1
00:00:00,270 --> 00:00:03,490
Ejaaz:
For three years, the AI trade has been one simple idea.

2
00:00:03,660 --> 00:00:07,880
Ejaaz:
Buy the chips. NVIDIA became the most valuable company on Earth by making these

3
00:00:07,880 --> 00:00:10,730
Ejaaz:
chips, and they've made countless of investors rich because of this.

4
00:00:11,060 --> 00:00:15,690
Ejaaz:
But this investment thesis now has become overcrowded. Not because AI is slowing

5
00:00:15,690 --> 00:00:18,260
Ejaaz:
down, but because there's another problem that has surfaced.

6
00:00:18,440 --> 00:00:22,660
Ejaaz:
No one can turn these GPUs on. Trillions of dollars being spent in AI,

7
00:00:22,850 --> 00:00:26,080
Ejaaz:
and all these GPUs are collecting dust in data centers.

8
00:00:26,080 --> 00:00:30,400
Ejaaz:
The problem, energy, electrical grids, wiring, networking, the transformers,

9
00:00:30,400 --> 00:00:34,720
Ejaaz:
the web of infrastructure that sits around a GPU that allows it to keep alive,

10
00:00:34,720 --> 00:00:38,310
Ejaaz:
to talk to each other, to transmit petabytes of data between each other.

11
00:00:38,310 --> 00:00:42,080
Ejaaz:
This is the next constraint that hasn't been solved yet and where the majority

12
00:00:42,080 --> 00:00:44,000
Ejaaz:
of the AI capital will eventually flow.

13
00:00:44,200 --> 00:00:48,070
Ejaaz:
Now, there's four physical things that keep a chip alive, essentially.

14
00:00:48,070 --> 00:00:51,130
Ejaaz:
The power to run it, the light to transmit all the data between them,

15
00:00:51,130 --> 00:00:55,120
Ejaaz:
the silicon wiring between them all, and then somewhere to rent it all from.

16
00:00:55,120 --> 00:00:58,120
Ejaaz:
And almost nobody is really talking about this. And on this episode,

17
00:00:58,120 --> 00:01:00,980
Ejaaz:
we're going to unpack that specific layer of the infrastructure stack,

18
00:01:00,980 --> 00:01:05,120
Ejaaz:
the power layer, and why it's so important going forwards in terms of capital.

19
00:01:05,120 --> 00:01:07,100
Josh:
Yeah, and this may feel a little bit different because we're always talking

20
00:01:07,100 --> 00:01:09,730
Josh:
about whose model is the smartest. That's the common conversation.

21
00:01:09,730 --> 00:01:14,410
Josh:
But once you strip away all that abstraction, a token is really just what,

22
00:01:14,960 --> 00:01:18,680
Josh:
is output and that is just a series of electrons flowing through chips like

23
00:01:18,680 --> 00:01:22,240
Josh:
flowing through fiber and then heat being pulled out of the system and all of

24
00:01:22,240 --> 00:01:25,150
Josh:
that requires a lot of energy and electricity uh

25
00:01:25,570 --> 00:01:29,370
Josh:
the money cycle that we're going to talk about has kind of flowed towards this

26
00:01:29,370 --> 00:01:33,110
Josh:
direction as well i mean everyone started with the crowded gpu trade then it

27
00:01:33,110 --> 00:01:36,650
Josh:
flowed to the semis that exist around them then it flowed to the memory trade

28
00:01:36,940 --> 00:01:39,710
Josh:
and now it's kind of moving over to the

29
00:01:40,070 --> 00:01:43,280
Josh:
seemingly the end game bottleneck, which is energy.

30
00:01:43,450 --> 00:01:47,770
Josh:
I mean, US data power center demand, I think is projected to roughly double

31
00:01:47,770 --> 00:01:53,750
Josh:
80 gigawatts in 2026 to 150 by 2028. So the grid's just not really built for that.

32
00:01:54,090 --> 00:01:58,020
Josh:
And there's no shortage of demand for that. And when we think about energy as

33
00:01:58,020 --> 00:02:02,900
Josh:
an idea, even if you believe that the AI bubble is towards the latter end of

34
00:02:02,900 --> 00:02:05,070
Josh:
it, energy is still something that's not going away.

35
00:02:05,070 --> 00:02:10,140
Josh:
We had energy problems prior to the LLM becoming a big deal from the chat GPT moment.

36
00:02:10,400 --> 00:02:13,810
Josh:
This is just an extension and an exaggeration on top of that.

37
00:02:14,110 --> 00:02:18,360
Josh:
And it creates a lot of interesting opportunities in the marketplace to actually

38
00:02:18,360 --> 00:02:23,200
Josh:
participate in this bottleneck that is now known as, I guess we call it the energy bottleneck.

39
00:02:23,310 --> 00:02:26,950
Ejaaz:
Yeah. And I think a lot of the reason why people don't talk about this is because

40
00:02:26,950 --> 00:02:31,680
Ejaaz:
it's sort of unsexy, you know, not everyone is an electrical grid expert.

41
00:02:31,680 --> 00:02:34,990
Ejaaz:
Not everyone knows how to network wires between these complex GPUs.

42
00:02:35,330 --> 00:02:39,010
Ejaaz:
I'm still trying to wrap my head around the GPU itself, So the fact that we're

43
00:02:39,010 --> 00:02:41,200
Ejaaz:
bringing in all these other things, it's quite complex.

44
00:02:41,570 --> 00:02:47,970
Ejaaz:
Now, if you were to look at a budget of spending for an AI lab or someone that's

45
00:02:47,970 --> 00:02:53,490
Ejaaz:
setting up a data center, only 10% of that budget gets allocated to the power

46
00:02:53,490 --> 00:02:54,710
Ejaaz:
and infrastructure side of things.

47
00:02:55,040 --> 00:03:00,730
Ejaaz:
The irony of it now is it's going close to 100% of the bottleneck of the entire

48
00:03:00,730 --> 00:03:03,830
Ejaaz:
thing. So even though it's not the majority of the cost, still 90% of the cost

49
00:03:03,830 --> 00:03:06,730
Ejaaz:
goes into the actual GPUs itself. That's the most expensive stuff,

50
00:03:06,730 --> 00:03:09,900
Ejaaz:
the memory and the silicon required to build those GPUs.

51
00:03:10,460 --> 00:03:15,690
Ejaaz:
Only 10% is allocated to power, but that is like the main bottleneck that we're facing today.

52
00:03:15,930 --> 00:03:20,010
Ejaaz:
Now, the lead time, if you are an AI lab, you could be the richest,

53
00:03:20,010 --> 00:03:21,830
Ejaaz:
most important person on earth right now.

54
00:03:22,010 --> 00:03:25,310
Ejaaz:
If you want to build a data center in the US, the lead time to get the necessary

55
00:03:25,310 --> 00:03:28,350
Ejaaz:
power to your GPUs is five years.

56
00:03:28,650 --> 00:03:31,560
Ejaaz:
And there are different sectors within like the power thing.

57
00:03:31,560 --> 00:03:34,350
Ejaaz:
It's not just the energy infrastructure grid you need to get access to.

58
00:03:34,630 --> 00:03:38,080
Ejaaz:
You need to get lead times for wiring and to get transformers designed,

59
00:03:38,080 --> 00:03:41,180
Ejaaz:
custom made and delivered to you. All of that takes a lot of time.

60
00:03:41,180 --> 00:03:43,650
Ejaaz:
And we're looking at basically like half a decade.

61
00:03:43,920 --> 00:03:48,570
Ejaaz:
So what we've seen a lot of these AI labs start to do now is figure out what

62
00:03:48,740 --> 00:03:53,300
Ejaaz:
alternatives they could potentially procure to help them get GPUs online.

63
00:03:53,470 --> 00:03:59,350
Ejaaz:
Now, you've all heard the news of Meta, Amazon, Microsoft, and Google spending

64
00:03:59,350 --> 00:04:01,200
Ejaaz:
trillions of dollars this year alone.

