Projects
![]()
No projects found
Investors
New
DAO Treasuries
DATs
Lend / Borrow APRs
Chains
Categories
Spaces
$NVDA's BIGGEST Event of the Year is Here | Why The Worlds Biggest Company is about to SKYROCKET
Recorded: March 12, 2026 Duration: 0:41:03Space Recording
Full Transcription
follow up on this Nvidia AMD semiconductor talk that we had last time. A lot of people enjoyed
it, a lot of questions, and there's big things coming up. So I figured we should dive right back
into it. Alex, how are you doing today? I'm doing great, man. Super excited to be here. Super excited
for GTC next week. Super excited about all the earnings calls that have happened so far. So yeah,
thanks for having me on and I'm excited to get into it. Absolutely. You mentioned GTC. Let's
start there. GTC is a massive event for Nvidia and this is their biggest GTC of the year that is going to be coming up.
It's much bigger than the one that they do on the East Coast side. Talk to me about what you're
looking at at this GTC, what you look to take away from it, and how investors in NVIDIA should
be approaching it. Yeah, great question. So basically, GTC, the one in March, is their
big main event. I think like you
just pointed out, they're actually at a lot of different conferences. So for example, they were
at the Consumer Electronics Show or the CES back in January. They're going to be in Paris. They're
going to be in Taiwan. I think they're going to Washington, DC later this fall. But GTC in March
is really the one at or near NVIDIA headquarters. and it's the one where they bring all their tech.
And it's the only conference that's fully dedicated to NVIDIA, their stack, their partners, and their ecosystem.
So what I'm really looking for there is, like, obviously the big next chip reveal,
whether that's, like, more of a deep dive into Vera Rubin and Rubin Ultra or a little more details on Feynman.
But hopefully we're getting more details on the Rubin stack and then the Rubin Ultra or a little more details on Feynman. But hopefully we're getting more details on the Ruben stack
and then the Ruben Ultra stack,
which is going to change many, many, many things from Ruben.
That's going to be a big jump.
And then the other two big things that I'm looking at
are the rack level memory solutions.
So last time I believe I was on your podcast,
we talked a little bit about how memory
is increasingly becoming a bottleneck.
And so at CES, Jensen, during his keynote, talked a little bit about how they're solving
that memory bottleneck. But he talked about it at a pretty high level and only for several
minutes out of his two-hour keynote. So getting more information on that, I think, will be great,
as well as any updates to their networking solutions and networking innovations, including co-packaged optics, photonics. I'm sure we'll get into this a
little bit, but NVIDIA has been investing in several optical networking companies and things
like that. So how that plays into their broader and larger strategy, I think is going to be really
interesting. And then for investors, it's going to be the same stuff, just tying that to where in the stack can we put our money to support that vision, get in early,
invest aggressively, and then enjoy the returns before sort of the rest of the market.
So for example, Coherent, what ended up being a pretty good company to invest in,
because that's one of the optical companies that NVIDIA is investing in, CoreWeave on the AI infrastructure side,
which we've been talking about early and often, and so on. And then the last thing I'll say
before I turn it back over to you is obviously the other special thing about GTC is that it's a
conference focused on NVIDIA's ecosystem, right? And their partners and the people who leverage
that ecosystem. So I'm hoping to see a lot of cool and exciting things in robotics, cool and exciting things in self-driving cars,
right? And all of the partners and part of the ecosystem in those areas as well. I think
physical AI, both on the road, in our homes, in industrial warehouses,
is a pretty exciting prospect right now. And that industry is moving fast. So I'm excited to see
what is actually happening on the bleeding edge there.
A lot of pieces there.
I want to dissect a couple of segments.
One that you mentioned was the investments
that NVIDIA is making themselves into companies
such as Nebius and others.
They have a little bit of their own portfolio working there.
And not everyone's aware of it,
but those that do see it,
I think have different takes on it.
Some are like, well, it's such a small portion of the cash that they have.
Other people are like, well, this is actually the start of something critical that they're going to build out.
How do you view the investments that they're making?
And how do you think investors should take that into account as they weigh NVIDIA as a whole?
Yeah, I actually, I'm probably going to give you a pretty non-intuitive take here.
So I think the investments say a lot less than people probably attribute them.
Like I think they should have a lot lower weight than people give them.
NVIDIA has a massive war chest.
They generate a lot of free cash flows and they need to find good ways of investing in
that money that also aligns with their vision.
I think that's the part that's like really critical here.
It would be very weird if NVIDIA just invested randomly and tried to get like returns on their money, right?
