Investing.com — Nvidia (NASDAQ:NVDA), the world’s most valuable company and the face of the AI revolution, is bringing its supercomputing power to the desktop.

The company’s DGX Spark, a pint-sized AI system that TIME Magazine named one of the “Best Inventions of 2025,” will go on general sale Oct. 15. The launch marks another step in Nvidia’s push to democratize high-performance computing, bringing data center–level performance to the desktop.

At its core is the GB10 Grace Blackwell Superchip, capable of delivering up to one petaFLOPS of performance, roughly a quadrillion floating-point operations per second, a term used to quantify compute power. The Spark integrates ConnectX-7 high-speed networking and Nvidia’s full AI software stack, effectively giving startups, researchers, and developers plug-and-play access to industrial-grade compute power.

Unveiled at Jensen Huang’s GTC 2025 keynote, the DGX Spark was presented as Nvidia’s answer to the growing demands of “agentic AI,” a new class of reasoning systems that think, plan, and can act autonomously. Holding the small box in the palm of his hand, Huang said DGX Spark and its workstation counterpart, DGX Station, were “created from the ground up to power the next generation of AI research and development.”

With 20 CPU cores and 128 GB of unified GPU memory, Spark’s hardware is tuned for real-world AI work. Nvidia says users can fine-tune models with up to 70 billion parameters, run inference locally, and keep sensitive data entirely on-premise, all without relying on cloud infrastructure.

It’s also designed to fit neatly into developers’ workflows: wired and wireless networking, Bluetooth peripherals, and even the option to link two Sparks into a mini-cluster. “You can literally create your own personal cloud,” said Allen Bourgoyne, Nvidia’s director of product marketing.

Nvidia frames the DGX Spark around four key use cases:

  • Prototyping next-generation AI agents and chatbots

  • Fine-tuning medium-to-large models locally

  • Inference and testing without external dependencies

  • Data security, keeping information private and on-site

The Spark’s release comes at a time when the boundary between personal and enterprise computing is blurring. As AI models evolve beyond text and vision to reasoning and autonomy, compute needs are scaling faster than cloud capacity can keep up. Nvidia’s bet is that bringing supercomputing closer to the user will be essential to sustain that pace.

“There’s a clear shift among consumers and enterprises to prioritize systems that can handle the next generation of intelligent workloads,” said Michael Dell, CEO of Dell Technologies, underscoring the shift in computational needs as AI rapidly advances.