65
00:04:01,200 --> 00:04:04,390
Ejaaz:
I think it was something along the lines of, actually, it might have hit a trillion

66
00:04:04,390 --> 00:04:08,400
Ejaaz:
dollars for this year, right? Because I think after the last quarterly earnings,

67
00:04:08,400 --> 00:04:09,450
Ejaaz:
they like upped their budget.

68
00:04:09,740 --> 00:04:13,350
Ejaaz:
They haven't set those data centers up. In fact, there have been delays.

69
00:04:13,770 --> 00:04:17,960
Ejaaz:
Stargate from OpenAI, which said that was their biggest project to kind of get

70
00:04:17,960 --> 00:04:20,290
Ejaaz:
compute online, has been delayed multiple times.

71
00:04:20,290 --> 00:04:24,200
Ejaaz:
And the point around this is because of compute, and not many people are talking about this yet.

72
00:04:24,200 --> 00:04:28,060
Ejaaz:
So I think on today's episode, as we walk down this infrastructure stack,

73
00:04:28,060 --> 00:04:31,430
Ejaaz:
and we did a previous episode before, where we covered all the layers of the

74
00:04:31,430 --> 00:04:32,470
Ejaaz:
AI infrastructure stack, we

75
00:04:33,270 --> 00:04:37,240
Ejaaz:
Talked about going from layer zero, which is the model labs down to layer one,

76
00:04:37,240 --> 00:04:41,750
Ejaaz:
which was the hyperscalers, GPU semiconductors, we didn't dig as much into layer

77
00:04:41,750 --> 00:04:43,890
Ejaaz:
five, which is the power and infrastructure side of things.

78
00:04:44,180 --> 00:04:46,370
Ejaaz:
And that's what we're going to do today. And we're going to do it through the

79
00:04:46,370 --> 00:04:50,260
Ejaaz:
lens of four specific companies, starting with Bloom Energy.

80
00:04:50,570 --> 00:04:53,500
Josh:
Yeah. And just before we get into Bloom, there's one thing. Well,

81
00:04:53,500 --> 00:04:56,500
Josh:
one of the things is, I think there's like these bottlenecks within bottlenecks.

82
00:04:56,500 --> 00:05:01,370
Josh:
You mentioned the lead time, but also the actual infrastructure is becoming much more demanding.

83
00:05:01,500 --> 00:05:05,100
Josh:
Vera Rubin, the new chips, I believe, those server racks take...

84
00:05:05,470 --> 00:05:10,060
Josh:
Over 100 kilowatts, which is a 10x increase from what these companies and these

85
00:05:10,060 --> 00:05:11,350
Josh:
data centers were preparing for.

86
00:05:11,660 --> 00:05:15,620
Josh:
And then also before that, I want to kind of loosely define the power for people

87
00:05:15,620 --> 00:05:18,630
Josh:
who are not familiar, because there's watts, there's watts hours,

88
00:05:18,630 --> 00:05:21,930
Josh:
there's gigawatts, it's just like a brief one on one quickly on what everything is.

89
00:05:21,930 --> 00:05:25,430
Josh:
A watt is a rate, it's how fast these data centers draw electricity,

90
00:05:25,790 --> 00:05:31,060
Josh:
a watt hour is an amount, which is the total you drew. So a gigawatt is a billion of those watts drawn.

91
00:05:31,370 --> 00:05:36,370
Josh:
One gigawatt is equal to roughly one large nuclear reactor or enough to power

92
00:05:36,550 --> 00:05:41,030
Josh:
750 to a million U.S. homes. So this is a tremendous amount of data that we're

93
00:05:41,030 --> 00:05:43,120
Josh:
talking about or a tremendous amount of energy that we're talking about.

94
00:05:43,350 --> 00:05:47,670
Josh:
And these data centers are demanding tens to hopefully soon hundreds of gigawatts.

95
00:05:47,920 --> 00:05:49,020
Josh:
Now back to Bloom Energy.

96
00:05:49,330 --> 00:05:53,870
Josh:
Bloom Energy has a solution for this. It is solving the most painful solution

97
00:05:53,870 --> 00:05:55,610
Josh:
in AI, which is this energy data source.

98
00:05:55,610 --> 00:05:59,670
Josh:
They use these kind of modular energy reactors that allow them to allow these

99
00:05:59,670 --> 00:06:03,450
Josh:
data centers to remove themselves off of the grid and generate a little bit

100
00:06:03,450 --> 00:06:05,370
Josh:
more energy closer to the data center.

101
00:06:05,530 --> 00:06:09,060
Ejaaz:
Yeah, actually, remember, the first time I heard about Bloom Energy at all was

102
00:06:09,060 --> 00:06:13,170
Ejaaz:
in Leopold Ashenbrenner's 13 filings at the end of last year.

103
00:06:13,170 --> 00:06:16,560
Ejaaz:
It was one of his biggest positions. I think in his recent one,

104
00:06:16,560 --> 00:06:20,540
Ejaaz:
I think it makes up about 12.7% of his entire portfolio.

105
00:06:20,540 --> 00:06:23,940
Ejaaz:
So he's very bullish, this particular company. And the reason why is,

106
00:06:24,750 --> 00:06:29,160
Ejaaz:
As I mentioned earlier, the lead times to getting power generation on site for

107
00:06:29,160 --> 00:06:33,130
Ejaaz:
your data center, if you go through the old traditional way, is around half a decade.

108
00:06:33,560 --> 00:06:39,030
Ejaaz:
Bloom energy will reduce that down to 90 days they have a product it's a gas

109
00:06:39,030 --> 00:06:42,880
Ejaaz:
turbine which they can custom build and deliver on site they kind of like fly

110
00:06:42,880 --> 00:06:44,120
Ejaaz:
it in and they reassemble it

111
00:06:44,430 --> 00:06:49,040
Ejaaz:
and it can generate enough power for you to run your gpus now obviously that's

112
00:06:49,520 --> 00:06:52,160
Ejaaz:
A very attractive thing a ton of companies have signed contracts with them now

113
00:06:52,160 --> 00:06:55,960
Ejaaz:
the one caveat that i will say is they haven't really delivered this at scale

114
00:06:55,960 --> 00:07:01,760
Ejaaz:
bloom energy has yet to prove themselves from a manufacturing capacity and delivery capacity, but it is

115
00:07:02,060 --> 00:07:06,150
Ejaaz:
a company that a lot of people like to talk about and kind of speak highly of.

116
00:07:06,400 --> 00:07:09,410
Ejaaz:
Now, the revenue kind of speaks for itself as well, because obviously there's

117
00:07:09,410 --> 00:07:11,640
Ejaaz:
a lot of lead time here, there's a lot of demand here. And as we mentioned,

118
00:07:11,640 --> 00:07:12,880
Ejaaz:
that money is going to flow down.

119
00:07:13,040 --> 00:07:17,150
Ejaaz:
As of Q1, 2026, their revenue hit $750 million.

120
00:07:17,150 --> 00:07:25,160
Ejaaz:
That is up 130% of revenue year over year, which is just like a crazy kind of swing.

121
00:07:25,160 --> 00:07:29,830
Ejaaz:
And of that, they're making around 70 million profit, which isn't large in the

122
00:07:29,830 --> 00:07:33,070
Ejaaz:
context of some of these big hyperscaling infrastructure companies.

123
00:07:33,070 --> 00:07:36,710
Ejaaz:
But for a relatively small company, this is pretty impressive.

124
00:07:36,990 --> 00:07:43,430
Ejaaz:
And Oracle is powering its 2.4 gigawatt data center, primarily using Bloom Energy.

125
00:07:43,430 --> 00:07:45,790
Ejaaz:
So they've signed some pretty major contracts as well.

126
00:07:45,980 --> 00:07:49,140
Ejaaz:
So Bloom Energy is one of these companies, and there's various different versions

127
00:07:49,140 --> 00:07:52,690
Ejaaz:
of this, where it's not just gas turbines, maybe it's solar energy or nuclear

128
00:07:52,690 --> 00:07:56,800
Ejaaz:
energy. But Bloom Energy is basically the first company that is having an impact

129
00:07:56,800 --> 00:08:00,250
Ejaaz:
almost immediately after this constraint has been identified.

