1
00:00:00,000 --> 00:00:02,880
These people are doing it because they're too stupid

2
00:00:02,880 --> 00:00:05,040
to realize that what they're in is kind of a scam.

3
00:00:05,190 --> 00:00:07,680
I think that they're scamming in the sense that they're selling something

4
00:00:07,680 --> 00:00:10,500
they don't really understand, but they're not like, I think this will die.

5
00:00:10,740 --> 00:00:12,630
I think that they think this will go forever.

6
00:00:12,690 --> 00:00:16,290
Kind of like crypto and when I, this is the other thing, people

7
00:00:16,290 --> 00:00:18,420
with my work, they're like, oh, you started like a year ago.

8
00:00:19,320 --> 00:00:20,490
I was writing in 2020.

9
00:00:20,580 --> 00:00:22,140
I was on crypto before.

10
00:00:22,140 --> 00:00:22,860
Everyone.

11
00:00:27,780 --> 00:00:29,640
Welcome to Screaming in the Cloud.

12
00:00:29,730 --> 00:00:33,690
I'm Cory Quinn, and I've been looking forward to this episode for a while.

13
00:00:33,990 --> 00:00:38,370
Ed Zitron is the host of Better Offline, the writer of Where's Your

14
00:00:38,370 --> 00:00:42,990
Ed at Newsletter and has often been referred to as an AI skeptic.

15
00:00:43,170 --> 00:00:47,760
I tend to view him as much more of an AI realist, but we'll get into that.

16
00:00:48,030 --> 00:00:49,770
Ed, thank you for joining me.

17
00:00:50,160 --> 00:00:51,060
Thank you for having me.

18
00:00:53,265 --> 00:00:57,585
So you have been very vocal about AI and I, I wanna say your

19
00:00:57,585 --> 00:01:01,155
skepticism of it, but that feels like it's an unfair characterization.

20
00:01:01,155 --> 00:01:03,615
I think you're, you're approaching this from the perspective

21
00:01:03,615 --> 00:01:06,465
of you are trying to sell me something, I'm not going to buy

22
00:01:06,465 --> 00:01:10,155
it without examination, and the rest sort of flows from there.

23
00:01:10,785 --> 00:01:14,505
Yes, and that's the thing right now, and I've said this

24
00:01:14,505 --> 00:01:17,535
a few times in a few interviews, I am being and have been

25
00:01:17,535 --> 00:01:20,835
characterized as a radical because I'm like, Hey, this company

26
00:01:20,835 --> 00:01:23,685
loses billions of dollars and has no path to profitability.

27
00:01:23,685 --> 00:01:28,365
Or, Hey, none of these companies appear to be making much revenue, let alone

28
00:01:28,365 --> 00:01:32,265
profit, but they're spending so much money on this, is that sustainable?

29
00:01:32,265 --> 00:01:33,914
And people like, ed, you are crazy.

30
00:01:34,365 --> 00:01:35,475
You are crazy.

31
00:01:35,625 --> 00:01:39,345
I irrational person believe that these companies will

32
00:01:39,345 --> 00:01:42,045
simply work it out in some way that I cannot describe.

33
00:01:42,645 --> 00:01:43,785
I have blind faith.

34
00:01:43,845 --> 00:01:45,165
You are the radical for not

35
00:01:45,255 --> 00:01:45,645
Right?

36
00:01:45,645 --> 00:01:47,985
Strangely, when I try and pitch people on things with blind

37
00:01:47,985 --> 00:01:50,175
faith in them, they, they tend to ask a whole bunch of

38
00:01:50,175 --> 00:01:52,335
uncomfortable questions that I'm not prepared to answer.

39
00:01:52,335 --> 00:01:54,315
But nowadays with ai, when you ask those

40
00:01:54,315 --> 00:01:56,205
questions, you're treated like you are the problem.

41
00:01:56,295 --> 00:01:56,805
Exactly.

42
00:01:56,805 --> 00:02:01,305
You, it's like a gifted child that you have to, it's what is that, um, Twilight

43
00:02:01,305 --> 00:02:04,485
zone episode with the child that can dream, dream, thinks of something.

44
00:02:04,635 --> 00:02:06,795
It's just everyone wants to coddle AI in case it

45
00:02:06,795 --> 00:02:09,285
hurts them or in case it breaks the narrative.

46
00:02:09,525 --> 00:02:14,475
It's just, it's very frustrating because disagreeing with me is one thing.

47
00:02:14,505 --> 00:02:18,315
I do not have trouble with that unless the person is just like, their

48
00:02:18,315 --> 00:02:23,595
argument is no, because that is most of the response to my work is just nah.

49
00:02:24,315 --> 00:02:24,945
No, it's not.

50
00:02:25,305 --> 00:02:26,955
Well, I'm trying to sell something and if what you're

51
00:02:26,955 --> 00:02:29,445
saying is true, that impedes my ability to do that.

