1
00:00:00,017 --> 00:00:03,717
Last week I went through Anthropic's financials and I was shocked.

2
00:00:03,717 --> 00:00:07,981
Turns out Anthropic makes a lot of money. For every $1 of compute they invest,

3
00:00:08,265 --> 00:00:10,407
they roughly make about $2 of revenue.

4
00:00:10,407 --> 00:00:16,957
Except today they released their Q3 financials and turns out all that profit is gone.

5
00:00:16,957 --> 00:00:22,207
Despite making $17 billion over the last three months, which is up 47% from

6
00:00:22,207 --> 00:00:29,617
the last quarter, Anthropic owes one of the largest compute bills ever, $895 billion.

7
00:00:30,440 --> 00:00:34,757
That is worse than the craziest bear prediction ever. And the worst part is

8
00:00:34,757 --> 00:00:38,898
$70 billion of that is owed over the next 12 months.

9
00:00:39,223 --> 00:00:42,647
So you have to ask the question, how is Anthropic going to pay for all of this?

10
00:00:42,647 --> 00:00:46,727
And it turns out some of the more well-esteemed financial creditors are also

11
00:00:46,727 --> 00:00:51,927
getting worried, assigning Anthropic a potential junk credit rating right before

12
00:00:51,927 --> 00:00:54,337
the biggest IPO that we've seen in history.

13
00:00:54,337 --> 00:00:58,787
Now, it's easy to argue the bear case of this, but what I want to present to

14
00:00:58,787 --> 00:01:01,887
you is something that is maybe slightly more optimistic.

15
00:01:01,887 --> 00:01:05,124
You see, the reason why Anthropic is losing money over the last three months

16
00:01:05,350 --> 00:01:10,058
is because they weren't able to meet customer demand in a suitable fashion.

17
00:01:10,354 --> 00:01:13,327
They had so many people trying to access their models that they didn't have

18
00:01:13,327 --> 00:01:14,951
enough compute to go around.

19
00:01:15,282 --> 00:01:18,997
So they to end up forking out a large premium to access that compute.

20
00:01:19,247 --> 00:01:22,717
Now, without getting too much of the details, the idea here is if Anthropic

21
00:01:22,717 --> 00:01:28,637
is able to secure enough compute over the next 12 to 24 months in these long-term

22
00:01:28,637 --> 00:01:31,037
leases, which capture around $900 billion,

23
00:01:31,497 --> 00:01:36,567
apparently, and demand stays sufficient, they should be able to theoretically

24
00:01:36,567 --> 00:01:37,717
make a decent chunk of money.

25
00:01:37,717 --> 00:01:41,467
That is still a huge amount of risk. And I want to dig into the bull and bear

26
00:01:41,467 --> 00:01:44,997
case of what's going on given these new financials. So let's get into it on

27
00:01:44,997 --> 00:01:46,540
today's episode brought to you by Qualcomm.

28
00:01:50,516 --> 00:01:54,281
So just to catch you up anthropic released their s1 last week which gave us

29
00:01:54,281 --> 00:01:58,916
the entire financials for 2025 up till about end of q2

30
00:01:59,090 --> 00:02:02,491
of this year so we got a peek under the hood as to whether anthropic was making

31
00:02:02,491 --> 00:02:06,881
money and a lot of people's opinions including myself was sort of skeptical

32
00:02:06,881 --> 00:02:10,891
about this i didn't think that they were making as much money as i eventually

33
00:02:10,891 --> 00:02:14,103
then found out plus i thought they were burning way too much cash

34
00:02:14,411 --> 00:02:19,241
turns out the actual opposite was true anthropic had around $132 billion.

35
00:02:19,914 --> 00:02:22,821
On their cash balance sheet, and they'd only burned around $7 to $9 billion.

36
00:02:23,401 --> 00:02:28,279
And most importantly, the money, the revenue that they were generating was actually quite lean.

37
00:02:28,476 --> 00:02:31,881
For every dollar of compute that they would invest, which is arguably the largest

38
00:02:31,881 --> 00:02:39,141
CapEx cost to create these models and serve these models, they were making around $2.08 of revenue.

39
00:02:39,141 --> 00:02:43,791
Inference margin on that is around 85%. So that's crazy for any company,

40
00:02:43,791 --> 00:02:46,971
irrespective of whether it's an AI lab or not. But there was some very fair

41
00:02:46,971 --> 00:02:51,713
criticisms about this, which was Q1 and Q2, everyone was super hyped about AI.

42
00:02:51,853 --> 00:02:55,781
Loads of the Fortune 500 companies were doing something known as token maxing.

43
00:02:55,781 --> 00:02:59,141
So trying to throw clawed models at anything and everything to try and generate

44
00:02:59,141 --> 00:03:02,731
ROI, but that ROI hasn't quite yet been seen.

45
00:03:02,731 --> 00:03:06,621
And so the main argument that was given was, if we see the Q3 results and it

46
00:03:06,621 --> 00:03:10,711
backs up the Q1 and Q2 results, then we'll believe that Anthropic is making money.

47
00:03:10,711 --> 00:03:14,601
The issue is, with the Q3 results that we just got, it shows that Anthropic

48
00:03:14,601 --> 00:03:18,281
is actually making a loss to the tune of around $554 million.