130
00:08:00,540 --> 00:08:03,110
Josh:
Yeah, well, in terms of gas turbines, I mean, that's one way of doing it.

131
00:08:03,490 --> 00:08:08,970
Josh:
And the line to get a gas turbine extends back to 2029. So if you actually want gas turbines.

132
00:08:09,440 --> 00:08:12,730
Josh:
A gas turbine at your data center and you place an order today.

133
00:08:13,160 --> 00:08:16,160
Josh:
And for those who aren't familiar, a gas turbine is basically a jet engine,

134
00:08:16,160 --> 00:08:17,610
Josh:
which is spinning a generator.

135
00:08:17,970 --> 00:08:21,080
Josh:
And it's so funny how this works. It's like the way we generate energy today

136
00:08:21,080 --> 00:08:23,430
Josh:
is the same way we did with like a steam engine way back in the day.

137
00:08:23,700 --> 00:08:27,020
Josh:
You're just basically creating energy to propel a spinning thing.

138
00:08:27,020 --> 00:08:28,120
Josh:
In this case, it's a jet engine.

139
00:08:28,280 --> 00:08:30,940
Josh:
If you want one of those, you're waiting at least three years,

140
00:08:30,940 --> 00:08:34,700
Josh:
probably the next decade. So if you want something in the 2020s, that's not for you.

141
00:08:34,950 --> 00:08:37,350
Josh:
The next option is nuclear or small nuclear reactors.

142
00:08:37,760 --> 00:08:41,740
Josh:
Those are amazing. They're incredible. But if you want them in this decade,

143
00:08:41,740 --> 00:08:46,520
Josh:
again, not possible. Those aren't probably coming online at scale until around 2030.

144
00:08:46,770 --> 00:08:49,200
Josh:
And the licensing and legislation around that is pretty slow.

145
00:08:49,440 --> 00:08:52,000
Josh:
So that leaves us with these fuel cells. And that's what Bloom has.

146
00:08:52,000 --> 00:08:57,070
Josh:
And they're these modular boxes you stand up on site in 90 days instead of three to five years.

147
00:08:57,340 --> 00:09:01,020
Josh:
And so far, people have really been liking them. We've seen that reflected in the stock price.

148
00:09:01,420 --> 00:09:04,820
Josh:
With people desperate for energy, anyone who's able to actually deliver it.

149
00:09:05,460 --> 00:09:09,420
Josh:
A accelerated timeline is going to win. I think about Elon all the time and

150
00:09:09,420 --> 00:09:13,180
Josh:
how they were able to build the Colossus data center and how much premium it

151
00:09:13,180 --> 00:09:15,400
Josh:
had to getting GPUs online quickly.

152
00:09:15,600 --> 00:09:18,940
Josh:
I mean, think about how much Anthropic and Google are paying for these GPUs.

153
00:09:19,100 --> 00:09:20,690
Josh:
The same thing is true with electricity.

154
00:09:21,090 --> 00:09:25,190
Josh:
If a data center is bottlenecked by this power, anyone who's able to provide

155
00:09:25,190 --> 00:09:29,120
Josh:
it for them is going to get paid a tremendous amount of money and a large margin for it.

156
00:09:29,450 --> 00:09:32,730
Josh:
And so far, Bloom has been one of those companies through their fuel cell program

157
00:09:32,730 --> 00:09:36,190
Josh:
that can be deployed in 90 days on site, they're collecting a lot of that value.

158
00:09:36,190 --> 00:09:39,760
Josh:
So I think that's kind of the bull case for Bloom. That's why people are mostly excited about it.

159
00:09:39,760 --> 00:09:44,950
Josh:
It is just a really difficult place for companies because it's hard to make

160
00:09:44,950 --> 00:09:48,500
Josh:
energy. And a company that has figured out a way, I'm not sure this is the end

161
00:09:48,500 --> 00:09:51,430
Josh:
state, but this is certainly a intermediary step.

162
00:09:51,690 --> 00:09:54,310
Josh:
And right now, that's all people need. It's just power today.

163
00:09:54,580 --> 00:09:58,630
Ejaaz:
Now, moving on, one of the other bottlenecks that is appearing,

164
00:09:59,170 --> 00:10:04,040
Ejaaz:
particularly with GPUs talking to each other, is data transfer specifically.

165
00:10:04,040 --> 00:10:10,550
Ejaaz:
Now, if we think of an AI model, typically it has grown exponentially in size

166
00:10:10,550 --> 00:10:14,340
Ejaaz:
and in quantity of data that is required to train a frontier AI model.

167
00:10:14,340 --> 00:10:19,030
Ejaaz:
We've gone from, if you remember, Josh, billions of parameters to trillions

168
00:10:19,030 --> 00:10:21,850
Ejaaz:
of parameters and now tens of trillions of parameters.

169
00:10:22,680 --> 00:10:27,740
Ejaaz:
So the model weights of these actual AI models are huge. People are carrying

170
00:10:27,740 --> 00:10:29,670
Ejaaz:
them around in briefcases.

171
00:10:29,880 --> 00:10:31,620
Ejaaz:
They're like incredibly dense.

172
00:10:31,870 --> 00:10:34,520
Ejaaz:
And when you're training a model, and when you're even inferencing a model,

173
00:10:34,520 --> 00:10:36,660
Ejaaz:
when you're trying to speak to it, when you're trying to send it a prompt and

174
00:10:36,660 --> 00:10:42,380
Ejaaz:
get a response, there is petabytes of data, which is an order of magnitude larger

175
00:10:42,380 --> 00:10:46,370
Ejaaz:
than anything we've ever really spoken about at scale in industry, at least, before.

176
00:10:46,880 --> 00:10:50,520
Ejaaz:
That are transferred between GPUs. So it's not just good enough to have the

177
00:10:50,520 --> 00:10:53,750
Ejaaz:
GPUs and wiring them up and having the power flowing to them.

178
00:10:53,940 --> 00:10:57,120
Ejaaz:
You need to transfer data between them. They need to talk to each other in rapid

179
00:10:57,120 --> 00:11:01,520
Ejaaz:
pace. If they talk to each other slowly, then they won't be able to work as

180
00:11:01,520 --> 00:11:04,160
Ejaaz:
quickly and as powerfully as you'd expect. Think of it like this.

181
00:11:04,440 --> 00:11:08,930
Ejaaz:
Every single way that a GPU requires to function, the lifeblood of it,

182
00:11:08,930 --> 00:11:12,790
Ejaaz:
the energy, the ability to communicate is a potential constraint.

183
00:11:12,790 --> 00:11:15,010
Ejaaz:
And this communication thing is another one.

184
00:11:15,240 --> 00:11:18,900
Ejaaz:
Now, typically, the way you achieve this is through copper wiring.

185
00:11:18,900 --> 00:11:23,110
Ejaaz:
Copper has a really interesting chemical property that allows you to transmit

186
00:11:23,110 --> 00:11:27,910
Ejaaz:
data at lightning speed without heating up, except we've run into a problem.

187
00:11:28,340 --> 00:11:32,330
Ejaaz:
We now have too much data. And so the copper wires are heating up,

188
00:11:32,550 --> 00:11:36,480
Ejaaz:
which causes energy to dissipate and energy efficiency and cost efficiency,

189
00:11:36,480 --> 00:11:41,270
Ejaaz:
therefore, to not really work in the favor of the AI capex that is being spent on this.

190
00:11:41,530 --> 00:11:44,220
Ejaaz:
So people have started scrambling around. They've started looking around for

191
00:11:44,220 --> 00:11:46,030
Ejaaz:
another type of material.

192
00:11:46,370 --> 00:11:51,780
Ejaaz:
And you wouldn't believe it. They've settled on light, optical fibers to transmit data using light.

193
00:11:51,780 --> 00:11:56,320
Ejaaz:
And one of the companies that is moving data as light is known as Lumenta.