So what are they doing? They're placing strategic bets on where they think the future of the AI
stack is going, right? So like Coherent and Lumentum, you know, these are strategic partnerships
to advance optical networking, but that also clearly signals that NVIDIA believes that optical
networking is the future of AI networking, right? Nebius and CoreWeave, these are two of the Neo
Clouds that definitely build all of their offerings on top of NVIDIA's chips. You know, I think over
90% of both of their chips are NVIDIA as opposed to AMD, as opposed to Grok, as opposed to like
other inference solutions. So this is more of a
strategic bet saying, hey, the new AI factory, the new type of AI data center probably looks a lot
like this, as opposed to different clusters inside big hyperscalers, right? Like why invest in
Nebius and CoreWeave and not Google and Microsoft, right? So I think the directionality of these
investments is really interesting to
investors, but I don't think it should be something like, oh, NVIDIA is investing in
Anthropic, go buy all the Amazon stock you can because of their relationship with Anthropic.
Oh, they're investing in OpenAI, go buy all the Microsoft stock. I don't think that's the way
investors should be interpreting this portfolio. It's a good way to put it. On that
point about OpenAI, that was something that was a little bit more up in the air the last time we
were chatting. Any additional thoughts on what's come out there? It's a really good and interesting
question. So the stuff with the Pentagon, I think, has given me the most clarity on where I want to put my money. So I was lucky enough to be an OpenAI investor,
I want to say in like 2023. I put out a video talking about, you know, we partnered with a
company at the time called Disraptor. They had secondary market shares of OpenAI.
Me and my audience were able to get in at a pretty low valuation. It was pretty nuts.
So from an investment point of view, I'm very happy with my personal tiny slice of OpenAI. But the direction the company has headed
since then is not a direction that I hope OpenAI would have gone down. I actually agree with a lot
of Elon Musk's sentiment around OpenAI. I think that they're clearly chasing as much profit and
as much user base as they can, which is very opposed to their original mission statement.
I actually find myself gravitating a lot more towards Anthropic these days just because of, A, how they handled the whole Pentagon thing.
We can talk about that if your audience is –
Yeah, you want to give a little bit of context for people?
Sure.
So the two-minute version is Anthropic has a big government contract or had at this point a big government contract with the Department of War specifically and different DoD subsections, right?
And they had two bright red lines that they weren't willing to cross as part of engaging with all these departments.
The two bright red lines are fully autonomous weapons.
So they don't want AI making end-to-end decisions about taking a human life domestically or abroad and mass surveillance on Americans. So, you know,
think collecting and aggregating lots of data specifically domestically on American citizens
and then using Anthropics AI to make decisions about people based on that data, right? So Dario
Amadei, the CEO of Anthropic, stood firm on those two red lines when
their contract was undergoing negotiations and changes with the Department of War. And they were
trying to put in wording that basically allowed them to overstep those two red lines at their
discretion. Dario said no. There was a whole bunch of tweets by Secretary of Defense, I believe,
There was a whole bunch of tweets by Secretary of Defense, I believe, Pete Hegseth,
as well as President Trump himself. I don't want to get political here, but end of the day,
there appears to be like overly harsh punitive measures being leveraged against Anthropic,
right? They're being put on a list that essentially says no government agency can
work with them. Definitely no Department of Defense agency that's involved in critical missions abroad, let's say. And that's a lot of contracts.
It's a lot of contracts. So it's obviously a punitive measure against Anthropic. And in
response, Anthropic actually held their ground and they said, hey, we get it. We're obviously
going to take this to court to the extent of our abilities that we feel this is unfair,
but we're not backing down on these two red lines. And then right after that, OpenAI got their Pentagon contract. And there
was a lot of backlash, in my opinion, rightfully so, because OpenAI basically in that sequence of
events said they don't have those two same red lines implicitly, right? They didn't explicitly
say that. But by taking that contract after Anthropik said no, they pretty much admitted, hey, we'll do whatever we can to make
money. I imagine Sam Alman was like, we would never surveil Americans. Wait, wait, how much
are you offering? Exactly. Exactly. Oh, money? Okay. Yeah, that changes my answer. So I think
Anthropik is really turning out to be the open AI we were hoping we would get, right? If that makes sense. If I remember this right,
this is a while ago, but Dario was originally, I believe, the head of OpenAI's safety team,
which is how it all started and got spun out. And then when OpenAI started going down this
unsafe AI path, that's when Dario split off and founded Anthropic.
Obviously, the Claude models are incredible.
Claude Code and Claude Cowork are incredible offerings.
So now I seem to be more of an Anthropic fanboy.
I have no investment in Anthropic.
I missed that boat completely.
So that's what's going on there.
Sorry, I forgot that was a tangent to an original.
Yeah, we were talking about the investments that NVIDIA has made, and then we were talking about the OpenAI investment specifically that they had made.
