1
00:00:03,160 --> 00:00:03,560
Speaker 1: Media.

2
00:00:05,200 --> 00:00:07,880
Speaker 2: Hi, I'm Edzetron, and this is the second part of

3
00:00:07,920 --> 00:00:22,479
Speaker 2: this week's better offline. As ever, go to the episode

4
00:00:22,520 --> 00:00:25,239
Speaker 2: notes by some merch get the challenge coin, subscribe to

5
00:00:25,239 --> 00:00:28,800
Speaker 2: my newsletter if you like it, go premium. Either way,

6
00:00:28,960 --> 00:00:31,000
Speaker 2: you've got this episode for free if you can stand

7
00:00:31,040 --> 00:00:32,360
Speaker 2: the ads, which many of you can.

8
00:00:32,840 --> 00:00:33,240
Speaker 3: Anyway.

9
00:00:33,320 --> 00:00:35,280
Speaker 2: The thrust of this two part is that the AI

10
00:00:35,400 --> 00:00:37,880
Speaker 2: trade is going to, as you've probably guessed by now,

11
00:00:38,240 --> 00:00:40,599
Speaker 2: end badly, and as I'll also explain, I believe we're

12
00:00:40,640 --> 00:00:43,720
Speaker 2: now set up for potentially economic and market wrecking consequences

13
00:00:44,000 --> 00:00:48,320
Speaker 2: of this hysterical investment bubble. Today we're going to go

14
00:00:48,560 --> 00:00:51,000
Speaker 2: beyond just startups and into how the poison of the

15
00:00:51,040 --> 00:00:53,720
Speaker 2: bubble is crept into our economy and how dangerous things

16
00:00:53,760 --> 00:00:56,640
Speaker 2: are getting as a result. In the first installment, we

17
00:00:56,680 --> 00:00:59,800
Speaker 2: talked about Cursor, the AI coding startup that, despite having

18
00:00:59,800 --> 00:01:03,000
Speaker 2: no mote and zero sustainability, has somehow earned a valuation

19
00:01:03,080 --> 00:01:05,440
Speaker 2: of ten billion dollars, and then asked why none of

20
00:01:05,440 --> 00:01:08,160
Speaker 2: these generative AI companies are being acquired, at least not

21
00:01:08,280 --> 00:01:11,560
Speaker 2: in the traditional sense. And no, I'm not including the

22
00:01:11,560 --> 00:01:14,160
Speaker 2: acquisitions where a big company snags the talent and leaves

23
00:01:14,200 --> 00:01:17,880
Speaker 2: barely breathing bodies of startups in their wake. This matters

24
00:01:17,920 --> 00:01:20,720
Speaker 2: because if these AI companies can't get an exit, go public,

25
00:01:20,800 --> 00:01:23,759
Speaker 2: or get bought, it raises serious questions about how any

26
00:01:23,800 --> 00:01:26,200
Speaker 2: of this ends. Cursor is the best example, and the

27
00:01:26,240 --> 00:01:29,160
Speaker 2: fact that Anthropic is hopelessly dependent on Cursor for revenue

28
00:01:29,319 --> 00:01:31,640
Speaker 2: despite doing everything they can to kill it, makes the

29
00:01:31,800 --> 00:01:36,200
Speaker 2: entire situation feel a little bit weird. Cursor's paths to

30
00:01:36,280 --> 00:01:39,200
Speaker 2: viability seem frankly to be non existent, and the collapse

31
00:01:39,200 --> 00:01:41,320
Speaker 2: of Cursor will inevitably result in a vast chunk of

32
00:01:41,400 --> 00:01:44,560
Speaker 2: anthropics revenue is going up in smoke, which again will

33
00:01:44,600 --> 00:01:47,120
Speaker 2: also make its collapse seem inevitable, as how can it

34
00:01:47,160 --> 00:01:49,960
Speaker 2: fundraise that evaluation over its previous one when its biggest

35
00:01:49,960 --> 00:01:53,400
Speaker 2: customer just died? So much of what I'm describing as

36
00:01:53,440 --> 00:01:57,160
Speaker 2: structural and few companies have bigger structural obstacles to survive

37
00:01:57,200 --> 00:02:00,160
Speaker 2: all than open AI. As a reminder, open aia I

38
00:02:00,160 --> 00:02:02,320
Speaker 2: appears to have burned at least ten billion dollars in

39
00:02:02,360 --> 00:02:05,120
Speaker 2: the last two months. It has just raised another eight

40
00:02:05,120 --> 00:02:07,560
Speaker 2: point three billion dollars after raising ten billion dollars in

41
00:02:07,640 --> 00:02:09,800
Speaker 2: June according to The New York Times, and intends to

42
00:02:09,800 --> 00:02:12,040
Speaker 2: receive around twenty two point five billion dollars by the

43
00:02:12,120 --> 00:02:14,400
Speaker 2: end of the year from SoftBank, and that is assuming

44
00:02:14,440 --> 00:02:16,280
Speaker 2: it becomes a for profit entity by the end of

45
00:02:16,360 --> 00:02:18,680
Speaker 2: the year, and if that doesn't happen, the route gets

46
00:02:18,720 --> 00:02:21,079
Speaker 2: cut to twenty billion dollars total, meaning that SoftBank would

47
00:02:21,080 --> 00:02:22,560
Speaker 2: only be on the hook for a further one point

48
00:02:22,600 --> 00:02:25,880
Speaker 2: seven billion dollars. I am repeating myself, and I need

49
00:02:25,919 --> 00:02:28,520
Speaker 2: you to really get this. Open Ai just got ten

50
00:02:28,560 --> 00:02:31,600
Speaker 2: billion dollars in June, just in June, and had to

51
00:02:31,680 --> 00:02:34,200
Speaker 2: raise another eight point three billion dollars in August. And

52
00:02:34,240 --> 00:02:37,920
Speaker 2: that is an unbelievable cash burn, one dwarfing any startup

53
00:02:37,919 --> 00:02:40,760
Speaker 2: in history, raveled only by Xai, makers of Grok the

54
00:02:40,840 --> 00:02:43,760
Speaker 2: races dollar m which loses over a billion dollars a month.

55
00:02:44,440 --> 00:02:46,720
Speaker 2: I should also be clear that if open ai does

56
00:02:46,760 --> 00:02:48,840
Speaker 2: not convert to a for profit there is no path

57
00:02:48,919 --> 00:02:52,000
Speaker 2: forward to continue raising capital. Open Ai must have the

58
00:02:52,000 --> 00:02:54,720
Speaker 2: promise of an IPO. It must go public because a

59
00:02:54,919 --> 00:02:57,880
Speaker 2: valuation of three hundred billion dollars or more. Open Ai

60
00:02:57,960 --> 00:03:00,360
Speaker 2: can no longer be acquired because nobody has that much money.

61
00:03:00,360 --> 00:03:03,600
Speaker 2: And if Let's be real, nobody actually believes open ai

62
00:03:03,680 --> 00:03:06,120
Speaker 2: is worth that much. The only way to prove that

63
00:03:06,160 --> 00:03:08,320
Speaker 2: anybody does is to take open ai public, and that

64
00:03:08,360 --> 00:03:11,800
Speaker 2: will be impossible if it cannot convert. And ironically, SoftBank's

65
00:03:11,840 --> 00:03:16,079
Speaker 2: large and late stage participation makes exits actually harder, as

66
00:03:16,120 --> 00:03:19,000
Speaker 2: early investors will see their holdings diluted as a percentage

67
00:03:19,000 --> 00:03:21,400
Speaker 2: of total equity or whatever the hell we're calling it.

68
00:03:21,400 --> 00:03:25,400
Speaker 2: Because these aren't real equity shares, their profit participation units.

69
00:03:25,560 --> 00:03:28,440
Speaker 3: They're a nonprofit. You can't do stock in them.

70
00:03:28,800 --> 00:03:31,079
Speaker 2: While a normal company could just issue equity and deal

71
00:03:31,120 --> 00:03:34,040
Speaker 2: with the dilution that way, open AI's structure necess states

72
00:03:34,040 --> 00:03:36,920
Speaker 2: the negotiation, where companies can obstruct the entire process if

73
00:03:36,920 --> 00:03:40,560
Speaker 2: they see fit. Speaking of companies doing that, let's talk

74
00:03:40,600 --> 00:03:43,440
Speaker 2: about Microsoft. As I asked in my premium newsletter a

75
00:03:43,480 --> 00:03:46,280
Speaker 2: few weeks ago, what if Microsoft does not want open

76
00:03:46,320 --> 00:03:49,640
Speaker 2: ai to convert. They own all the IP, they own

77
00:03:49,680 --> 00:03:52,240
Speaker 2: all the access to open AI's research, and already run

78
00:03:52,320 --> 00:03:55,600
Speaker 2: most of their infrastructure, while assuming the best case scenario

79
00:03:55,720 --> 00:03:57,440
Speaker 2: that it would end up owning a massive junk of

80
00:03:57,440 --> 00:03:59,360
Speaker 2: the biggest tech stile up of all time. And I'm

81
00:03:59,400 --> 00:04:02,360
Speaker 2: talking about equity not open AI's current profit sharing units.

82
00:04:02,600 --> 00:04:05,400
Speaker 2: Microsoft might also believe that it stands to gain more

83
00:04:05,600 --> 00:04:08,160
Speaker 2: by letting open ai die and assuming its role in

84
00:04:08,200 --> 00:04:11,960
Speaker 2: the AI ecosystem. But let's assume that open ai converts

85
00:04:12,000 --> 00:04:15,240
Speaker 2: and now has to continue raising money at a rate

86
00:04:15,280 --> 00:04:18,000
Speaker 2: that will require it allegedly to only need to raise

87
00:04:18,040 --> 00:04:21,440
Speaker 2: seventeen billion dollars in twenty twenty seven. That number doesn't

88
00:04:21,440 --> 00:04:23,960
Speaker 2: make any sense considering the open ai already had to

89
00:04:23,960 --> 00:04:26,400
Speaker 2: bring forward its eight point three billion dollar fund raised

90
00:04:26,400 --> 00:04:29,240
Speaker 2: by at least three months. But let's stick with that idea.

91
00:04:29,400 --> 00:04:32,480
Speaker 2: Open ai believes it will be profitable somehow by twenty thirty,

92
00:04:32,520 --> 00:04:34,480
Speaker 2: and even if we assume that that means it intends

93
00:04:34,480 --> 00:04:37,279
Speaker 2: to burn over one hundred billion dollars total to get there,

94
00:04:37,760 --> 00:04:40,680
Speaker 2: is the plan to take open ai public and then

95
00:04:40,760 --> 00:04:45,279
Speaker 2: dumper toxic acx onto the public markets. Then it will

96
00:04:45,320 --> 00:04:48,320
Speaker 2: just flounder and convulse and die for everybody to seek.