194
00:11:56,320 --> 00:12:00,780
Ejaaz:
They're using optical fibers and laser transmitters to move data between chips

195
00:12:00,780 --> 00:12:06,710
Ejaaz:
as light because copper melts at a certain level where data becomes too burdensome, essentially.

196
00:12:07,070 --> 00:12:10,030
Josh:
Yeah, the bottleneck was very much copper and it might still be to some extent

197
00:12:10,030 --> 00:12:13,300
Josh:
because copper is the fastest way of transferring data just because it has so

198
00:12:13,300 --> 00:12:15,540
Josh:
much bandwidth. It's so dense and you could transfer.

199
00:12:15,710 --> 00:12:19,930
Josh:
I mean, now we have clusters of hopefully soon half a million GPUs.

200
00:12:19,930 --> 00:12:21,300
Josh:
I know that's what people are working on.

201
00:12:21,600 --> 00:12:25,330
Josh:
All those gpus need to communicate they need to do so as fast as possible copper

202
00:12:25,330 --> 00:12:30,230
Josh:
has enough bandwidth unfortunately there's this direct correlation between the amount of data

203
00:12:30,550 --> 00:12:34,970
Josh:
and the length it has to travel it's physically impossible to put that many

204
00:12:34,970 --> 00:12:39,060
Josh:
gpus that close to each other to make it feasible there's just not enough space

205
00:12:39,060 --> 00:12:42,850
Josh:
inside of a room to make it happen so as these data centers get larger the

206
00:12:43,170 --> 00:12:46,690
Josh:
distance in which jane needs to travel gets further copper becomes a worse and

207
00:12:46,690 --> 00:12:51,720
Josh:
worse option because i mean every single bit that has to flow through that it causes heat.

208
00:12:52,270 --> 00:12:55,240
Josh:
It causes energy and then like you mentioned the copper it's going to start

209
00:12:55,240 --> 00:12:58,160
Josh:
melting and it's really expensive to cool and it's really difficult so what's

210
00:12:58,160 --> 00:13:02,590
Josh:
the next best thing is photonics you just use light and you use fiber optics

211
00:13:02,590 --> 00:13:05,250
Josh:
and nothing travels faster than the speed of light so

212
00:13:05,620 --> 00:13:09,190
Josh:
it's just a matter of how much bandwidth we could squeeze into those rays of

213
00:13:09,190 --> 00:13:14,110
Josh:
light and by stacking this fiber and by using this glass fiber we're able to actually transfer,

214
00:13:14,590 --> 00:13:18,940
Josh:
information from one ship to another at the speed of light and lumentum is a

215
00:13:18,940 --> 00:13:21,980
Josh:
company that's working on that so i think that's why people are really excited

216
00:13:21,980 --> 00:13:26,490
Josh:
about this company in particular as these data centers grow larger in footprint

217
00:13:26,840 --> 00:13:27,980
Josh:
they are going to require

218
00:13:28,370 --> 00:13:31,990
Josh:
a longer distance in which data has to travel and there's going to be a lot

219
00:13:31,990 --> 00:13:36,650
Josh:
more data which means copper probably won't be ideal and unless we solve this

220
00:13:36,650 --> 00:13:38,490
Josh:
material problem the next best thing

221
00:13:38,750 --> 00:13:43,570
Josh:
is photonics it's just sending it right over a fiber optic line at the speed of light or i should say,

222
00:13:44,080 --> 00:13:48,130
Josh:
many fiber optic lines to store as much data as possible in this transfer.

223
00:13:48,450 --> 00:13:51,580
Josh:
And that's why Lumentum is, I think, one of the more exciting companies.

224
00:13:51,580 --> 00:13:55,440
Josh:
This is part of Jensen's portfolio, if I am not mistaken, right?

225
00:13:55,780 --> 00:13:59,400
Ejaaz:
Yeah, yeah. He's put in about $2 billion into Lumentum Holdings.

226
00:13:59,400 --> 00:14:03,680
Ejaaz:
But to your point, it's important to mention that this is a relatively new type

227
00:14:03,680 --> 00:14:09,160
Ejaaz:
of technology. It hasn't quite hit that exponential wave that NVIDIA GPUs did two years ago.

228
00:14:09,400 --> 00:14:12,710
Ejaaz:
And so we're waiting on that. The bet is that that is going to happen for these

229
00:14:12,710 --> 00:14:18,520
Ejaaz:
particular types of transmitters and receivers. Now, their unit of sales is

230
00:14:18,520 --> 00:14:20,360
Ejaaz:
going up pretty aggressively.

231
00:14:20,910 --> 00:14:23,510
Ejaaz:
Last year, they sold, I think, around like 20 million units.

232
00:14:23,510 --> 00:14:26,950
Ejaaz:
This year, it's now 60 million units. So it's kind of like a two and a half

233
00:14:26,950 --> 00:14:28,930
Ejaaz:
to three X from what they were last year.

234
00:14:29,160 --> 00:14:35,150
Ejaaz:
But they have competitors in the form of Corning, GLW, who NVIDIA also invested $3.2 billion.

235
00:14:35,510 --> 00:14:39,280
Ejaaz:
Now, they're not direct competitors, but Corning creates the actual optical

236
00:14:39,280 --> 00:14:43,730
Ejaaz:
glass fibers that will be used to kind of like transmit light and data for this.

237
00:14:44,040 --> 00:14:47,120
Ejaaz:
And there are many other kind of like companies that are like highly focused on this.

238
00:14:47,360 --> 00:14:52,820
Ejaaz:
It's worth mentioning though that scaling any of these material goods,

239
00:14:52,820 --> 00:14:55,790
Ejaaz:
whether it's the glass fibers, whether it's transmitters, whether it's the power

240
00:14:55,790 --> 00:15:00,660
Ejaaz:
switches or whatever that might be, takes an incredible amount of lead time.

241
00:15:00,660 --> 00:15:05,920
Ejaaz:
Like one of the major critics of the memory bottleneck that everyone is quite

242
00:15:05,920 --> 00:15:09,260
Ejaaz:
familiar with right now is Apple just recently announced that they're raising

243
00:15:09,260 --> 00:15:12,960
Ejaaz:
the prices of all their devices because they can't get their hands on enough memory.

244
00:15:13,150 --> 00:15:16,730
Ejaaz:
NVIDIA is facing the same thing. And when they go to SK Hynix and the memory

245
00:15:16,730 --> 00:15:21,030
Ejaaz:
makers and they say, hey, can you just build more, please? They say, it takes time.

246
00:15:21,330 --> 00:15:25,850
Ejaaz:
The machinery is incredibly detailed. Like it takes a while and we're gonna

247
00:15:25,850 --> 00:15:29,530
Ejaaz:
be in this for a few years. And the optical fiber side of things,

248
00:15:29,530 --> 00:15:34,440
Ejaaz:
the material side of power infrastructure in itself is where this is also playing

249
00:15:34,440 --> 00:15:36,230
Ejaaz:
out. nothing's really changed in that sense.

250
00:15:36,730 --> 00:15:40,880
Josh:
Also, I just sent you a custom link of a post that I'd like to share because

251
00:15:40,880 --> 00:15:44,580
Josh:
this reminded me of an idea that I heard that I think is probably relevant because

252
00:15:44,580 --> 00:15:47,770
Josh:
the metric underneath all of this is joules per bit. It's the energy cost to

253
00:15:47,770 --> 00:15:51,020
Josh:
move one bit. And at scale, light wins decisively.

254
00:15:51,290 --> 00:15:55,210
Josh:
Like there is nothing that flows smoother than light. And it reminded me of

255
00:15:55,210 --> 00:15:56,510
Josh:
this post from John Carmack.

256
00:15:56,880 --> 00:15:59,770
Josh:
He is the guy who created Dune. He's like this internet legend.

257
00:16:00,030 --> 00:16:05,680
Josh:
And basically, John was saying that you can store data within this light if

258
00:16:05,680 --> 00:16:07,760
Josh:
it is transferred over a certain.

259
00:16:08,100 --> 00:16:09,990
Ejaaz:
Length of space. Oh, Elon responded to this, right?