Yeah, sorry. So that's right. At some point, Jensen floated the term of he could invest up to
$100 billion in OpenAI. It seems that investment is going to be substantially smaller, you know,
and I think it was never going to be the full $100 billion.
I think where the clear miscommunication is
and like headlines versus reality and stuff is like those up-tos.
Intuitively, when I hear I get to invest up to X amount of dollars
in a company like OpenAI,
and I obviously have that money like NVIDIA does,
the implication is I'm going to invest as
much as I can, right? Open AI is clearly a rocket ship. Whatever you believe about their ethics and
their moral compass, clearly their valuation is growing fast, right? And they're a premier AI lab.
So when he was like, hey, we're not going to invest $100 billion just because we were given
the privilege to invest up to that amount. I think people took
that as a sign of like, oh, he's lost faith in OpenAI and blah, blah, blah. But the reality,
I think, is more of what we were talking about earlier, right? Where it's like,
NVIDIA needs to make investments that align with their principles. OpenAI is definitely
going to be a slice of that. I would be surprised if they don't invest a huge chunk. I think they
already have tens of billions invested in it, right?
It's like one of their bigger private investments.
But I think it would be smart for them to spread that money around different AI labs,
different AI infrastructure companies, different robotics companies, self-driving car.
Like there's just a lot in the NVIDIA stack that's not just OpenAI, not just Premier,
Frontier large language models. So it
would be pretty crazy if there was a $100 billion investment there, as opposed to like a $2 billion
investment in networking, right? Those are pretty big differences. Yeah, massive difference. And
that makes a lot of sense to me. And I want to rotate a little bit here into some of the competition area.
But before I do that, just kind of some callbacks from our last chat that we had.
One thing that you talked about was a inference-specific chip.
You talked about calling an inference-specific chip.
And then you said, I wouldn't be surprised if NVIDIA came out with an inference-focused version of their GPU.
And then, of course, Ruben CPX announced in that area. So I
wanted to give you some kudos on that area. And also I think that you called the Spectrum X
Ethernet pivot before that announcement as well, saying that back in the day, they were exclusively
on Quantum InfantBand. So just before we talk about some of the competition, there has been
some updates from NVIDIA just in terms of technology and other pieces that you've, I think,
been looking forward to.
So maybe you can speak to that first.
Sure.
You know, like one of the big critical.
So let's back up a bit.
Those two chips are part of a larger ecosystem, right?
So like the Rubin CPX is a chip designed for a specific part of the inference step of AI.
And they also have the traditional Rubin chip, right? So the Rubin
GPU itself, not the CPX version, which is great for training and still really good for inference,
right? So one is much more flexible and one is much more catered towards a specific slice of
like the AI stack, but is really high performant in that stack. I think we're going to see a lot
more of that sort of all over the place. So I don't think, for example, NVIDIA is going to come out with its own high bandwidth memory,
right? As opposed to relying on SK Hynix, Samsung, and Micron, right? What I do think will happen is
we'll get more things like their rack level memory solution, you know? So a different way to manage
memory that leverages that resource and gets like a lot more efficiency out of it, let's say.
Or a dynamic memory stack that says, okay, I'm only going to really use the expensive memory when I have to.
And I'll divert to cheaper memory when I can use cheaper memory, which is kind of what that rack level solution already is.
I think the same will happen for networking. You know, on networking, I wouldn't be surprised to see instead of like one network compute fabric, which is like, we're just connecting every GPU
together, and then we're just connecting every rack together. I wouldn't be surprised to see
a fast and slow network where it's like the stuff that needs to be real time and that bottlenecks
the whole AI calculation is one layer of the network. And then all the metadata, health and
status of racks, all the, you know, sort of the command and control of the network. And then all the metadata, health and status of racks,
all the, you know, sort of the command and control of the data center that's not critical to the AI
workloads just gets passed on like a completely different network, right? So you can imagine
there's all of these things that go into maintaining a data center beyond just the
calculations and the workloads and the things going on within them. I think being able to take that out, put it on its own separate management layer and do that,
and then obviously using that freed up space for more calculations, more power, more things like
that is going to be a big part of the game plan. I think they're going to try to do that in every
area like they have with CPX, with the memory solution, with their multiple networking solutions and so on. So that's
specifically what I'm looking for. Makes sense. And then just tying that back to GTC as well as
earnings and what we heard from there, how do you think this all continues to progress? Are
you taking that mindset of what you're seeing them release and going to GTC and then digging
to GTC and then digging further into that, taking the details that they shared on earnings,
further into that, taking the details that they shared on earnings, putting all the pieces together?