49
00:03:18,661 --> 00:03:20,311
And this ties into the general sentiment

50
00:03:20,311 --> 00:03:23,371
that we're seeing not just with Anthropic, but other major AI labs.

51
00:03:23,371 --> 00:03:26,931
If you look at OpenAI, the rumor has it that the most recent funding round that

52
00:03:26,931 --> 00:03:31,721
they've been trying to raise, $30 billion, had to get slashed to $20 billion.

53
00:03:31,721 --> 00:03:35,221
They were trying to raise money enough to fund the compute and capex spend over

54
00:03:35,221 --> 00:03:37,514
the next couple of months before they eventually IPO, I think,

55
00:03:37,647 --> 00:03:39,818
early 2027 is what the rumor has it.

56
00:03:40,184 --> 00:03:43,731
And they had to slash that. But worse, they had to slash the valuation from

57
00:03:43,731 --> 00:03:49,681
$1.5 trillion, which compared to Anthropik's $2.2 trillion is fairly conservative.

58
00:03:50,105 --> 00:03:54,607
They had to cut that to $1.25 trillion because investors didn't quite believe

59
00:03:54,607 --> 00:03:56,637
that their models were worth that much.

60
00:03:56,637 --> 00:04:00,567
Now, putting that aside and looking at the crazy market dynamics that we've

61
00:04:00,567 --> 00:04:01,837
seen in stocks recently, we've

62
00:04:01,837 --> 00:04:06,047
seen semiconductor stocks draw down around 35 to 40% with near clouds.

63
00:04:06,047 --> 00:04:10,676
And then we've seen Intel, Amazon, Microsoft, Google suffer major stock losses

64
00:04:10,810 --> 00:04:14,055
over the last week or so, despite record earnings.

65
00:04:14,357 --> 00:04:18,197
It kind of gives you the general sentiment that people are getting sort of concerned.

66
00:04:18,197 --> 00:04:23,377
And indeed, like lending rates for all of this humongous CapEx build-out is getting too expensive.

67
00:04:23,377 --> 00:04:26,287
So without getting too much in the financials, what does this mean for Anthropics

68
00:04:26,287 --> 00:04:28,937
IPO, which is rumored to go live in like two weeks?

69
00:04:28,937 --> 00:04:33,037
Well, first, the headline numbers. They made $17 billion over the last three

70
00:04:33,037 --> 00:04:39,114
months. That is up 1,127% from that period last year.

71
00:04:39,341 --> 00:04:44,487
And it's up 47% from just three months ago. So they are still rapidly increasing

72
00:04:44,487 --> 00:04:47,617
their revenue rate, however way you want to look at it. And that means in the

73
00:04:47,617 --> 00:04:49,907
last nine months, they made around $33 billion.

74
00:04:50,475 --> 00:04:53,107
Over the same period last year, they made around $2.5 billion.

75
00:04:53,107 --> 00:04:55,107
So we've seen a 13x increase.

76
00:04:55,107 --> 00:04:59,917
By any means, the claim that Anthropic is still one of the, if not the fastest

77
00:04:59,917 --> 00:05:04,037
growing company that we've ever seen in history still stands.

78
00:05:04,037 --> 00:05:08,219
But the question mark then becomes, is that even sustainable in the first place?

79
00:05:08,643 --> 00:05:13,417
Last quarter, or rather last week when we saw the S1 in Q2, we saw them make

80
00:05:13,417 --> 00:05:16,867
around an operating profit of, I think it was $974 million.

81
00:05:17,407 --> 00:05:20,297
Fast forward to today where we see their Q3 results and we see that they've

82
00:05:20,297 --> 00:05:24,067
actually got an operating loss of around $564 million.

83
00:05:24,740 --> 00:05:27,712
And there are three big reasons as to why they suffered that loss.

84
00:05:28,014 --> 00:05:33,970
Number one starts with $5 billion of short-term compute payments that they had to make.

85
00:05:34,382 --> 00:05:39,297
This wasn't estimated previously in the S1 because they didn't expect the overwhelming

86
00:05:39,297 --> 00:05:43,167
demand that they receive for their model. So let me kind of put something into

87
00:05:43,167 --> 00:05:49,231
perspective here. Anthropic and any other AI company needs to purchase access to compute and,

88
00:05:49,720 --> 00:05:53,185
before they know how much demand they're going to get. Think about it.

89
00:05:53,463 --> 00:05:56,876
If you try and predict the demand that you're going to have for your models,

90
00:05:57,114 --> 00:05:59,401
it's token consumption-based.

91
00:05:59,581 --> 00:06:02,860
What I mean by that is you release a new model and you hope everyone likes it

92
00:06:02,860 --> 00:06:05,793
and you hope everyone wants to consume as much of it as they can.

93
00:06:06,077 --> 00:06:10,820
And that applies for the most expensive model, but it also applies for the most cheapest model.

94
00:06:10,820 --> 00:06:14,483
And if you look at OpenAI, if you look at Anthropic, you realize that not only

95
00:06:14,657 --> 00:06:18,170
do they just release the latest, greatest, most expensive model that is the

96
00:06:18,490 --> 00:06:22,340
most intelligent, they also release much cheaper versions of that either at

97
00:06:22,340 --> 00:06:24,800
the same time or a few weeks later.

