NVIDIA DGX Spark: Petaflop AI Supercomputer Delivered to SpaceX

NVIDIA launches DGX Spark, the world's smallest AI supercomputer with 1 petaflop performance and 128GB memory, first delivered to Elon Musk at SpaceX.

by HowAIWorks Team
ainvidiahardwaresupercomputerdgx-sparkspacexedge-aidevelopersai-infrastructureartificial-intelligence

Introduction

NVIDIA has unveiled DGX Spark, the world's smallest AI supercomputer that brings petaflop-scale performance to developers, researchers, and creators everywhere. In a symbolic gesture marking the beginning of AI democratization, NVIDIA CEO Jensen Huang personally delivered the first DGX Spark system to Elon Musk at SpaceX's Starbase facility in Texas, where the world's most powerful rockets launch.

This compact powerhouse packs 128GB of unified memory and delivers a petaflop of AI performance—enough to run models with 200 billion parameters locally. The launch represents a significant milestone in taking AI beyond the data center and into the hands of those building the future.

Historic Delivery at SpaceX

Starbase Handoff

The DGX Spark launch began with a memorable moment at SpaceX's Starbase facility. Jensen Huang arrived at the Texas site amid towering rocket engines and gleaming steel structures, walking past rows of engineers who waved and grinned in anticipation.

Huang met with Elon Musk in the SpaceX cafeteria, recounting the story of delivering the first DGX-1 system to OpenAI nine years ago and explaining how DGX Spark takes that mission further. "Imagine delivering the smallest supercomputer next to the biggest rocket," Huang said with a laugh, perfectly capturing the contrast between SpaceX's massive Starship and the compact DGX Spark.

The handoff came as SpaceX prepared for the 11th test of Starship, the world's most powerful launch vehicle, symbolizing how cutting-edge AI infrastructure will power the next generation of ambitious projects—from space exploration to advanced robotics.

Nine Years of Innovation

The delivery echoes NVIDIA's first DGX-1 delivery to OpenAI nine years ago, when the company bet on the future of AI. Today, that bet has evolved beyond the data center, bringing supercomputer-class performance to a device you can grab and go. From robotics labs to creative studios, DGX Sparks are landing where ideas happen—putting petaflop AI within arm's reach of everyone.

DGX Spark Technical Specifications

Core Hardware

DGX Spark isn't just compact—it's dense with possibility. Inside its 1.2 kg chassis is a full-blown AI supercomputer built to take AI beyond the data center:

Size and Form Factor:

  • Dimensions: The size of a piece of origami paper
  • Thickness: Comparable to a hardcover book
  • Weight: Just 1.2 kg (2.6 lbs)
  • Portability: Truly portable AI supercomputing power

Processing Power:

  • NVIDIA GB10 Grace Blackwell Superchip: Delivering up to 1 petaflop of AI performance at FP4 precision
  • Unified memory: 128GB of unified CPU-GPU memory for seamless development workflows
  • Performance: Capable of running models with 200 billion parameters locally
  • Efficiency: Optimized for both development and inference workloads

Connectivity and Storage

Networking:

  • NVIDIA ConnectX networking: For clustering multiple DGX Spark systems
  • NVIDIA NVLink-C2C: Provides 5x PCIe bandwidth for ultra-fast data transfer
  • Cluster capabilities: Ability to connect multiple units for larger-scale projects

Storage and Display:

  • NVMe storage: High-speed storage for models and data
  • HDMI output: Direct visual output for development and demonstration
  • Expansion options: Support for external storage and peripherals

Software Stack

DGX Spark comes with the complete NVIDIA AI software ecosystem:

Included Software:

  • NVIDIA AI frameworks: Full suite of deep learning frameworks
  • Pretrained models: Access to NVIDIA's library of optimized models
  • NVIDIA NIM microservices: Ready-to-deploy AI services
  • Development tools: Complete toolkit for AI development

Supported Workflows:

  • Image generation: Customizing models like FLUX.1 for creative applications
  • Vision AI: Building vision search and summarization agents with NVIDIA Cosmos
  • Chatbots: Deploying optimized conversational AI using models like Qwen3
  • Robotics: Running simulations and training robotic systems at the edge
  • Code generation: AI-assisted software development and debugging

Industry Partners and Ecosystem

OEM Partnerships

NVIDIA has partnered with leading PC manufacturers to bring DGX Spark to market:

Launch Partners:

  • Acer: Desktop and workstation implementations
  • ASUS: Gaming and creator-focused systems
  • Dell Technologies: Enterprise and professional workstations
  • GIGABYTE: High-performance computing solutions
  • HP: Business and creator workstations
  • Lenovo: Professional and research systems
  • MSI: Gaming and content creation systems

These partners deliver more than hardware—they deliver possibility, bringing NVIDIA's full AI stack in a compact form factor that accelerates agentic and physical AI development everywhere ideas happen.