putting all the pieces together? Yeah. So I think like what really
NVIDIA is doing is they're broadening out their ecosystem, right? So let's say I'm a company that
already is invested in my own chip. I'm probably not interested in using too many Rubin GPUs,
but I really want the rest of the stack. Well, NVIDIA is coming out with solutions to do that,
right? Like for example, they have something called NVLink Fusion, which allows companies to bring
their own processors, either their GPU or their CPU, use NVIDIA's for one and their own as the
other, right? So let's just use Intel as an example. Intel makes fabulous CPUs, but not GPUs,
right? So they would be using NVIDIA's GPUs, and instead of the Grace CPU,
they would use an Intel CPU, right? So that's an example of using NVIDIA's stack, but not one of
their core six chips, right? And NVIDIA is making that available both to fiber optic data centers
with InfiniBand, like you said earlier, but now Ethernet data centers as well through their
Spectrum X Ethernet solutions. Same thing we'll see with memory, I'm sure. Hey, I'm using this type of memory stack,
but I want to make as much use of it as possible. So I want to integrate NVIDIA's memory solutions,
so on and so forth. My point in saying that is, if you look at NVIDIA's earnings, tying that back
to their actual earnings, their revenue is actually accelerating, right? Why? It's not just because they're making more chips that cost more money each. It's because
they're able to broaden out and sell networking. I don't know if you caught this from their latest
earnings call and actually the couple before it, but $11 billion of their revenue last quarter,
I believe, actually came from networking, not GPUs. So that's an insane amount. I believe that was
something like 15% of their revenues, a really meaningful amount, came from just not GPUs at
all, right? Networking. So I wouldn't be surprised in the future if we see multiple segments like
that, where it's like networking, compute, memory, right? Like power heating and cooling,
you know, not that they make those things, but that they figure out how to tightly integrate those things. So that's, sorry,
connecting that all together is exactly what I'm looking for, the offerings and how they show up
in the earnings. Yeah, it's wild, right? Not even their main offering or their main focus necessarily,
and it's still producing over eight figures or I guess at billions. I mean, I don't even know
what numbers are looking at there. Yeah, it's 10 figures, I guess. billions. I mean, I don't even know what numbers are looking at there. Yeah, 10 figures, I guess?
10 figures of revenue continuing to come in
within these pieces and areas.
So just since we're on the topic right now,
just taking a look for a second at the numbers
from this last earnings call for anybody that didn't catch it.
They just continued to blow people away.
We talked about it in advance.
You were very confident that they were going to be able
to beat these numbers and, of course, beat they did. So for those, gosh, I got to pull people away. We talked about it in advance. You were very confident that they were going to be able to beat these numbers and of course beat they did. So for those, gosh, I got
to pull back up the data for a second here. I just want to make sure that I'm accurate with everything
that we're sharing. But for those that missed it, the expectation for the quarter Q4 2026 was 65.9
billion. They reported 68.13 billion for a nice beat there and then beat on EPS as well. $1.9 billion. They reported $68.13 billion for a nice beat there. And then beat on EPS as well,
$1.50 expectation, $1.62 of the actual report. And all of the metrics underlying that seemed
really healthy as well. People do typically just see mostly the top line items. But what did you
care about that was under the hood during that earnings report? I actually think we're understating the importance of their revenue beat. If you
actually draw a line of their revenue growth over time, it was pretty linear for the last,
I would say, call it two and a half years, like 10 quarters maybe. So sorry. Yeah. If you could
just show a chart of their revenue over time. That's revenue last four years. That's revenue last 10 years.
Perfect.
Yeah.
Like maybe four years is probably like 10 years is obviously like they're, they blew
it out of the water.
Right.
But if you look at their last four years, I'm specifically like, you know, you can ignore
probably the first quarter of the chart.
And then if you can somehow zoom in on that, whatever your audience wants, right?
Like, yeah.
Yeah.
But I don't know if I can zoom in more than four years right here, But yeah, I get what you're saying. Exactly. Like just how much. Yeah. So,
so in the last two quarters specifically, it's been accelerating. Right. And so like,
I was hesitant to call that acceleration last quarter because like one point doesn't,
doesn't align make. Right. But now that we've kind of gotten to, we can clearly see, Hey,
something different is happening,
even at the top line here. Obviously, if you look at their operating income, sorry,
I'm plotting it myself. It all flows downstream, right? They have incredibly high operating margins,
net margins, and gross margins. But that top line revenue number, I think, is really telling. And
that acceleration, I think, should be surprising a lot more people than it has. Like, the fact that the biggest company on
earth is accelerating revenues when all of the financial models say those revenues should be
slowing down. There's a clear mismatch between performance and expectations here that no one
besides me seems to be talking about. I'm not really sure. Like, I'm not trying to toot my
own horn here,
but it's like every time I make an NVIDIA video,
people are like, this is old news.
