complex AI models while still being compact enough for a desktop setting.
Nvidia announced that Spark can be ordered via nvidia.com beginning Wednesday, October 15th, along with select partners and retailers in the United States. The initial cost was stated to be $3,000 when they unveiled Spark earlier this year, but it seems the DGX Spark will now be priced at $3,999, based on an infographic included in Nvidia’s press release. Many PC manufacturers are offering their own tailored versions, such as the Acer Veriton GN100, as one example, which also retails for $3,999.
Spark delivers performance that previously relied on high-cost, energy-intensive data centers. It may help make AI more widely accessible and could be particularly advantageous for researchers. When first discussing Spark earlier this year (then referred to as Digits), Nvidia CEO Jensen Huang remarked, “placing an AI supercomputer on the desks of every data scientist, AI researcher, and student empowers them to engage and shape the age of AI.”
Purchasers can anticipate seeing a range of similar models on the market as Nvidia has expressed that third-party manufacturers are welcome to produce their own iterations. Brands such as Acer, Asus, Dell, Gigabyte, HP, Lenovo, and MSI are all set to launch their personalized versions of Spark, which Nvidia confirmed today.
Spark includes Nvidia’s GB10 Grace Blackwell Superchip, 128GB of consolidated memory, and up to 4TB of NVMe SSD storage. Nvidia claims it can achieve a petaflop of AI performance — which translates to performing a million billion calculations per second — and can accommodate AI models containing up to 200 billion parameters. Additionally, it is compact, easily fitting on a desk and operating from a standard power supply. Nvidia refers to it as “the world’s smallest AI supercomputer.”
We concur: it truly is quite small. Spark also has a larger counterpart, Station, although details on when or if that will become available to the public are not yet known.
Correction, October 13th: An earlier version of this article incorrectly stated that the DGX Spark was “now available to buy.” It will actually become available on October 15th.
