• NVIDIA CEO Jensen Huang frames AI as the third major computing paradigm shift, positioning the company as essential infrastructure for an emerging 'token economy.'
  • The company is advancing 'agentic AI' systems and robotics, while major partners like Samsung and South Korea commit to deploying hundreds of thousands of NVIDIA GPUs.
  • Enterprise adoption accelerates as industry leaders, including Disney, Eli Lilly, and TSMC, transition to NVIDIA's AI-focused infrastructure solutions.

At the heart of Jensen Huang's keynote at Computex 2025 was a fundamental reimagining of the data center. The NVIDIA CEO declared that the era of general-purpose computing is giving way to a new paradigm where data centers function as "AI factories," manufacturing intelligence as their primary output. This vision, which Huang described as the third major shift after personal computing and the internet, reframes computational investment directly to business output through a nascent "token economy." Companies are beginning to measure their AI output by tokens produced, much like a factory measures widgets, according to Huang's presentation.

The strategic implications are vast, and NVIDIA is building the hardware and software stack to enable it. The recently unveiled Blackwell GPU architecture is designed for adaptability across the entire AI lifecycle, from pre-training massive models to running complex healthcare diagnostics. For enterprises, the shift is already underway. NVIDIA RTX PRO Servers are being deployed by a roster of global giants—from Disney and Foxconn to pharmaceutical leader Eli Lilly and chipmaker TSMC—to replace general-purpose clusters with AI-optimized infrastructure. "We are transitioning from IT infrastructure to AI infrastructure," said an executive at a major adopter, who spoke on condition of anonymity because the deployment details are private.

Partnerships underscore the scale of this industrial shift. NVIDIA and Samsung are jointly building an AI Factory powered by 50,000 NVIDIA GPUs, focused on intelligent manufacturing and optimizing chip production itself. In a broader national commitment, South Korea plans to deploy over 250,000 NVIDIA GPUs to bolster its AI capabilities, people familiar with the matter said. This demand is fueled by what Huang terms "agentic AI"—systems capable of perception, reasoning, and action that act as digital collaborators. NVIDIA's Nemo framework and Cosmos synthetic data platform are tools to build and train these agentic systems and the next generation of robots, aiming to solve the critical bottleneck of limited real-world training data.

For the average developer, the barrier to entry is also falling. Project Digits, NVIDIA's compact AI supercomputer, is designed to democratize access to high-performance AI development. Meanwhile, the new GeForce RTX 5090 brings staggering performance gains for running popular generative AI models locally. The cumulative effect, Huang argues, is that AI will become as fundamental as electricity, with organizations across every sector—from healthcare and logistics to entertainment and drug discovery—needing to rearchitect their computational foundations to remain competitive. The race to build and leverage these 'AI factories' is no longer speculative; it's the central project of the next decade.

Correction: An earlier version of this article misstated the number of GPUs in the Samsung AI Factory partnership. It is 50,000 GPUs.