Nobody cares.
And I'm like, how are people not caring
that a $5 trillion company is accelerating revenues?
I think people get in their head about,
hey, it's so big, $4 or $5 trillion.
Is the opportunity there the same
as when you look at some of these companies
that are $10 billion, $ 10 billion, 100 billion, right?
And they have a much more potential 10X in those areas.
But to your point, some of the numbers go a little bit deeper.
That revenue number is up 73% year over year, right?
Data center is up 75%.
Year over year, yeah.
Yeah, $62 billion.
Non-gap profit margins or gross margins are 75%, right? You have free cash
flow of $35 billion just in a quarter. Dude, yeah. This is a hyper, hyper,
hyperscaler, right? I think what people don't realize, so like back in the day, I used to
cover ARK Invest a lot. And one of the really brilliant things that Cathie Wood repeatedly called out,
even when people refuse to listen, is like each S-curve is much bigger than the last, right?
So if you go back to, let's just say the car versus the horse-drawn carriage,
the car didn't just replace the horse-drawn carriage. The market for automobiles, mechanics,
painting, like the whole total addressable market of the car was
much bigger than the horse-drawn carriage. So if you, why am I saying that? Because like,
if you're just comparing NVIDIA to other companies today that are worth three, four,
five trillion dollars, of course, it looks like they hit their ceiling. But what they're replacing
is actually a much bigger portion of the economy, the compute industry,
data centers, and all that. I think we're halfway through a transition that sees NVIDIA becoming
the first $10 trillion company, right? Which implies a roughly 100% upside from here, right?
And you got to remember, from here is pretty recent. Like, wasn't too long ago, NVIDIA was
a $2 trillion company, a $3 trillion company,
where if you're talking about $10 trillion, that's a 5X, a 3X. You know what I mean? These are still
big multiples. And the thing, the last thing I'll say about this is like, obviously it's great to
like want to invest for a 10X, right? But with high rewards often comes high risk, right? So
it's very easy for a $10 billion company to lose $10
billion in value, right? To lose $4 billion in value. That's like not even the noise for NVIDIA,
right? Like NVIDIA can lose $100 billion a day in value. Like that's just the size of the company,
right? And so if you want something big, stable, safe everywhere, and that has a lot of room to grow, I'm hard-pressed to find a better investment than NVIDIA over, like, risk to reward, not just
thinking about the upside only, right? Yeah, great point about them increasing the
TAM of the entire market rather than comparing, you know, hey, they're just a new player within
the same market. Yeah. It's a very different area. Well, speaking of the markets, I want to talk a
bit about the competition here. Sure. And just pulling up a couple different area. Well, speaking of the markets, I want to talk a little bit about the competition here.
Sure.
And just pulling up a couple pieces from my notes.
One thing that you published recently was a Broadcom deep dive.
And make sure I'm quoting you correctly, but you said Broadcom is the only real competitor
to NVIDIA because they're going after different parts of the stack.
Now, you own both of them.
And so you're covered in this area, no matter
which way the AI market goes. Broadcom's AI revenue hit $8.4 billion last quarter, up 106%
year over year. And you noted that Broadcom controls roughly 70% of the custom AI accelerator
market and 80% of data center Ethernet switch chips. So a very interesting opportunity here.
And we talked to AMD last time. So let's talk a little bit about Broadcom now. Sure, yeah. So, I mean, I've been pounding the table on Broadcom for years
now. The thing that Broadcom is most famous for probably within the AI area is they are the key
co-designer of Google's TPUs. Google's seventh generation TPUs, Ironwood, are obviously incredible
chips. Those are the chips that are powering all of Google Gemini.
And Gemini is obviously an incredible model.
So Google has an end-to-end stack for AI at every level, right?
A lot of that is thanks to Broadcom.
Thanks to years and years and years of innovation specifically in ASICs,
application-specific integrated circuits for your audience.
Basically, exactly what we were talking about with Ruben CPX, right? These are highly specialized chips that focus on one
or two things and do them extremely well. And what I mean by extremely well is very high performance
per watt of power that they consume, right? Efficiency, yeah.
