NVDA将GPU定位为可持续产生收入的资产，并广泛拓展至机器人与主权AI生态，这一策略支撑其长期看涨逻辑。

This past week, we had a lot of great insight and updates from the biggest AI company itself, Nvidia. And I believe they continue continue to change the AI ecosystem.

Jensen also shared some information earlier this week. Retweeted that the H100 is a three-year-old training chip. Its rental price is up 22% on the month to $328 an hour.

Nvidia CEO says Nvidia Compute is fungeable, durable, and highly rentable. it is a productive revenue generating asset.

The first thing is Nvidia is really trying to showcase that their GPUs are a different type of asset compared to other types of compute.

They're showing it with data. They're saying the H100, look, it's a multi-year old chip. It's still generating more revenue than it was in the past.

So again, here is another data point that Nvidia is really trying to bring up to the world that says, look, this is a market asset that holds value. Not only does it hold value, but it also provides a revenue source for you. But Nvidia is really hounding on this.

What we saw here right is and this is why Nvidia wants to make sure open source also continues to develop and becomes a key player because with open source this open source model can run extremely well on this AI chip opposed to like let's say there were a frontier companies in a lot of these private AIS there's a world for both but one of the benefits for open for for openweight for Nvidia is in theory in theory a frontier model can say I'm done training if they wanted to right I don't think we're going to see it but In theory, they can say, "I'm done training or developing on an Nvidia chip, and this model is going to be amazing on this XY chip that's not Nvidia." Again, I don't think we're going to do that, and Nvidia continues to improve all their software stacks. It would be extremely dumb for anybody to do that just because of the developments that are happening within Nvidia itself.

a few core thesis that came out for me is new Nvidia generations don't automatically make older GPUs economically obsolete what matters is whether an older GPU can still run or useful workloads competitively and openweight models are what let operators you can develop and make sure that your model is running equivalent or strong in older generations.

is another reason why Nvidia is really trying to push into this open- source world because again it provides more demand and it also continues to allow them to show that world where older GPUs still hold value

So this is just another data point here that is showcasing that Nvidium has this asset that can continue to last over time and then it makes the financial community a lot more comfortable lending out money when you're building on Nvidia ecosystem.

there are reports by Reuters saying that Nvidia mos a 10 billion anthropic IPO backing. We're just going to go and make sure Enthropic is also surviving and doing extremely well. And there's going to be extremely competitive pressure that this competitive pressure, no matter what type of AI chip, you're going to come back to us because with competition, you need to make sure you have the best and we are going to have the best product out there.

Now, the next thing I want to jump into with Nvidia is they did announce that Dmatrix adopts Nvidia NDLink Fusion for rack scale XPU deployment.

So Nvidia is licensing or selling out, however you want to do it, selling out and allowing you to integrate whatever XPU you want to Nvidia's ecosystem. So here Raptor is created by the XPUs, but the whole AI system is pretty much Nvidia. You're going to be using Nvidia's great CPU. You're going to be using Nvidia's connect connector uh for for Nyx and all the networking solutions spectrum X you're going to use NV link you're going to use so much of Nvidia solutions that Nvidia is going to win no matter what you do if you want to make your own XPU that's great you still need again XPU is only the first step the AI acceleration is only the first step you need to come with us because you need to have the rack architecture the power the cooling the software and like more importantly the supply chain, right?

So, even if Nvidia doesn't sell the chips, they're still going to be able to sell everything else, which is pretty pretty damn good. Outside of Nvidia really shake changing the financial world what they're trying to also do is secure as much land power and shell. We talked about this a few about a week ago where Nvidia kind of really explains that land power shell is a massive constraint in the overall space.

So Nvidia expands their AI infrastructure capacity in partnerships with Australia's data center ecosystem. Some of the few names that they mentioned were firm Sharon AI and I rent and they're building up to 2 gawatt built out by 2027.

Now 2 gawatt for for whole year of 2027 Jensen mentioned is not that crazy. It's not that much, but still 2 gigawatts is about $80 billion in revenue potentially for Nvidia.

This is why Nvidia is trying to become a platform of AI. And what they mean with the platform is they help you do what you need to do. You're not going to worry about Nvidia competing against you.

So Nvidia says, "Look, we have this Nvidia DGX compute infrastructure solution and this DSX platform that showcases you how to do all this. Take it, run with it. Now you're going to be able to build massive data centers all over Australia."

Now the other one I thought was also pretty cool. Nvidia and Palanteer brings sovereign intelligence to critical supply chains. and custom Nvidia Neotron open models for complex supply chain operations.

And what better place to start than with Nvidia's own operations. Nvidia works with thousands and thousands of supply chain suppliers of thousands and thousands of pieces and parts of equipments that Nvidia says, "Look, we're going to use Palunteer's AIP and you're going to help us make our supply chain better.