260
00:16:09,990 --> 00:16:14,710
Josh:
Yes. It is amazing. So when we think about memory and we think about data transfer

261
00:16:14,710 --> 00:16:19,440
Josh:
in general, I think fiber optics is a really interesting industry to consider

262
00:16:19,440 --> 00:16:22,680
Josh:
that can go much further than I think what we imagine it can do today.

263
00:16:23,050 --> 00:16:27,710
Josh:
Because during the time in which this light or these bits are being transferred at the speed of light

264
00:16:28,130 --> 00:16:32,520
Josh:
you can transfer a lot of data and if that's spun over a long enough distance

265
00:16:32,760 --> 00:16:35,390
Josh:
it can actually store data in a way that is

266
00:16:35,730 --> 00:16:40,480
Josh:
very fast and accessible way faster than we get in memory and RAM because it

267
00:16:40,480 --> 00:16:43,870
Josh:
is actually moving at the speed of light and there's a lot of interesting science

268
00:16:43,870 --> 00:16:46,920
Josh:
experiments and I think general experimentation as it relates to,

269
00:16:47,520 --> 00:16:51,240
Josh:
fiber in general not only as a way of transferring data but as a way of storing

270
00:16:51,240 --> 00:16:54,510
Josh:
data should you stack it and extend it over a long period of time and i just

271
00:16:54,510 --> 00:16:55,380
Josh:
want to highlight this is

272
00:16:55,680 --> 00:16:59,860
Josh:
like lumentum is hot right now but i think the industry of fiber and fiber optics

273
00:16:59,860 --> 00:17:03,030
Josh:
and moving things around at the speed of light is ultimately where we're going

274
00:17:03,030 --> 00:17:05,170
Josh:
to land on we talk about data centers in space

275
00:17:05,600 --> 00:17:07,900
Josh:
They're going to be traveling at the speed of light. It has the lowest amount

276
00:17:07,900 --> 00:17:10,190
Josh:
of friction. It has the highest velocity of information.

277
00:17:10,460 --> 00:17:13,030
Josh:
If we can experiment and kind of iterate on the way that we do that,

278
00:17:13,030 --> 00:17:14,780
Josh:
I think that's something worth noting.

279
00:17:14,780 --> 00:17:18,670
Josh:
And Lumentum being the fiber optic company may stand to benefit over a much

280
00:17:18,670 --> 00:17:21,990
Josh:
longer period of time than I think people think based on this current bottleneck.

281
00:17:21,990 --> 00:17:24,320
Josh:
I just want to highlight that because I was like, huh, that's pretty cool.

282
00:17:24,320 --> 00:17:26,060
Josh:
That reminded me of something I saw back in the day.

283
00:17:26,370 --> 00:17:31,440
Ejaaz:
Oh, I love it. Now, you would think that the entire problem with the power infrastructure

284
00:17:31,440 --> 00:17:34,190
Ejaaz:
stack is just generating the power, right, Josh?

285
00:17:34,580 --> 00:17:39,400
Ejaaz:
But there's also delivering the power, making sure the right amounts of power

286
00:17:39,400 --> 00:17:42,520
Ejaaz:
go between the chips at this time. And then when they are really in demand,

287
00:17:42,520 --> 00:17:44,210
Ejaaz:
delivering even more power, right?

288
00:17:44,410 --> 00:17:49,770
Ejaaz:
That takes incredibly complex design patterns. Now, most people probably aren't

289
00:17:49,770 --> 00:17:53,060
Ejaaz:
aware of this, but NVIDIA isn't a chip manufacturing company.

290
00:17:53,060 --> 00:17:54,980
Ejaaz:
They're a chip design company.

291
00:17:55,210 --> 00:17:58,440
Ejaaz:
They design what the silicon wafer is going to look like. And then they send

292
00:17:58,440 --> 00:18:00,490
Ejaaz:
that over to their friends in Taiwan and TSMC.

293
00:18:00,700 --> 00:18:04,800
Ejaaz:
And they actually manufacture the thing for them. Now, one key component of

294
00:18:04,800 --> 00:18:09,380
Ejaaz:
manufacturing this design is figuring out how the power is transmitted between

295
00:18:09,380 --> 00:18:11,590
Ejaaz:
these chips and within the chip itself.

296
00:18:11,900 --> 00:18:16,640
Ejaaz:
And this is a huge role that is played within the power infrastructure because

297
00:18:16,640 --> 00:18:22,120
Ejaaz:
it determines where the power gets dispersed. It's an incredibly hard thing to figure out. Now...

298
00:18:22,720 --> 00:18:27,830
Ejaaz:
About three weeks ago, there was this very popular compute conference called

299
00:18:27,830 --> 00:18:34,120
Ejaaz:
Computex, which was basically NVIDIA GTC 2.0. Jensen was like the bell of the bell there.

300
00:18:34,350 --> 00:18:37,140
Ejaaz:
So he goes on stage and he talks about this company, Marvel,

301
00:18:37,140 --> 00:18:41,580
Ejaaz:
and he says, this is the next trillion dollar company. And this isn't investment

302
00:18:41,730 --> 00:18:45,620
Ejaaz:
advice, by the way. This is literally verbatim what he said in Taiwan.

303
00:18:45,850 --> 00:18:50,720
Ejaaz:
And the stock subsequently went up 76% after he mentioned this.

304
00:18:51,000 --> 00:18:53,610
Ejaaz:
And I was looking up on this just to see, you know, have they like,

305
00:18:53,610 --> 00:18:54,980
Ejaaz:
you know, rescinded since then?

306
00:18:55,700 --> 00:19:00,140
Ejaaz:
Josh, they broke into the S&P 500 based off of this single use.

307
00:19:00,480 --> 00:19:05,720
Ejaaz:
And the reason the TLDR of why Marvel is attracting so much attention and capital right now

308
00:19:05,940 --> 00:19:10,550
Ejaaz:
is because they are experts in designing that part of the chip,

309
00:19:10,550 --> 00:19:14,390
Ejaaz:
which determines where power gets dispersed, how it gets dispersed and where

310
00:19:14,390 --> 00:19:16,990
Ejaaz:
it gets used within the chip and between chips.

311
00:19:17,260 --> 00:19:21,610
Ejaaz:
And if you're wondering, oh, okay, well, they're just solely reliant on NVIDIA, the answer is no.

312
00:19:21,840 --> 00:19:26,400
Ejaaz:
Broadcom and every other single AI company that is trying to design their own

313
00:19:26,400 --> 00:19:30,640
Ejaaz:
chip and diversify away from NVIDIA is using Marvel.

314
00:19:31,450 --> 00:19:37,690
Josh:
Wow, it's up 80% on the month I'm seeing in one month since Jensen mentioned this company.

315
00:19:38,280 --> 00:19:41,230
Josh:
I actually don't know much about Marvel and I feel bad about it because clearly

316
00:19:41,230 --> 00:19:44,780
Josh:
I missed out. But it's one of those things where it's been very lucrative to

317
00:19:44,780 --> 00:19:46,530
Josh:
copy trade the people in authority.

318
00:19:46,810 --> 00:19:51,680
Josh:
It's like if someone who is running the largest company in the world says this

319
00:19:51,680 --> 00:19:54,330
Josh:
company is going to a trillion dollars, people are going to react.

320
00:19:54,530 --> 00:19:57,450
Josh:
And you have to just trust the fact that he has a good reason to do this.

321
00:19:57,450 --> 00:20:02,160
Josh:
I mean, Jensen seems like he holds pretty strong generally on his morals.

322
00:20:02,380 --> 00:20:04,940
Josh:
He's not trying to pump and dump. He genuinely believes this.

323
00:20:05,170 --> 00:20:08,330
Josh:
He has made a sizable investment. What is the total amount? It's a couple of

324
00:20:08,330 --> 00:20:09,220
Josh:
billion dollars, right?

325
00:20:09,220 --> 00:20:14,260
Ejaaz:
I did $2 billion. Yeah. So he's up, he's up pretty big now.