98
00:06:24,800 --> 00:06:28,699
And the idea here is it should cover all cost bases and customer types.

99
00:06:28,908 --> 00:06:33,670
If you wanted to code a company or business from scratch, you shouldn't need

100
00:06:33,670 --> 00:06:37,337
to have to spend so much money or use the most intelligent model all the time.

101
00:06:37,586 --> 00:06:42,236
So the idea here is it's very unpredictable to figure out how much demand you're going to get.

102
00:06:42,381 --> 00:06:46,515
So Anthropic invested a lot of money to get access to compute for Q3.

103
00:06:46,944 --> 00:06:51,188
But they underestimated it. And so the idea here is they needed to get emergency

104
00:06:51,188 --> 00:06:54,278
access to compute, of which they had to fork out $5 billion.

105
00:06:54,278 --> 00:06:58,054
Now, to give you an easy analogy of this, imagine that you run,

106
00:06:58,205 --> 00:07:02,938
I don't know, hottest restaurant in town, right? And everyone wants to book

107
00:07:02,938 --> 00:07:05,008
a table and you're booked out day in, day out.

108
00:07:05,298 --> 00:07:08,230
But one night you run out of ingredients. What are you going to do?

109
00:07:08,404 --> 00:07:10,938
Are you going to turn people away and say, sorry, we can't serve you?

110
00:07:10,938 --> 00:07:15,798
Or are you going to run down to the local supermarket and buy the ingredients

111
00:07:15,798 --> 00:07:19,462
off the shelf at retail price, even though it's eating into your revenue margins,

112
00:07:19,683 --> 00:07:21,355
just so that you can serve up the product.

113
00:07:21,622 --> 00:07:24,498
That is effectively what Anthropic had to do. And if I had to guess,

114
00:07:24,498 --> 00:07:29,478
they were probably paying SpaceX, which is their primary short-term lease compute

115
00:07:29,478 --> 00:07:32,396
provider, a very hefty price to get access to this compute.

116
00:07:32,500 --> 00:07:37,888
So just because of this one maneuver, it meant that the compute costing around

117
00:07:37,888 --> 00:07:44,238
48% of their entire revenue, which gave them the $1 in, $2 and eight cents of

118
00:07:44,238 --> 00:07:46,648
revenue out changed to 58%.

119
00:07:46,648 --> 00:07:53,055
So roughly for every dollar of compute spend that Anthropic invested in Q3 resulted in around $1.72.

120
00:07:53,287 --> 00:07:56,928
Still pretty good, but still notably declining. And you want to make sure that

121
00:07:56,928 --> 00:08:00,768
that isn't a continual decline. Now, the second thing that Anthropic is paying

122
00:08:00,768 --> 00:08:03,550
for that is going out of pocket is a little weirder.

123
00:08:03,724 --> 00:08:10,290
It's $2 billion of charitable donations matching for their employees. Now.

124
00:08:10,956 --> 00:08:14,457
I think about two weeks ago, I made an episode on how Anthropik is basically

125
00:08:14,457 --> 00:08:19,027
preparing for mass unemployment because they created this new charitable foundation,

126
00:08:19,027 --> 00:08:21,423
which is committing, I think, upwards of 10%

127
00:08:21,736 --> 00:08:26,177
of the entire company's equity valuation for the public good.

128
00:08:26,177 --> 00:08:30,547
The idea here is they want to distribute their shares and value of the company

129
00:08:30,547 --> 00:08:35,657
that it generates to the wider public because they're worried that people won't

130
00:08:35,657 --> 00:08:38,567
necessarily be able to earn a livable income.

131
00:08:38,567 --> 00:08:42,507
And so they wanted to create some kind of universal basic income or universal

132
00:08:42,507 --> 00:08:45,786
high income to distribute the goods and wealth that AI is going to generate.

133
00:08:45,995 --> 00:08:50,397
This is that first step towards that. And it's going to the tune of $2 billion.

134
00:08:51,032 --> 00:08:56,866
The reason why this is significant is $2 billion on any company's balance sheet is notable, right?

135
00:08:57,017 --> 00:09:00,167
The fact that this is a charitable donation that is eating up against their

136
00:09:00,167 --> 00:09:01,823
margins should be noted.

137
00:09:02,322 --> 00:09:06,547
But it should be kind of considered separately, I suppose, which is the sense

138
00:09:06,547 --> 00:09:10,217
that if Anthropic is likely going to be doing this consistently,

139
00:09:10,217 --> 00:09:15,777
but probably not to that large of an extent, we probably won't see that as a consistent line item.

140
00:09:15,777 --> 00:09:18,780
So we should probably put that aside and just focus on the compute costs,

141
00:09:19,093 --> 00:09:22,807
which brings me to the third thing, the estimate of how much they're going to

142
00:09:22,807 --> 00:09:28,257
spend over the next five to 10 years, which is roughly around 895 billion.

143
00:09:28,257 --> 00:09:32,397
We were discussing rumors, I remember, on this show as to what this spend might

144
00:09:32,397 --> 00:09:38,261
be like. And the initial rumors was $500 billion, half a trillion dollars of spend.