97
00:04:48,720 --> 00:04:51,640
Speaker 2: Kind of sucks. Can you imagine open aiss one? How

98
00:04:51,680 --> 00:04:54,240
Speaker 2: old do you think this company would do when they

99
00:04:54,360 --> 00:04:57,200
Speaker 2: had a true financial audit from a major accounting firm,

100
00:04:57,320 --> 00:04:59,599
Speaker 2: and you can claim they already have one. I simply

101
00:04:59,600 --> 00:05:02,600
Speaker 2: don't belie leave you, this burn does not seem like

102
00:05:02,640 --> 00:05:05,720
Speaker 2: an accountancy firm would be good with them. And if

103
00:05:05,760 --> 00:05:07,880
Speaker 2: you want to know what all that looks like Google

104
00:05:07,880 --> 00:05:10,279
Speaker 2: we Work, which went from tech industry darling to joke

105
00:05:10,320 --> 00:05:12,200
Speaker 2: in a matter of days in part because it was

106
00:05:12,240 --> 00:05:14,560
Speaker 2: forced to disclose how bad things actually were when they

107
00:05:14,560 --> 00:05:16,880
Speaker 2: filed to go public. No really are linked to an

108
00:05:16,960 --> 00:05:19,200
Speaker 2: article I've linked in the spreadsheet called the Strangest and

109
00:05:19,240 --> 00:05:22,080
Speaker 2: most Alarming things in we Work's IPO filing. It's a

110
00:05:22,080 --> 00:05:24,719
Speaker 2: bad sign if your IPO filing is described as strange

111
00:05:24,720 --> 00:05:28,080
Speaker 2: and alarming. Those are not even terms of art, they're

112
00:05:28,080 --> 00:05:30,880
Speaker 2: just worrisome. With that in mind, I feel the same

113
00:05:30,880 --> 00:05:33,640
Speaker 2: way about Anthropic. Nobody is buying this company a one

114
00:05:33,720 --> 00:05:35,680
Speaker 2: hundred and seventy billion dollars, which is how much it's

115
00:05:35,720 --> 00:05:37,680
Speaker 2: raising it right now, and thus the only way to

116
00:05:37,720 --> 00:05:40,599
Speaker 2: access liquidity would be to take Anthropic public and show

117
00:05:40,600 --> 00:05:42,920
Speaker 2: the world how a company that made seventy two million

118
00:05:42,960 --> 00:05:45,599
Speaker 2: dollars in January and more than four hundred million dollars

119
00:05:45,600 --> 00:05:48,320
Speaker 2: in July twenty twenty five and also plans to lose

120
00:05:48,320 --> 00:05:51,680
Speaker 2: three billion dollars or more, and then you show them

121
00:05:51,720 --> 00:05:54,280
Speaker 2: that and let the market sign a fair price, especially

122
00:05:54,279 --> 00:05:56,440
Speaker 2: if the S one includes a bunch of strange and

123
00:05:56,480 --> 00:06:00,000
Speaker 2: alarming stuff that raises questions about whether anthropics value is

124
00:06:00,000 --> 00:06:06,800
Speaker 2: actually justified. I'm I'm stuttering here because everyone's acting like

125
00:06:06,839 --> 00:06:09,680
Speaker 2: these irrational things like yeah, our open ai, which burns

126
00:06:09,839 --> 00:06:12,520
Speaker 2: five billion dollars in twenty twenty four, probably ten or

127
00:06:12,600 --> 00:06:15,880
Speaker 2: more in twenty twenty five. If if not fifteen or twenty,

128
00:06:16,000 --> 00:06:17,960
Speaker 2: like that's fine, Like this is all good.

129
00:06:17,960 --> 00:06:20,719
Speaker 3: They'll just go public one day, will they really?

130
00:06:20,920 --> 00:06:21,240
Speaker 2: Really?

131
00:06:21,520 --> 00:06:22,200
Speaker 3: Will they? No?

132
00:06:22,360 --> 00:06:25,960
Speaker 2: Really, what's the plan? What's the plan here? Because the

133
00:06:26,040 --> 00:06:28,360
Speaker 2: arguments against my work always come down to costs will

134
00:06:28,400 --> 00:06:31,120
Speaker 2: go down and these products will become essential. Outside of

135
00:06:31,200 --> 00:06:33,520
Speaker 2: chat GPT, there's no real proof that these products are

136
00:06:33,520 --> 00:06:37,080
Speaker 2: anything remotely essential, And I'd argue there's very little about

137
00:06:37,120 --> 00:06:40,120
Speaker 2: chat GPT that Microsoft can't provide with rate limits via

138
00:06:40,200 --> 00:06:42,840
Speaker 2: co pilot. I'd also argue that essential is a very

139
00:06:42,880 --> 00:06:46,280
Speaker 2: subjective term, essentially in the sense that some people use

140
00:06:46,320 --> 00:06:49,080
Speaker 2: it as search doesn't mean it's useful for enterprises or

141
00:06:49,120 --> 00:06:51,480
Speaker 2: the majority of people, or could not be replaced by

142
00:06:51,520 --> 00:06:55,520
Speaker 2: anything else. And I guess chat GPT somehow makes one

143
00:06:55,600 --> 00:06:58,080
Speaker 2: billion dollars a month in revenue selling access to premium

144
00:06:58,160 --> 00:07:01,400
Speaker 2: versions of chat GPT, though I'm not one hundred percent

145
00:07:01,440 --> 00:07:04,760
Speaker 2: sure how Assuming it has twenty million customers paying twenty

146
00:07:04,800 --> 00:07:07,080
Speaker 2: bucks a month, that's four hundred million dollars a month.

147
00:07:07,160 --> 00:07:09,640
Speaker 2: Then five million business customers paying an average of one

148
00:07:09,720 --> 00:07:12,680
Speaker 2: hundred dollars a month each, that's nine hundred million dollars.

149
00:07:12,960 --> 00:07:15,760
Speaker 2: And is that is that really indicative? Is the average

150
00:07:15,760 --> 00:07:19,000
Speaker 2: really that good? Are there that many people paying thirty

151
00:07:19,000 --> 00:07:21,560
Speaker 2: five or fifty or two hundred bucks a month? How

152
00:07:21,560 --> 00:07:24,760
Speaker 2: many of those versus one hundred bucks? I mean, most

153
00:07:24,800 --> 00:07:27,080
Speaker 2: don't pay one hundred bucks. It's either thirty five, fifty

154
00:07:27,120 --> 00:07:29,720
Speaker 2: or two hundred. Open ai doesn't break out the actual

155
00:07:29,720 --> 00:07:31,960
Speaker 2: revenues behind these numbers for a reason. I believe that

156
00:07:32,240 --> 00:07:34,880
Speaker 2: reason is they don't look as good. I also, by

157
00:07:34,920 --> 00:07:37,480
Speaker 2: the way, when I was fucking around trying to play

158
00:07:37,480 --> 00:07:40,880
Speaker 2: with GPT five, I got offered a team subscription for

159
00:07:40,960 --> 00:07:45,200
Speaker 2: a dollar for one month. How how's how many people?

160
00:07:45,400 --> 00:07:47,680
Speaker 2: How many of those business customers got a month for

161
00:07:47,720 --> 00:07:50,920
Speaker 2: a dollar? How long does open ai counter customer? What

162
00:07:51,080 --> 00:07:54,240
Speaker 2: is open AI's churn like? And does it really as

163
00:07:54,240 --> 00:07:56,000
Speaker 2: I wrote in my news there, how much money do

164
00:07:56,040 --> 00:07:58,560
Speaker 2: open AI and Anthropic make in the year making more

165
00:07:58,600 --> 00:08:01,280
Speaker 2: than Spotify at one point five billion dollars a month.

166
00:08:01,720 --> 00:08:03,960
Speaker 2: We don't know an open AI, much like Anthropic, has

167
00:08:03,960 --> 00:08:07,320
Speaker 2: never shared actual revenues other than in June when they

168
00:08:07,360 --> 00:08:10,680
Speaker 2: said ten billion annualized revenue, choosing instead to leak to

169
00:08:10,680 --> 00:08:12,880
Speaker 2: the media and hope to wfiscape the actual amounts of

170
00:08:12,920 --> 00:08:16,240
Speaker 2: money being spent on its services. Anyway, long story short,

171
00:08:16,280 --> 00:08:18,760
Speaker 2: these companies are unprofitable with no end in sight, don't

172
00:08:18,800 --> 00:08:20,960
Speaker 2: even make that much money in most cases, and have

173
00:08:21,040 --> 00:08:23,560
Speaker 2: valued more than anybody would ever buy them for. And

174
00:08:23,600 --> 00:08:25,400
Speaker 2: they don't even have that much in the way of

175
00:08:25,480 --> 00:08:28,280
Speaker 2: valuable IP. And on top of that, the two biggest

176
00:08:28,280 --> 00:08:31,440
Speaker 2: players burn billions of dollars more than they make serving

177
00:08:31,520 --> 00:08:36,400
Speaker 2: these companies that also burn billions of dollars. I wish

178
00:08:36,480 --> 00:08:39,760
Speaker 2: someone would make this make sense because it doesn't. But

179
00:08:40,080 --> 00:08:42,559
Speaker 2: ed the government will give them more money forever, They'll

180
00:08:42,559 --> 00:08:45,400
Speaker 2: bail them out. I'm tired of this fucking point. I'm

181
00:08:45,440 --> 00:08:47,640
Speaker 2: so tired of hearing about bailouce. I'm so tired of

182
00:08:47,679 --> 00:08:50,440
Speaker 2: people saying they'll bail them out. I get that you're scared,

183
00:08:50,480 --> 00:08:53,080
Speaker 2: I get that things are grim, things are absolutely fucking grim,

184
00:08:53,240 --> 00:08:56,320
Speaker 2: But engage with reality here. You can't bail this out.

185
00:08:56,720 --> 00:08:59,240
Speaker 2: You can't bail it out even if this were going

186
00:08:59,320 --> 00:09:02,280
Speaker 2: to happen, will not. Who would they give the money

187
00:09:02,320 --> 00:09:02,840
Speaker 2: to and.

188
00:09:02,760 --> 00:09:05,800
Speaker 3: For how long would they give it to? All AI startups?

189
00:09:06,200 --> 00:09:08,880
Speaker 2: Is every startup going to get a paycheck protection program

190
00:09:08,920 --> 00:09:11,839
Speaker 2: for generative AI? How would that play out in rural,

191
00:09:11,880 --> 00:09:14,400
Speaker 2: red districts where big tech has never been popular, which

192
00:09:14,440 --> 00:09:16,720
Speaker 2: are being hit with both massive cuts to welfare as

193
00:09:16,760 --> 00:09:18,520
Speaker 2: well as the shock waves of betrayed war that has

194
00:09:18,520 --> 00:09:22,240
Speaker 2: made American agricultural exports like feedstocks, which previously went to

195
00:09:22,280 --> 00:09:26,840
Speaker 2: China by shiploads, less appealing worldwide. So they just, in

196
00:09:26,880 --> 00:09:29,160
Speaker 2: this scenario, by the way, bail out open AI then

197
00:09:29,360 --> 00:09:31,679
Speaker 2: stuff it full of government contracts to the tune of

198
00:09:32,120 --> 00:09:35,720
Speaker 2: what fifteen billion dollars a year? Right just in this example,

199
00:09:36,280 --> 00:09:38,080
Speaker 2: just to be clear, that's the low end of what

200
00:09:38,120 --> 00:09:39,679
Speaker 2: this would take to do. And they'll have to keep

201
00:09:39,720 --> 00:09:42,040
Speaker 2: doing it forever until Sam Wultman can build enough data

202
00:09:42,080 --> 00:09:45,720
Speaker 2: centers to what keep burning billions of dollars. There's no

203
00:09:45,880 --> 00:09:49,240
Speaker 2: plan to make this profitable. Say this happens, it won't

204
00:09:49,240 --> 00:09:54,000
Speaker 2: be Sam, Now what America has a bullshit generative AI

205
00:09:54,080 --> 00:09:56,920
Speaker 2: company attached to it, attached to the state that doesn't

206
00:09:56,920 --> 00:09:59,280
Speaker 2: really innovate. It doesn't really matter in any meaningful way

207
00:09:59,320 --> 00:10:01,840
Speaker 2: except that it owns a bunch of data centers, which

208
00:10:01,840 --> 00:10:04,520
Speaker 2: it doesn't yet. By the way, Microsoft owns all their infrastructure,

209
00:10:04,520 --> 00:10:07,080
Speaker 2: the core stuff is barely happening. Fucking hell, I just

210
00:10:07,120 --> 00:10:09,400
Speaker 2: don't think this happens. I think it's a silly idea,

211
00:10:09,440 --> 00:10:11,720
Speaker 2: and the most likely situation would be that Microsoft would

212
00:10:11,800 --> 00:10:16,080
Speaker 2: unhinge its daw and swallow open AI in its customer's whole. Hey, hey, hey,

213
00:10:16,120 --> 00:10:18,560
Speaker 2: did you know, by the way, the Microsoft's data center

214
00:10:18,600 --> 00:10:20,880
Speaker 2: construction is down year every year, and it's basically signed

215
00:10:20,920 --> 00:10:22,200
Speaker 2: no new data center leases.

216
00:10:22,520 --> 00:10:24,000
Speaker 3: I wonder why it isn't building these new.