326
00:20:14,260 --> 00:20:16,600
Josh:
Man, my God, exactly. That's the point. And it's like, yeah,

327
00:20:16,600 --> 00:20:20,390
Josh:
he, he actually has the ability to influence the success of this company,

328
00:20:20,390 --> 00:20:24,310
Josh:
not just by talking about it, but by actually entering and the technology into his company.

329
00:20:24,520 --> 00:20:27,730
Josh:
Same thing happens when you talk about the person in charge of the United States

330
00:20:27,730 --> 00:20:31,040
Josh:
of America. Like when Trump says something, it generally, it goes up generally

331
00:20:31,040 --> 00:20:34,170
Josh:
as a good reason. People are looking for projects to invest in,

332
00:20:34,170 --> 00:20:35,230
Josh:
for companies to invest in.

333
00:20:35,530 --> 00:20:40,200
Josh:
And Marvel has been one of the most recent winners. Now, there is another section

334
00:20:40,200 --> 00:20:44,800
Josh:
of this stack in which you have to rent out the compute. Say that you are not,

335
00:20:45,290 --> 00:20:49,030
Josh:
you don't have hundreds of billions of dollars of CapEx at your disposal this year.

336
00:20:49,360 --> 00:20:51,670
Josh:
Sorry to hear that. Chances are you're not winning the AI race,

337
00:20:51,770 --> 00:20:54,920
Josh:
but you can at least make an impact. You could do something novel,

338
00:20:54,920 --> 00:20:58,040
Josh:
something interesting in AI. And if that's the case, you're going to need to

339
00:20:58,040 --> 00:20:59,310
Josh:
rent that compute from someone.

340
00:20:59,610 --> 00:21:03,600
Josh:
There's a bunch of companies that we've talked about on the show in which you can rent compute from.

341
00:21:03,810 --> 00:21:07,230
Josh:
We talked about Iren, a lot of the NeoClouds, but we have a new one to talk

342
00:21:07,230 --> 00:21:08,800
Josh:
about today, which is called Nebius.

343
00:21:09,120 --> 00:21:12,760
Josh:
Ejaz, can you explain to everyone what Nebius is, what makes them so different and unique?

344
00:21:13,900 --> 00:21:16,620
Ejaaz:
It's actually not what makes them different or unique. It's just,

345
00:21:16,620 --> 00:21:18,680
Ejaaz:
it's a scarce service that they supply.

346
00:21:18,680 --> 00:21:25,360
Ejaaz:
So if you've heard of CoreWeave or if you've heard of Iron that Josh just mentioned,

347
00:21:25,740 --> 00:21:28,330
Ejaaz:
they're known as a neocloud.

348
00:21:28,330 --> 00:21:33,010
Ejaaz:
A neocloud is basically a company that will come in, they will construct the

349
00:21:33,010 --> 00:21:36,590
Ejaaz:
data center for you, they will handle all the wiring, they'll make sure that

350
00:21:36,590 --> 00:21:40,090
Ejaaz:
the GPUs talk to each other, they'll make sure that the networking and transformers

351
00:21:40,390 --> 00:21:43,010
Ejaaz:
and receivers, they'll make sure that the copper doesn't melt.

352
00:21:43,250 --> 00:21:47,330
Ejaaz:
And they'll make sure that these beautiful, precious GPUs that you just spend

353
00:21:47,330 --> 00:21:50,450
Ejaaz:
billions of dollars on function optimally.

354
00:21:50,730 --> 00:21:54,810
Ejaaz:
The reason why someone will go to a Nebius or a CoreWeave is simple.

355
00:21:55,150 --> 00:21:58,980
Ejaaz:
Google, Amazon, and all of those companies, they don't want to focus and spend

356
00:21:58,980 --> 00:22:03,450
Ejaaz:
time operating and maintaining and designing the flow of a data center.

357
00:22:03,450 --> 00:22:06,310
Ejaaz:
They want someone else to handle them. That's what Nebius does.

358
00:22:06,580 --> 00:22:10,580
Ejaaz:
And they do it pretty damn well. I believe Nebius's background is they operated

359
00:22:10,580 --> 00:22:14,940
Ejaaz:
a bunch of data centers that focus primarily on, maybe it was mining

360
00:22:15,580 --> 00:22:18,690
Ejaaz:
in the Web3 side of things, maybe it was something else, but they pivoted pretty

361
00:22:18,690 --> 00:22:22,950
Ejaaz:
aggressively to focus on AI specifically and their expertise of managing hardware.

362
00:22:23,220 --> 00:22:27,110
Ejaaz:
And most importantly, the software that talks to each of these GPUs and the

363
00:22:27,110 --> 00:22:30,560
Ejaaz:
different hardware components is their expertise. So the answer,

364
00:22:30,560 --> 00:22:34,050
Ejaaz:
truthfully, is they do exactly what CoreWeave does. They do exactly what Iron

365
00:22:34,050 --> 00:22:36,030
Ejaaz:
does, but they do it at a massive scale.

366
00:22:36,280 --> 00:22:40,800
Ejaaz:
And their Q1 revenue is up almost 700% on the year. It's just insane.

367
00:22:41,010 --> 00:22:44,850
Josh:
Yeah, it's crazy. I saw it at $400 million dollars and this is the thing with the demand

368
00:22:45,220 --> 00:22:48,160
Josh:
issues is that like if you build it they will come it doesn't matter who you

369
00:22:48,160 --> 00:22:53,640
Josh:
are one fun fact about nebius is they were previously known as yandex or this

370
00:22:53,640 --> 00:22:55,150
Josh:
is what was built from the remains of yandex

371
00:22:55,420 --> 00:22:58,820
Josh:
if you are from russia you probably know this because that was very much the

372
00:22:58,820 --> 00:23:03,350
Josh:
google of russia up until 2024 when they severed and they started to build a new cloud

373
00:23:03,690 --> 00:23:06,360
Josh:
so that's a yeah they are familiar with

374
00:23:06,780 --> 00:23:11,680
Josh:
managing data and storing data and uptime with servers wow yeah so that was

375
00:23:11,680 --> 00:23:14,180
Josh:
interesting and then i think one of the things that makes this company special

376
00:23:14,180 --> 00:23:16,890
Josh:
again nvidia is a backer nvidia has placed

377
00:23:17,130 --> 00:23:20,750
Josh:
two billion dollars of direct equity into this company nvidia should just like

378
00:23:20,870 --> 00:23:24,000
Josh:
man if they weren't making gpus jensen could just make it as a vc because this.

379
00:23:24,000 --> 00:23:24,770
Ejaaz:
Dude is making

380
00:23:25,170 --> 00:23:26,160
Josh:
Pretty unbelievable,

381
00:23:28,820 --> 00:23:32,610
Josh:
yeah two billion yeah maybe we just screenshot this portfolio and i just send

382
00:23:32,610 --> 00:23:35,530
Josh:
it to my broker and say can you buy all of these for me just split it across

383
00:23:35,530 --> 00:23:36,550
Josh:
my whole portfolio because,

384
00:23:37,070 --> 00:23:40,730
Josh:
clearly jensen knows what he's doing but the the company is doing remarkably

385
00:23:40,730 --> 00:23:44,910
Josh:
well and because they have contracts from companies that you actually have heard

386
00:23:44,910 --> 00:23:50,300
Josh:
of they have 20 billion dollar contract from microsoft they have a 27 billion dollar contract from.

387
00:23:51,140 --> 00:23:55,200
Josh:
Meta and then they have a backlog that's approaching 50 billion dollars for

388
00:23:55,200 --> 00:24:01,790
Josh:
the years 2027 all the way out to 2031 and this is the thing that we talk about a lot too is that the,

389
00:24:02,290 --> 00:24:05,890
Josh:
the useful life of a gpu has extended so far

390
00:24:06,220 --> 00:24:09,810
Josh:
that i think these companies are getting revalued and re-underwritten based

391
00:24:09,810 --> 00:24:11,090
Josh:
on that because a lot of the time

392
00:24:11,400 --> 00:24:17,670
Josh:
people would say that oh if you have a gb100 today it's going to be worth 10

393
00:24:17,670 --> 00:24:21,680
Josh:
of what it is today five years from now and it's a five-year depreciation schedule

394
00:24:21,860 --> 00:24:24,400
Josh:
the reality is is that gpu from three to four years ago

395
00:24:24,640 --> 00:24:27,080
Josh:
is worth more today per hour than it was back then

396
00:24:27,450 --> 00:24:31,770
Josh:
and so long as this demand curve continues these gpus the useful life of them

397
00:24:31,770 --> 00:24:35,570
Josh:
is going to continue to extend out over longer and longer durations,

398
00:24:35,790 --> 00:24:38,610
Josh:
making the inventory of these companies more and more valuable.

