Nvidia will be releasing two AI personal computers under a new brand called DGX, formerly Project Digits, to help researchers, scientists and developers, the company said at its GPU Technology Conference on Tuesday.
The smaller DGX Spark and the desktop-size DGX Station will feature Nvidia’s Blackwell Ultra platform, the company said. These computers are meant for developers, researchers, robotics developers, data scientists and students to tune AI models locally. It’ll also be possible to use the power of the Nvidia DGX Cloud to accelerate development if the power on the machines isn’t enough. The DGX Spark is priced at $4,000 and reservations are now open. Pricing and availability for the DGX Station will come later this year and the device will be manufactured by Nvidia’s hardware partners, which include Asus, Boxx, Dell, HP, Lambda and Supermicro.
The DGX Spark will use the Nvidia GB10 Grace Blackwell Superchip and fifth-generation Tensor Cores, “delivering up to 1,000 trillion operations per second of AI compute for fine-tuning and inference,” according to a company press release. Nvidia said this machine can handle the latest reasoning models, like DeepSeek R1.
The DGX Station will use a more powerful GB300 Grace Blackwell Ultra Desktop Superchip with a huge 784GB of coherent memory space for large training and inference.
Watch this: Watch Nvidia’s GTC 2025 Keynote: All the Highlights in 16 Minutes
As AI has become integral to product development, more companies and individuals are looking for on-site solutions. When tapping into services from OpenAI, Google, Anthropic and others, there’s a cost associated with using AI on their servers. After a lot of prolonged use, those costs quickly add up. For some, like researchers and creators who are constantly iterating on a project, having a device that can do all that legwork on-site can lead to major cost savings. Plus, there’s less concern about servers slowing down due to high load. Data stored on local machines is also more secure. For financial institutions or hospitals that prefer to have sensitive data on-site, local AI computers are preferred.
Read the full article here