Nvidia GTC Conference: NemoClaw, Autonomous AI Cars + Vera Rubin

NVIDIA's trillion dollar AI vision at their GTC conference. Here's a recap of everything that happened. So Jensen Huang walked onto a stage in San Jose, looked out at 30,000 people, packed into the SAP center and said NVIDIA expects a trillion dollars in purchase orders for its AI chips through 2027. A trillion dollars.

Then he spent the next two hours explaining why that number might be too low. I want you to sit with that for a second because when I tell you about trillion dollar forecasts, it can sound kind of like a number that just gets thrown around in tech, right? Like, oh, another big number from a big company. But this is different.

This is NVIDIA, the company that already supplies the backbone of basically every AI system you've ever used. Chachapati runs on NVIDIA. Claude runs on NVIDIA. Google's AI runs on NVIDIA chips, right?

And Jensen Huang just said demand is so high, so overwhelming, that customers are literally constrained only by how fast NVIDIA can build the hardware. If they could just get more chips, they would generate more revenue. That is the world that we're living in right now. And what I want to do today is just break down everything that happened at their GCC conference that just happened a few hours ago.

NVIDIA's annual developer conference in plain English, because there are about nine major announcements. I'm sure more will come out. And most of the coverage is so technical that a normal person reads it and feels lost. I want to tell you what it actually means, not just for tech companies, but for you, for your job, for your business, and for the next five years of your life.

Let's start at the very beginning with a chip. So Vera Rubin is the NVIDIA's new AI supercomputer platform, right? It's named after the astronomer Vera Rubin, who discovered dark matter by looking at things other people missed. Jensen Huang loves naming his hardware after scientists, and the names matter.

They tell you something about what technology and what that specific technology is supposed to do. Vera Rubin, the astronomer, revealed hidden structure in the universe. Vera Rubin, the chip, is designed to power the era of what NVIDIA calls agentic AI. Autonomous systems capable of complex reasoning, planning, and physical interaction.

Here's what that means in plain language. The last generation of NVIDIA chips, the Blackwell chips, were built to train and run large AI models, right? So big, powerful, but basically doing one thing at a time. Vera Rubin is quite different.

So it's built for AI agents. AI that doesn't just answer a question, it actually goes out into the world, takes actions, completes tasks, talks to other AI systems, runs for hours or days without stopping. That's fundamentally a different kind of computing than we've ever had before. Vera Rubin uses NVIDIA's 3NM process from TSMC.

It has 336 billion transistors, nearly double the density of its predecessor, and it includes HBM4 memory with 22 terabytes per second of memory bandwidth. Now, that sounds complicated. It sounds like technical numbers. Here's a translation.

Think of memory like and a memory bandwidth like a highway, right? The old chips had a very high, very wide highway, but there were traffic jams when too many cars tried to get through at once. HBM4 is like quadrupling the number of lanes and doubling the speed limit at the same time. The AI can think faster, run bigger models, and handle more complexity all at the same time.

Vera Rubin is about to deliver and able to deliver 700 million tokens per second compared to just 2 million on the previous hopper generation. 700 million tokens per second versus 2 million. It's not an upgrade. That's a different category of machine.

And here's what that actually means for you. Every time you run an AI tool, chat GPT, perplexi, quad, whatever, the AI is generating tokens, words, ideas, code. The faster it can do that, the more it can do, the cheaper it gets, and the more useful it becomes. When NVIDIA says 700 million tokens per second, what they're really saying is this.

The AI of 2026 is about to be dramatically more capable than the AI of 2025. And the AI of 2025 already felt like it changed everything. Vera Rubin is a new full stack computing platform comprising seven chips, five rack scale systems, and one supercomputer for AI. Seven chips with five rack scale systems and one complete supercomputer.

And the first one is already running. The first Vera Rubin system is already running in Microsoft's Azure cloud. It's not coming soon. It's here right now.

Now let's talk about the 20 billion bet, the 20 billion dollar bet that most people are sleeping on. Late last year, NVIDIA spent 20 billion dollars, its largest deal ever to acquire Grok. Not the Grok AI from Elon Musk's ex, Grok as in G-R-O-Q, a chip company founded by the people who built Google's in-house AI processors. The Grok 3 LPU is built to enhance NVIDIA's technology, with one core optimized for speeding up the GPU.

NVIDIA introduced a full rack dedicated to housing the new Grok accelerators. The Grok 3 LPX rack holds 256 LPUs and is meant to sit beside the Vera Rubin rack scale system. Why would that matter so much? Well, because until now NVIDIA owned the training C-device, so they had the best chips for building AI models.

But there's another side of this, which is inference. Inference is when you actually use AI. When you type a question, it answers back. Inference has a different challenge.