Very efficient. Perfect. Yeah. So Broadcom is a key
designer in those things. Broadcom has since won many other contracts. So Broadcom is heavily
involved in Meta's MTIA chips. Meta is one of the few companies that serves AI inference today at
massive billion person scales, right? Like you can find Meta's AI in Instagram, in Facebook, in WhatsApp. So like all that AI is powered by these MTIA chips
as well as NVIDIA's GPUs. And then they make chips for OpenAI. I don't know that they make
them yet, but they have a contract to do that. And for Anthropic, whether that's their own chips
or whether Anthropic just ends up buying Google TPUs now that Google is offering TPUs to third-party
companies outside of Google Cloud. My point is Broadcom is the NVIDIA, in quotes, I'm making
big air quotes here, of ASICs, right? So why is that different? Because no GPU maker is going to
be NVIDIA. NVIDIA was founded making GPUs. They've been making GPUs for over 30 years. They're the king
of GPUs, full stop, period. We're done talking about GPUs. What's going to displace Nvidia is
going to be a fundamentally different kind of chip, whether that's an inference-only chip,
whether that's an ASIC. I don't know what it is, but the way, like, horse-drawn carriages didn't
get displaced by faster horses or bigger carriages, right?
They got displaced by the internal combustion engine.
When Tesla came in, they didn't displace the internal combustion engine with a better ice engine, right?
They said, hey, there's a completely different way of powering a car.
EVs, this is a relatively unexplored space at the time when Tesla was like
innovating and coming up, right?
These electric vehicles, we can do better. EVs started displacing ICE vehicles, right?
ASICs have a reasonable chance of displacing GPUs in some data centers in some contexts, right?
Like I'm Google. I care a lot about a very specific type of inference, maybe like let's say
search engine, right? I just want this AI to focus on helping me refine my answers for AI overviews, right?
I'm going to build a chip because it's going to be much more efficient at that thing.
The hardware and the software are closely tied together.
Boom, the TPU is born, right?
That...
Oh, go ahead. Sorry.
And sorry, just say, I also see more potential for ASICs as more people get driven into AI and research and people get a better idea because right now it's very broad, right? There's
so much you can do and so many opportunities, but people are going to find their lane and they're
going to say, all right, I just want to build this. Totally. And if that lane is big enough,
it'll get its own ASIC, right? So imagine, you know, I'm Google DeepMind. I care a lot about
protein folding. I care a lot about like genomics, gene sequencing,
a lot of the like that type of stuff that comes out of DeepMind already, right? Alpha fold,
for example. I'm going to get a chip that just does that because that problem looks so different
from search, so different from chatbots, so different from everything else. It really needs
its own chip to work at scale, right? Well, if there's enough of a market out
there, i.e. the whole pharmaceutical industry, the drug industry, et cetera, that chip will get made.
And that's a set of workloads that will slowly move away from GPUs.
The real thing I think we're not saying that we probably should is the total addressable market
for AI is growing much, much, much, much faster than either of these companies can serve it alone, right? GPUs on
the general AI side, ASICs on the application-specific side. So right now, what you're
seeing is both companies are just winning. They're 75% year-over-year growth for NVIDIA,
106% AI revenue growth for Broadcom, right?
The pie is getting bigger faster than these two companies can slice it up.
So for right now, it's a great way to have your cake and eat it too.
The cake is they're both doing great.
They eat it too is you're actually diversifying at the same time as you're
picking two winning stocks.
If GPUs do better, you're invested in NVIDIA.
But if ASICs do better, you're also invested in Broadcom and the two connect a lot, right?
For example, in networking, in Ethernet-based data centers, et cetera.
Yeah, I don't know if anyone actually knows what the TAM is transparently at the moment.
It's impossible to put a number on.
Gajillions, you know what I mean?
Just, you know, infinity trillion dollars, let's just say, or, you know, some made up
number where somebody will put a cap on
it at some point. But realistically, it could be tens of trillions of dollars because just the
replacement of current regular non-AI infrastructure is like several trillion dollars, right?
Makes sense. Pulling back to your portfolio. So the portfolio here, a couple of the names we've
already talked about, NVIDIA,
AVGO, right? Those aren't going to surprise people that they're in here. I think a couple
that compliment those very nicely are Taiwan Semi, Micron, people have been hearing those. But
one thing which you've posted a good amount, a good amount of bull posting, I would call it,
is IRON lately. So I believe you tweeted and you said Wall Street just made a huge mistake,
all caps, on IRON and that you were buying it aggressively and you framed it as a power and
an optionality play in the NVIDIA ecosystem. So I want to talk about IRON here and how it fits
into that NVIDIA ecosystem as part of your portfolio. Yeah, so IRON is another one of
those NeoClouds,
right? Along with CoreWeave and Nebius. And the special thing about IREN, there's two. One that
I think is great and one that I think is like a little bit risky, depending on how you look at it,
right? The risky one is right now they're primarily a Bitcoin miner. Most of their revenue comes from
Bitcoin. But just like many other Bitcoin miners right now, they're transitioning their
infrastructure from Bitcoin mining to AI, right? So these are companies with large,
large data centers. They're already managing tens or hundreds of thousands of chips. Those chips are
specific right now for Bitcoin mining. But you can imagine a lot of the power, cooling,
infrastructure, the facilities overhead, right? Like is the same kind of thing you want to leverage
for AI clusters, right? So the special thing about IREN is they've already secured 4.5 gigawatts of
power. I strongly believe that what's happening right now is like the equivalent of like the
gold rush in the 1800s, right? Where it's like people were trying to claim land
and figure out what to do with that land
and use up their natural resources
and get there before the next guy.