399
00:24:38,610 --> 00:24:41,710
Josh:
And if that trend continues, everyone's going to have to continue to re-underwrite

400
00:24:41,710 --> 00:24:44,280
Josh:
these companies based on this new depreciation scale.

401
00:24:44,560 --> 00:24:47,430
Josh:
And that is going to continue to bring the value of the company up.

402
00:24:47,430 --> 00:24:50,500
Josh:
They have the scarce resource, which is the ability to actually build these

403
00:24:50,500 --> 00:24:53,630
Josh:
data centers, which means they have the land, they have the power, they have the shell.

404
00:24:53,890 --> 00:24:58,800
Josh:
And this is another company that is doing it pretty well. So Nebius is a good

405
00:24:58,800 --> 00:25:01,050
Josh:
example. Neoclads in general are a good.

406
00:25:03,250 --> 00:25:07,790
Josh:
Insane demand that is insatiable it's like no one can actually provide enough

407
00:25:07,790 --> 00:25:12,860
Josh:
energy enough gpus that are powered online and any company that's able to contribute

408
00:25:12,860 --> 00:25:16,470
Josh:
to that is just going to continue to grow so long as this demand stays there

409
00:25:16,470 --> 00:25:17,930
Josh:
and there's no signs of it's slowing down.

410
00:25:18,470 --> 00:25:22,850
Ejaaz:
You know as we talk about this it's so interesting to see this sort of

411
00:25:23,750 --> 00:25:30,830
Ejaaz:
It's like this entire ecosystem that has built around the gpu as gpus are basically

412
00:25:30,830 --> 00:25:33,530
Ejaaz:
the new gold bar right there's this whole economy around this.

413
00:25:34,430 --> 00:25:40,410
Ejaaz:
This is news that I was seeing earlier this week that the Chicago Mercantile Exchange, CMC,

414
00:25:40,990 --> 00:25:47,640
Ejaaz:
I believe, is setting up a compute futures trading index so that you can hedge

415
00:25:47,640 --> 00:25:52,010
Ejaaz:
trades and do trades based off of compute and GPU infrastructure.

416
00:25:52,260 --> 00:25:57,860
Ejaaz:
It is becoming a financial instrument that you can eventually borrow money off

417
00:25:57,860 --> 00:26:00,580
Ejaaz:
of. That's what NVIDIA and a ton of other companies are doing.

418
00:26:00,820 --> 00:26:06,290
Ejaaz:
And so it's this kind of like pseudo or proxy investment vehicle where it's just based around GPUs.

419
00:26:06,290 --> 00:26:10,040
Ejaaz:
And this whole ecosystem is now blossoming around it, whether it's neoclouds,

420
00:26:10,040 --> 00:26:13,560
Ejaaz:
whether it's optical light fibers, whether it's the silicon itself,

421
00:26:13,560 --> 00:26:15,080
Ejaaz:
whether it's the chip design that helps

422
00:26:15,370 --> 00:26:19,360
Ejaaz:
transmit all this kind of power, it is formed around the gravitational pull

423
00:26:19,360 --> 00:26:21,890
Ejaaz:
of these GPUs is pretty insane.

424
00:26:22,160 --> 00:26:25,350
Ejaaz:
One final note that I wanted to mention for Nebius specifically,

425
00:26:25,590 --> 00:26:29,670
Ejaaz:
And they have the same advantage that CoreWeave and Iron has is permitting and

426
00:26:29,670 --> 00:26:32,540
Ejaaz:
licenses, like yawn, right? Like who the hell cares?

427
00:26:32,940 --> 00:26:37,920
Ejaaz:
As we mentioned earlier, a lot of the five-year lead time to get like power

428
00:26:37,920 --> 00:26:41,110
Ejaaz:
and energy to your GPUs is regulatory.

429
00:26:41,490 --> 00:26:46,820
Ejaaz:
It's red tape. It's trying to jump through the government hoops to get all of this figured out.

430
00:26:47,220 --> 00:26:53,030
Ejaaz:
Nebius, Iron, Coreweave all have this permitting in advanced stages.

431
00:26:53,030 --> 00:26:56,860
Ejaaz:
And that's why they're signing these multi-billion dollar deals with Meta,

432
00:26:57,850 --> 00:27:01,390
Ejaaz:
Google, and all the other hyperscalers because they have this advantage.

433
00:27:01,390 --> 00:27:05,790
Ejaaz:
It's all a regulatory game. It's all a hardware game. Whoever can figure out

434
00:27:06,020 --> 00:27:12,190
Ejaaz:
a way to arbitrage the red tape that they're currently facing will end up winning pretty hugely.

435
00:27:12,480 --> 00:27:16,430
Ejaaz:
And if you look at what leopold ashenbrenner is investing and if you look at

436
00:27:16,430 --> 00:27:19,850
Ejaaz:
what uh jen so puang is investing the people who are on the ground that are

437
00:27:19,850 --> 00:27:23,630
Ejaaz:
seeing these bottlenecks in real time that can kind of give you an idea of where

438
00:27:23,630 --> 00:27:25,700
Ejaaz:
that ai capital is eventually going to flow

439
00:27:25,970 --> 00:27:29,680
Josh:
Yeah and then one one final point on the neoclabs which i think is fun and underrated

440
00:27:29,680 --> 00:27:33,080
Josh:
and a lot of people don't recognize what the incentive is to go with them as

441
00:27:33,080 --> 00:27:36,090
Josh:
if you don't need the gpus but if you're a large company like meta for example

442
00:27:36,380 --> 00:27:38,500
Josh:
who has that 27 billion dollar deal with nebius

443
00:27:38,860 --> 00:27:43,260
Josh:
you can actually write off that spend as an expense spread over years instead

444
00:27:43,260 --> 00:27:46,570
Josh:
of just like hammering your free cash flow with this increased capex today.

445
00:27:46,940 --> 00:27:50,630
Josh:
And for a lot of earnings reports, we see the first numbers we go to look at

446
00:27:50,630 --> 00:27:54,070
Josh:
are the capex. How much are they going to be spending on building out compute?

447
00:27:54,340 --> 00:27:57,210
Josh:
And if that number is too high based on what the market believes they can actually

448
00:27:57,210 --> 00:27:59,070
Josh:
build, then they get crushed for it.

449
00:27:59,320 --> 00:28:02,960
Josh:
So a way to kind of obfuscate that is by investing in these neoclouds,

450
00:28:03,300 --> 00:28:07,090
Josh:
signing these deals that can then be extended out over the years and don't immediately

451
00:28:07,090 --> 00:28:10,270
Josh:
crush your balance sheet like we're seeing a lot of these companies doing um

452
00:28:10,440 --> 00:28:13,820
Josh:
google being one nvidia being one where they just raised 25 billion dollars

453
00:28:13,820 --> 00:28:17,260
Josh:
so it's an interesting financial instrument as well that these companies are using but i think

454
00:28:17,610 --> 00:28:20,330
Josh:
that is mostly it in the hardware energy

455
00:28:20,710 --> 00:28:23,050
Josh:
stack that's where we are um how do we.

456
00:28:23,570 --> 00:28:27,140
Ejaaz:
How do we do how do we do i'm curious like people have been asking for this

457
00:28:27,140 --> 00:28:31,610
Ejaaz:
episode for a while i i feel like we did a good job kind of shepherding the

458
00:28:31,610 --> 00:28:34,270
Ejaaz:
story between these companies? Josh, do you feel the same?

