Introduction
Nvidia has announced it will begin taking orders for the DGX Spark, a $4,000 desktop AI computer that delivers one petaflop of computing performance and 128GB of unified memory in a compact form factor that fits on a desk. Its primary selling point is the large integrated memory capacity that enables running larger AI models than consumer GPUs can handle.
Orders for the DGX Spark opened on October 15, 2025, through Nvidia's website, with systems also available from manufacturing partners and select US retail stores.
A New Category of AI Workstation
The DGX Spark, previewed as "Project DIGITS" in January and formally named in May, represents Nvidia's attempt to establish a new category of desktop workstation specifically designed for AI development.
With the Spark, Nvidia addresses a common challenge facing AI developers: many AI tasks exceed the memory and software capabilities of standard PCs and workstations, forcing professionals to shift their work to cloud services or data centers. However, the actual market demand for a desktop AI workstation remains uncertain, particularly considering the upfront cost versus cloud alternatives that allow developers to pay as they go.
Capabilities and Practical Applications
The DGX Spark includes sufficient memory to run larger-than-typical AI models for local tasks, supporting up to 200 billion parameters and fine-tuning models containing up to 70 billion parameters without requiring remote infrastructure. Potential applications include running large open-weights language models and media synthesis models such as AI image generators.
According to Nvidia, users can customize Black Forest Labs' Flux.1 models for image generation, build vision search and summarization agents using Nvidia's Cosmos Reason vision language model, or create chatbots using the Qwen3 model optimized for the DGX Spark platform.
Technical Specifications: Big Memory in a Tiny Box
Nvidia has packed substantial capabilities into a 2.65-pound box measuring 5.91 x 5.91 x 1.99 inches that uses 240 watts of power. The system runs on Nvidia's GB10 Grace Blackwell Superchip, includes ConnectX-7 200Gb/s networking, and utilizes NVLink-C2C technology that provides five times the bandwidth of PCIe Gen 5.
The system features 128GB of unified memory shared between system and GPU tasks. The DGX Spark is an ARM-based system running Nvidia's DGX OS, an Ubuntu Linux-based operating system built specifically for GPU processing. It comes with Nvidia's AI software stack preinstalled, including CUDA libraries and the company's NIM microservices.
Market Positioning and Performance Comparison
The DGX Spark starts at $3,999. While this may seem expensive, considering the cost of high-end GPUs with ample video RAM like the RTX Pro 6000 (approximately $9,000) or AI server GPUs (such as $25,000 for a base-level H100), the DGX Spark may represent a far less expensive option overall, though it's not nearly as powerful.
According to The Register, the GPU computing performance of the GB10 chip is roughly equivalent to an RTX 5070. However, the 5070 is limited to 12GB of video memory, which constrains the size of AI models that can run on such a system. With 128GB of unified memory, the DGX Spark can run far larger models, albeit at a slower speed than, for example, an RTX 5090 (which typically ships with 24GB of RAM).
For instance, running the 120 billion-parameter larger version of OpenAI's recent gpt-oss language model requires about 80GB of memory, which is far more than available in consumer GPUs.
Conclusion
Nvidia's DGX Spark positions itself as an intermediate solution between consumer GPUs and cloud or data center infrastructure for AI development. It offers an interesting compromise for developers and researchers who need to run large models locally without investing in hardware costing tens of thousands of dollars or constantly relying on paid cloud services. Commercial success will depend on actual market demand for this new product category and Nvidia's ability to demonstrate concrete return on investment compared to existing alternatives.
FAQ
What is the Nvidia DGX Spark and who is it for?
The DGX Spark is a compact $4,000 AI desktop workstation with 128GB of unified memory, designed for developers and researchers who need to run large AI models locally without resorting to cloud services or data centers.
What AI models can the DGX Spark run?
The DGX Spark can run AI models with up to 200 billion parameters and fine-tune models up to 70 billion parameters, including open-weights language models and AI image generators like Flux.1, Cosmos Reason, and Qwen3.
How much does the DGX Spark cost compared to other AI solutions?
The DGX Spark starts at $3,999, significantly less expensive than professional GPUs like the RTX Pro 6000 ($9,000) or server GPUs like the H100 ($25,000), representing an economical alternative for those requiring high memory capacity.
What are the technical specifications of the DGX Spark?
The system weighs 2.65 pounds, measures 5.91 x 5.91 x 1.99 inches, uses 240W of power, is based on the GB10 Grace Blackwell Superchip, includes 200Gb/s networking, and features NVLink-C2C technology with five times the bandwidth of PCIe Gen 5.
Is the DGX Spark more powerful than an RTX 5090?
No, the GB10 in the DGX Spark has GPU performance comparable to an RTX 5070. However, its 128GB of unified memory exceeds the 5090's 24GB, enabling it to run much larger models albeit at lower speeds.
What operating systems does the DGX Spark support?
The DGX Spark is an ARM-based system running Nvidia's DGX OS, an Ubuntu Linux-based operating system optimized for GPU processing, with preinstalled AI software stack including CUDA libraries and NIM microservices.