So let's back up and talk about
what data centers actually need.
They need a lot of area, right?
Physical land.
They need a lot of access to power, right?
And those things are both very finite resources, right?
Like the earth only has so much land,
only some of that, only so much of that land has
easy access to gigawatts of power, right?
The infrastructure to support that.
So every area, every piece of land, every facility that gets built by one company is
a prime piece of land that another company couldn't capture.
So if we're living in a power limitedlimited world, which every single data center company
says we are, what are we limited by? The amount of power we can run through these GPUs.
Which company are you going to bet on? The company with the most available power to power GPUs,
right? So NVIDIA, where they play, is like they obviously make their GPUs much more power
efficient, right? You get a lot more compute per unit power. But I mean, that still
limits you to how much power you have to run through them, right? IREN's 4.5 gigawatts
of power and their ability to manage these data centers at scale already proven through
their Bitcoin mining, now transitioning to like a $10 billion Microsoft contract is exactly
why I invested in them.
Yeah, people hear about it. And sometimes I think the
Bitcoin mining part throws them off a little bit, right? But there is a lot of similarities. What
are the differences? Like what are the things that they're going to have to rip out and say,
like, all right, we were doing this before, we have to change that if we want to be successful?
Yeah. So just at a high level, ASICs for Bitcoin mining are fundamentally different from the
GPUs that they're trying to serve and install, right?
So there is a big rip, replace, and depreciation component to their stuff.
For example, they already have a lot of assets, and we saw this in their last couple earnings
calls, where they have to say, hey, we bought these Bitcoin miners.
We're obviously not using them to their intended lifetime.
We have to depreciate these assets and take a loss because we're pivoting to this different
set of infrastructure, which now we need to spend even more money on, right?
So that's a bitter pill to swallow for people who are IREN investors back when IREN was
Iris Energy and like they were focused on Bitcoin mining exclusively.
And that was the value proposition, right?
So that sucks.
It's a hard transition.
It's big. It's capital intensive. They're sort of fighting two wars at once. They're competing
with people who are trying to do this full-time very successfully while depreciating things that
they just bought to use for a different application. So there's going to be one of
the things they'll definitely have to prove is like their investing discipline. Are they spending
capex super efficiently?
How are they getting the money without diluting shareholders?
Which, by the way, they probably are going to dilute shareholders.
So there's like a lot of those sort of more, I'd say, fiscal engineering questions like,
hey, diluting shareholders, issuing new stock, raising new capital,
as opposed to like the technical risks of just building data centers and running them,
which we know they can do just for this different application, right?
Yeah.
Sorry, did that answer your question?
I didn't mean to.
No, it's perfect.
Yeah, yeah, yeah.
It literally just outlined exactly what I was trying to get you to say, which was, hey,
they have all this stuff that they're going to have to rip out and depreciate off and
they're going to have to buy a whole bunch of new things and put them in.
And this is not a $4 trillion company. This is a $13.5 billion company with unstable revenue,
one might say, in this area. So it is a very different picture.
And you're seeing like just to be just to drive it home in their earnings, you're seeing that
show up as big losses, which means big losses per share, right? In their in sorry, in their
earnings report, in their earnings per share, you're seeing huge one-time losses
associated with these things. So it really materially impacts the stock in a big way.
Exactly. How do you position size in your portfolio something like NVIDIA versus IRON?
I actually, so just for context, I'm 38. So I don't retire for like hopefully many years, right?
I actually am much more conservative than I think my channel lets on.
I have a lot more in NVIDIA than in something like IREN,
even though NVIDIA does have, I believe, much less upside than a company like IREN, right?
It's not hard to imagine a world where IREN doubles.
It's already hard to imagine a world where NVIDIA doubles, right?
That puts them in the like $9 trillion category relatively soon, right? Oh, yeah.
Yeah. What do I actually care about as like a long term investor? I think I forget who said
this, it might have been Munger, it might have actually been Warren Buffett himself.
The goal isn't to be an above average investor, the goal is to be an average investor for an above average
period of time, right? So I don't care about making 200, 300, a gajillion percent on my money.
I think that's actually a fairly unreasonable expectation. And then when it doesn't happen,
you know, emotions run high, everyone's upset, you're not getting what you wanted, right?