217
00:10:23,960 --> 00:10:26,920
Speaker 2: Data centers or open AI could be anything. By the way,

218
00:10:26,960 --> 00:10:30,480
Speaker 2: Stargate isn't saving them either. As I've written recently, Stargate

219
00:10:30,480 --> 00:10:32,800
Speaker 2: doesn't actually exist beyond the media hype it generated, and

220
00:10:32,880 --> 00:10:37,199
Speaker 2: every single person who wrote big articles about Stargate being

221
00:10:37,280 --> 00:10:40,800
Speaker 2: real and Stargate being connected to Abilene Texas should be

222
00:10:40,800 --> 00:10:44,920
Speaker 2: fucking ashamed of themselves. It's deeply unprofessional anyway. That also

223
00:10:45,040 --> 00:10:49,920
Speaker 2: counts the hint of the involvement. There are several articles

224
00:10:49,920 --> 00:10:53,920
Speaker 2: that hint without directly saying that Abilene Texas was part

225
00:10:53,960 --> 00:10:57,120
Speaker 2: of the soft bank stargate thing. You can't wriggle out

226
00:10:57,120 --> 00:10:59,560
Speaker 2: of this one when the time comes or will be

227
00:10:59,600 --> 00:11:01,679
Speaker 2: ad by which I mean, I'm going to be pointing

228
00:11:01,679 --> 00:11:05,079
Speaker 2: at the articles aggressively. We have other stuff to do anyway.

229
00:11:05,160 --> 00:11:07,240
Speaker 2: By the way, does the government do this for everybody?

230
00:11:07,240 --> 00:11:10,280
Speaker 2: This very important question is everyone getting bailed out? Is

231
00:11:10,320 --> 00:11:13,360
Speaker 2: it just anthropical out and open AI? What about Google

232
00:11:13,440 --> 00:11:15,320
Speaker 2: and Microsoft? They're going to be burning and come on

233
00:11:15,440 --> 00:11:17,040
Speaker 2: a bunch of money. Why do they get the money

234
00:11:17,080 --> 00:11:19,080
Speaker 2: no one else? While their industries are going to turn

235
00:11:19,120 --> 00:11:20,840
Speaker 2: around and go wait wait, wait, wait, wait wait. Why

236
00:11:20,840 --> 00:11:23,240
Speaker 2: do I have to fucking make money when what's the

237
00:11:23,240 --> 00:11:25,760
Speaker 2: point is? Because AI is important, You're going to get

238
00:11:25,760 --> 00:11:27,920
Speaker 2: a bunch of fraud if this happens, it's not going

239
00:11:27,960 --> 00:11:30,840
Speaker 2: to Just to be clear, and what is the limit?

240
00:11:30,880 --> 00:11:33,959
Speaker 2: By the way, do they subsidize or compute for companies

241
00:11:34,000 --> 00:11:36,760
Speaker 2: like Cursor to what end? Where's the limit? How does

242
00:11:36,800 --> 00:11:38,640
Speaker 2: it end? What is the bailout? What are you bailing out?

243
00:11:38,640 --> 00:11:41,400
Speaker 2: Where does the bailout go? These are all questions that

244
00:11:41,400 --> 00:11:46,080
Speaker 2: would get answered. You can. You can sit there and say, oh,

245
00:11:46,200 --> 00:11:48,560
Speaker 2: the government's corrupt. Oh I feel this way and that way.

246
00:11:48,800 --> 00:11:51,959
Speaker 2: Here's the thing top, which is the thing that was

247
00:11:52,040 --> 00:11:54,199
Speaker 2: used to bail out banks and such and hedge ones,

248
00:11:54,200 --> 00:11:57,200
Speaker 2: and during that whole situation at least made more sense

249
00:11:57,240 --> 00:12:00,160
Speaker 2: because if a bank fails, people lose actual money. This

250
00:12:00,160 --> 00:12:02,680
Speaker 2: stuff fails, what do people lose? What is the service

251
00:12:02,679 --> 00:12:04,600
Speaker 2: that people lose access to? Chat GPT?

252
00:12:05,480 --> 00:12:06,000
Speaker 3: That's it.

253
00:12:06,520 --> 00:12:08,920
Speaker 2: I don't even think that Donald Trump could explain what

254
00:12:09,040 --> 00:12:12,240
Speaker 2: chat GPT is. And even then, he al really has

255
00:12:12,280 --> 00:12:14,200
Speaker 2: people to give him money, like these people don't have

256
00:12:14,240 --> 00:12:16,560
Speaker 2: anything to offer other than just this vague sense of

257
00:12:16,640 --> 00:12:20,240
Speaker 2: AI is important. I just don't buy it. And I

258
00:12:20,280 --> 00:12:23,120
Speaker 2: think that people if you're worried to get hopeful.

259
00:12:23,400 --> 00:12:23,760
Speaker 3: I don't know.

260
00:12:24,160 --> 00:12:27,080
Speaker 2: I'm hopeful that something's going to explode within this terrible

261
00:12:27,160 --> 00:12:31,160
Speaker 2: fucking industry. And I'm kind of hopeful because I think

262
00:12:31,200 --> 00:12:34,439
Speaker 2: it's time to ask a basic question, which is what

263
00:12:34,520 --> 00:12:38,080
Speaker 2: if generative AI just is not profitable? And I think

264
00:12:38,120 --> 00:12:40,160
Speaker 2: this is a question that we have to seriously consider

265
00:12:40,160 --> 00:12:43,760
Speaker 2: at this point because it's ramifications are significant. If I'm honest,

266
00:12:43,800 --> 00:12:45,760
Speaker 2: I think the future of large language models will be

267
00:12:45,840 --> 00:12:50,160
Speaker 2: largely client side on egregiously expensive personal setups or enthusiasts,

268
00:12:50,200 --> 00:12:53,440
Speaker 2: and in a handful of niche enterprise roles. Large language

269
00:12:53,440 --> 00:12:56,360
Speaker 2: models do not scale profitably, and their functionality is not

270
00:12:56,400 --> 00:12:59,040
Speaker 2: significant enough to justify the costs of running them. By

271
00:12:59,080 --> 00:13:01,800
Speaker 2: immediately applying old economics the idea that you would pay

272
00:13:01,840 --> 00:13:05,160
Speaker 2: a monthly fee to have relatively unlimited access, companies like

273
00:13:05,200 --> 00:13:07,840
Speaker 2: open ai and Anthropic immediately trained users to use their

274
00:13:07,880 --> 00:13:10,280
Speaker 2: products in a way that was antithetical to their costs.

275
00:13:10,840 --> 00:13:13,360
Speaker 2: Then again, had these models been served in that way,

276
00:13:13,720 --> 00:13:15,560
Speaker 2: had they been mindful of the cost and charged what

277
00:13:15,600 --> 00:13:17,840
Speaker 2: they actually were, there would likely have been no way

278
00:13:17,880 --> 00:13:20,240
Speaker 2: to get this far. If open ai is making billions

279
00:13:20,280 --> 00:13:22,600
Speaker 2: of dollars a month, or a billion dollars a month,

280
00:13:22,679 --> 00:13:25,040
Speaker 2: or whatever the fuck it is, it's possibly losing that

281
00:13:25,160 --> 00:13:27,680
Speaker 2: much or more after revenue, and that's the money it

282
00:13:27,679 --> 00:13:29,640
Speaker 2: can get selling the product in a form that can

283
00:13:29,679 --> 00:13:33,080
Speaker 2: never turn profitable itself. If open ai charged in line

284
00:13:33,080 --> 00:13:35,960
Speaker 2: with its actual costs, would even be able to justify

285
00:13:36,000 --> 00:13:38,720
Speaker 2: a free version of chat GPT outside of a few

286
00:13:38,760 --> 00:13:41,800
Speaker 2: requests a month. The revenue you see today is what

287
00:13:41,800 --> 00:13:44,160
Speaker 2: people are willing to pay for a product that loses money,

288
00:13:44,280 --> 00:13:46,400
Speaker 2: and I cannot imagine they would pay as much as

289
00:13:46,400 --> 00:13:49,520
Speaker 2: companies in question charge their costs. If I'm wrong, Curse

290
00:13:49,640 --> 00:13:51,760
Speaker 2: will be just fine. And that's assuming that Cursor's current

291
00:13:51,800 --> 00:13:54,520
Speaker 2: hobbled form is even profitable, which it has not said

292
00:13:54,559 --> 00:13:57,280
Speaker 2: it is. So you've got an entire industry of companies

293
00:13:57,320 --> 00:13:59,199
Speaker 2: that struggle to do anything other than lose a lot

294
00:13:59,200 --> 00:14:02,400
Speaker 2: of money. Great, and now we have a massive expansion

295
00:14:02,440 --> 00:14:04,760
Speaker 2: of data centers the likes of which we've never seen

296
00:14:04,960 --> 00:14:07,360
Speaker 2: or to capture demand for a product that nobody makes

297
00:14:07,440 --> 00:14:10,480
Speaker 2: much money selling. This naturally leads to an important question,

298
00:14:10,920 --> 00:14:13,960
Speaker 2: how do these people building data centers actually make money.

299
00:14:24,640 --> 00:14:26,760
Speaker 2: At the start of August, the Wall Street Journal published

300
00:14:26,760 --> 00:14:28,360
Speaker 2: one of the most worrying facts I've seen in the

301
00:14:28,440 --> 00:14:31,800
Speaker 2: last two years. The Journal quoted investor Paul Kodrowski is

302
00:14:31,840 --> 00:14:34,520
Speaker 2: saying that, and I quote as a percentage of gross

303
00:14:34,520 --> 00:14:37,840
Speaker 2: domestic products, spending on AI infrastructure has already exceeded spending

304
00:14:37,840 --> 00:14:40,560
Speaker 2: on telecom and internet infrastructure from the dot com boom,

305
00:14:40,600 --> 00:14:44,720
Speaker 2: and it's still growing. Kudrowski described this insane capex spending

306
00:14:44,760 --> 00:14:48,160
Speaker 2: as a sort of private sector stimulus program. The Journal

307
00:14:48,200 --> 00:14:51,560
Speaker 2: also quotes analyst Neil Dhuta as saying that CAPEX spending

308
00:14:51,600 --> 00:14:54,280
Speaker 2: for AI contributed more to growth in the US economy

309
00:14:54,320 --> 00:14:57,800
Speaker 2: in the past two quarters than all consumers spending. The

310
00:14:57,880 --> 00:15:00,440
Speaker 2: massive build out of data centers and the associate sheated

311
00:15:00,440 --> 00:15:03,160
Speaker 2: physical gear like chips, service some raw materials to building

312
00:15:03,160 --> 00:15:06,240
Speaker 2: them has become a massive dominant economic force, building capacity

313
00:15:06,280 --> 00:15:07,440
Speaker 2: for an industry that is yet.

314
00:15:07,280 --> 00:15:08,720
Speaker 3: To prove it can make real money.