It needs to be fast, low latency, real time. The Grok LPU is purpose built for exactly that. NVIDIA says the addition of Grok's hardware allows them to deliver 35 times more performance at extremely high token per second rates. The Grok LPUs are deterministic.

Everything is scheduled in software. NVIDIA uses Grok as an accelerator for Vera Rubin, with Grok handling the decode, while Vera Rubin handles the pre-fill. In simple English, NVIDIA now owns both sides of the equation, training and inference, building the AI and running the AI. That is a level of virtual integration that no other company on earth has right now.

And they just demonstrated it with a $20 billion hardware purchase. I keep saying that the AI infrastructure race is the most important thing happening in business right now. Most people don't see it happening because it's going on in data centers and server racks, not in apps on their phone. But the companies building this infrastructure are making bets that will pay off or not over the next decade.

And NVIDIA just made the single biggest bet in the history of this race. AWS and NVIDIA are deploying more than 1 million NVIDIA GPUs starting this year across AWS's global cloud regions. NVIDIA was the first hyperscaler to power up Vera Rubin. 1 million GPUs through Amazon Web Services alone.

Let that land for a second. AWS is one cloud provider, just one. And they're committing to over a million chips from NVIDIA. That tells you something and pretty much everything about where the demand is going.

Now, let's talk about the thing that made Jensen Huang look slightly crazy to some people in the room and honestly kind of thrilled to everyone else. NVIDIA is going to space. Huang said Vera Rubin is headed to space. The goal is to start data centers with Vera Rubin space one.

Now, this is quite an interesting one. So what this means essentially is, and this was just announced at the GTC as well, is that they're working on how to deal with radiation right now. So data centers in orbit around Earth. Now, I want to acknowledge something here.

Your first instinct might be to laugh this off. Jensen Huang announced space data centers. Sounds like a joke, right? Sounds like something from a sci-fi movie.

And I get that, but here's why I take it seriously. Think about what drove the last wave of data centers. It was the internet. The internet created demand for compute so large that we had to build massive warehouses full of servers around the world to keep it up, right?

We put them near rivers for cooling. We put them in cold countries to save on energy. Optimized every inch of those buildings to run as efficiently as possible. AI is doing the same thing to compute right now, but at a scale that makes the internet wave look small.

And one of the biggest limiting factors for AI data centers right now is power. You need enormous amounts of electricity and global data centers are struggling with the dual pressures of skyrocketing AI demand and fixed power grids. Space solves a lot of those problems. You have near unlimited solar power.

You have the vacuum of space for cooling. You have no ground level infrastructure constraints. The physics of running a data center in orbit are genuinely better than running on the ground. If you can solve the engineering challenges.

NVIDIA says they're solving the engineering challenges and given their track record, I'm not going to bet against it, right? Now, let's talk about the announcement that's going to affect your life the most, even if you've never heard of it before today. OpenCore. OpenCore is an open source platform built by a developer named Peter Steinberger.

It lets you create agents. They go off and implement stuff, right? Now, roughly two hours into Jensen Huang's keynote, Huang turned to the phenomenon of OpenCore, which was launched in January and it surged to millions of uses across the world, right? It's one of the biggest open source projects of all time.

Jensen Huang called OpenCore the operating system of personal AI. He compared it to what Windows was for personal computers. Think about that comparison. Before Windows, computers were for engineers and programmers.

After Windows, your grandma could use one. Jensen is saying OpenCore is doing the same thing for AI agents. He's taking something that used to require deep technical knowledge and made it accessible to anyone. Now, NVIDIA has an enterprise version of it.

They're calling it NemoCore. NemoCore is enterprise secure, helping protect sensitive information. AI agents can communicate externally and execute without intervention. NemoCore provides a reference software stack for businesses to keep OpenCore secure.

It's why that matters so much. The version of OpenCore that went viral is great, but it had a problem, right? Security. When your AI agent is running around your computer, connecting to external systems, reading emails, you need to know that information isn't leaking out somewhere, right?

For a big company, that's a huge risk to take. Now, NemoCore solves that. It's the enterprise safe version. The version a company can actually deploy at scale without their legal team worrying, right?

Jensen said every company in the world needs to have an OpenCore strategy similar to a Linux focus or a HTTP, HTML focus. He said OpenCore has open-sourced the operating system of agentic computers. Every company. Not every tech company, every company.

I want you to sit with that because Jensen Huang is not a hype merchant. He's one of the most technically credible CEOs in the world. When he says every company needs a strategy for this, he's not making a sales pitch. He's describing what he sees when he checks the data, the demand curves, the customer conversations, the infrastructure build out.

