NVIDIA GTC just wrapped in San Jose. Over 30,000 people from 190 countries descended on the SAP Center for four days. Jensen Huang delivered a two-hour-plus keynote to a packed arena. The event has been called the "Woodstock of AI" and the "Super Bowl of the Tech Industry," and at this point, both labels undersell it.

This isn't a developer conference anymore. GTC has become the most important annual event in tech, full stop. And what that says about where real power sits in the AI era is worth paying attention to.

The conference that eclipsed all others

Think about the events that used to define the tech calendar. Apple's WWDC. Google I/O. CES. These were the moments when the industry held its breath. Steve Jobs pulling an iPhone out of his pocket. Sundar Pichai demoing a new Android version. The spectacle was software, and software was king.

GTC has quietly overtaken all of them. Not in flash, but in gravity. When Jensen Huang takes the stage, the audience isn't just developers. It's heads of state, Fortune 500 CEOs, Wall Street analysts, and automotive executives. MotorTrend named Huang their 2026 Person of the Year. A car magazine. For a chip company CEO.

That shift tells you something fundamental. The center of gravity in tech has moved from the application layer to the infrastructure layer. The conferences that matter most are no longer about what you can build on top of the platform. They're about the platform itself.

Jensen Huang is not a software founder

The most influential person in tech right now didn't build a social network, a search engine, or a smartphone. Jensen Huang co-founded NVIDIA in 1993 at a Denny's in San Jose with a bet that dedicated graphics processing would be central to computing's future. That bet took decades to fully pay off.

Today, NVIDIA is worth roughly $4.5 trillion, making it the most valuable public company in the world. Huang has collected TIME Person of the Year, the IEEE Medal of Honor, the Queen Elizabeth Prize for Engineering, and a Financial Times Person of the Year nod. He's not just a CEO giving a keynote. He's become the face of an entire technological era.

What makes this remarkable is that Huang's power comes from selling infrastructure, not consumer products. He doesn't have a device in your pocket or an app on your home screen. NVIDIA's influence is one layer removed from the end user, and yet it's arguably more decisive than anything happening at the application layer. Every major AI model, every autonomous vehicle stack, every agentic system runs on NVIDIA's platform. When Huang says "we are a vertically integrated computing company that runs all of AI," it's not hyperbole. It's closer to a statement of fact.

Own the picks and shovels, then own the narrative

NVIDIA's playbook has been remarkably consistent: own the foundational layer, then use that position to shape the story.

At GTC 2026, Huang announced that purchase orders for Blackwell and Vera Rubin systems are on track to reach $1 trillion through 2027, doubling NVIDIA's earlier projection of $500 billion. The company has now posted eleven consecutive quarters of revenue growth above 55%. Revenue this quarter alone is expected to hit roughly $78 billion, up 77% year over year.

But the numbers are only part of the story. What NVIDIA does better than almost any company in tech is frame what those numbers mean. Every keynote isn't just a product launch. It's a narrative about where computing is headed, and NVIDIA positions itself as the inevitable center of that future.

This is the playbook Apple perfected in the Jobs era: don't just ship products, define the category. The difference is that Apple did it with consumer hardware people could hold. NVIDIA is doing it with data center infrastructure most people will never see.

The "AI factory" is genius positioning

Perhaps the most revealing move in NVIDIA's narrative strategy is the rebranding of data centers as "AI factories."

"These are not data centers of the past," Huang has said. "They are, in fact, AI factories. You apply energy to it, and it produces something incredibly valuable, and these things are called tokens."

This isn't just marketing. It's a politically astute reframing that aligns Silicon Valley's infrastructure ambitions with the language of industrial policy. The word "factory" evokes physical production, jobs, and national competitiveness. It makes a GPU cluster sound like a steel mill. And in an era where governments are pouring billions into AI infrastructure subsidies, that framing matters enormously.

As Nathan Benaich of Air Street Capital has pointed out, the metaphor is deliberate. Traditional factories produce physical goods. Power plants, fiber optic networks, and legacy data centers don't get called factories. But by adopting industrial language, NVIDIA and its allies give policymakers cover to treat AI infrastructure spending as a form of domestic manufacturing investment.

The framing also captures something real about how the economics of AI are shifting. GTC 2026's central theme was the "inference inflection point," the moment when AI moves from training models to deploying them at scale. In this framing, AI factories don't just store data. They produce intelligence, continuously generating the tokens that power agentic systems, autonomous vehicles, and enterprise tools. Analysts estimate that for every dollar spent on NVIDIA chips, an $8 to $10 multiplier effect ripples across the ecosystem. That's factory economics.

What happens when a chip company commands this kind of attention

There's something worth sitting with here. We're living in a moment where a semiconductor company commands more cultural attention than any software company on Earth. The last time anything like this happened was Intel in the 1990s, and even "Intel Inside" never reached this level.

The comparison to Apple's peak keynote era is instructive. In the early 2010s, Apple events were cultural phenomena because the products Apple announced directly changed how hundreds of millions of people lived their daily lives. You could hold the new iPhone in your hand.

GTC operates on a different register. Most people will never touch an NVIDIA GPU or walk through an AI factory. But the products built on that infrastructure, the AI agents, the autonomous vehicles, the reasoning systems, will reshape daily life just as profoundly. The power has moved upstream, from the thing you use to the thing that makes the thing you use.

This is what makes GTC feel so consequential. It's the event where the actual constraints on what's possible get redefined. When Huang unveils Vera Rubin's 10x performance-per-watt improvement over Grace Blackwell, or demonstrates how the new Feynman architecture integrates CPUs, DPUs, and LPUs into a unified agentic computing stack, he's not announcing a feature. He's raising or lowering the ceiling for every AI company on the planet.

The question nobody wants to ask

Is this permanent?

NVIDIA's dominance is real, but the history of tech is littered with companies that owned a critical layer and then watched it get commoditized. Intel dominated for decades until it didn't. Cisco owned the network layer until cloud abstractions made the underlying hardware less relevant.

The bull case for NVIDIA is that its moat isn't just hardware. It's the full stack: CUDA, the software ecosystem, the developer community, the networking (Mellanox), and now even custom silicon through the Groq acquisition. NVIDIA isn't selling chips. It's selling a vertically integrated platform that's increasingly difficult to replicate.

The bear case is that every dominant platform eventually faces disaggregation. Custom silicon from hyperscalers (Google's TPUs, Amazon's Trainium, Microsoft's Maia) is advancing. Open-source software stacks could erode CUDA lock-in. And if inference workloads eventually become more commoditized than training workloads, the premium NVIDIA commands could narrow.

For now, though, there's no real alternative at scale. And GTC 2026 made that abundantly clear.

What this means for builders

If you're a developer, a founder, or anyone building in the AI space, GTC's ascendance carries a practical message: the platform you build on matters more than what you build.

In the software era, the platform was relatively neutral. You could build a great app on iOS or Android or the web, and switching costs were manageable. In the AI era, your choice of compute infrastructure shapes what's even possible. Model architecture, inference latency, cost per token, these are all downstream of hardware decisions. And right now, those hardware decisions overwhelmingly lead back to NVIDIA.

That's not necessarily comfortable. Dependency on a single vendor rarely is. But it reflects the reality of where we are in the AI infrastructure cycle. The companies that understand this, and build their strategies around infrastructure constraints rather than pretending they don't exist, will be better positioned than those that don't.

GTC is NVIDIA's Super Bowl. But more importantly, it's become the event where the boundaries of the AI era get drawn. Whether those boundaries end up being permanent or temporary, right now, nobody else is in a position to draw them.

References