315
00:15:09,200 --> 00:15:11,920
Speaker 2: And no Microsoft talking about it's a zero revenue it's

316
00:15:12,000 --> 00:15:13,880
Speaker 2: last quarterly earnings for the first time is not the

317
00:15:13,920 --> 00:15:17,200
Speaker 2: same thing as it stopped explicitly stating its AI revenue

318
00:15:17,240 --> 00:15:19,760
Speaker 2: in January when it was thirteen billion dollars annualized, so

319
00:15:19,800 --> 00:15:23,440
Speaker 2: about a billion a month. And by the way, that's revenue,

320
00:15:23,480 --> 00:15:27,280
Speaker 2: not profit. Just to remind you, I need to every

321
00:15:27,320 --> 00:15:31,160
Speaker 2: time remind anyone who ever mentions AI revenue that that's

322
00:15:31,280 --> 00:15:36,080
Speaker 2: not profit anyway. Aikapex allegedly, though I've heard some questions

323
00:15:36,120 --> 00:15:38,480
Speaker 2: about this figure, accounts to one point two percent of

324
00:15:38,480 --> 00:15:40,480
Speaker 2: the US GDP in the first half of the year,

325
00:15:40,640 --> 00:15:42,720
Speaker 2: and it accounted for more than half of the to

326
00:15:42,840 --> 00:15:45,400
Speaker 2: quote the Wall Street Journal, already sluggish one point two

327
00:15:45,440 --> 00:15:47,920
Speaker 2: percent growth of the US economy in the first half

328
00:15:47,960 --> 00:15:50,720
Speaker 2: of the year. Another Wall Street Journal piece published a

329
00:15:50,760 --> 00:15:53,400
Speaker 2: few days later discussed how data center development is souring

330
00:15:53,440 --> 00:15:55,800
Speaker 2: the free cash flow for big tech, turning them from

331
00:15:55,800 --> 00:15:58,440
Speaker 2: the kind of asset light businesses that the markets love

332
00:15:58,600 --> 00:16:01,520
Speaker 2: into entities burdened by physics, real estate and their associated

333
00:16:01,520 --> 00:16:05,040
Speaker 2: costs and I quote. For years, investors loved those companies

334
00:16:05,080 --> 00:16:07,280
Speaker 2: because they were asset lighting. They earned their profits or

335
00:16:07,320 --> 00:16:10,800
Speaker 2: on intangible assets such as intellectual property, software, and digital

336
00:16:10,800 --> 00:16:13,640
Speaker 2: platforms with network effects. This can be seen as a

337
00:16:13,640 --> 00:16:16,320
Speaker 2: metric called free cash flow, roughly defined as cash flow

338
00:16:16,360 --> 00:16:20,720
Speaker 2: from operations minus capital expenditures. This is arguably the purest

339
00:16:20,760 --> 00:16:25,200
Speaker 2: measure of business's underlying cash generating potential. From twenty sixteen

340
00:16:25,240 --> 00:16:27,800
Speaker 2: through twenty twenty three, free cash flow and net earnings

341
00:16:27,800 --> 00:16:32,280
Speaker 2: of Alphabet, Amazon, Meta, and Microsoft grew roughly in tandem,

342
00:16:32,400 --> 00:16:34,760
Speaker 2: but since twenty twenty three the two have diverged. The

343
00:16:34,800 --> 00:16:37,040
Speaker 2: four companies can by net income is up seventy three

344
00:16:37,080 --> 00:16:39,400
Speaker 2: percent to ninety one billion in the second quarter from

345
00:16:39,400 --> 00:16:42,080
Speaker 2: two years earlier, while free cash flow is down thirty

346
00:16:42,080 --> 00:16:44,880
Speaker 2: percent to forty billion. According to fact Sheet Data, Apple

347
00:16:45,000 --> 00:16:48,920
Speaker 2: a relative pika on capital spending has also seen free

348
00:16:48,960 --> 00:16:52,560
Speaker 2: cash flow lag behind. These numbers are all very scary,

349
00:16:52,600 --> 00:16:54,680
Speaker 2: and I mean that sincerely, but they also failed to

350
00:16:54,720 --> 00:16:58,600
Speaker 2: explain why how much was actually spent on AI capex

351
00:16:58,640 --> 00:17:00,840
Speaker 2: in the US. One would think two different articles on

352
00:17:00,880 --> 00:17:03,560
Speaker 2: the subject would include that number versus a single quarter's worth.

353
00:17:03,720 --> 00:17:07,000
Speaker 2: But from my estimates, I expect capital expenditures from the

354
00:17:07,080 --> 00:17:09,600
Speaker 2: Magnificent seven alone to crest two hundred billion dollars in

355
00:17:09,640 --> 00:17:12,440
Speaker 2: the first half of twenty twenty five, with Axios estimating

356
00:17:12,480 --> 00:17:15,879
Speaker 2: they'd spend around four hundred billion total this year. Most

357
00:17:15,960 --> 00:17:18,160
Speaker 2: articles are drafting off of a blog from Paul Kodrowski,

358
00:17:18,200 --> 00:17:20,560
Speaker 2: who estimates that total AI capex would be somewhere in

359
00:17:20,560 --> 00:17:22,560
Speaker 2: the region of five hundred and twenty billion dollars for

360
00:17:22,600 --> 00:17:25,600
Speaker 2: the total in the year, which felt conservative for me,

361
00:17:25,680 --> 00:17:28,760
Speaker 2: so I did the smart thing and asked him. Kudrowski

362
00:17:28,880 --> 00:17:31,000
Speaker 2: noted that these numbers focus entirely on the four big

363
00:17:31,040 --> 00:17:34,119
Speaker 2: spenders Microsoft, Google, Meta, and Amazon, and his own estimated

364
00:17:34,160 --> 00:17:36,520
Speaker 2: three hundred and twelve billion dollar capex and one point

365
00:17:36,560 --> 00:17:38,920
Speaker 2: two percent number came from the assumption that the US

366
00:17:39,000 --> 00:17:41,240
Speaker 2: GDP in twenty twenty five will be around twenty eight

367
00:17:41,240 --> 00:17:44,480
Speaker 2: trillion dollars, which is lower than other forecasts which put

368
00:17:44,520 --> 00:17:48,520
Speaker 2: it around thirty trillion. Even Kujoski, in his own words,

369
00:17:48,560 --> 00:17:50,760
Speaker 2: was trying to be conservative using public data and then

370
00:17:50,760 --> 00:17:54,040
Speaker 2: building his analysis from there. I personally believe his estimate

371
00:17:54,080 --> 00:17:56,040
Speaker 2: is too conservative because it does in factor in the

372
00:17:56,080 --> 00:17:58,840
Speaker 2: capital expenditures Maracle, which along with Crusoe is building the

373
00:17:58,960 --> 00:18:02,200
Speaker 2: vast abilenei Exis data center for Open AI or any

374
00:18:02,240 --> 00:18:05,200
Speaker 2: other private data center developers thinking cash into ar capex.

375
00:18:05,600 --> 00:18:08,280
Speaker 2: When I asked him to elaborate, he estimated that AI

376
00:18:08,400 --> 00:18:12,359
Speaker 2: spending all in was around half of three percent Q

377
00:18:12,520 --> 00:18:15,640
Speaker 2: two real GDP growth. He added that it could be

378
00:18:15,880 --> 00:18:19,919
Speaker 2: half of US GDP four year GDP growth if.

379
00:18:19,760 --> 00:18:22,200
Speaker 3: Certain conditions are out. That's so cool.

380
00:18:22,720 --> 00:18:25,120
Speaker 2: Half of the US economy's growth came from building data

381
00:18:25,160 --> 00:18:27,399
Speaker 2: centers for generative AI, which is the combined revenue of

382
00:18:27,400 --> 00:18:29,600
Speaker 2: a little more than the fucking smartwatch industry did in

383
00:18:29,640 --> 00:18:32,680
Speaker 2: twenty twenty four. Another troubling point is that big tech

384
00:18:32,720 --> 00:18:34,879
Speaker 2: doesn't just buy data centers and uses them, but in

385
00:18:34,880 --> 00:18:37,400
Speaker 2: many cases, pace for a construction company to build them,

386
00:18:37,560 --> 00:18:39,679
Speaker 2: fills them full of GPUs, then leases them from a

387
00:18:39,720 --> 00:18:42,679
Speaker 2: company that runs them, meaning that they themselves don't have

388
00:18:42,720 --> 00:18:45,239
Speaker 2: to personally staff up and maintain them. This creates an

389
00:18:45,240 --> 00:18:47,679
Speaker 2: economic boom for construction companies in the short term, as

390
00:18:47,720 --> 00:18:51,360
Speaker 2: well as lucrative contracts for ongoing support as long as

391
00:18:51,400 --> 00:18:54,560
Speaker 2: the company in question still wants them. While Microsoft or

392
00:18:54,560 --> 00:18:57,399
Speaker 2: Amazon might use a data center and indeed act as

393
00:18:57,440 --> 00:18:59,800
Speaker 2: if they own it, ultimately somebody else is holding the

394
00:18:59,800 --> 00:19:03,719
Speaker 2: bag and the ultimate responsibility for said data center. One

395
00:19:03,800 --> 00:19:06,480
Speaker 2: such company is QTS, the data center developer that leases

396
00:19:06,480 --> 00:19:08,560
Speaker 2: both to both Amazon and Meta according to The New

397
00:19:08,640 --> 00:19:11,560
Speaker 2: York Times, which was acquired by Blackstone in twenty twenty

398
00:19:11,600 --> 00:19:14,439
Speaker 2: one for ten billion dollars. Since then, Blackstone has used

399
00:19:14,480 --> 00:19:18,040
Speaker 2: commercially mortgage backed securities I know it's so great to

400
00:19:18,119 --> 00:19:21,320
Speaker 2: raise over eight point seven billion dollars since since then

401
00:19:21,359 --> 00:19:24,240
Speaker 2: to sink into qts's expansion, and as of mid July

402
00:19:24,359 --> 00:19:27,080
Speaker 2: said it'd be investing twenty five billion dollars in AI

403
00:19:27,160 --> 00:19:30,600
Speaker 2: data centers and energy. Blackstone, according to The New York Times,

404
00:19:30,600 --> 00:19:33,320
Speaker 2: see strong demand from tech companies who apparently are willing

405
00:19:33,359 --> 00:19:36,080
Speaker 2: to sign what they describe as airtight leases for fifteen

406
00:19:36,080 --> 00:19:38,640
Speaker 2: to twenty years to rent out data center space. As

407
00:19:38,640 --> 00:19:42,080
Speaker 2: a reminder that Microsoft has over one thousand lawyers in

408
00:19:42,119 --> 00:19:45,439
Speaker 2: the company anyway. Yet the Times also names another problem,

409
00:19:45,480 --> 00:19:48,560
Speaker 2: the unanswered question of how these private equity firms actually

410
00:19:48,600 --> 00:19:52,480
Speaker 2: exit these situations. Blackstone, KKR and other asset management firms

411
00:19:52,480 --> 00:19:55,119
Speaker 2: do not buy companies with the intention of siphoning off revenue,

412
00:19:55,160 --> 00:19:57,560
Speaker 2: but to pump them and sell them to another company

413
00:19:57,680 --> 00:20:00,280
Speaker 2: or dump them onto the public markets. Of course, much

414
00:20:00,320 --> 00:20:03,000
Speaker 2: like AI startups, it isn't obvious who would buy qts,

415
00:20:03,000 --> 00:20:04,600
Speaker 2: and what I imagine would be a twenty five or

416
00:20:04,640 --> 00:20:08,000
Speaker 2: thirty billion dollar valuation mean that black Stone would, as

417
00:20:08,000 --> 00:20:11,040
Speaker 2: I've mentioned, have to take them public. Similarly, KKR is

418
00:20:11,040 --> 00:20:14,359
Speaker 2: supposed fifteen billion dollar partnership with investment firm Energy Capital

419
00:20:14,400 --> 00:20:17,240
Speaker 2: Partners to build data centers and their associated utilities does

420
00:20:17,240 --> 00:20:19,280
Speaker 2: not appear to have much of an exit plan either.

421
00:20:20,040 --> 00:20:23,040
Speaker 2: And let's not forget Oracle Open AI and Cruso's abominable

422
00:20:23,080 --> 00:20:25,800
Speaker 2: mess in Abilene, Texas, where Oracle is paying for forty

423
00:20:25,800 --> 00:20:29,359
Speaker 2: billion dollars worth of GPUs and Crusoe is spending fifteen

424
00:20:29,359 --> 00:20:32,159
Speaker 2: dollars billion dollars raised from Blue Man, Owl Capital and

425
00:20:32,200 --> 00:20:35,240
Speaker 2: primary digital infrastructure. Though I think JP Morgan is involved

426
00:20:35,240 --> 00:20:37,719
Speaker 2: too to build data centers for open Ai, a company

427
00:20:37,760 --> 00:20:39,359
Speaker 2: that loses billion dollars.

428
00:20:39,400 --> 00:20:41,320
Speaker 3: If yeah, why would they do that?