He's telling you where this is going. Here's what this means practically. Right now, AI automation-builder agents that do the work for you is mostly happening in tech companies and startups. The people who know how to build them are doing it.

The people who don't know how to build them are watching it happen. NemoCore is a bridge that lets any company in any industry start to play agents without needing a team of engineers. And that gap between the people who know how to use it and the people who don't is the most important gap in business right now. It's widening every single month.

This is exactly why I built the AI Profitable Boarding because the honest truth is that most people are never going to figure this out on their own, right? Not because they're not smart, but because technology is moving so fast that even full-time AI people have to sprint to keep up. The boardroom is where I break down exactly what's actionable, what's noise, and what you should be building right now. And if you want to get ahead of this instead of scrambling to catch up, the link is in the comments description.

Now, let's go back to the announcement because we're just getting started here. Let's talk about autonomous vehicles because Jensen Huang made some announcements here that I think got lost in all the chip news and they're going to matter enormously. Huang said NVIDIA has four new partners when it comes to autonomous cars and declared the chat-cheapity moment for autonomous driving is here. He also announced a new partnership with Uber.

The chat-cheapity moment for self-driving cars. If you remembered what happened when chat-cheapity launched in late 2022, that moment when AI went from a tech industry to a mainstream thing, when your parents started asking you what it was, when every company in the world had to suddenly have an AI strategy, Jensen is saying that's happening right now for self-driving vehicles. Collaborations with automotive giants like BYD, Hyundai, and Nissan enhanced the Brobo taxi platform. Partnerships with Uber and robotics companies aim to deploy physical AI models and BYD, the biggest electrical vehicle company in the world, Hyundai, Nissan, Geely, these are not startup partnerships.

These are the companies that build tens of millions of vehicles every year and they're all building on NVIDIA's platform. Think about what that means. When you talk about AI coming for jobs, people usually think about software, programmers, writers, analysts, but the biggest job category in most countries is driving. Truck drivers, delivery drivers, taxi drivers, rideshare drivers, and they just announced that Uber is deploying NVIDIA-powered robotaxis across 28 cities by 2028.

28 cities by 2028. That's two years from now. I'm not here to tell you how to feel about that, but I think you should know it's coming because if you have family members or friends whose livelihood depends on driving, this is not an abstract future concern. This is a very specific near-term change and here's the thing that I keep coming back to.

The people who understand this technology, who understand what's coming, who can help companies and individuals navigate it, are going to be in extremely high demand. The people who ignore it are going to be caught completely off guard. Now let's talk about what comes after Vera Rubin because NVIDIA is not stopping. Feynman systems are on track for 2028.

If you don't know what that means, I'll explain it in a second, but essentially that's coming pretty soon as well. So named after Richard Feynman, the physicist who made quantum mechanics understandable to normal people, Jensen Huang's chip naming convention is never an accident, right? So Feynman was a person who could take the most complex ideas in physics and explain them so simply that anyone could grasp them. And Jensen announcing his chip architecture named after him, well that tells you something about where he thinks AI is going.

It's getting more powerful, it's getting more accessible. And NVIDIA's next major architecture is Feynman, which will include a new GPU, NVIDIA Rosa, named for Rosalind Franklin, whose x-ray crystallography revealed the structure of DNA and reshaped modern biology, right? Rosalind Franklin, the woman whose work on DNA was one of the most important scientific discoveries of the 20th century, and who didn't get nearly enough credit for it in her own lifetime. NVIDIA naming their next CPU after her is not just a tribute, it's a signal.

The CPU is built to move data, tools, and tokens efficiently across the full stack of AI infrastructure, revealing the hidden architecture of the system, just like Franklin revealed the hidden architecture of life. I find this stuff genuinely fascinating. Not just the tech, the story Jensen Huang is telling with his choices. Now let's zoom out for a second because I want to give you the real picture of what just happened yesterday.

One year ago, NVIDIA's projection for Blackwell and the previous generation of chips was $500 billion in revenue opportunity through 2027, a $500 billion forecast. That was considered enormous, almost unbelievable. At GCC 2026, Huang took the stage and said he expects purchase orders between Blackwell and Vera Rubin to reach $1 trillion through 2027. Last year, the company had projections for a $500 billion revenue opportunity.

Following NVIDIA's earnings report last month, Finance Chief Colette Kress said the company expects growth this year to exceed what was included in that estimate. They doubled the forecast and then said demand is constrained only by supply. Let me translate what demand constrained by supply means. It means customers are lined up with money in their hands waiting to buy NVIDIA chips and NVIDIA literally can't make them fast enough to satisfy the demand.

