It’s that time of year again. The eyes of the world were affixed on downtown San Jose to pay homage to Jensen Huang and the team at Nvidia. The four-day conference marked a victory lap for the company with the world’s largest market cap, with talk of partners and cost savings largely outweighing the tech at times.

Where the event was once Christmas for computing nerds, GTC has since morphed into an effective showcase for investors. Huang himself even made reference to this during his over two-hour-long keynote, when he said that the largest percentage of GTC attendees were from the financial services industry, adding, “I'm hoping it's developers, not traders.”

Top of line in terms of announcements was our first look at the Nvidia Groq 3 language processing unit (LPU), marking the first joint offering since its licensing deal come acqui-hire struck last Christmas Eve.

In hardware terms, the LPU is not a replacement for its graphics processing unit (GPU) line but instead a complementary card designed to dramatically lower latency for AI inference workloads.

With just one-five-hundredth the amount of capacity per chip compared to its GPU line, per data center head Ian Buck, it provides up to 150 Tb/s of bandwidth, with Nvidia positioning it as a means to perform decoding operations for GPUs.

Nvidia has tried to hype up the chip, with Huang going as far in the weeks leading up to the event saying the company had a chip that would “surprise people.”

But despite being its flagship unveiling, the LPU was ultimately an underwhelming showcase. Everyone knew it was coming, though that didn’t stop a few in Reddit forums and Discord channels from speculating about a potential silicon photonic chip not seen before. But for the majority, this was all but expected.

And in a way, to the CEO’s point on traders, GTC has now become this safe space for Nvidia. With its annual roadmap or "rhythm" laid out for all to see years in advance, the excitement for actual announcements was replaced by clapping for stuff we’ve already seen before, like Kyber.

Jensen really loves OpenClaw

If the ultimate end-game of showcasing the Groq 3 LPU was a safe flagship product reveal, then its industry position makes a bit more sense when you consider what themes were dominant throughout GTC: agentic AI.

Nvidia went all in on the latter, culminating in a security wrap-around offering for OpenClaw, dubbed NemoClaw.

Huang was overly complimentary of OpenClaw – which enables autonomous AI agents to run continuously on local devices or servers – going as far to claim in his keynote that it’s impact is “no different than how Windows made it possible for us to create personal computers.”

“This is the new computer now,” the Nvidia CEO said, even claiming it would revolutionize both enterprise computing and the whole IT industry.

“Every single IT company, every single company, every SaaS (software-as-a-service) company will become a GaaS company, no question about it. A ‘generate-as-a-service’ company. And what's amazing is this now OpenClaw gave us, gave the industry exactly what it needed at exactly the time, just as Linux gave the industry exactly what it needed exactly the time, just as Kubernetes," Huang said.

Jensen’s GTC love-in with OpenClaw was the flagship part of his company’s wider agentic efforts, unveiling a reworked Agent Toolkit suite to all firms to build AI agents from a host more easily, and open models to provide the underlying brains for those agents, such as Nemotron 3 Omni.

In fact, during his keynote speech, "agentic" was mentioned only a humble 18 times, while agents was said 45 times – more than inference at 29 mentions.

The decision to double down on the hot new thing the industry only really began talking seriously about last summer makes sense for Nvidia. The more agents there are being deployed and the more folks using its related software, the more hardware Nvidia can sell and the more it can then invest in companies designing systems to power those agents, and the cycle continues.

“Every single enterprise company, every single software company in the world, needs an agentic strategy. You need to have an OpenClaw strategy,” Huang said in his keynote. “Every single software company of the future will be agentic, and they will be token manufacturers. There'll be token users for their engineers, and there'll be token manufacturers for all of their customers.”

“The OpenClaw event cannot be understated," Huang added. "This is as big a deal as HTML. This is as big a deal as Linux.”

From chip designer to ‘vertically integrated & horizontally open’

The end of the keynote was almost the antithesis of this year’s GTC. It saw Jensen sit down on the floor, gazing up at video footage of all the company has achieved – while next to a robotic Olaf from Frozen.

If the semantics of that image are lost on you, just remind that build up to the keynote was labeled as the pre-game, while repeatedly throughout the four days the moniker of the "Super Bowl of AI" was bandied about.

As Jensen sat on that stage, he knew fully well that Nvidia are kings of the world. He even proclaimed they were the “top networking company in the world.”

But to be the best, you’ve got to be the best at everything. And the conference covered the entirety of the AI stack. Not just chips, but platforms.

“We are now a computing platform that runs all of AI,” Huang told attendees. “Our business already started to show that, 60% of our business is hyperscalers, the top five hyperscalers.”

But instead of calling themselves a platform company, he likened his business to being “the world's first vertically integrated, but horizontally open company.”

“We'll work and integrate Nvidia's technology into whatever platform you would like us to integrate into,” Huang said. “We offer you the software, we offer you libraries, we integrate with your technology so that we can bring accelerated computing to everybody in the world.”

Ultimately, the optics of GTC served to cement Huang and co’s core message: Nvidia is no longer just a chip designer, but the indispensable, integrated, and open computing platform for the entire AI ecosystem, led by a man whose strategic vision has eclipsed the need for once-simple hardware surprises.