145
00:09:38,656 --> 00:09:42,247
That alone, I remember reacting to at the time, was huge.

146
00:09:42,247 --> 00:09:46,277
It is humongous by any means. And I had no idea as to how Anthropic was going

147
00:09:46,277 --> 00:09:47,197
to be able to pay for this.

148
00:09:47,197 --> 00:09:51,927
Now, the worst bear case prediction was around $700 to $750 billion.

149
00:09:51,927 --> 00:09:55,917
Well, it turns out that the figure is way larger. At $895 billion,

150
00:09:55,917 --> 00:09:57,410
you have to ask the question,

151
00:09:57,962 --> 00:10:01,757
has Anthropic kind of lost control of this? So I guess the way to look at this

152
00:10:01,757 --> 00:10:06,457
is let's actually break down how much of that is going to who and how much of

153
00:10:06,457 --> 00:10:09,427
that needs to be paid up front, because the immediate payment over the next

154
00:10:09,427 --> 00:10:11,227
12 months is $70 billion.

155
00:10:11,888 --> 00:10:15,907
Now, the first thing I want to call out is that the total figure is actually not $900 billion.

156
00:10:16,347 --> 00:10:21,527
It's probably more around $1.05 trillion, because there's around $150 billion.

157
00:10:22,122 --> 00:10:25,407
That haven't been priced in yet. These are deals that they have signed in the

158
00:10:25,407 --> 00:10:29,402
early part of October. So Q4 Financials is probably going to go crazy.

159
00:10:29,680 --> 00:10:34,893
But the idea here is it's a very large sum of money. Now, $320 billion of that

160
00:10:35,166 --> 00:10:39,087
is signed contracts where Anthropik is like, listen, I want to get access to

161
00:10:39,087 --> 00:10:42,457
this compute with the likes of SpaceX, with the likes of Google,

162
00:10:42,457 --> 00:10:44,489
with the likes of other neoclouds.

163
00:10:44,743 --> 00:10:49,287
Now, around $415 billion of that are like leases, which is, OK,

164
00:10:49,607 --> 00:10:52,967
if you build the data centers and you're able to kind of deliver that compute,

165
00:10:52,967 --> 00:10:54,826
we will pay you for this service.

166
00:10:55,169 --> 00:10:59,052
Now, I want to point a very obvious thing out, which maybe a lot of people don't understand.

167
00:10:59,465 --> 00:11:04,074
This is Anthropic renting compute. It's not them owning the compute.

168
00:11:04,347 --> 00:11:06,787
Owning the compute would mean that there's like a lot of debt on the balance

169
00:11:06,787 --> 00:11:11,327
sheet. This is them renting it. So it's only dependent on if these companies,

170
00:11:11,327 --> 00:11:16,369
these neoclouds, these hyperscalers, such as SpaceX AI, are able to deliver on the compute.

171
00:11:16,491 --> 00:11:20,377
But a lot of this is still meant to be paid for upfront.

172
00:11:20,377 --> 00:11:24,017
I think the figure is around 40 to 50%, which is gargantuan.

173
00:11:24,017 --> 00:11:32,036
And $70 billion of that is due over the next 12 months and $447 billion is due by 2030.

174
00:11:32,367 --> 00:11:36,087
Now let's address the $70 billion first. That sounds like a big number.

175
00:11:36,087 --> 00:11:38,654
But when I look at that cash balance sheet from last week's S1,

176
00:11:38,927 --> 00:11:40,247
they have around $132 billion.

177
00:11:40,567 --> 00:11:44,287
Okay, so they should be able to cover that. And then if they IPO over the next

178
00:11:44,287 --> 00:11:47,807
couple of weeks, the rumors say that they're trying to raise around $100 billion,

179
00:11:48,167 --> 00:11:51,801
that should cover that with a bit of extra room, certainly, right?

180
00:11:52,365 --> 00:11:56,218
But they keep signing more expensive leases and crazy compute deals.

181
00:11:56,218 --> 00:11:58,977
So the massive bet that Anthropic is taking is

182
00:11:59,407 --> 00:12:02,977
if we sign enough of these long-term compute deals

183
00:12:03,284 --> 00:12:07,476
and we keep generating enough demand for our products, for our models,

184
00:12:07,777 --> 00:12:11,284
and so we generate increasing amounts of revenue quarter upon quarter,

185
00:12:11,557 --> 00:12:16,508
we will end up being able to net a much higher margin for the amount that we're investing.

186
00:12:16,508 --> 00:12:21,318
So it seems like a big deal up front over here, but it's eventually going to

187
00:12:21,318 --> 00:12:26,118
end up in us earning way more profit than if we were to buy the compute in response

188
00:12:26,118 --> 00:12:29,378
to demand, which they did over the last three months and ended up eating into that profit.

189
00:12:29,378 --> 00:12:32,158
So they're trying to work against that scenario actually happening.

190
00:12:32,158 --> 00:12:37,656
Okay, but where is this money going? Well, you have around $121 billion going to SpaceX.

191
00:12:37,917 --> 00:12:41,660
And as you know, Anthropik's paying them around $3.1 billion per month.