In most industries, that's a temporary situation. You ramp up production, you meet the demand, the gap closes. But NVIDIA's chips are so specialized, so complex and so dependent on a specific supply chain, TSMC in Taiwan for manufacturing, SK hynix and Samsung for memory, that ramping up production takes years. You can't just flip a switch and make more.

This is why the demand forecast keeps going up, because if demand is real and the supply is genuinely limited. NVIDIA currently holds 90% of the high-end AI accelerator market and NVIDIA hardware has turned into a de facto standard that software engineers and developers can't ignore. 90%. In business school professors talk about companies with pricing power market dominance as theoretical ideals.

NVIDIA is running a live demonstration of what that actually looks like and when you have 90% of the market for something that every major company in the world desperately needs, you can set the price, you can set the terms, you can dictate the roadmap and here's what that means for everyone else including you. Companies that are building on top of NVIDIA's infrastructure are making huge bets that it will keep getting cheaper and better. OpenAI, Anthropic, Google, every major AI company in the world and NVIDIA just showed them a roadmap that goes from 700 million tokens per second today to whatever Feynman delivers in 2028. Every single one of those token improvements translates directly into better AI deals for you.

So it was a big announcement for all of this. So much happened at that event that is absolutely wild. I don't know where to start with all this but I would say let's bring it home okay because I want to give you the clearest picture of what yesterday and the GCC conference means right and there's still more announcements coming. Basically Jensen Huang stood in front of 30,000 people in San Jose plus hundreds of thousands watching online made the following claims to a straight face.

NVIDIA's customers want a trillion dollar of chips and can't get them fast enough. NVIDIA is building data centers in Orbit. They just acquired a chip company for 20 billion dollars specifically to make AI faster and cheaper for real-time use and every company on earth needs an AI agent strategy. The chat cheapity moment for self-driving cars has arrived and the architecture they're building right now, Feynman, arriving in 2028 will make today's heart look like a toy.

The remarkable thing is that none of those claims are like wild speculation. Every single one of them is backed by purchase orders, partnerships, hardware that's already shipping and a company with a decade-long track record of delivering exactly what it says. NVIDIA has built and installed base across every cloud and computer in every single industry. Their CUDA software platform is celebrating its 20th anniversary at this year's GCC.

20 years of CUDA right 20 years of building software layer that AI runs on. The mode NVIDIA has built is not just hardware, it's not just chips, it's the entire ecosystem. The software, the developer tools, the partnerships, the supply chain that every AI company in the world has built their products on top of. And that's not something a competitor can replicate in a year or five years or maybe ever.

As Jensen Huang said, AI is no longer a single breakthrough application, it's a central infrastructure. Every company will use it, every nation will build it, every company in every nation. I've been covering AI for a long time now, I've seen a lot of announcements, a lot of keynotes, a lot of big numbers and bold claims. What I can tell you is that GCC 2026 felt different, not because of the individual announcements, although they were genuinely impressive, but because of the convergence.

The hardware and the software and the partnerships and the use cases are all coming together at the same time, at the same scale, in a way that doesn't look like a trend. It looks like infrastructure. The way electricity became infrastructure, the way the internet became infrastructure, the way the smartphone became infrastructure. These things don't just change how they work, they change how the world works.

They become invisible because they're everywhere and AI has become an infrastructure. NVIDIA is building the power grid. The question for you is what are you going to do with the electricity? Because the people who figured out electricity early didn't just use it, they built things on top of it that nobody had imagined before.

The people who figured out the internet early didn't just browse websites, they built businesses that changed every industry on the planet. The people who figure out AI agents early, who understand nemaclore or who understand what NVIDIA is building or who know how to direct and deploy these systems are going to build things we can't even imagine right now. That's the opportunity on the other side of this and the other side of everything Jensen Huang said yesterday. So the world is not getting simpler, it's not going back, the trillion dollar forecast isn't going to shrink, the chips are getting faster, the agents are getting smarter and the companies are deploying it at scale.

The only question is whether you're standing and where you're standing when all of it arrives. And I keep saying this, the time to get ready is not when the wave is already on top of you, the time to get ready is now. When you still have the time to learn, to build, to position yourself, to understand what's happening and what it means for your work, your business and your future. Don't wait until your company announces an AI initiative and you're scrambling to catch up right.

Don't wait until the robotaxi is already in your seat. Don't wait until nemaclore is running inside every competitor in your industry and you're wondering why your business is falling behind. Start now, learn now, build now, that's what I'm trying to help you do every single episode, every single day. I'll see you tomorrow and if you want to learn AI automation, if you want to get help with this, if you want to connect with me personally, join the AL Profit Boardroom, link in the comments description or go to thealprofitboardroom.com.

I appreciate you watching and I will see you on the next one.