429
00:20:41,520 --> 00:20:44,479
Speaker 2: Right? Why why would they do that? Well, it's simple,

430
00:20:44,520 --> 00:20:47,840
Speaker 2: you roube, you moron, you buffoon. It's so that open

431
00:20:47,840 --> 00:20:50,439
Speaker 2: Ai can allegedly, starting in twenty twenty eight, pay Oracle

432
00:20:50,520 --> 00:20:53,320
Speaker 2: thirty billion dollars a year for compute. And I am

433
00:20:53,359 --> 00:20:57,280
Speaker 2: being fully serious to be clear, open Ai, by my estimates,

434
00:20:57,280 --> 00:21:00,000
Speaker 2: has made only five point twenty six billion dollars this year,

435
00:21:00,480 --> 00:21:03,200
Speaker 2: and we'll have trouble meeting their twelve point seven billion

436
00:21:03,240 --> 00:21:05,800
Speaker 2: dollar revenue projection for twenty twenty five and will likely

437
00:21:05,840 --> 00:21:09,240
Speaker 2: lose more than ten billion dollars doing so. Oracle will,

438
00:21:09,280 --> 00:21:12,560
Speaker 2: according to the information, oh Crusoe one billion dollars in

439
00:21:12,600 --> 00:21:15,399
Speaker 2: payments across a fifteen year span of release. How does

440
00:21:15,440 --> 00:21:17,960
Speaker 2: Crusoe afford to pay back it's fifteen billion dollars in lungs?

441
00:21:18,160 --> 00:21:18,560
Speaker 3: Beats me.

442
00:21:19,080 --> 00:21:21,560
Speaker 2: The information says that Crusoe is raising a billion dollars

443
00:21:21,600 --> 00:21:24,199
Speaker 2: to take on the cloud giants by earning construction management

444
00:21:24,240 --> 00:21:26,320
Speaker 2: fees and rent and you can sell it's stake in

445
00:21:26,359 --> 00:21:29,480
Speaker 2: the project upon reaching certain completion milestones, while also building

446
00:21:29,520 --> 00:21:32,520
Speaker 2: its own Ai Compute, making the assumption that the demand

447
00:21:32,560 --> 00:21:35,960
Speaker 2: is there outside of hyperscalers. Then there's corp Eve, my

448
00:21:36,080 --> 00:21:38,960
Speaker 2: least favor company in the world. As I discussed a

449
00:21:38,960 --> 00:21:41,880
Speaker 2: few months ago, core Wave is burned by obscene debt

450
00:21:41,920 --> 00:21:44,439
Speaker 2: and a horrifying cash burn and has seen its stock spikee

451
00:21:44,600 --> 00:21:46,159
Speaker 2: to a high of one hundred and eighty three on

452
00:21:46,280 --> 00:21:48,800
Speaker 2: June twentieth, twenty twenty five, which has led to an

453
00:21:48,800 --> 00:21:51,520
Speaker 2: all stock attempt to acquire develop a core scientific for

454
00:21:51,640 --> 00:21:54,880
Speaker 2: nine billion dollars. And they're doing that with all stock,

455
00:21:54,960 --> 00:21:57,000
Speaker 2: and now that's falling apart because core Wave, as my

456
00:21:57,080 --> 00:21:58,680
Speaker 2: last check, and who knows where it will be by

457
00:21:58,680 --> 00:22:02,080
Speaker 2: the time this episode runs, is below one hundred bucks.

458
00:22:02,080 --> 00:22:04,520
Speaker 2: Because the IPO lockiles has gone.

459
00:22:05,080 --> 00:22:05,800
Speaker 3: All the people that.

460
00:22:05,800 --> 00:22:08,280
Speaker 2: Invested seem to be getting out. We'll see where they

461
00:22:08,320 --> 00:22:11,280
Speaker 2: go from here. Cor we've by the way, has since

462
00:22:11,320 --> 00:22:14,080
Speaker 2: gone public and had to borrow billions of dollars to

463
00:22:14,119 --> 00:22:17,399
Speaker 2: fund their obscene capital expenditures. And they are also by

464
00:22:17,440 --> 00:22:19,399
Speaker 2: the way, going to have to handle in October at

465
00:22:19,400 --> 00:22:22,240
Speaker 2: the beginning of their DDTL two point zero, their biggest loan,

466
00:22:22,840 --> 00:22:25,080
Speaker 2: and that's also when open Ai is allegedly going to

467
00:22:25,119 --> 00:22:29,040
Speaker 2: start paying them for their compute and by the way,

468
00:22:28,840 --> 00:22:32,080
Speaker 2: they lose hundreds of millions. They lose so much money

469
00:22:32,080 --> 00:22:36,560
Speaker 2: every quarter. They have no part to profitability. Ah, every

470
00:22:36,600 --> 00:22:39,440
Speaker 2: time I think about core Weave feel crazy and to

471
00:22:39,560 --> 00:22:41,360
Speaker 2: be honest, I'm gonna be honest when to put out

472
00:22:41,359 --> 00:22:44,400
Speaker 2: that blog, and people sent me some very unfair emails

473
00:22:44,560 --> 00:22:47,040
Speaker 2: about that. They said, there's no way that corl We've

474
00:22:47,040 --> 00:22:49,440
Speaker 2: can run out of money. If Corwe've runs out of money,

475
00:22:49,440 --> 00:22:51,240
Speaker 2: by the way, and they collapse, which I think is

476
00:22:51,320 --> 00:22:54,520
Speaker 2: very possible. I'm taking some fucking names on this one.

477
00:22:54,520 --> 00:22:56,840
Speaker 2: A lot of people very unfair to miss. The ztrum

478
00:22:57,240 --> 00:22:59,800
Speaker 2: corwave I had is also pretty much reliant on Microsoft

479
00:22:59,800 --> 00:23:02,760
Speaker 2: as one of its primary customers. While this relationship has

480
00:23:02,760 --> 00:23:05,840
Speaker 2: been fairly smooth so far as far as we know,

481
00:23:06,240 --> 00:23:08,640
Speaker 2: this dependence also presents the next eistential threat to core

482
00:23:08,680 --> 00:23:10,040
Speaker 2: we even is part of the reason why I'm so

483
00:23:10,160 --> 00:23:14,119
Speaker 2: pessimistic about their survival. Microsoft has its own infrastructure and

484
00:23:14,119 --> 00:23:16,520
Speaker 2: has every incentive to cut out middlemen when it's able

485
00:23:16,560 --> 00:23:19,440
Speaker 2: to meet supply with the demand itsself owns or lease

486
00:23:19,480 --> 00:23:22,760
Speaker 2: it's rather than subcontracts out simply because middlemen add costs

487
00:23:22,760 --> 00:23:23,720
Speaker 2: and shrink margins.

488
00:23:23,920 --> 00:23:25,480
Speaker 3: If Microsoft walks, what's left?

489
00:23:25,600 --> 00:23:29,640
Speaker 2: How does it service its its ongoing obligations and the debt.

490
00:23:29,960 --> 00:23:32,200
Speaker 2: In all of these cases, data center developers seem to

491
00:23:32,240 --> 00:23:35,040
Speaker 2: have very few options as to making actual money. We

492
00:23:35,119 --> 00:23:37,760
Speaker 2: have companies spending billions of dollars to vastly expend their

493
00:23:37,840 --> 00:23:40,199
Speaker 2: data center footprint, but very little evidence that doing so

494
00:23:40,320 --> 00:23:43,280
Speaker 2: results in revenue, let alone some sort of payoff event.

495
00:23:43,600 --> 00:23:46,560
Speaker 2: And similarly, the actual capital expenditures they're making are much

496
00:23:46,560 --> 00:23:50,159
Speaker 2: smaller than those of big tech. Digital Realty Trust, one

497
00:23:50,200 --> 00:23:52,760
Speaker 2: of the largest developers with over three hundred data centers

498
00:23:52,800 --> 00:23:55,280
Speaker 2: worldwide and five point five billion dollars in revenue in

499
00:23:55,280 --> 00:23:58,080
Speaker 2: twenty twenty four, only spend three point five billion dollars

500
00:23:58,080 --> 00:24:00,960
Speaker 2: in Capex last quarter, and Equinix eight point seven billion

501
00:24:00,960 --> 00:24:03,159
Speaker 2: dollars of revenue in twenty twenty four total, which has

502
00:24:03,200 --> 00:24:06,040
Speaker 2: two hundred and seventy data centers, but Capex at three

503
00:24:06,040 --> 00:24:09,080
Speaker 2: point five billion dollars for a quarter. Two NTT Global

504
00:24:09,160 --> 00:24:11,199
Speaker 2: Data Centers, which has over one hundred and sixty of

505
00:24:11,240 --> 00:24:14,080
Speaker 2: the fuckers, has dedicated ten billion dollars in Capex through

506
00:24:14,119 --> 00:24:16,040
Speaker 2: twenty twenty seven to build out more of them.

507
00:24:16,520 --> 00:24:17,280
Speaker 3: Yeah, in many of.

508
00:24:17,240 --> 00:24:19,520
Speaker 2: These cases, it's because these companies are, to quote a

509
00:24:19,560 --> 00:24:22,960
Speaker 2: source of mind, functionally obsolete for this cycle because legacy

510
00:24:23,000 --> 00:24:25,280
Speaker 2: data centers are not plug and play ready for GPUs

511
00:24:25,320 --> 00:24:28,480
Speaker 2: to slot into. And remember GPUs and the way they

512
00:24:28,480 --> 00:24:31,479
Speaker 2: work with GENERATIVEAI is just atypical. They burn them, they

513
00:24:31,560 --> 00:24:34,760
Speaker 2: run them hot constantly, They're running full fucking till the

514
00:24:34,800 --> 00:24:38,000
Speaker 2: whole fucking time. It's absolutely ridiculous, leading to a bunch

515
00:24:38,000 --> 00:24:41,600
Speaker 2: of replacements. But it just to be specific, requires a

516
00:24:41,640 --> 00:24:45,920
Speaker 2: bunch of specialized cooling, much more energy. And yeah, there's

517
00:24:46,040 --> 00:24:48,080
Speaker 2: just it's not a thing where you just take out

518
00:24:48,119 --> 00:24:50,960
Speaker 2: one computer and you put in another. And also any

519
00:24:51,000 --> 00:24:53,239
Speaker 2: investment in CAPEX by these companies would have to be

520
00:24:53,480 --> 00:24:56,879
Speaker 2: for both GPUs and said retrofitting. By the way, massive

521
00:24:56,880 --> 00:24:59,600
Speaker 2: amounts of money. And this means that the money flowing

522
00:24:59,640 --> 00:25:02,400
Speaker 2: into a ID data centers, it's predominantly going to neo

523
00:25:02,440 --> 00:25:05,360
Speaker 2: clouds like Corweven Crusoe, and all seems to flow back

524
00:25:05,359 --> 00:25:07,399
Speaker 2: to private equity firms that never thought about where the

525
00:25:07,400 --> 00:25:10,560
Speaker 2: cash out might be. Blackstone led corby seven point five

526
00:25:10,600 --> 00:25:13,040
Speaker 2: billion dollar a line with Magnetal Capital, who, by the way,

527
00:25:13,080 --> 00:25:15,680
Speaker 2: was involved with a bunch of subprime mortgages, and Crusoe

528
00:25:15,720 --> 00:25:18,200
Speaker 2: signed a deal a week ago with infrastructure firm and

529
00:25:18,200 --> 00:25:21,160
Speaker 2: they're also owned by Blackstone called tall Grass, and they're

530
00:25:21,160 --> 00:25:23,920
Speaker 2: building a data center in Wyoming, all of which seems

531
00:25:23,960 --> 00:25:25,760
Speaker 2: very good for Blackstone, unless you think, how does it

532
00:25:25,800 --> 00:25:28,919
Speaker 2: actually make money here? As private equity firms generally not

533
00:25:29,080 --> 00:25:32,080
Speaker 2: like to hold assets longer than five years, and the

534
00:25:32,119 --> 00:25:35,840
Speaker 2: timer is running. Even if it did, it's capital expenditures

535
00:25:35,840 --> 00:25:37,879
Speaker 2: that are dropping the bucket in the grand scheme of things.

