英伟达相对于同行被低估，这得益于黄仁勋解决供应链瓶颈和维持高增长的能力，使其成为更优的长期投资标的。

there is a lot of execution risk. Like I understand that what they're trying to do, you know, to be able to sustain growth, you know, they we just saw 100% year-over-year guiding to 70% year-on-year growth next year. That's insane when you have a $5 trillion market cap and, you know, you're at nearly a trillion dollars of revenue run rate um per year.

But he has like the market has basically underrated him in this way consistently for the last decade until they just can't deny it anymore and then it you may overshoot for a while. but I think we're in another one of those periods with Nvidia that he really is focused on solving all of the bottlenecks that are going to prevent them from being able to grow at, you know, sustained probably at least 50% rate in an area that is just deeply like working with atoms and supply chains.

this is the field that people respect Elon for so much. I think Jensen is just as good as Elon is in this regard, and he just doesn't get the credit for it. So, yeah, that's why I would go with Nvidia.

one thing I can say that's good about your Nvidia choice is you're balancing risk and reward. So, they are kind of safe.

The Terafab story we'll talk about in a minute too could threaten the throne of Nvidia. I am probably heavily allocated to Tesla and like you been building my SpaceX position much larger you know position in my portfolio than my Nvidia position

300,000 Vera Rubins running at the same time in a coherent manner. That is the moat.

it used to be that there was sort of a zero sum mentality between Elon and Jensen. Elon felt like I'm going to have to build dojo and there wasn't that trust uh between them and that has been something that we've seen erode over time is that Tesla and SpaceX are both using more and more Nvidia hardware over time. They feel like they can trust Jensen to be able to deliver what it is that they need to accomplish their roadmap.

cuz I think there was some question like is Nvidia going to be a like we see this giant need for compute and that's what you see right now with the terafab. It's like we see this giant need for compute and we want to be able to accomplish these grand plans that require more compute than what we're seeing in the road map for players like Nvidia.

And I think that Jensen is able to keep up with anyone in terms of just growing capacity for compute per watt. and yeah, they're going to be they're going to be in the driver's seat next year. I think they're going to bring on well more power than anyone else.

Elon Musk industries, both Tesla and SpaceX have secured 30 to 40% of the entire Vera Rubin supply for the next couple of years. They're also sticking Nvidia chips into space first later 2027 which is also huge.

and they're co-designing those chips with Nvidia. So it's you know there is a deep integration of the engineering teams between Nvidia and SpaceX that are helping to make that a reality.

We know Nvidia is like running 85% margins to for Elon to do things at scale the way he wants. He can pay that now. But when does that end and when is the tipping point for Terraab coming?

Elon's best strategic move is to bring on the most Nvidia compute possible because as long as he's bringing on Nvidia compute, he can serve any of those. If he makes specific inference chips that don't live inside of the CUDA ecosystem, the ability to provide the type of open openly usable infrastructure that any of these different inference providers can run services off of. Um, that's severely limited.

He won't care when it comes to the time. He won't have enough chips anyway for his own Grock and his mess in his cars. So >> as long as they are using their inference, if the thing that they're limited by is inference for their own products, then I think that's where you do get back into a position where maybe you want to have your own inference, but maybe you also work with Nvidia to to produce that as well. and maybe you do keep it inside of the ecosystem.

Like I think there is a lot of surface area for Elon and Jensen to work together to accelerate each other's road maps because I think there is so much overlap in what it is that they're trying to build.

Um so I don't know that the right framing for investors to think about this is Elon versus Jensen or you know Tesla and SpaceX versus Nvidia over the long term.

I think that the the overlap in what it is that they're trying to build creates a lot of surface area for them to collaborate on the scaling and I think memory is going to be you know the huge constraining factor from a silicon perspective for all of this buildout for a considerable amount of time and so my expectation for Terraab is that it's going to focus on you know specific inference chips for Tesla with the for the Optimus bots and the cyber cabs memory and then the Starlink satellites and the Star Mine satellites. But I think that that memory piece is going to be the general purpose compute that Terraab focuses on. And I think that they will consume a lot of that in their other chips, but the excess capacity that they can build up there could also be sold to companies like Nvidia or you know other other people.

I think that's one of the reasons why Nvidia is so heavily discounted right now is that people see Jensen strategically using capital in order to accelerate the buildout of this ecosystem and they are afraid that that just isn't sustainable and it really depends, you know, what is the cash conversion cycle that Nvidia has. As long as they are able to just bring in

And this is where I come back to just the extreme amount of respect that I have developed for Jensen over the last couple years. And he is fully aware of all of that. and he has been thinking in ecosystems that are far larger than anyone in the market appreciates since the founding of Nvidia and he has executed against those goals against all odds.

Who's going to catch most of those trillions? Like a caveman I kind of reverse engineer find out the players you know the AMDs the Marvels the Nvidas the Microns blah blah blah all the usual ASMLs TSMC's all the usual suspects.