NVIDIA’s new DGX Spark brings cloud-level AI power to your desk

BY Michael Anthony Bitoon

Published 15 Oct 2025

Why trust Greenbot

We maintain a strict editorial policy dedicated to factual accuracy, relevance, and impartiality. Our content is written and edited by top industry professionals with first-hand experience. The content undergoes thorough review by experienced editors to guarantee and adherence to the highest standards of reporting and publishing.

NVIDIA started selling its DGX Spark on Wednesday, a desktop computer small enough to fit beside your keyboard but powerful enough to run artificial intelligence (AI) models that previously required expensive cloud services or data centers.

The machine went on sale on October 15 for $3,999. CEO Jensen Huang marked the occasion by personally delivering one of the first units to Elon Musk at SpaceX’s Texas facility, recreating a similar handoff he made nine years ago.

The Nvidia DGX Spark is the tiny gold box sitting on the desk to the left of the monitor.

Source: Nvidia

“In 2016, we built DGX-1 to give AI researchers their own supercomputer. I hand-delivered the first system to Elon at a small startup called OpenAI — and from it came ChatGPT, kickstarting the AI revolution,” Huang said in a statement. “With DGX Spark, we return to that mission — placing an AI computer in the hands of every developer to ignite the next wave of breakthroughs.”

The Spark measures just 5.91 by 5.91 by 1.99 inches and weighs 2.65 pounds. Despite its tiny size, the system delivers one petaflop of AI performance, meaning it can handle a million billion calculations each second.

What sets the DGX Spark apart from consumer graphics cards is its 128GB of unified memory shared between its processor and graphics chip. That’s ten times more than NVIDIA’s RTX 5070 consumer card, which tops out at 12GB.

Memory advantage is a big deal for AI developers. Running a 120 billion-parameter language model requires about 80GB of memory, well beyond what any consumer GPU offers. The Spark can handle models up to 200 billion parameters for testing and fine-tune models up to 70 billion parameters.

NVIDIA’s new DGX Spark brings cloud-level AI power to your desk 3

Source: Nvidia

The computer runs on NVIDIA’s GB10 Grace Blackwell Superchip, which combines a 20-core ARM processor with a Blackwell graphics processor. The chip was developed with MediaTek and connects the processor and GPU using NVLink-C2C technology, running five times faster than standard PCIe connections.

NVIDIA ships the Spark with its full AI software stack preinstalled on a custom Ubuntu Linux operating system. Developers can immediately access CUDA libraries and NIM microservices for tasks like customizing image generation models or building AI chatbots.

Many AI tasks now exceed the capabilities of standard computers, forcing teams to migrate work to cloud platforms or data centers. Local development can be faster and more secure for sensitive projects.

“DGX Spark allows us to access peta-scale computing on our desktop,” said Kyunghyun Cho, professor at the NYU Global AI Frontier Lab. “This new way to conduct AI research and development enables us to rapidly prototype and experiment with advanced AI algorithms and models — even for privacy- and security-sensitive applications, such as healthcare.”

Companies, including Google, Meta, Microsoft, and Hugging Face, received early units for testing. The system is available through NVIDIA.com and from partners including Acer, ASUS, Dell, Gigabyte, HP, Lenovo, and MSI. Select Micro Center stores in the U.S. will also stock the machines.