192
00:12:42,123 --> 00:12:45,712
Elon Musk is rubbing his hands and smiling gleefully because he's making a ton

193
00:12:45,712 --> 00:12:50,337
of money. It's what led to the first positive revenue earnings of SpaceX a few weeks ago.

194
00:12:50,714 --> 00:12:54,812
Now, another $120 billion is going to Google for access to their TPUs.

195
00:12:54,812 --> 00:13:00,082
Remember, Anthropic has one of the most diverse GPU clusters,

196
00:13:00,082 --> 00:13:04,272
meaning they aren't just reliant on NVIDIA. They're reliant on Google's TPUs,

197
00:13:04,272 --> 00:13:07,602
which is their own versions of GPUs, as well as a number of different ASICs,

198
00:13:07,602 --> 00:13:09,680
which are specialized versions of GPUs.

199
00:13:09,922 --> 00:13:13,295
They're also working on their own chip in partnership with Broadcom.

200
00:13:13,498 --> 00:13:17,962
So the idea here is if they have this diversified stack they should be able

201
00:13:17,962 --> 00:13:23,162
to continue training Claude in as many different ways with as much flexibility

202
00:13:23,162 --> 00:13:24,714
as they can versus someone like.

203
00:13:25,086 --> 00:13:28,824
OpenAI which arguably has a lot more concentration risk on NVIDIA.

204
00:13:29,102 --> 00:13:34,062
Then you have another $110 billion going to Amazon and you have around another

205
00:13:34,062 --> 00:13:38,242
let's call it $100 billion going to say NeoClouds and then the Broadcom deal,

206
00:13:38,242 --> 00:13:41,272
which I mentioned just now, is around $125 billion.

207
00:13:41,592 --> 00:13:47,015
So all big figures concentrated to around, I want to say, four to six companies.

208
00:13:47,288 --> 00:13:50,892
And the circular economy sort of keeps on going there. I want to call that out,

209
00:13:50,892 --> 00:13:53,342
right? Because a lot of these companies are invested in Anthropic.

210
00:13:53,342 --> 00:13:59,229
You've got Amazon and Google, which I think collectively own around 25% of Anthropic pre-IPO.

211
00:13:59,531 --> 00:14:03,345
And then you've got Google owning a significant chunk of SpaceX.

212
00:14:03,565 --> 00:14:07,452
You've got SpaceX kind of in bed with NVIDIA and Jensen Huang securing all the

213
00:14:07,452 --> 00:14:11,202
compute. You've got NVIDIA and Jensen Huang backstopping a lot of the NeoClouds,

214
00:14:11,202 --> 00:14:15,522
such as Nscale, which is providing $45 billion worth of compute to Anthropic.

215
00:14:15,522 --> 00:14:18,350
So again, a lot of circular motion going on.

216
00:14:18,617 --> 00:14:21,752
And the question is, can Anthropic sustainably pay for all of this?

217
00:14:21,752 --> 00:14:24,902
Well, I answered the short-term, near-term thing, right? Which is they have

218
00:14:24,902 --> 00:14:27,151
around $130 billion in their cash balance sheet.

219
00:14:27,657 --> 00:14:31,790
Probably they'll be fine for the next $70 billion that's owed over the next 12 months.

220
00:14:32,214 --> 00:14:38,830
But what about the remaining $810 billion over the next couple of years?

221
00:14:39,022 --> 00:14:42,859
Well, that's the bet that Anthropic is making, and there's no easy way to say this.

222
00:14:43,097 --> 00:14:46,816
They are betting that the continual investment in compute, in training better

223
00:14:46,816 --> 00:14:52,356
models, in achieving recursive self-improvement, aka the model that is so far

224
00:14:52,356 --> 00:14:55,896
exceeding of human intelligence that it is able to automate a bunch of things

225
00:14:55,896 --> 00:14:58,085
and generate so much wealth for humanity,

226
00:14:58,497 --> 00:15:02,286
that model, if they are able to build that successfully, then they're going

227
00:15:02,286 --> 00:15:04,181
to generate endless amounts of money.

228
00:15:04,419 --> 00:15:07,774
That is a big, risky, and some might say far-fetched bet.

229
00:15:08,064 --> 00:15:12,516
Now, the bare case is they aren't able to build that in any short time span.

230
00:15:12,516 --> 00:15:17,340
And we're already seeing kind of these AI model labs restricting the release of their AI models.

231
00:15:17,793 --> 00:15:21,236
Rumor has it that Anthropik's been sitting on Fable 5.5, which is the codename

232
00:15:21,236 --> 00:15:22,600
of their next big model drop.

233
00:15:22,832 --> 00:15:26,186
OpenAI has been doing the same thing because it hasn't been able to sufficiently

234
00:15:26,186 --> 00:15:29,746
satisfy alignment procedures, which means that if you release the model right

235
00:15:29,746 --> 00:15:33,616
now, it would go out and do a bunch of crazy shit like cyber hacks,

236
00:15:33,616 --> 00:15:35,936
bioweapons, exploits, and stuff like that.

237
00:15:35,936 --> 00:15:39,916
So the idea is there's this push and pull. Regulations are getting involved.

238
00:15:39,916 --> 00:15:42,916
And most importantly, data centers are getting harder to stand up.

