The subsequent AI revolution begins the place rockets launch. NVIDIA DGX Spark’s first cease: Starbase, Texas.
NVIDIA founder and CEO Jensen Huang arrived on the SpaceX facility — amid towering engines and gleaming metal — to hand-deliver the corporate’s just-launched DGX Spark to Elon Musk.
Huang arrived strolling previous rows of engineers who waved and grinned. Moments later, Musk appeared within the cafeteria, greeting employees and opening donuts and chips for teenagers earlier than grabbing a slice of pizza.
Huang joined him, recounting the story of delivering the primary DGX system to OpenAI and explaining how Spark takes that mission additional.
“Think about delivering the smallest supercomputer subsequent to the most important rocket,” Huang stated with amusing.
The handoff got here as SpaceX ready for the eleventh check of Starship, the world’s strongest launch automobile.
DGX Spark packs 128GB of unified reminiscence and delivers a petaflop of AI efficiency, sufficient to run fashions with 200 billion parameters regionally.
Constructed for builders, researchers and creators, DGX Spark brings supercomputer-class efficiency past the information heart — able to seize and go.
9 years in the past, NVIDIA wager on the way forward for AI with NVIDIA DGX-1. Right this moment, that wager goes past the information heart with the handoff to Musk coming amid the eleventh check of SpaceX’s Starship, the world’s strongest launch automobile.
Constructed for builders, researchers and creators who need supercomputer-class efficiency that’s able to seize and go.
From robotics labs to artistic studios, DGX Sparks are touchdown the place concepts occur… placing petaflop AI inside arm’s attain of everybody.
This weblog shall be up to date as DGX Spark programs land from Ollama in Palo Alto to Arizona State’s robotics lab, from Refik Anadol’s studio to the arms of Jo Mardall at Zipline. Every supply is a brand new chapter within the story of AI.
What’s Inside DGX Spark? A Supercomputer within the Palm of Your Hand
It’s the scale of a bit of origami paper, and the thickness of a hardcover guide. It acts like a rocket engine for AI.
DGX Spark isn’t simply compact — it’s dense with risk. Inside its 1.2 kg chassis is a full-blown AI supercomputer constructed to take AI past the information heart and into the arms of those that create.
At its core:
- NVIDIA GB10 Grace Blackwell Superchip — delivering as much as 1 petaflop of AI efficiency at FP4 precision.
- 128GB of unified CPU-GPU reminiscence — so builders can prototype, fine-tune and run inference regionally with out bouncing between machines or cloud cases.
- NVIDIA ConnectX networking for clustering and NVIDIA NVLink-C2C for 5x PCIe bandwidth.
- NVMe storage for velocity and HDMI out for visuals.
And it’s not simply {hardware}. DGX Spark comes with the total NVIDIA AI software program stack — frameworks, libraries, pretrained fashions and NVIDIA NIM microservices, able to energy workflows like:
- Customizing image-generation fashions resembling FLUX.1
- Constructing imaginative and prescient search and summarization brokers with NVIDIA Cosmos
- Deploying optimized chatbots utilizing Qwen3
This isn’t a dev field. It’s a launchpad… A petaflop of AI efficiency inside arm’s attain for builders, researchers and creators in every single place.
Companions Energy Up: DGX Spark Lands Past the Knowledge Middle
From PC giants to AI pioneers, the diminutive DGX Spark is bound to start out one thing massive.
DGX Spark is already within the arms of innovators — from ISVs optimizing their instruments to researchers pushing the boundaries of robotics, artwork and edge AI.
DGX Spark isn’t only a breakthrough in dimension and efficiency… it’s a platform constructed on collaboration. Acer, ASUS, Dell Applied sciences, GIGABYTE, HP, Lenovo and MSI are rolling out programs that put petaflop AI in your desk, reworking the desktop into an AI launchpad.
These companions ship greater than {hardware}. They ship risk — NVIDIA’s full AI stack in a compact type issue that accelerates agentic and bodily AI improvement in every single place concepts occur.
And the rollout doesn’t cease at OEMs. DGX Sparks are already lighting up the AI ecosystem. Some highlights:
- Ollama in Palo Alto, rewriting how builders run giant language fashions regionally.
- NYU International Frontier Lab, the place researchers prototype algorithms for privacy-sensitive functions.
- Zipline, pushing the boundaries of autonomous supply.
- Arizona State College, operating robotics simulations and imaginative and prescient fashions on the edge.
- Refik Anadol’s studio, mixing artwork and AI with petaflop efficiency.
From artistic studios to robotics labs, DGX Sparks are touchdown the place creativeness meets engineering — taking AI past the information heart and into the arms of these constructing what’s subsequent.
DGX Spark shall be typically obtainable beginning Wednesday, Oct. 15, on NVIDIA.com and thru companions worldwide.