NVIDIA debuts DGX Spark, the AI supercomputer you can fit on your desk

NVIDIA has launched the DGX Spark, the world’s smallest AI supercomputer, which will start shipping this week. The device integrates GPUs, CPUS, networking, CUDA libraries, and AI software. 

The new release has been integrated with the GB10 Grace Blackwell Superchip. It can deliver up to 1 petaflop of AI performance and 128 GB of unified CPU-GPU memory in a compact desktop form factor. The supercomputer uses ConnectX-7 200 Gb/s networking and NVLink-C2C technology, which provides five times the bandwidth of the fifth-generation PCIe. 

Huang says DGX Spark will democratize AI computing power

NVIDIA’s CEO, Jensen Huang, revealed that the DGX Spark returns the firm to its original mission of democratizing AI computing power. 

“In 2016, we built DGX-1 to give AI researchers their supercomputer. I hand-delivered the first system to Elon at a small startup called OpenAI, and from it came ChatGPT. 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.”

–Jensen Huang, NVIDIA CEO

The DGX Spark enables developers to run inference on models with up to 200 billion parameters and fine-tune models of up to 70 billion parameters locally, eliminating the need to rely solely on cloud or data center services. The latest blog revealed that the DGX Spark has been preloaded with the full NVIDIA AI software stack, including CUDA libraries and NIM microservices. The software stack enables immediate deployment of AI workflows, including image generation, vision search, and language model tuning.

Early adopters include Anaconda, Google, Hugging Face, Meta, Microsoft, JetBrains, Docker, and LM Studio, who have validated and optimized their tools for Spark. Research institutions, such as NYU’s Global Frontier Lab, have also begun testing the unit. Kyunghyun Cho, professor of computer and data science at NYU, said that the DGX Spark allows them to access petascale computing on a desktop. He added that it enables rapid prototyping and experimentation with advanced AI models, even for privacy-sensitive applications such as healthcare.

DGX Spark runs on NVIDIA’s Linux-based DGX OS and is purpose-built for developing and training AI models, rather than consumer applications. Users can also link two Spark systems to support models of up to 405 billion parameters.

Huang personally delivered one of the first DGX Spark units to Elon Musk at SpaceX’s Starbase facility in Texas to commemorate the launch, echoing the 2016 delivery of the original DGX-1 to OpenAI.

NVIDIA stock jumps 2.88% today following the announcement

NVIDIA announced that the DGX Spark will be available for order starting October 15 through NVIDIA’s website and select partner manufacturers, including Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, and MSI. It will also be available at Micro Center retail outlets in the U.S. and will be made globally available through the AI lab’s distribution network later.

🚨NVIDIA and its partners will start shipping DGX Spark — the world’s smallest AI supercomputer.⁰⁰Early recipients are testing, validating and optimizing their tools, software and models for DGX Spark.

Built on the NVIDIA Grace Blackwell architecture, DGX Spark integrates… pic.twitter.com/5VPBDvwXFd

— NVIDIA Newsroom (@nvidianewsroom) October 13, 2025

The launch follows NVIDIA’s ongoing GPU supply deals with OpenAI, xAI, Amazon, Microsoft, Meta, Google, and other AI companies. The company recently announced a $100 billion GPU supply agreement with OpenAI and a $6.3 billion deal with CoreWeave.

Cryptopolitan reported that the AI chip maker will invest $100 billion into OpenAI to help the AI lab build massive data centers centered on NVIDIA’s processors. The companies revealed that OpenAI will deploy NVIDIA systems that will require 10 gigawatts of power collectively. The gigawatt measure, used more often now to better describe large AI chip clusters, sets the scale of the project. Jensen Huang told CNBC’s Jon Fortt in San Jose, California, that 10 gigawatts equals around 4 to 5 million GPUs. That matches what the AI chip maker expects to ship this year, twice as much as last year. 

According to Google Finance data, NVIDIA stock was trading at $188.32 at the time of publication, representing a 2.88% increase over the past 24 hours. The stock has also maintained a positive YTD of +40.23%, showing sustained positive investor confidence in the AI chip maker’s performance. 

Claim your free seat in an exclusive crypto trading community - limited to 1,000 members.