459
00:28:34,560 --> 00:28:37,810
Josh:
I feel pretty good about it. I think there's this like,

460
00:28:38,440 --> 00:28:42,310
Josh:
the layers go so deep and you could even go deeper than this with like the actual

461
00:28:42,560 --> 00:28:47,910
Josh:
um manufacturing of this and like who is supplying all the raw materials in

462
00:28:48,160 --> 00:28:50,170
Josh:
but i think this is about as low as you need to get

463
00:28:50,460 --> 00:28:53,280
Josh:
because this is the this is the interesting part past this you start getting

464
00:28:53,280 --> 00:28:57,020
Josh:
into all the science and physics that i'm not sure it's super interesting but i feel good about this

465
00:28:57,320 --> 00:28:59,900
Josh:
i actually learned a lot in preparing and recording for this episode

466
00:29:00,270 --> 00:29:03,680
Josh:
as it relates to just like where the bottlenecks are who's responsible for them

467
00:29:04,020 --> 00:29:07,100
Josh:
i didn't even know what Marvel was until Jensen talked about them on stage.

468
00:29:07,100 --> 00:29:10,270
Josh:
So we're learning a lot about them. There's a lot of alpha and coming across

469
00:29:10,270 --> 00:29:12,020
Josh:
these early. I mean, all of the companies so far,

470
00:29:12,500 --> 00:29:15,600
Josh:
they've done pretty well. It's no lie. Like people know about this.

471
00:29:15,600 --> 00:29:19,720
Josh:
This is not this like deep alpha, but there is a really great opportunity.

472
00:29:19,720 --> 00:29:24,020
Josh:
And these companies are building valuable goods and services for companies that

473
00:29:24,200 --> 00:29:28,310
Josh:
need them. So yeah, I think that's probably it. That covers the whole thing.

474
00:29:29,250 --> 00:29:32,950
Josh:
We went through the stack all the way down to the bottom. We previously covered

475
00:29:32,950 --> 00:29:36,370
Josh:
the higher parts of the stack. If you're interested in that episode, you can find it.

476
00:29:36,730 --> 00:29:39,610
Josh:
We'll link it in the description. It's also in our channel. You may have heard

477
00:29:39,610 --> 00:29:40,250
Josh:
it if you've been around.

478
00:29:40,480 --> 00:29:43,880
Josh:
But yeah, I think that's pretty much it. That is covering the base layer,

479
00:29:43,880 --> 00:29:47,660
Josh:
the energy infralayer of this AI stack.

480
00:29:47,660 --> 00:29:52,470
Ejaaz:
There is one more layer, Josh, that I, not today, not next week,

481
00:29:52,470 --> 00:29:55,810
Ejaaz:
but maybe in a few months time that we need to prep hard for,

482
00:29:55,810 --> 00:29:57,700
Ejaaz:
which is the substrate layer.

483
00:29:57,700 --> 00:30:01,820
Ejaaz:
That is the one at the bottom of the bottom. That is what we're talking.

484
00:30:01,820 --> 00:30:04,770
Ejaaz:
We're going back to old school days. We're talking about raw materials here

485
00:30:05,170 --> 00:30:10,360
Ejaaz:
that actually help form the materials, the silicon itself, the wiring that we've

486
00:30:10,360 --> 00:30:12,670
Ejaaz:
been speaking about, where do those materials come from?

487
00:30:12,670 --> 00:30:17,090
Ejaaz:
And like fun little nugget, an Easter egg before you even hear anything about

488
00:30:17,090 --> 00:30:20,660
Ejaaz:
this episode is the majority of those materials, they're in Japan.

489
00:30:21,070 --> 00:30:24,780
Ejaaz:
And Japan has a company that manufactures toilets.

490
00:30:25,020 --> 00:30:27,480
Ejaaz:
It's known as Toto. In fact, if you look at your toilet right now,

491
00:30:28,940 --> 00:30:30,350
Ejaaz:
we filmed an episode about Toto.

492
00:30:30,670 --> 00:30:35,520
Ejaaz:
They happen to have machines which get access to this really funky material

493
00:30:35,720 --> 00:30:40,700
Ejaaz:
that is only accessible in Japan or primarily based in Japan that is used to make GPUs.

494
00:30:41,070 --> 00:30:44,440
Ejaaz:
It is an essential component in NVIDIA's manufacturing stack.

495
00:30:44,630 --> 00:30:47,240
Ejaaz:
Just a little fun fact for you. We're going to make an episode on that eventually.

496
00:30:47,680 --> 00:30:51,630
Josh:
There's a teaser. What do toilet bowls and AI have in common? Ceramic is one answer.

497
00:30:52,130 --> 00:30:53,890
Ejaaz:
Billions of dollars of market cap.

498
00:30:54,570 --> 00:30:58,690
Josh:
Yeah, maybe we could do this again. We'll do it for the furthest down the stack.

499
00:30:58,910 --> 00:31:02,380
Josh:
It's amazing how granular it gets. Like when we talk about these two nanometer

500
00:31:02,590 --> 00:31:06,030
Josh:
chips that are coming down the line, the difference is a matter of atoms.

501
00:31:06,030 --> 00:31:09,520
Josh:
Like they are literally constructing on the atomic level, five atoms.

502
00:31:09,520 --> 00:31:12,350
Josh:
And if one of them is in the wrong spot, then these things don't work.

503
00:31:12,350 --> 00:31:16,170
Josh:
So the precision required to make this happen is is like it's unbelievable there's

504
00:31:16,170 --> 00:31:20,950
Josh:
nothing else in the history of humanity that has ever been so complex as this

505
00:31:21,180 --> 00:31:23,950
Josh:
there's just nothing that humans have made at least and,

506
00:31:24,400 --> 00:31:27,760
Josh:
going down the stack it's just unbelievable to see and the resulting output

507
00:31:27,980 --> 00:31:31,660
Josh:
is you could type into your chat box and you get intelligence out and it seems

508
00:31:31,660 --> 00:31:36,830
Josh:
imperceivable from magic so that is the ai energy stack episode i hope you enjoyed

509
00:31:36,830 --> 00:31:38,820
Josh:
if you did please be sure to

510
00:31:39,200 --> 00:31:41,790
Josh:
leave a comment share with a friend who you think might be interested be like

511
00:31:41,790 --> 00:31:44,080
Josh:
yo bro you should check out this company.

512
00:31:44,080 --> 00:31:47,220
Josh:
These guys on Limitless talked about it and they made a pretty interesting case.

513
00:31:47,430 --> 00:31:50,780
Josh:
Again, none of this is investment advice. I'm not exposed to half these companies

514
00:31:50,780 --> 00:31:55,570
Josh:
for better or worse, but we should start following your own book.

515
00:31:56,430 --> 00:31:57,080
Ejaaz:
We should just copy

516
00:31:57,080 --> 00:31:57,580
Josh:
Trade our own.

517
00:31:57,580 --> 00:32:00,970
Ejaaz:
Episodes because everything that we mentioned has been going nuclear.

518
00:32:01,330 --> 00:32:04,850
Josh:
That's the current state of the union. So if you invest, listen, that's on you.

519
00:32:05,130 --> 00:32:09,280
Josh:
We're too stupid to take our own advice sometimes, But that is the episode.

520
00:32:09,280 --> 00:32:10,570
Josh:
Thank you all so much for watching.

521
00:32:11,210 --> 00:32:15,620
Josh:
Don't forget to share, like, subscribe, leave a comment on something we missed.

522
00:32:15,620 --> 00:32:17,940
Josh:
Is there another one? Is there another layer of the stack that we're missing?

523
00:32:17,940 --> 00:32:20,770
Josh:
Is there another company that is going to go nuclear?

524
00:32:21,170 --> 00:32:24,610
Josh:
That'd be good to know. Just out of curiosity. But yeah, I think that's it.

525
00:32:24,610 --> 00:32:25,640
Josh:
So we'll see you in the next one.

526
00:32:25,780 --> 00:32:26,280
Ejaaz:
See you guys.