NVIDIA’s GTC 2025 A glimpse into our AI-powered future

I just got back from NVIDIA’s GTC event in Washington, DC (Oct 26-29), and my head is still spinning. If you’re trying to understand where the tech world is headed, this was the place to be. There weren’t a ton of flashy new product reveals, but what I saw felt even more significant. It was a masterclass in strategy, showing how NVIDIA is cementing itself not just as a chip company, but as the bedrock of the entire modern economy.

The energy was different this year, especially with Jensen Huang making his first keynote appearance at the DC edition. The message was clear: the AI revolution is here, it’s massive, and it’s being built right here in America.

Let me break down what stood out to me.

Raul Leite

The staggering scale of it all

The most mind-boggling number tossed out was $500 billion. That’s the cumulative revenue Jensen expects from their Blackwell and next-gen Rubin platforms through next year. Let that sink in for a moment. He mentioned they’re planning on shipping about 20 million of these GPUs by the end of 2026. What really put it in perspective for me was his comment that this level of financial foresight is unheard of in tech. And get this, that colossal figure doesn’t even include the Chinese market due to export controls. This is just a demand from “the West.” It’s a clear signal that the hunger for AI compute is nowhere near satisfied.

They’re also sticking to a relentless annual release cadence. While Blackwell Ultra is already out the door, Rubin is already on the horizon for next year. The scale is moving from single servers to entire racks, like the Vera Rubin NVL144, which essentially crams 144 powerful GPUs into one cohesive unit. To keep all that data flowing, they also unveiled the BlueField-4 DPU, which is a monster of a data processor. The infrastructure being built today is almost hard to comprehend.

Raul Leite

The partnerships defining the next decade

This event was less about a solo act and more about a symphony of alliances. NVIDIA is planting its flag everywhere.

The US government is all-in: The collaboration with Oracle and HPE to build seven new AI supercomputers for the Department of Energy is a huge deal. The crown jewel is “Solstice,” which will be a 100,000-GPU Blackwell beast. This isn’t just for science; it’s a direct investment in national security and technological sovereignty.

A $1 billion bet on 6G: The telecom world is next. NVIDIA is investing a cool billion in Nokia to fundamentally reshape cellular networks for the 6G future. They’re working with T-Mobile to start trials next year, which tells me this isn’t just a distant dream — it’s a concrete plan.

Enterprise AI gets real: I saw how companies like CrowdStrike and Palantir are building deeply integrated AI agents for everything from cybersecurity to managing global supply chains. And in a “wow” moment, Uber announced plans for a 100,000-strong fleet of robotaxis using NVIDIA’s platform, starting in 2027. This stuff is moving from lab to reality, fast.

Raul Leite

Invented in America. Built in America

A powerful, recurring theme was the rebirth of American manufacturing. Jensen was very direct, praising the pro-energy and pro-manufacturing focus of the current administration. He argued that to win in tech, we need the energy and the industrial base to support it.

He then laid out a concrete US supply chain:

Blackwell chips will be fabricated at TSMC’s new plant in Arizona.

Assembly will happen at a Foxconn facility in Texas.

The high-bandwidth memory chips will come from a factory in Indiana.

This is a monumental shift from the old model. They’re also using their Omniverse technology to help partners design the hyper-efficient, robotic factories of the future. It feels like a true industrial rebirth, powered by AI.

Raul Leite

A quantum computer

Where my world fits in: The open-source foundation

As someone deeply invested in the open-source ecosystem, I was particularly focused on Red Hat’s role at NVIDIA GTC this year. Seeing our presence there really reinforced something I’ve believed for a long time: All this powerful hardware needs an equally powerful, flexible, and secure software foundation. Our message was simple and reinforced as usual: “We’re better together.” And it couldn’t be truer. There’s a greater than 95% chance that our AI solutions will run on NVIDIA hardware, so our mission is to make that partnership seamless and secure.

As one of the sponsors at GTC DC 2025, Red Hat showcased its commitment to the AI ecosystem and positioned itself as a leader in the modern AI and computing infrastructure space. Our strategy centers around the message “Red Hat & NVIDIA: Better Together” focusing on complementing NVIDIA’s innovation and accelerating adoption by providing a faster, more reliable path to production.

Our goals for the event were clear: increase awareness of our AI strategy, strengthen relationships with NVIDIA customers and partners, and demonstrate how open source can be the foundation for this new wave of intelligent infrastructure. It’s evident that for the new American industrial base to thrive, it needs a robust, open, and certified software platform. That’s the role we’re committed to playing, ensuring that from government supercomputers to the factory floor, the foundation is solid, secure, and ready for what’s next.

Raul Leite

During the GTC week, Red Hat announced two key collaborations with NVIDIA aimed at making AI development simpler and more secure. First, the NVIDIA CUDA Toolkit is now available directly through Red Hat platforms RHEL, OpenShift and Red Hat AI, giving developers a single, trusted source for essential GPU tools. Second, Red Hat introduced the STIG-hardened Universal Base Image (UBI-STIG), which NVIDIA is using to build a government-ready GPU Operator, helping agencies accelerate secure AI and ML deployments.

Walking out of the convention center, one thing was unmistakable: We’re no longer just talking about AI, we’re building it. And the scale and speed of what’s happening are unlike anything I’ve seen before. It’s going to be a fascinating few years ahead.

This article is published as part of the Foundry Expert Contributor Network.Want to join?