536
00:25:38,040 --> 00:25:40,960
Speaker 2: Assuming Crusoe burns as the information suggests, it will as

537
00:25:41,000 --> 00:25:43,760
Speaker 2: much as four billion dollars in twenty twenty five. Corweave

538
00:25:43,800 --> 00:25:46,560
Speaker 2: spends as much as twenty billion dollars. Digital Realty Crust

539
00:25:46,600 --> 00:25:49,679
Speaker 2: spends as much as fourteen billion dollars. Global data centers

540
00:25:49,720 --> 00:25:52,600
Speaker 2: three point three three billion, that's ten billion over three years.

541
00:25:52,760 --> 00:25:56,400
Speaker 2: Equiniz spends fourteen billion. That's about fifty five point three

542
00:25:56,480 --> 00:25:58,880
Speaker 2: three billion dollars in AI CAPEK spent in twenty twenty

543
00:25:58,880 --> 00:26:01,639
Speaker 2: five from the largest developer of data centers in the world.

544
00:26:01,840 --> 00:26:04,119
Speaker 2: For some context, one hundred and two billion dollars were

545
00:26:04,160 --> 00:26:07,439
Speaker 2: spent by Meta Alphabet, Microsoft and Amazon in the last quarter.

546
00:26:09,000 --> 00:26:11,720
Speaker 2: Private equity may face the same problem as AI startups.

547
00:26:11,720 --> 00:26:14,680
Speaker 2: In the end, though, there is no real exit strategy

548
00:26:14,720 --> 00:26:17,800
Speaker 2: for any of these investments. Furthermore, the supposed AI capex

549
00:26:17,840 --> 00:26:20,320
Speaker 2: boom that's driving the US economy is not, as reported,

550
00:26:20,600 --> 00:26:24,439
Speaker 2: driven by the massive interest in driving AI infrastructure up

551
00:26:24,440 --> 00:26:25,720
Speaker 2: for a variety of customers.

552
00:26:25,960 --> 00:26:29,280
Speaker 3: The reality is simple. The majority of AI capex.

553
00:26:28,920 --> 00:26:31,440
Speaker 2: Is from big tech, which is a massive systemic weakness

554
00:26:31,440 --> 00:26:49,280
Speaker 2: for our economy. Now, while some might say that AI

555
00:26:49,359 --> 00:26:51,919
Speaker 2: capex has swallowed the US economy, I think it's more

556
00:26:51,960 --> 00:26:54,280
Speaker 2: appropriate to say that big tech capex has swallowed the

557
00:26:54,400 --> 00:26:57,080
Speaker 2: US economy. I also want to be clear that the economy,

558
00:26:57,119 --> 00:26:59,800
Speaker 2: which is the overall state of the country's production and

559
00:26:59,840 --> 00:27:02,679
Speaker 2: can asumption of stuff and the flow of money between

560
00:27:02,680 --> 00:27:05,280
Speaker 2: participants in said economy, and the markets as in the

561
00:27:05,280 --> 00:27:07,679
Speaker 2: stock market, well, those are two very different things. But

562
00:27:07,720 --> 00:27:10,520
Speaker 2: the calculations from Paul Kodrowski and others have now allowed

563
00:27:10,600 --> 00:27:13,280
Speaker 2: us to see where one might hit the other. You see,

564
00:27:13,320 --> 00:27:16,880
Speaker 2: the markets do not actually represent reality. While Microsoft, Amazon,

565
00:27:16,880 --> 00:27:18,520
Speaker 2: Google and Meta might want you to think there's a

566
00:27:18,600 --> 00:27:20,720
Speaker 2: ton of money in AI, their growth is mostly from

567
00:27:20,720 --> 00:27:23,960
Speaker 2: selling further iterations and contracts of their existing stuff, or

568
00:27:24,000 --> 00:27:26,480
Speaker 2: in the case of Meta further increasing its ad revenue.

569
00:27:26,680 --> 00:27:28,760
Speaker 2: The economy is where things are actually bought and sold,

570
00:27:28,800 --> 00:27:31,280
Speaker 2: representing the economic effects of both the things that happen

571
00:27:31,560 --> 00:27:34,040
Speaker 2: to build out the AI boom or whatever you call it,

572
00:27:34,280 --> 00:27:36,919
Speaker 2: and selling access to the services and the models themselves.

573
00:27:37,560 --> 00:27:39,920
Speaker 2: I recognize this is simplistic, but I'm laying it out

574
00:27:39,920 --> 00:27:43,680
Speaker 2: for a reason. As I discussed in the three part

575
00:27:43,720 --> 00:27:46,080
Speaker 2: Hater's Guide to the AI Bubble in video, is the

576
00:27:46,080 --> 00:27:48,919
Speaker 2: weak point in the US stock market, representing ninety percent

577
00:27:48,920 --> 00:27:51,000
Speaker 2: of the value of the Magnificent Seven, which in turn

578
00:27:51,040 --> 00:27:53,159
Speaker 2: makes up about thirty five percent of the value of

579
00:27:53,160 --> 00:27:56,440
Speaker 2: the U S stock market. The associated Magnificent Seven stocks

580
00:27:56,560 --> 00:27:58,920
Speaker 2: have seen a huge boom through their own growth, which

581
00:27:58,960 --> 00:28:01,800
Speaker 2: has been mistakenly in incorrectly attributed to revenue from AI,

582
00:28:01,840 --> 00:28:04,359
Speaker 2: which I've laid out previously is about forty billion dollars

583
00:28:04,760 --> 00:28:06,680
Speaker 2: of revenue in what in the last year or two,

584
00:28:07,040 --> 00:28:10,080
Speaker 2: maybe a little bit more. Nevertheless, the markets can continue

585
00:28:10,080 --> 00:28:12,040
Speaker 2: to be irrational because all they care about is number

586
00:28:12,080 --> 00:28:14,199
Speaker 2: going up. As the value of a stock is often

587
00:28:14,440 --> 00:28:17,680
Speaker 2: times disconnected from the value of the company itself, instead

588
00:28:18,080 --> 00:28:21,600
Speaker 2: associated with its kind of propensity for growth. It's fishy,

589
00:28:21,640 --> 00:28:24,520
Speaker 2: it sucks, and it's going to end badly. GDP and

590
00:28:24,560 --> 00:28:26,560
Speaker 2: other measurements of the economy aren't really something you can

591
00:28:26,640 --> 00:28:29,280
Speaker 2: fudge quite as easily, nor can you say a bunch

592
00:28:29,320 --> 00:28:31,320
Speaker 2: of fancy words to make people feel better in the

593
00:28:31,320 --> 00:28:33,160
Speaker 2: event that growth stores are the clients.

594
00:28:33,480 --> 00:28:36,320
Speaker 3: This leads me to my principal worry that AI.

595
00:28:36,119 --> 00:28:39,120
Speaker 2: Capex is actually a term for the expenditures of four companies,

596
00:28:39,440 --> 00:28:43,360
Speaker 2: namely Microsoft, Amazon, Google, and Meta, with Nvidia's GPU sales

597
00:28:43,360 --> 00:28:46,480
Speaker 2: being part of that capex too. What we include you

598
00:28:46,640 --> 00:28:49,560
Speaker 2: can if you want, include others like Oracle, XAI and

599
00:28:49,640 --> 00:28:53,400
Speaker 2: various neoclouds like Corweven, Crusoe, or who. According to DA

600
00:28:53,520 --> 00:28:56,280
Speaker 2: Davidson's gil Luria will account for about ten percent of

601
00:28:56,320 --> 00:28:59,400
Speaker 2: GPU sales from Nvideo in twenty twenty five. The reality

602
00:28:59,480 --> 00:29:02,600
Speaker 2: is that whatever our economic forces being driven by AI investment,

603
00:29:02,840 --> 00:29:05,840
Speaker 2: it's really just four companies building and leasing data centers

604
00:29:05,880 --> 00:29:08,440
Speaker 2: to burn on generatorvai a product that makes it relatively

605
00:29:08,480 --> 00:29:11,080
Speaker 2: small amount of money before losing a great deal more. Also,

606
00:29:11,120 --> 00:29:14,360
Speaker 2: want to be clear, Anthropic runs on Amazon and Google,

607
00:29:15,000 --> 00:29:18,400
Speaker 2: Open AI runs on Microsoft. These companies don't really own

608
00:29:18,440 --> 00:29:21,280
Speaker 2: their own infrastructure. That's what Open AI is allegedly building

609
00:29:21,320 --> 00:29:25,160
Speaker 2: with Oracle. Now forty two percent of Nvidia's revenue comes

610
00:29:25,160 --> 00:29:28,040
Speaker 2: from Magnificent seven per Laura Brand at Yahoo Finance, Big

611
00:29:28,080 --> 00:29:29,840
Speaker 2: up to Laura one of the best out there, which

612
00:29:29,920 --> 00:29:32,320
Speaker 2: naturally means that big tech is the lynchpin of investment

613
00:29:32,360 --> 00:29:35,320
Speaker 2: in data centers. I'll put it far more simply. If

614
00:29:35,360 --> 00:29:37,920
Speaker 2: AI capex represents such a large part of our GDP

615
00:29:38,040 --> 00:29:40,960
Speaker 2: and economic growth, our economy does on some level rest

616
00:29:40,960 --> 00:29:43,440
Speaker 2: on the back of Microsoft, Google, Meta, and Amazon and

617
00:29:43,480 --> 00:29:46,920
Speaker 2: their continued investment in AI. What should worry everybody is

618
00:29:46,920 --> 00:29:49,400
Speaker 2: that Microsoft, which makes up eighteen point nine percent of

619
00:29:49,520 --> 00:29:52,880
Speaker 2: Nvidia's revenue, assigned basically no leases in the last twelve months,

620
00:29:52,880 --> 00:29:55,920
Speaker 2: and it's committed data center construction and lamp purchases are

621
00:29:55,920 --> 00:29:59,560
Speaker 2: down year over year. Also, their data center capex is

622
00:29:59,600 --> 00:30:02,280
Speaker 2: not all to be AI, and all of these data

623
00:30:02,280 --> 00:30:05,840
Speaker 2: centers won't necessarily be used for that. In fact, nothing

624
00:30:05,880 --> 00:30:09,480
Speaker 2: they're doing suggests that they're super into it.

625
00:30:08,480 --> 00:30:10,200
Speaker 3: It's not good.

626
00:30:10,880 --> 00:30:13,000
Speaker 2: And while its CAPEX may not have dipped yet, in

627
00:30:13,040 --> 00:30:15,320
Speaker 2: part because the chip heavy nature of generative AI means

628
00:30:15,360 --> 00:30:19,680
Speaker 2: that CAPEX isn't exclusively dominated by property or construction. It's

629
00:30:19,720 --> 00:30:22,600
Speaker 2: now obvious that if it does, there will be direct

630
00:30:22,640 --> 00:30:24,880
Speaker 2: effects on both the US economy and the stock market,

631
00:30:25,040 --> 00:30:27,240
Speaker 2: as Microsoft is part of what amounts to a stimulus

632
00:30:27,240 --> 00:30:31,000
Speaker 2: packaged propping up America's economic growth. And not to repeat

633
00:30:31,000 --> 00:30:32,680
Speaker 2: the point too much, but Big Tech is yet to

634
00:30:32,680 --> 00:30:35,960
Speaker 2: actually turn anything resembling a profit on these data centers

635
00:30:36,000 --> 00:30:38,400
Speaker 2: and isn't making much revenue at all out of generative AI.

636
00:30:39,120 --> 00:30:41,040
Speaker 2: How exactly does this end?

637
00:30:41,240 --> 00:30:42,080
Speaker 3: What is the plan?

638
00:30:42,240 --> 00:30:44,680
Speaker 2: Here's Big Tech going to spend hundreds of billions of

639
00:30:44,720 --> 00:30:47,640
Speaker 2: dollars a year in CAPEX on generative AI in perpetuity?