Early Adopters

DGX Spark systems are already landing with innovators across multiple sectors:

Research and Education:

  • NYU Global Frontier Lab: Prototyping algorithms for privacy-sensitive applications
  • Arizona State University: Running robotics simulations and vision models at the edge
  • Academic institutions: Enabling cutting-edge AI research without data center infrastructure

Development and ISVs:

  • Ollama (Palo Alto): Rewriting how developers run large language models locally
  • Independent developers: Empowering solo developers and small teams with supercomputer access

Innovation and Creative Applications:

  • Refik Anadol's studio: Blending art and AI with petaflop performance
  • Creative professionals: Enabling artists and designers to run AI models locally

Enterprise and Specialized Applications:

  • Zipline: Pushing the boundaries of autonomous delivery with edge AI
  • Robotics companies: Developing and testing robotic systems in real-world environments

Use Cases and Applications

DGX Spark enables entirely new workflows across multiple industries:

Development and Research:

  • Rapid prototyping: Test AI models without cloud dependencies
  • Algorithm development: Prototype new AI algorithms with full hardware access
  • Privacy-sensitive applications: Keep data on-premises for secure AI development
  • Multi-modal AI: Experiment with vision, language, and multimodal models

Robotics and Edge AI:

  • Simulation: Run complex robotics simulations locally
  • Autonomous systems: Develop perception and planning for vehicles and drones
  • Industrial automation: Create smart factory solutions
  • Real-time inference: Test edge deployment scenarios

Creative and Enterprise Applications:

  • Generative AI: Run models like FLUX.1 for image generation and content creation
  • Business intelligence: Run AI-powered analytics and document processing locally
  • Quality control: AI-powered visual inspection systems
  • Content production: Accelerate video, image, and audio production workflows

Market Impact and Future Implications

AI Democratization

DGX Spark represents a fundamental shift in AI accessibility, breaking down traditional barriers:

Breaking Down Barriers:

  • No cloud dependency: Run advanced AI without ongoing cloud costs
  • Instant access: Get started immediately without infrastructure setup
  • Data privacy: Keep sensitive data on-premises
  • Cost predictability: One-time hardware cost versus ongoing cloud expenses

Empowering Innovation:

  • Solo developers: Access to supercomputer power without enterprise budgets
  • Researchers: Enable cutting-edge research at any institution
  • Edge AI acceleration: Enables new categories of edge AI applications
  • Global reach: Bringing AI capabilities to underserved markets

Technology Evolution

The launch signals important shifts in AI infrastructure:

Key Trends:

  • Edge intelligence: Moving AI inference and training to the edge
  • Performance density: Continued miniaturization of AI hardware
  • Unified memory: Simplifying AI development workflows
  • Complete ecosystems: Integrated software stacks for AI development

Availability

Launch Details:

  • Official launch: October 15, 2025
  • Distribution: NVIDIA.com and partners (Acer, ASUS, Dell, GIGABYTE, HP, Lenovo, MSI)
  • Target users: AI developers, researchers, creative professionals, robotics engineers

Conclusion

NVIDIA DGX Spark marks a pivotal moment in the democratization of artificial intelligence. By delivering a petaflop of AI performance in a compact, portable form factor, NVIDIA has created a platform that brings supercomputer-class capabilities to anyone with an idea and the ambition to build.

Key Takeaways:

  • Revolutionary form factor: The world's smallest AI supercomputer at just 1.2 kg
  • Petaflop performance: 1 petaflop of AI performance with 128GB unified memory
  • Complete ecosystem: Full NVIDIA AI software stack with frameworks, models, and tools
  • Broad accessibility: Available through major OEM partners starting October 15, 2025
  • Symbolic launch: First delivery to Elon Musk at SpaceX Starbase

The hand-delivery to SpaceX symbolizes something profound: AI infrastructure is no longer confined to massive data centers. Like SpaceX democratized space access, DGX Spark democratizes AI supercomputing—putting transformative technology within reach of everyone building the future.

From creative studios to robotics labs, from university research centers to solo developer workstations, DGX Spark is landing where ideas happen. This isn't just a new product—it's the beginning of a new era where petaflop AI performance is as accessible as a laptop, enabling the next generation of AI innovation from anywhere.

Sources


Want to learn more about AI hardware and infrastructure? Explore our AI glossary for key terms like neural processing unit, check out our AI fundamentals courses to understand how AI models work, or browse our AI tools catalog to discover other AI development platforms.

Frequently Asked Questions

DGX Spark is NVIDIA's smallest AI supercomputer, delivering 1 petaflop of AI performance with 128GB unified memory in a compact form factor the size of origami paper, capable of running models with 200 billion parameters locally.
DGX Spark features the NVIDIA GB10 Grace Blackwell Superchip delivering up to 1 petaflop of AI performance at FP4 precision, 128GB unified CPU-GPU memory, ConnectX networking, NVLink-C2C for 5x PCIe bandwidth, NVMe storage, and HDMI output.
DGX Spark will be generally available starting October 15, 2025, on NVIDIA.com and through partners including Acer, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, and MSI.
The first DGX Spark system was hand-delivered by NVIDIA CEO Jensen Huang to Elon Musk at SpaceX's Starbase facility in Texas, marking the beginning of the DGX Spark rollout.
DGX Spark is designed for developers, researchers, and creators to prototype, fine-tune, and run AI inference locally. It supports customizing image-generation models, building vision agents, deploying chatbots, and running robotics simulations.

Continue Your AI Journey

Explore our lessons and glossary to deepen your understanding.