NVIDIA DGX Spark Delivery Begins: Jensen Huang Appears at SpaceX to Deliver First Desktop AI Supercomputer
Abstract
NVIDIA announced on October 13, 2025 (ET) the commencement of deliveries for DGX Spark, a desktop AI supercomputer. CEO Jensen Huang personally delivered the first unit to Elon Musk at SpaceX's Starbase in Texas. The system will be globally available for purchase starting October 15 via NVIDIA's official website and partners, priced at $3,999.
On October 13, 2025 (ET), NVIDIA CEO Jensen Huang appeared at SpaceX's Starbase facility in Texas to personally deliver the company's newly launched DGX Spark to Elon Musk. This delivery ceremony marks the official market entry of NVIDIA's latest generation desktop AI supercomputer.
The DGX Spark is NVIDIA's smallest AI supercomputer globally, featuring a compact desktop form factor. The system boasts 128GB of unified memory and delivers 1 petaflop of AI performance, sufficient to run models with up to 200 billion parameters locally.
Technical Specifications and Performance
The DGX Spark is built on NVIDIA's Grace Blackwell GB10 superchip and weighs only 1.2 kilograms. The system integrates NVIDIA ConnectX networking capabilities and NVLink-C2C technology, which provides 5 times the bandwidth of standard PCIe Gen5.
This system is capable of performing inference on AI models with up to 200 billion parameters and can fine-tune models with up to 70 billion parameters locally. The DGX Spark employs a hybrid CPU-GPU design, achieving 128GB of unified memory through NVIDIA's NVLink-C2C interconnect technology, allowing both CPU and GPU to access the same data pool without slow data transfers via the PCIe bus.
Market Positioning and Price
Priced at $3,999, the DGX Spark will be available for order starting October 15, 2025 (ET) through NVIDIA's official website. Partners including Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, and MSI will also launch systems featuring the DGX Spark globally.
NVIDIA initially announced the product at CES in January 2025, with an anticipated price of $3,000 and an original release target of May. After several months of delays, the final price was adjusted upwards to $3,999.
Historical Significance
At the delivery event, Jensen Huang stated, "In 2016, we built the DGX-1, giving AI researchers their own supercomputer. I personally delivered the first system to Elon, whose then-small startup, OpenAI, went on to create ChatGPT and ignite the AI revolution."
This delivery coincided with SpaceX's 11th Starship test flight, with Starship being the world's most powerful launch vehicle. Huang quipped, "Imagine delivering the smallest supercomputer next to the biggest rocket."
Application Scenarios
The DGX Spark comes pre-installed with the complete NVIDIA AI software stack, including frameworks, libraries, pre-trained models, and NVIDIA NIM microservices. Supported workflows include customizing image generation models like FLUX.1, building visual search and summarization agents using NVIDIA Cosmos, and deploying optimized chatbots.
The first DGX Spark systems have been delivered to various institutions and teams, including Ollama in Palo Alto, NYU's Global Frontier Lab, autonomous delivery company Zipline, Arizona State University's Robotics Lab, and artist Refik Anadol's studio.
Industry Impact
Kyunghyun Cho of NYU's Global Frontier Lab commented, "The DGX Spark gives us access to peta-scale computing power right on our desktop." The system targets a wide range of users, from independent researchers to small AI startups, as well as enterprise teams requiring dedicated, secure machines to prototype and fine-tune proprietary models.
The launch of the DGX Spark signifies NVIDIA's strategic shift from a cloud-first model towards edge computing, aiming to attract market segments dissatisfied with data egress fees, security vulnerabilities, and unpredictable cloud billing. By localizing computing power, NVIDIA empowers individuals to innovate without the latency, data privacy concerns, or recurring costs associated with cloud services.