Around halfway through Jensen Huang’s nearly two-hour COMPUTEX keynote, a certain sense of scale began to settle in.
Not just scale in the usual Nvidia sense (ever larger GPUs and sprawling datacenters) but scale in the form of an entirely new kind of infrastructure.
Jensen’s talk of "AI factories" wasn't simply hyperbole or marketing flash, it marked a meaningful pivot in computing’s trajectory.
He took the audience through decades, tracing Nvidia’s evolution from graphics chips to CUDA-enabled accelerated computing, onward to AI-centric workloads that demand entirely new ways of thinking about compute, storage, and networking.
He likened Nvidia’s AI revolution to historical moments—electricity, the internet—new foundational layers that started abstractly but became indispensable.
Now, he argued, AI infrastructure itself is becoming essential, "an infrastructure of intelligence," embedded everywhere, transforming everything.
But beneath these broad strokes, there emerged a subtler narrative about how these lofty ambitions land inside real businesses.
Around 75 minutes into the keynote, Jensen turned to something a bit more grounded, practical: enterprise data infrastructure.
While more sober than the cool ray-traced graphics or robotic gymnasiums, the enterprise AI topic carries huge, immediate weight. And at its heart sits Nvidia’s new AI-Q platform, purpose-built for the messy world of enterprise data, specifically unstructured and designed for sprawling datasets.
Here, Jensen underscored a critical pivot: traditional SQL databases are neatly organized but AI demands something messier and richer—semantic search, GPU acceleration, and continuous embedding and indexing.
"Humans query structured databases like SQL, but AI wants to query unstructured data. They want semantic. They want meaning. And so we have to create a new type of storage platform."
The data that businesses need most desperately today is anything but neat and structured. It's in emails, PDFs, multimedia repositories, and customer logs, all waiting for meaning to be mined.
Nvidia’s IQ, powered by its NeMo Retriever and Llama NeMoTron reasoning models, promised to address precisely this.
To ground this sweeping vision in practical reality, Jensen highlighted VAST Data, which is already known for upending traditional storage architectures with flash-driven speed and scale.
He explained how, in just three days, VAST leveraged Nvidia’s AI-Q blueprint to build an AI-driven sales research agent. Without flashy demos or grandstanding, he detailed how the system continuously pulled from CRM systems, internal tools, and multimodal repositories, embedding and indexing data on-the-fly for rapid semantic search.
From there, Nvidia’s Llama NeMoTron model stepped in, automatically converting a rough AI-generated outline into a detailed sales execution plan.
Tasks previously consuming days of manual labor suddenly shrank into a matter of minutes. Jensen presented this without drama. “With VAST accelerated AI data platform, organizations can create specialized agents for every employee,” he told the crowd.
Underneath VAST’s rapid AI agent buildout lies a sophisticated data platform specifically architected around Nvidia’s IQ framework.
This isn’t just off-the-shelf hardware acceleration—it's a fully integrated stack leveraging Nvidia’s DGX systems and GPU Direct Storage, fused with VAST's disaggregated shared-everything flash architecture.
Nvidia's NeMo models run in-place on data stored within VAST, allowing semantic queries and vector embedding at speed and scale—without intermediate data copying or network bottlenecks.
And at the heart of it all, Nvidia’s Llama NeMoTron and NeMo Retriever models do the heavy lifting of reasoning and semantic extraction directly where data resides, turning large, cumbersome enterprise datasets into immediately actionable insights.
"Your future storage is no longer CPUs sitting on top of a rack of storage. It's going to be GPUs sitting on top of a rack of storage. And the reason for that is because you need the system to embed, find the meaning in the data, in the unstructured data, in the raw data," Jensen Huang explained to the COMPUTEX crowd.
Underlying the larger narrative of Nvidia’s vision, this example illustrated precisely what "AI infrastructure" might look like in practice—embedded directly into routine tasks, reshaping work, one quietly automated process at a time.
Yet beyond the immediacy of faster sales research, Jensen's keynote also signaled deeper changes coming from the rise of AI agents themselves.
Toward the close of his talk, he explicitly framed agentic AI as more than mere automation tools—they’re emerging as new digital coworkers capable of thinking, planning, reasoning, and collaborating.
This shift is profound because it moves beyond simple automation and into the realm of meaningful collaboration. Jensen reinforced this vision, adding a layer of practicality, explaining that digital agents will help address looming labor shortages by becoming active team members:
"The world has a shortage of labor. We have a shortage of workers by 2030, by about 30 to 50 million shortage. It's actually limiting the world's ability to grow. And so now we have these digital agents that can work with us,” he explained.
"100% of NVIDIA software engineers now have digital agents working with them so that they can help them, assist them in developing better code and more productively,” he added.
And therein lies the essential core of the keynote: the future of computing—far from abstract—is increasingly tangible, immediate, and personal.
The broad themes of AI factories and vast infrastructures inevitably distill down to small, daily revolutions—like VAST’s quietly transformative sales research agent. J
Jensen’s words underscored how these practical examples point toward a profound and ubiquitous change in how we all will soon work, create, and collaborate.
And perhaps that subtle but steady shift—the integration of AI into the texture of daily life—is exactly the quiet revolution Jensen aimed to convey.