So I have a little bit of money in private equity.
Like it's substantial today because things have like gone up in value, but the amount I actually invested is, you know, fairly small compared to its value today.
And then most of my money is in things that I think will just normally beat the market,
right?
Google, Nvidia, Broadcom, Meta platforms.
These are not like high risk, high reward names to any stretch
of the imagination. But what these all companies have in common is they've continuously beaten the
S&P 500. And I believe they're fairly low risk investments. So when Meta tanked, not to 100,
I missed that ride. But like there have been several other times they've tanked due to tariffs.
I missed that ride, but like there have been several other times they've tanked due to tariffs.
I bought aggressively, right? When NVIDIA goes down 10, 15, 20, 30%, I buy aggressively. These
are big, stable names. And that I would say makes up probably two thirds of my portfolio, right?
The other one third, maybe even one quarter is names like IREN, CoreWeave, Nebius, Vertiv,
IREN, CoreWeave, Nebius, Vertiv, Micron, right? These names that are not, sorry, they're core to
the AI ecosystem, but they're not the center of it, right? Like NVIDIA and Broadcom and TSMC are
the center. And then there's like the second layer, hey, Micron just does memory. Vertiv does
power and cooling. These other three companies, CoreWeave, IREN, and Nebius, they take NVIDIA chips and they
build cloud services on top of them that they rent out, right? They're like auxiliary to the core
players, right? So less of my money goes in there knowing, hey, there's more risk, but there's also
more upside. So overall, I tend to outperform the market by a factor of two or three. So if the
market, just to be clear, that means if the market goes up 15%, I'll probably do 30 to 45%, not 100%, not 1,000%. I don't do options. I'm not trying to claim I'm
like the best investor ever. But if you outperform the market by a factor of two or three for a
decade, right, you're doing pretty well. So I think that's the part that a lot of people miss.
Sorry, go ahead. Yeah, most hedge funds underperform the SB500. So yeah. Yes. I don't charge fees, right? Like
people watch my content for free. They take it or they don't, but you know.
Yeah. I don't think we're gonna have a ton of time to get into TSM today, but it's another
one that people should really be looking at. We'll talk about it on our next one, but that's,
you know, up 9% year to date in a market like this was up 24% year to date at one point,
even in this market,
something people should definitely be paying attention to.
NVIDIA GTC is going to be happening this coming week.
So I encourage people to check that out.
There's going to be a keynote from Jensen on Monday.
I believe it's the 16th.
The people can go watch and just get more insights
and pieces along those lines.
Obviously, people can follow you on social media,
both on Twitter and on YouTube to get a ton
of great information. Anywhere else that you're posting? That's it. YouTube mainly and then
Twitter. YouTube and Twitter. Go check it out so you can get kind of that real-time flow of thoughts.
Anything else you want to share on this one just to kind of wrap up the thoughts and anything else
you're excited for in terms of GTC? No, I'm super happy to be here. I'm super happy to cover GTC.
I'm obviously going to be covering it live, like through Twitter, especially during the keynote. And then I'm going to do a big
post-GTC wrap-up video being like, hey, here's all the signal I found. Here's what I think is
still noise. Here are the industries I think are much closer than we expected. And just going over
like my general analysis of the whole AI ecosystem from NVIDIA's vantage point sort of at GTC.
Perfect. I encourage people to watch that. It's always smart with these companies,
you know, trust, but verify continuously, right? You want to make sure that there's nothing out
of the order that's happening. And transparently, sometimes you even find something that gives you
more confidence, right? And leads you to potentially making additional investments,
right? Buying those dips, doubling down in areas where there is a sell-off. That just comes from staying in the know
and knowledge with these things.
You can't just look at it one time,
do the research one time and forget.
It's not my personal approach,
unless maybe you're buying MSP 500, right?
But if you're going to buy individual stocks,
please, you know, trust, but verify continuously.
I think that that's the right way to go, right?
Totally, especially when the industry
you're talking about is moving so fast, right? Like the stuff that you knew about AI three months ago is already obsolete today. So
yeah, put as they say. All right. Thanks so much, everyone for tuning in and watching here on X. This
is also going to be uploaded to YouTube. Drop any questions you have in the comments. We'll try to
get to those and maybe insert them into our next video with Alex. Alex, it's always a pleasure
having you on. Wealth of knowledge. I come out of these feeling so much more educated. I know the
audience does too. So we appreciate you and wishing you a great trip to GTC. Thank you. Thanks so much
for having me. Looking forward to it. Absolutely. Take care, everybody else. Have a great rest of
your Thursday and an awesome weekend. We'll see you on the next one. I'm going to go ahead and end
this one here.
