Artificial Intelligence (AI)

Nvidia’s Personal AI Supercomputers Are Here

Ken Metral

Ken Metral

— 2 min read
Cosmico - Nvidia’s Personal AI Supercomputers Are Here
Credit: DGX Spark/DGX Station/Nvidia, Inc.

At GTC 2025, Nvidia CEO Jensen Huang introduced two groundbreaking systems that redefine what personal computing looks like in the age of AI: DGX Spark and DGX Station. Dubbed “personal AI supercomputers,” these systems are powered by Nvidia’s cutting-edge Grace Blackwell platform and represent a new category of desktop machines built specifically for running large-scale AI models locally.

“AI has transformed every layer of the computing stack. It stands to reason a new class of computers would emerge—designed for AI-native developers and to run AI-native applications,” said Huang during the keynote.

Built for AI, From the Ground Up

These systems aren’t just high-end desktops—they’re AI-native machines purpose-built for:

  • Developers
  • Researchers
  • Data scientists

Anyone looking to prototype, fine-tune, and deploy large AI models locally can now do so without relying solely on cloud infrastructure.

The DGX architecture also makes it easy to scale: developers can build models on these machines and move them seamlessly to DGX Cloud or other AI cloud platforms with minimal changes.

DGX Spark: Compact Power

The DGX Spark is the smaller, more accessible of the two:

  • Powered by the GB10 Grace Blackwell Superchip
  • Features Blackwell GPUs and 5th-gen Tensor Cores
  • Capable of 1,000 trillion operations per second

This system delivers serious performance in a desktop-friendly form—ideal for those who want to start building with AI without setting up a full data center.

DGX Station: Desktop Beast Mode

For those who need even more muscle, there’s the DGX Station:

  • Equipped with the GB300 Grace Blackwell Ultra Superchip
  • Includes 784GB of unified memory
  • Networking powered by ConnectX-8 SuperNIC, supporting up to 800Gbps

This machine is designed for high-throughput AI workloads, model training, and real-time inference at the cutting edge.

Built by the Biggest Names

Nvidia will license the DGX architecture to top-tier manufacturers, including:

  • Asus
  • Dell
  • HP
  • Lenovo
  • BOXX, Lambda, and Supermicro (for DGX Station)

DGX Spark reservations are open now, while DGX Station units are expected to ship later in 2025.

Pricing

While Nvidia didn’t disclose pricing during the keynote, earlier estimates suggest that entry-level DGX Spark systems may start around $3,000. Expect DGX Station to come in at a significantly higher price, given its specs and target market.

The Future of Local AI Development

With DGX Spark and DGX Station, Nvidia is laying the foundation for a future where AI computing isn’t just in the cloud—it’s on your desk. These systems promise to democratize access to powerful AI tools, enabling developers and researchers to iterate faster, own their workflows, and build AI-native apps locally.

It’s a bold move—and one that could reshape how and where AI innovation happens.

Read more