239
00:15:42,916 --> 00:15:46,606
I mentioned on an episode last week that there's hundreds of millions of dollars

240
00:15:46,606 --> 00:15:50,956
worth of GPUs collecting dust in warehouses because we haven't got the power

241
00:15:50,956 --> 00:15:54,956
or the energy grid or just the technical expertise to stand these up.

242
00:15:55,188 --> 00:15:58,736
So the idea here is things might potentially slow down, which means that Anthropic

243
00:15:58,736 --> 00:16:01,526
makes less revenue, which means that they aren't able to meet their compute

244
00:16:01,526 --> 00:16:05,300
bills, which means that this whole thing, this house of cards ends up collapsing.

245
00:16:05,521 --> 00:16:10,652
Now, I want to point out that this is a very big risk to take and there are

246
00:16:10,809 --> 00:16:14,376
great arguments to argue why Anthropic won't be able to pull this off and great

247
00:16:14,376 --> 00:16:16,898
arguments to say that they will. It's my inclination to.

248
00:16:17,531 --> 00:16:20,969
If we're able to continue this acceleration, and we're seeing it kind of recently

249
00:16:20,969 --> 00:16:24,999
on the periphery where you've got internal open-air models solving a lot of

250
00:16:24,999 --> 00:16:27,336
hard, complex challenges in math and physics.

251
00:16:27,585 --> 00:16:29,693
You've got Anthropik models doing the same in medicine.

252
00:16:30,029 --> 00:16:35,979
Imagine a world where these models are able to do so at scale to many other people right now.

253
00:16:35,979 --> 00:16:38,349
You look at the spending budgets of Anthropik as well, right?

254
00:16:38,349 --> 00:16:47,611
Where you see that it was less than 0.1% of the Fortune 500 customers that Anthropik has, which is.

255
00:16:48,169 --> 00:16:52,865
Quite a lot of them, have allocated that much to Claude Models. So less than 0.1%.

256
00:16:53,242 --> 00:16:54,862
That is a tiny amount.

257
00:16:55,338 --> 00:17:00,619
Imagine if that scales up. By the way, what that 0.1% looks like is around 11,900

258
00:17:00,619 --> 00:17:05,102
individuals or companies paying around $100,000 a year.

259
00:17:05,351 --> 00:17:11,145
And then you have around, I think it's like $200,000 to $300,000 paying $10 million per year.

260
00:17:11,429 --> 00:17:18,587
Now, that was about a 600% jump up from that same evaluation from three to five months earlier.

261
00:17:18,825 --> 00:17:22,199
So can you imagine what that looks like, I don't know, a year,

262
00:17:22,199 --> 00:17:23,417
a year and a half from now?

263
00:17:23,556 --> 00:17:27,169
You can imagine that that 0.1% budget spend is going to increase to,

264
00:17:27,169 --> 00:17:32,342
let's say, 1% or conservatively 5% of a company's spend, which I don't think is too crazy.

265
00:17:32,621 --> 00:17:38,199
And if a lot of that accrues to Anthropik, that's going to be a large amount of revenue increase.

266
00:17:38,199 --> 00:17:41,759
So that's the optimistic take. The negative take is they won't to be able to

267
00:17:41,759 --> 00:17:44,779
pull this off and we're going to grind to a halt because we can't get the power

268
00:17:44,779 --> 00:17:47,819
to supply any of these gpus in the first place which i guess brings us to

269
00:17:48,239 --> 00:17:51,709
the final part of this episode which is the bull and bear case for anthropics

270
00:17:51,709 --> 00:17:53,774
ipo is this a buy is this a sell,

271
00:17:54,227 --> 00:17:57,708
is this something to be optimistic about or something to be concerned about

272
00:17:57,871 --> 00:18:00,384
let's get into it on today's episode brought to you by qualcomm.

273
00:18:04,343 --> 00:18:08,601
My first point is the question has sort of changed from last week.

274
00:18:08,601 --> 00:18:12,046
Last week, the quick question was, does AI make money?

275
00:18:12,546 --> 00:18:16,639
It was answered pretty explicitly. Yes, Anthropic is making a heck ton of money,

276
00:18:16,900 --> 00:18:19,489
and they're doing so at a crazy growing rate.

277
00:18:19,965 --> 00:18:24,051
But now the question is, given this operating loss, and given the explanation

278
00:18:24,051 --> 00:18:27,547
of why they're making that loss, aka a lot of compute expense,

279
00:18:27,825 --> 00:18:33,481
can the demand come first before Anthropic has to pay the compute bill.

280
00:18:33,481 --> 00:18:38,239
In other words, can they generate enough demand for their clawed models before

281
00:18:38,460 --> 00:18:39,661
they have to pay the compute bill?

282
00:18:39,661 --> 00:18:43,261
So they need to make enough money to be able to pay it off before that bill arrives.

283
00:18:43,261 --> 00:18:46,111
Otherwise, you know, they run out of money. And it reminds me of something that

284
00:18:46,111 --> 00:18:50,911
Dario mentioned earlier this year, which he said, and he was kind of boasting, I guess, in a way,

285
00:18:51,331 --> 00:18:54,702
where I think it was, I forgot what podcast it was, but he basically said,

286
00:18:54,987 --> 00:19:00,911
I pride myself and Anthropics self on the ability to not go all in on compute.