If you can do this for us, whatever you're going to be able to do with other players is going to be insane." And this is why I like Nvidia a lot, right? Nvidia Nvidia goes and and and tries to build the market, right? It has done that with everything. It goes and builds the market. It says, "You want to see where AI use cases are there? My supply chain is crazy. Me and Palanteer are going to show you that we're going to be able to improve it dramatically because then your supply chain, you can come to Palunteer and be able to do whatever you need to do. And then if Palanteer wins, you're most likely going to be using Palanteer's model selection that they have. And most likely you're going to be partnering up with Nebius, which means you're going to be partnering up with Nvidia one way or another. So Nvidia is really trying to become this overall platform.

Next, they do talk about some of the robotics things happening in the industry right now. They shared some solutions that we are seeing physical AI reach a new front of level. uh they talk about all different types of solutions from kind of the consumer market how AI's helping the factory work and how Nvidia is part of the whole ecosystem from training from inference from even prior to that with simulation the software stack and the gly goes on and on nothing that I would say is revolutionary right now but is a long-term thesis

he talks about Whimo He talks about their partnership with Mercedes and how Nvidia is part of this whole ecosystem from training computer.

You have Nvidia's DGX system. You also have Nvidia's Alpamo which is their open reasoning vision language action model that they showcase and has improved improved errors dramatically.

Then for simulations they have the Omniverse software solutions. They have the Cosmos. They have Nvidia's RTX Pro which are great for validation and simulations of systems and then invests with the Nvidia drive Hyperion with drive AGX and whatever kind of cockpit and software solutions they also have in there and kind of safety solutions like the Nvidia halos.

So robo taxi leaders are are winning and Nvidia is going to win with them. Nvidia partners with Whimo, they partner with Uber. They partner with boat. They partner with Lyft.

They even partner with Tesla in a way where Tesla uses their technology for certain training and inference solutions. Nvidia is just showcasing that the robo taxi world is here.

Now the way I want to end this episode actually two things. Nvidia did have a Goldman Sachs technology conference on September 10th. I want to showcase some of the things that I thought were important because it was Jensen who did come out for this community.

First, he he kind of really pushes on this what I started off at the beginning of the episode where Nvidia wants to make Nvidia compute from a technology to an investable asset.

And he thinks this is going to be a huge needle mover for people to recognize that their computing systems have a long-term value and has durable value. At the moment, none of that is being valued and that is a huge untapped opportunity for them.

Let's say they sell 1 gawatt of data sensor for around $60 billion. You're going to break that over five six years. So $10 billion a year. Obviously a little bit more with all the other expenses, but let's say $10 billion a year for some simple math. That 10 billion is currently being rented out at around $50 billion as you guys know. So the revenue per gigawatt right now is about $50 billion.

If this number is true, you're going to see an explosion of data centers being built out. Because if you're telling me I can put something for $10 a year and I'm going to get $50 every year on it, I'm going to do that over and over and over and over again.

And if the man for AI continues to grow, I expect this solution to continue to to continue to be needed. And then they mentioned that you can still rent out Voltus. Volt is a 10-year-old product. And right now the Miers are still also selling out which is over a 5 six year old product.

Now the other thing Jensen talks about is the supply chain. the supply chain is super critical for them. Just think about all his partners. How many NeoClouds are reporting back to them? How many OEMs are reporting back to them? How many clouds are reporting back to them? How many AI native companies are reporting back to them? They are working with everybody. So, they know everything. It's pretty amazing.

And this is what gets me excited. This is why I follow Nvidia because they have ears and eyes in the whole ecosystem in the AI story. So if they tell me one thing is happening, I am truly truly going to put a lot of weight to what they say because they even if we as the consumer don't see it, they've already seen it somewhere and for them to say something about it, it but it mentions that it's going to happen pretty soon.

The AI agents, right, that came out of nowhere, but Nvidia already had the products, chat GPT, Nvidia already had the the solutions and we continue to see that, right?

The Vera Rubin is perfect for tool calling, which is what we're seeing right now. So to me the excitement for Nvidia and the growth in innovation here is there because Nvidia sees the whole ecosystem and I think that is a massive massive asset for them.

Another thing that they mentioned this is maybe for for investor excitement they mentioned that in the Grace Blackwell solution the envy link they saw about a 72% increase month overmonth on the Grace Blackwell not the Vera Rubin. So that's massive.

They do give us a shout out of what's the next solution. Jensen believes that the cyber security will likely be the next major use case of AI and is going to run continuously.

I got to make sure I bought myself a DGX Spark. I need to compute in my house as much as possible because I believe local, not only open-source models are going to be crazy.

But I also believe there's going to be a huge explosion in local solutions in the strength of local demand. So I ended up picking up a Dichi X Park working on some pretty cool stuff right now.