640
00:30:47,840 --> 00:30:50,120
Speaker 2: Will they continue to buy more and more in video

641
00:30:50,200 --> 00:30:52,360
Speaker 2: chips as they do so, because it's always got to

642
00:30:52,400 --> 00:30:54,240
Speaker 2: be more. It can't be the same. They must get

643
00:30:54,280 --> 00:30:57,680
Speaker 2: more otherwise Nvidia doesn't grow At some point. Surely these

644
00:30:57,680 --> 00:31:01,000
Speaker 2: companies are built in enough data centers. Surely at some

645
00:31:01,160 --> 00:31:03,040
Speaker 2: point they will run out of space to put these

646
00:31:03,080 --> 00:31:05,640
Speaker 2: GPUs in. Is the plan to by then make so

647
00:31:05,720 --> 00:31:08,360
Speaker 2: much money from MAI that it won't matter? What does

648
00:31:08,480 --> 00:31:10,720
Speaker 2: Nvidia do at that point? How does the US economy

649
00:31:10,760 --> 00:31:12,720
Speaker 2: rebound from the loss of activity that follows?

650
00:31:12,920 --> 00:31:14,960
Speaker 3: I don't fucking know. And it's worrying me.

651
00:31:16,640 --> 00:31:19,400
Speaker 2: As I've said again and again, the genera IFAI bubble

652
00:31:19,520 --> 00:31:22,880
Speaker 2: is and always has been fundamentally irrational and inherently gothic,

653
00:31:22,960 --> 00:31:25,840
Speaker 2: playing in the ruins, patterns, and pathways of previous tech booms,

654
00:31:25,840 --> 00:31:28,840
Speaker 2: despite this one having little or no resemblance to them.

655
00:31:29,520 --> 00:31:31,520
Speaker 2: Though the tech industry loves to talk about building a

656
00:31:31,560 --> 00:31:34,960
Speaker 2: glorious future, its presence is one steeped in rituals of

657
00:31:35,000 --> 00:31:37,560
Speaker 2: death and decay, with the virtues of value creation and

658
00:31:37,560 --> 00:31:40,200
Speaker 2: productivity take a back seat to burning billions of dollars

659
00:31:40,200 --> 00:31:42,520
Speaker 2: and lying to the public again and again and again.

660
00:31:43,000 --> 00:31:45,080
Speaker 2: The way in which the media has participated in these

661
00:31:45,120 --> 00:31:49,200
Speaker 2: lives is disgusting. Venture capital, still drunk off the fumes

662
00:31:49,200 --> 00:31:52,120
Speaker 2: of twenty twenty one, keeps running the same old planebook,

663
00:31:52,600 --> 00:31:54,680
Speaker 2: shove as much money into a company as possible in

664
00:31:54,680 --> 00:31:56,360
Speaker 2: the hopes you can dump it onto an acquirer of

665
00:31:56,360 --> 00:31:58,840
Speaker 2: the public markets, only to get high on their own supply,

666
00:31:59,120 --> 00:32:01,920
Speaker 2: pushing valuations to the point that there's no possible liquidity

667
00:32:01,960 --> 00:32:04,360
Speaker 2: event for the majority of big private AI companies as

668
00:32:04,360 --> 00:32:07,960
Speaker 2: a result of their overstuffed valuations, burdens and business models,

669
00:32:08,040 --> 00:32:11,320
Speaker 2: and a lack of any real intellectual property, and like

670
00:32:11,360 --> 00:32:13,600
Speaker 2: the rest of the AI and bubble, Silicon Valley's only

671
00:32:13,680 --> 00:32:16,880
Speaker 2: liquidity path out of the bubble is big tech itself.

672
00:32:17,320 --> 00:32:20,400
Speaker 2: Without Google, character Dot AI, and WINDSORFS founders would likely

673
00:32:20,440 --> 00:32:22,680
Speaker 2: have been left for dead, and the same ghosts for Inflection,

674
00:32:22,800 --> 00:32:25,680
Speaker 2: And I'd even argue Scale AI, who's fourteen point three

675
00:32:25,720 --> 00:32:29,360
Speaker 2: billion dollar investment doing air quotes there for Meta effectively

676
00:32:29,360 --> 00:32:33,040
Speaker 2: decapitated the company, removing its CEO, Alexander Wang, leaving the

677
00:32:33,040 --> 00:32:36,080
Speaker 2: rest of the company to die playing are fourteen percent

678
00:32:36,080 --> 00:32:38,400
Speaker 2: of its staff on five hundred contractors mere weeks after

679
00:32:38,440 --> 00:32:42,680
Speaker 2: its CEO and investors cashed in. In fact, Generative AI

680
00:32:42,840 --> 00:32:45,680
Speaker 2: is turning out to be a fever dream entirely made up.

681
00:32:45,680 --> 00:32:46,400
Speaker 3: Big tech.

682
00:32:46,800 --> 00:32:48,800
Speaker 2: OpenAI would be dead if it wasn't for the massive

683
00:32:48,840 --> 00:32:51,800
Speaker 2: infrastructure provided by Microsoft at cost in returns for its

684
00:32:51,880 --> 00:32:54,320
Speaker 2: rights to its IP research and the ability to sell

685
00:32:54,360 --> 00:32:56,440
Speaker 2: its models, on top of tens of billions of dollars

686
00:32:56,480 --> 00:32:59,960
Speaker 2: of venture capital thrown into its billion dollar cache. Incinerator

687
00:33:00,600 --> 00:33:03,000
Speaker 2: Anthropic would be dead if both Google and Amazon, the

688
00:33:03,080 --> 00:33:06,240
Speaker 2: latter of which provides most of its infrastructure, hadn't invested

689
00:33:06,280 --> 00:33:08,200
Speaker 2: billions in keeping it alive. So that it can burn

690
00:33:08,280 --> 00:33:10,800
Speaker 2: three billion dollars or more in twenty twenty five while

691
00:33:10,840 --> 00:33:13,720
Speaker 2: fucking over its enterprise customers and rate limiting the rest.

692
00:33:14,560 --> 00:33:17,719
Speaker 2: The generative AI industry is, at its core unnatural. It

693
00:33:17,760 --> 00:33:20,680
Speaker 2: does not make significant revenue compared to its unbelievable costs,

694
00:33:20,720 --> 00:33:24,080
Speaker 2: nor does it have much revenue potential itself. It requires,

695
00:33:24,160 --> 00:33:27,880
Speaker 2: unlike just about every other software revolution, an unbelievable amount

696
00:33:27,920 --> 00:33:30,880
Speaker 2: of physical infrastructure to run because nobody but big tech

697
00:33:30,880 --> 00:33:33,880
Speaker 2: can afford to build the infrastructure necessary, creates very little

698
00:33:33,920 --> 00:33:37,520
Speaker 2: opportunity for competition or efficiency. As the markets are in

699
00:33:37,560 --> 00:33:39,680
Speaker 2: the throes of the growth of or costs rot economy,

700
00:33:39,840 --> 00:33:42,160
Speaker 2: they failed to keep big tech in line, conflating Big

701
00:33:42,200 --> 00:33:45,000
Speaker 2: text ability to grow with growth driven as a result

702
00:33:45,000 --> 00:33:48,720
Speaker 2: of the capital expenditures. Sensible reasonable markets would notice the

703
00:33:48,760 --> 00:33:51,800
Speaker 2: decay of free cash flow or ridiculousness of big text

704
00:33:51,800 --> 00:33:54,680
Speaker 2: capex bonanza, but instead they clap and squeal on oink

705
00:33:54,720 --> 00:33:57,520
Speaker 2: whenever Satching the Della jingles his keys. And I got

706
00:33:57,520 --> 00:33:59,200
Speaker 2: told the other day by someone that Satch in a

707
00:33:59,280 --> 00:34:01,360
Speaker 2: Della is not a is the citiot, that he's secretly

708
00:34:01,400 --> 00:34:03,880
Speaker 2: this big tech guy. And here's here's what I'd like

709
00:34:03,920 --> 00:34:07,040
Speaker 2: to say to them, it's fucking stupid. It's fucking dumb.

710
00:34:07,120 --> 00:34:10,680
Speaker 2: If this guy's actually technical, he's just a craven liar.

711
00:34:11,000 --> 00:34:14,400
Speaker 2: I actually think he's a bozo. Anyway, put in that side,

712
00:34:14,640 --> 00:34:18,560
Speaker 2: what's missing is any real value generation. Again, I tell you,

713
00:34:18,600 --> 00:34:21,120
Speaker 2: put aside any feelings you may have about generative AI

714
00:34:21,160 --> 00:34:24,320
Speaker 2: itself and focus on the actual economic results of this bubble.

715
00:34:24,600 --> 00:34:27,120
Speaker 2: How much revenue is there? Why is there no profit?

716
00:34:27,520 --> 00:34:29,640
Speaker 2: Why are there no exits? Why is big tech, which

717
00:34:29,640 --> 00:34:32,120
Speaker 2: has sunk hundreds of billions of dollars into generative AI,

718
00:34:32,440 --> 00:34:35,200
Speaker 2: not talk about the money they're making from it? Why

719
00:34:35,320 --> 00:34:38,359
Speaker 2: for three straight years have we been asked just wait

720
00:34:38,400 --> 00:34:40,360
Speaker 2: and see and for how long are we going to

721
00:34:40,400 --> 00:34:42,560
Speaker 2: have to wait? And when are we going to see it?

722
00:34:43,239 --> 00:34:46,200
Speaker 2: What's incredible is that the inherently compute intensive nature of

723
00:34:46,280 --> 00:34:50,160
Speaker 2: generative AI basically requires the construction of these facilities without

724
00:34:50,200 --> 00:34:52,680
Speaker 2: actually representing whether they are contributing to the revenues of

725
00:34:52,680 --> 00:34:55,719
Speaker 2: the companies that operate the models, like Anthropic or Open AI,

726
00:34:55,920 --> 00:34:58,560
Speaker 2: or any other business that builds on top of them.

727
00:34:58,880 --> 00:35:01,080
Speaker 2: As the models get more complet and hungary, more data

728
00:35:01,080 --> 00:35:03,800
Speaker 2: centers get built, which Hyperscaler's book as long term revenue,

729
00:35:03,840 --> 00:35:06,920
Speaker 2: even though it's subsidized by Hyperscalers or funded by VC money.

730
00:35:07,160 --> 00:35:10,240
Speaker 2: This in turn stimulates even more CAPEX spending, and without

731
00:35:10,280 --> 00:35:13,920
Speaker 2: having to answer any basic questions about longevity or market

732
00:35:13,960 --> 00:35:16,480
Speaker 2: fit longevity longevity.

733
00:35:16,040 --> 00:35:17,319
Speaker 3: I don't know yet.

734
00:35:17,360 --> 00:35:20,680
Speaker 2: The worst part of this financial farce is that we've

735
00:35:20,760 --> 00:35:23,560
Speaker 2: now got a built in economic breaking point in the

736
00:35:23,600 --> 00:35:27,120
Speaker 2: form of capex from AI. At some point, Kapex has

737
00:35:27,160 --> 00:35:29,239
Speaker 2: the slow if not because the lack of revenues are

738
00:35:29,280 --> 00:35:31,840
Speaker 2: massive costs associated, but because we live in a world

739
00:35:31,880 --> 00:35:35,840
Speaker 2: with finite space. And when said CAPEX slow happens, so

740
00:35:35,960 --> 00:35:38,640
Speaker 2: will the purchase of Nvidio GPUs, which in turn has

741
00:35:38,680 --> 00:35:42,480
Speaker 2: proven by Kadrowski and others slow America's economic growth. And

742
00:35:42,520 --> 00:35:44,480
Speaker 2: that growth is pretty much based on the whims of

743
00:35:44,520 --> 00:35:47,640
Speaker 2: four companies, which is an incredibly risky and scary proposition.

744
00:35:48,080 --> 00:35:50,120
Speaker 2: I haven't even dug into the wealth of private credit

745
00:35:50,120 --> 00:35:52,640
Speaker 2: deals that underpin build outs from private AI neo clouds

746
00:35:52,719 --> 00:35:55,239
Speaker 2: like core Wave, Cruise, So, Nebuis, and Lambda, in part

747
00:35:55,239 --> 00:35:57,440
Speaker 2: because their economic significance is so much smaller than the

748
00:35:57,440 --> 00:36:00,359
Speaker 2: big text ugly meanland the sprawl, and because I don't

749
00:36:00,400 --> 00:36:04,200
Speaker 2: like saying their names very much. To quote Paul Kodrowski,

750
00:36:05,160 --> 00:36:08,239
Speaker 2: we live in a historically anomalous moment. Regardless of what

751
00:36:08,320 --> 00:36:10,759
Speaker 2: one thinks about the merits of AI or explosive data

752
00:36:10,760 --> 00:36:14,000
Speaker 2: center expansion, the scale and pace of capital deployment into

753
00:36:14,000 --> 00:36:18,520
Speaker 2: a rapidly depreciating technology is remarkable. These are not railroads.