287
00:19:00,911 --> 00:19:04,151
And he said this at the start of the because he said that is quite a costly

288
00:19:04,151 --> 00:19:10,115
expense and we have saved ourselves a lot of money and damage because we weren't able to do that.

289
00:19:10,636 --> 00:19:13,799
Of course, he ate his words a few weeks later, a few months later,

290
00:19:14,071 --> 00:19:17,883
when it was very much proven that scaling laws for training these models were

291
00:19:17,883 --> 00:19:21,923
very much intact, and that he needed way more compute in order to catch up.

292
00:19:21,923 --> 00:19:25,902
And that's, in fact, what allowed OpenAI, who was back then behind on models,

293
00:19:26,320 --> 00:19:29,383
allowed them to catch up to clawed models at the same proficiency because they

294
00:19:29,383 --> 00:19:31,543
invested so much in compute upfront.

295
00:19:31,543 --> 00:19:35,173
So what seemed like a pretty bad decision for OpenAI three months ago ended

296
00:19:35,173 --> 00:19:39,023
up becoming a really good decision back then in January, February time.

297
00:19:39,023 --> 00:19:42,609
So Anthropic, in the same vein, is making a very similar bet.

298
00:19:42,748 --> 00:19:46,313
And the answer to that will only be seen over the next couple of quarters.

299
00:19:46,313 --> 00:19:47,810
And it's something to definitely track.

300
00:19:48,042 --> 00:19:51,339
I actually think the best metric to track for this is what I mentioned earlier,

301
00:19:51,618 --> 00:19:56,796
the number of companies that are spending around $100,000 to $10 million per year.

302
00:19:57,214 --> 00:20:00,529
Are there more companies like that every single quarter?

303
00:20:00,744 --> 00:20:04,443
If that number keeps going up, then that tells us a very obvious thing,

304
00:20:04,443 --> 00:20:08,483
which is there's insatiable demand for this thing and the enterprise customer

305
00:20:08,483 --> 00:20:12,614
base, which is Anthropix bread and butter, is increasing in the way that you want it to.

306
00:20:12,869 --> 00:20:16,403
Now, the second thing to point out is, I guess, in favor of both the bull and

307
00:20:16,403 --> 00:20:18,883
bear case. On the bullish side of things, you could argue that.

308
00:20:19,325 --> 00:20:23,001
Yes, Anthropic made an operating loss, but it was because they had so much demand

309
00:20:23,001 --> 00:20:26,111
that they had to pay out of pocket for more expensive ingredients,

310
00:20:26,111 --> 00:20:28,294
if you want to use my restaurant analogy earlier.

311
00:20:28,712 --> 00:20:32,971
The bad case is Anthropic didn't even have a cushion or figure out a way to

312
00:20:32,971 --> 00:20:37,841
kind of satiate this demand or anticipate this demand, and that might end up with them blowing up.

313
00:20:37,841 --> 00:20:41,401
And like, listen, this is a high stakes game that Anthropic is playing.

314
00:20:41,401 --> 00:20:46,311
They don't have alternative revenue earnings from any other parts of a business.

315
00:20:46,311 --> 00:20:50,081
They are an AI-focused company, and they are betting the entire house on it.

316
00:20:50,081 --> 00:20:55,211
If you compare them to Google, Meta, Microsoft, they have more flexibility.

317
00:20:55,211 --> 00:20:57,961
Google has search revenue, which I know is going down slightly,

318
00:20:57,961 --> 00:20:59,693
but they're making billions of dollars from that.

319
00:20:59,919 --> 00:21:03,507
And they have so many other product lines. You've got Meta with the same scenario as well.

320
00:21:03,762 --> 00:21:07,332
And so they should be able to kind of keep this expense afloat.

321
00:21:07,611 --> 00:21:11,061
Anthropic does not have the same sort of flexibility, neither does OpenAI.

322
00:21:11,061 --> 00:21:14,951
And the other thing I want to tack on to that is market sentiment is probably

323
00:21:14,951 --> 00:21:18,371
at its lowest that it's been since this entire AI hype started,

324
00:21:18,487 --> 00:21:19,963
I don't know, a few years ago.

325
00:21:20,211 --> 00:21:23,711
So the idea here is, have we reached the breaking point where no matter how

326
00:21:23,711 --> 00:21:28,251
much good news a company can sort of surface, it's still going to be bearish

327
00:21:28,251 --> 00:21:30,567
from a public investment side of things? At the end of the day,

328
00:21:30,881 --> 00:21:35,136
Anthropic needs to raise $100 billion in roughly around two weeks.

329
00:21:35,612 --> 00:21:39,596
Will they be able to do it in the light of their second biggest competitor,

330
00:21:39,596 --> 00:21:42,576
OpenAI, not even being able to raise $30 billion?

331
00:21:43,016 --> 00:21:47,066
And we heard from Elon Musk last week that he even struggled to raise the $40

332
00:21:47,066 --> 00:21:50,626
billion that he needed to buy more compute. And he has a very obvious business

333
00:21:50,626 --> 00:21:52,563
line where he just sells it at a premium.

