Nvidia Has the Most Important Stage in Tech. GTC’s Keynote Has Not Caught Up.

GTC used to be a developer conference. A gathering of engineers, researchers, and builders who spoke GPU fluently and didn’t need a translator. Jensen Huang was in his element, deep in the stack, talking to his people.

That version of GTC is gone.

The signal is right there on the conference floor. Lenovo, Dell, Cisco, Micron, HPE to name a few of the brands that all showed up to GTC 2026 not just as attendees but as participants, making announcements of their own tied directly to the Nvidia platform. Server lineups, AI factory infrastructure, Vera Rubin-based systems, all of it timed to the GTC stage because that is where the audience now is. When the biggest names in enterprise hardware treat your developer conference as a launch platform, you are no longer running a developer conference. You are running the most important stage in tech.

Nvidia is now the most watched brand in tech when it comes to AI. Not one of them. The one. That changes what the GTC keynote is and, more importantly, what it needs to be. The audience in that room and the far larger one watching from outside it includes policymakers, investors, enterprise buyers, journalists, and a broader public trying to understand what this technology is going to mean for their jobs, their cities, and their lives. The stage is bigger. The responsibility that comes with standing on it is bigger too.

Jensen delivered. On the tech, he almost always does. The Vera Rubin platform, the Groq 3 LPU integration, the roadmap out to Feynman, it was dense, ambitious, and genuinely significant. But two and a half hours is not the same thing as depth, and throughput is not the same thing as thought leadership.

Here is the gap: Nvidia is shaping the trajectory of AI more than almost any other company on the planet. The chips, the software stack, the developer ecosystem, it all runs through Santa Clara. That kind of structural influence carries an obligation to engage with what the technology is actually doing in the world, not just what it can do in a benchmark.

That obligation was largely absent from the keynote.

Take one concrete example. The integration of the Groq 3 LPU into the Vera Rubin platform came with an eye-catching efficiency claim: the combined system delivers up to 35 times higher inference throughput per megawatt compared to Blackwell. Jensen framed it as proof of Nvidia’s engineering discipline. And it is a real gain. Performance per watt is a legitimate metric and Nvidia deserves credit for the progress.

But performance per watt as an efficiency argument only holds if you stop the analysis at the chip level. The broader picture, the one that matters to anyone tracking what AI infrastructure is doing to energy grids, water systems, and carbon commitments, tells a more complicated story. AI’s appetite for compute is growing faster than efficiency gains are shrinking it. Jensen himself forecast at least one trillion dollars in orders for Blackwell and Vera Rubin systems through 2027. That is not a number that points toward a smaller footprint. More efficient per unit, yes. Smaller in aggregate, no.

This is not a hyperscaler problem to manage in isolation, and it should not be treated as one. It is a systemic challenge, and a company that supplies the engines of that system has standing, and arguably a responsibility, to engage with it honestly. If you are going to lead with tokens per megawatt as a selling point, you owe the audience a reckoning with what a trillion dollars of infrastructure buildout actually means for the grid.

The familiar deflection is some version of “we build the technology, others decide how to use it.” That framing worked in a different era. It does not hold when you are the foundational layer of an industry reshaping every other industry. The picks and shovels provider in a gold rush still has a relationship with what the rush produces. Neutrality is a position, and at Nvidia’s scale, it is an increasingly costly one to maintain.

None of this diminishes what Nvidia has built or what GTC represents as a technical showcase. The point is that the moment is calling for something more than a showcase. The companies that will define the next decade of tech are not just the ones building the most powerful tools. They are the ones that demonstrate they understand the weight of what they have built and are willing to engage with the world it is creating.

The fact that Dell, HPE, Lenovo, and Cisco now orbit Nvidia’s stage says everything about the platform Huang has built. The question from this keynote is whether Nvidia is ready to lead in proportion to it.

Nvidia has the platform. The question is whether it has the appetite.