754
00:36:18,760 --> 00:36:22,600
Speaker 2: We aren't building century long infrastructure. AI data centers are

755
00:36:22,640 --> 00:36:27,360
Speaker 2: short lived, asset intensive facilities, writing declining cost technology curves,

756
00:36:27,400 --> 00:36:32,080
Speaker 2: requiring frequent hardware replacement to preserve margins, to which I say,

757
00:36:32,120 --> 00:36:35,000
Speaker 2: what margin's for? But nevertheless, you can't bail this out

758
00:36:35,040 --> 00:36:38,600
Speaker 2: because there is nothing to bail out. Microsoft, Meta, Amazon,

759
00:36:38,640 --> 00:36:40,600
Speaker 2: and Google have plenty of money and have proven they

760
00:36:40,600 --> 00:36:43,360
Speaker 2: can spend it in video is already doing everything it

761
00:36:43,400 --> 00:36:46,319
Speaker 2: can to justify people spending more on their GPUs. There's

762
00:36:46,360 --> 00:36:48,719
Speaker 2: little more it can do here other than soak up

763
00:36:48,760 --> 00:36:52,239
Speaker 2: the growth before the party ends. That CAPEX reduction will

764
00:36:52,239 --> 00:36:55,080
Speaker 2: bring with it a reduction in expenditures on those GPUs,

765
00:36:55,320 --> 00:36:57,680
Speaker 2: which will take a chunk out of the US stock market.

766
00:36:57,840 --> 00:37:00,360
Speaker 2: Although the stock market isn't the economy, the two are

767
00:37:00,440 --> 00:37:03,440
Speaker 2: inherently linked, and the popping of the AI bubble will

768
00:37:03,440 --> 00:37:06,879
Speaker 2: have down downstream ramifications, just that the dot com bubble did.

769
00:37:06,880 --> 00:37:10,000
Speaker 2: With the wider economy, expect to see an acceleration in

770
00:37:10,080 --> 00:37:12,600
Speaker 2: layoffs and offshoring in pop driven by a need for

771
00:37:12,640 --> 00:37:15,080
Speaker 2: tech companies to show for the first time in living memory,

772
00:37:15,320 --> 00:37:18,720
Speaker 2: fiscal restraint. For cities where tech is a major sector

773
00:37:18,719 --> 00:37:21,480
Speaker 2: of the economy thinks Seattle and San Francisco, there will

774
00:37:21,480 --> 00:37:23,799
Speaker 2: be knock on effects to those companies and the individuals

775
00:37:23,800 --> 00:37:27,560
Speaker 2: that support the tech sector, like restaurants, construction companies, building apartments,

776
00:37:27,800 --> 00:37:31,040
Speaker 2: uber drivers, and so on. We'll see a drying up

777
00:37:31,040 --> 00:37:33,960
Speaker 2: of VC funding, which is kind of already starting. Pension

778
00:37:34,000 --> 00:37:35,759
Speaker 2: funds will take a hit, which will affect how much

779
00:37:35,800 --> 00:37:37,320
Speaker 2: people have to spend in retirement.

780
00:37:37,760 --> 00:37:39,000
Speaker 3: It'll be grim.

781
00:37:39,280 --> 00:37:41,120
Speaker 2: Worse than that is the fact that these companies will

782
00:37:41,160 --> 00:37:44,000
Speaker 2: be by definition, non performing assets and ones that inflicted

783
00:37:44,040 --> 00:37:47,160
Speaker 2: an opportunity cost that will be impossible to calculate. While

784
00:37:47,160 --> 00:37:49,920
Speaker 2: a house, once built and sold, technically falls into that category,

785
00:37:50,160 --> 00:37:54,360
Speaker 2: doesn't add to any economic productivity. Obviously, people at least

786
00:37:54,400 --> 00:37:57,439
Speaker 2: need somewhere to live. Shelter is an essential component of life.

787
00:37:57,480 --> 00:37:59,400
Speaker 2: You can live without a data center the size of

788
00:37:59,440 --> 00:38:02,839
Speaker 2: Manhattan what would happen if companies like Microsoft and Meta

789
00:38:02,880 --> 00:38:05,440
Speaker 2: instead spent the money on things that actually drug productivity

790
00:38:05,520 --> 00:38:08,640
Speaker 2: or created a valuable competitive business that drove economic activity.

791
00:38:08,680 --> 00:38:11,160
Speaker 2: For example, Hell, even if they just gave everyone a

792
00:38:11,160 --> 00:38:13,440
Speaker 2: ten percent raise, it would likely be better for the

793
00:38:13,480 --> 00:38:15,840
Speaker 2: economy than this if we're factoring in things like consumer

794
00:38:15,880 --> 00:38:21,800
Speaker 2: spending instead. It's just waste, ugly, pointless waste. In summary,

795
00:38:21,840 --> 00:38:24,680
Speaker 2: we're already facing the prospect of a recession, and though

796
00:38:24,719 --> 00:38:28,080
Speaker 2: I'm not an economist, I can imagine being a particularly

797
00:38:28,160 --> 00:38:30,719
Speaker 2: nasty one. Given that the Magnificent seven account for forty

798
00:38:30,760 --> 00:38:33,280
Speaker 2: seven point eight seven percent of the Rustle one thousand

799
00:38:33,400 --> 00:38:36,719
Speaker 2: indexes returns in twenty twenty four. Even if big Tech

800
00:38:36,800 --> 00:38:39,759
Speaker 2: somehow makes this crap profitable, they won't. It's hard to

801
00:38:39,800 --> 00:38:42,560
Speaker 2: imagine that they'll be able to counterbalance any CAPEX reduction

802
00:38:42,640 --> 00:38:44,480
Speaker 2: with revenue, because there doesn't seem to be that much

803
00:38:44,520 --> 00:38:47,440
Speaker 2: revenue in generative AI to begin with. This is what

804
00:38:47,520 --> 00:38:49,799
Speaker 2: happens when you allow the rot economy to run wild,

805
00:38:49,920 --> 00:38:52,200
Speaker 2: building the stock market and tech industry on growth over

806
00:38:52,280 --> 00:38:55,560
Speaker 2: everything else. This is what happens when the tech media

807
00:38:55,600 --> 00:38:58,320
Speaker 2: repeatedly fails to hold the powerful to account catering to

808
00:38:58,360 --> 00:39:01,799
Speaker 2: their narratives and making excuses from are abominable billion dollar

809
00:39:01,880 --> 00:39:05,520
Speaker 2: losses and mediocre, questionably useful products. Well, for all you

810
00:39:05,560 --> 00:39:08,720
Speaker 2: want about the so called agentic era or anyonized revenues

811
00:39:08,719 --> 00:39:10,920
Speaker 2: that make you hot under the collar, I see no

812
00:39:11,040 --> 00:39:13,720
Speaker 2: reason for celebration about an industry with no exit plans

813
00:39:13,719 --> 00:39:15,920
Speaker 2: and needless capital expenditures that appeared to be one of

814
00:39:15,920 --> 00:39:19,040
Speaker 2: the few things keeping the American economy growing for effectively

815
00:39:19,080 --> 00:39:23,680
Speaker 2: no reason. I've been writing about the tech industry's obsession

816
00:39:23,719 --> 00:39:26,319
Speaker 2: with General IVAI for two years, and I've really never

817
00:39:26,360 --> 00:39:29,799
Speaker 2: felt more grim before. This was an economic uncertainty, a

818
00:39:29,800 --> 00:39:33,000
Speaker 2: way that our markets might contract, the big tech might

819
00:39:33,040 --> 00:39:35,440
Speaker 2: take a big haircunt, that a big a bunch of

820
00:39:35,440 --> 00:39:37,520
Speaker 2: money might have been wasted. But otherwise the world could

821
00:39:37,560 --> 00:39:42,359
Speaker 2: keep turning, right, it felt before Kutrowski said that thing,

822
00:39:42,360 --> 00:39:45,000
Speaker 2: and this was all sitting in public as well. I

823
00:39:45,000 --> 00:39:47,200
Speaker 2: don't know how I missed it. It felt until then

824
00:39:47,239 --> 00:39:49,000
Speaker 2: that it was like, Oh, this will really fuck up

825
00:39:49,040 --> 00:39:51,080
Speaker 2: big tech. It will hit their stocks. It'll be bad

826
00:39:51,120 --> 00:39:55,120
Speaker 2: for them now because the economy has slowed so much. Otherwise,

827
00:39:56,000 --> 00:39:58,640
Speaker 2: this data center, Capex and these four companies is holding

828
00:39:58,640 --> 00:40:02,080
Speaker 2: everything up and it won't lost. There's just no way

829
00:40:02,120 --> 00:40:04,680
Speaker 2: it can last forever, and there's not really anything to

830
00:40:04,840 --> 00:40:08,439
Speaker 2: bulk it up or and Microsoft's already pulling back.

831
00:40:08,960 --> 00:40:09,799
Speaker 3: How does this end?

832
00:40:09,840 --> 00:40:11,960
Speaker 2: I keep asking that the original title of this was

833
00:40:12,000 --> 00:40:14,000
Speaker 2: called how does this end? And I really don't know

834
00:40:14,040 --> 00:40:17,520
Speaker 2: how this ends. It feels as if everything is the

835
00:40:17,560 --> 00:40:20,480
Speaker 2: learning for disaster, and I really feel there's nothing that

836
00:40:20,480 --> 00:40:21,440
Speaker 2: can be done to avert in.

837
00:40:22,200 --> 00:40:25,000
Speaker 3: I will be here to tell you what's happening. Is

838
00:40:25,000 --> 00:40:25,600
Speaker 3: it happens?

839
00:40:26,200 --> 00:40:27,880
Speaker 2: And your third part for this week is going to

840
00:40:27,920 --> 00:40:30,360
Speaker 2: be about GPT five, which I'm going to be honest

841
00:40:31,120 --> 00:40:34,360
Speaker 2: doesn't really feel me full of hope or promise either.

842
00:40:42,719 --> 00:40:44,400
Speaker 2: Thank you for listening to Better Offline.

843
00:40:44,520 --> 00:40:46,960
Speaker 1: The editor and composer of the Better Offline theme song

844
00:40:47,040 --> 00:40:49,680
Speaker 1: is Matasowski. You can check out more of his music

845
00:40:49,719 --> 00:40:53,359
Speaker 1: and audio projects at Mattasowski dot com, m A T

846
00:40:53,360 --> 00:40:57,799
Speaker 1: T O S O W s ki dot com. You

847
00:40:57,840 --> 00:41:00,520
Speaker 1: can email me an easy at Better Offline dot or

848
00:41:00,560 --> 00:41:03,080
Speaker 1: visit Better Offline dot com to find more podcast links

849
00:41:03,080 --> 00:41:06,239
Speaker 1: and of course, my newsletter. I also really recommend you

850
00:41:06,239 --> 00:41:08,480
Speaker 1: go to chat dot where'soead dot at to visit the

851
00:41:08,520 --> 00:41:10,560
Speaker 1: Discord and go to our slash.

852
00:41:10,239 --> 00:41:13,400
Speaker 2: Better Offline to check out our reddit. Thank you so

853
00:41:13,480 --> 00:41:14,240
Speaker 2: much for listening.

854
00:41:15,080 --> 00:41:17,560
Speaker 3: Better Offline is a production of cool Zone Media.

855
00:41:17,680 --> 00:41:20,520
Speaker 2: For more from cool Zone Media, visit our website cool

856
00:41:20,600 --> 00:41:23,919
Speaker 2: Zonemedia dot com, or check us out on the iHeartRadio app,

857
00:41:24,000 --> 00:41:26,680
Speaker 2: Apple Podcasts, or wherever you get your podcasts.