334
00:21:52,830 --> 00:21:57,410
So it's obvious that the market is questioning whether this thing has reached its breaking point.

335
00:21:57,694 --> 00:21:59,796
But on the positive side, so

336
00:21:59,796 --> 00:22:03,676
little of the world has actually been penetrated by AI in the first place.

337
00:22:03,676 --> 00:22:07,966
I think you could feasibly estimate that if they remain the top two AI labs

338
00:22:07,966 --> 00:22:12,356
in the world, they'll be able to capture a humongous amount of this revenue

339
00:22:12,356 --> 00:22:14,456
that would be able to pay for a lot of this cost.

340
00:22:14,456 --> 00:22:19,006
Listen, when I step back from all of this, I want to point out that I am probably

341
00:22:19,006 --> 00:22:22,376
one of the bigger or biggest AI bulls out there.

342
00:22:22,747 --> 00:22:27,836
But even I am a little skeptical as to how much money is going into this thing.

343
00:22:27,836 --> 00:22:30,282
And I want to acknowledge that it is a humongous bet.

344
00:22:30,613 --> 00:22:35,257
Now, I naturally lean on the side of thinking, if you are a long-term investor,

345
00:22:35,652 --> 00:22:37,716
then this is something that is obvious for you.

346
00:22:37,716 --> 00:22:41,036
Directionally, AI is going to go out there and change so many things.

347
00:22:41,036 --> 00:22:45,096
Even if you're not a massive believer in centralized labs, if you're a believer

348
00:22:45,096 --> 00:22:47,813
in open source or whatever that might be, the idea here is simple.

349
00:22:48,323 --> 00:22:51,597
AI is going to improve in intelligence and it's going to decrease in cost,

350
00:22:51,702 --> 00:22:55,047
which means that eventually it's going to be able to run on any and all of our devices.

351
00:22:55,047 --> 00:22:58,107
We're going to have personal agents that we can interact with at any single

352
00:22:58,107 --> 00:23:02,367
point in time that can do myriad different things for us. It's not just going

353
00:23:02,367 --> 00:23:05,088
to be work-related, it's going to be leisure-related as well.

354
00:23:05,355 --> 00:23:08,796
And so in that world, are you going to be net-net more productive,

355
00:23:08,901 --> 00:23:12,247
net-net more successful, and make more money? What are your odds of that?

356
00:23:12,587 --> 00:23:16,383
Or are you going to lose more money and consume many of these products.

357
00:23:16,697 --> 00:23:19,537
It could go either way. I tend to think that this technology is going to generate

358
00:23:19,537 --> 00:23:24,457
an enormous amount of wealth, and we need to be tentative in how we're spending

359
00:23:24,457 --> 00:23:25,892
and make sure we don't over leverage.

360
00:23:26,141 --> 00:23:29,949
But we do need to fund the build out. It is now more important than ever.

361
00:23:30,187 --> 00:23:34,277
But equally so, if you are maybe more short term minded, and you're concerned

362
00:23:34,277 --> 00:23:38,167
about the inflated valuations of a lot of these NeoClouds, the amount of money

363
00:23:38,167 --> 00:23:40,294
that's being spent on computer, most importantly, and this is,

364
00:23:40,747 --> 00:23:45,527
very understandable and valid, that a lot of these GPUs and compute aren't able

365
00:23:45,527 --> 00:23:49,936
to actually come online as fast as we wanted to. And so there's naturally going to be a slowdown.

366
00:23:50,471 --> 00:23:53,449
Then that's fair. And you should sort of invest and trade accordingly.

367
00:23:53,449 --> 00:23:56,589
Again, like none of this is financial advice. It's just my kind of thoughts on this.

368
00:23:56,589 --> 00:24:00,659
I'm more long-term minded. I know a lot of our recent episodes have been market-oriented,

369
00:24:00,659 --> 00:24:02,729
but hey, the market's sort of calling for it.

370
00:24:02,729 --> 00:24:05,819
Stocks are down pretty massively. And I want to inject a bit of optimism,

371
00:24:05,819 --> 00:24:09,779
I guess, into this entire thing. We are going to see one of the biggest IPOs,

372
00:24:09,779 --> 00:24:13,667
whether you like it or not, in a few weeks' time, if I had to guess.

373
00:24:13,899 --> 00:24:20,429
My bet is a lot of the numbers that we are speculating to be bad a year from

374
00:24:20,429 --> 00:24:23,663
now is actually going to be a lot more surprising.

375
00:24:23,918 --> 00:24:27,799
I think it's going to move towards the upside. And if I had to bet where those

376
00:24:27,799 --> 00:24:31,669
numbers of Fortune 500 companies, number one, spending a larger percentage of

377
00:24:31,669 --> 00:24:36,569
their budget, certainly greater than 0.1%, but also two, spending tens of millions

378
00:24:36,569 --> 00:24:37,899
of dollars versus just $10 million

379
00:24:38,199 --> 00:24:40,787
a year on Anthropik models or just AI models in general,

380
00:24:41,095 --> 00:24:44,979
I will bet you that it's going to be much, much higher a year from now.

381
00:24:45,164 --> 00:24:48,366
And yeah, that's it. Thank you for listening to today's episode brought to you by Qualcomm.