52
00:02:29,445 --> 00:02:30,525
So shut up, please.

53
00:02:30,825 --> 00:02:33,675
Yeah, could you, but even then, it's like it goes beyond that.

54
00:02:33,675 --> 00:02:35,355
'cause you've got tons of journalists as

55
00:02:35,355 --> 00:02:37,305
well who are just like, no, it's not true.

56
00:02:37,305 --> 00:02:38,115
They're gonna work it out.

57
00:02:38,115 --> 00:02:39,375
This is all part of the plan.

58
00:02:39,495 --> 00:02:40,425
What's the plan?

59
00:02:40,425 --> 00:02:41,715
Can you tell me the rest of it?

60
00:02:41,715 --> 00:02:45,885
Because it doesn't seem like a good plan unless the plan is raise as much

61
00:02:45,885 --> 00:02:50,565
money and get as much awareness as possible, which they've succeeded at.

62
00:02:50,595 --> 00:02:51,555
They've not made any money.

63
00:02:51,555 --> 00:02:53,715
In fact, they're not making much money at all.

64
00:02:53,715 --> 00:02:55,155
No one is making any money.

65
00:02:55,245 --> 00:02:59,085
I think there's like two profitable companies within this space.

66
00:02:59,595 --> 00:03:01,065
Maybe, maybe three.

67
00:03:01,740 --> 00:03:06,240
If we go back in time and we look at the standard generative AI chatbot or

68
00:03:06,240 --> 00:03:09,960
coding assistance and you offer it to someone in a vacuum, they would spend

69
00:03:09,960 --> 00:03:13,260
thousands of dollars a month for this based upon the perceived utility.

70
00:03:13,260 --> 00:03:16,800
Not everyone, but enough people would now though that has

71
00:03:16,800 --> 00:03:21,150
been anchored at about $20 a month per user is the perception.

72
00:03:21,270 --> 00:03:24,540
So I don't see a path to suddenly multiplying each price by

73
00:03:24,540 --> 00:03:28,200
a hundred to wind up generating some sort of economic return.

74
00:03:28,350 --> 00:03:31,410
They have anchored their pricing at a point where I don't see a

75
00:03:31,410 --> 00:03:35,190
path to significant widespread industry defining profitability.

76
00:03:35,400 --> 00:03:38,370
I'm gonna be honest, I don't think they would've paid thousands of dollars.

77
00:03:38,370 --> 00:03:40,440
You had code auto complete things.

78
00:03:40,440 --> 00:03:45,000
I talked to a client about this before COVID, like I was remember in a WeWork,

79
00:03:46,290 --> 00:03:48,570
and they were telling me, yeah, it was kind like auto complete for coding.

80
00:03:48,570 --> 00:03:50,010
I'm like, oh, seems useful.

81
00:03:50,010 --> 00:03:52,380
I bet people would like that They, I think

82
00:03:52,380 --> 00:03:54,150
they sold part of it to Microsoft or something.

83
00:03:54,150 --> 00:03:54,930
I, I forget.

84
00:03:55,260 --> 00:03:56,940
But also, if they would've paid thousands

85
00:03:56,940 --> 00:03:58,500
for it, wouldn't they have paid thousands of.

86
00:03:58,875 --> 00:04:00,135
Dollars for it already.

87
00:04:00,279 --> 00:04:04,274
'cause GitHub copilot basically got out there before everyone.

88
00:04:04,424 --> 00:04:07,035
Would people not have paid thousands for that?

89
00:04:07,424 --> 00:04:11,924
That's the thing, Microsoft launching this product, they are the pricing gods.

90
00:04:11,924 --> 00:04:13,515
If they thought they could get a thousand

91
00:04:13,515 --> 00:04:16,305
dollars a month per head, they would do it.

92
00:04:16,305 --> 00:04:17,265
They would do it today.

93
00:04:17,265 --> 00:04:20,385
They would've done it yesterday or several years ago, but they didn't.

94
00:04:20,445 --> 00:04:24,615
And there was a story that came out in like 2023, I think that said that, um,

95
00:04:25,034 --> 00:04:29,085
Microsoft was losing like 20 bucks a customer with GitHub copilot as well.

96
00:04:29,085 --> 00:04:34,305
But putting that aside, I think that people overestimate, or sorry,

97
00:04:35,025 --> 00:04:38,200
underestimate how much software engineers love automating shit.

98
00:04:39,390 --> 00:04:42,030
Like there, there was the whole platform as a service era.

99
00:04:42,330 --> 00:04:45,780
Uh, the, the, I mean when that with containerization, when it came

100
00:04:45,780 --> 00:04:49,350
in, I remember talking to a client at the time PR firm who was

101
00:04:49,350 --> 00:04:51,780
just like, yeah, there are people who just do containerization.

102
00:04:51,780 --> 00:04:54,099
'cause their boss said they'd heard it at a conference.

103
00:04:54,430 --> 00:04:58,830
Nick Suresh Esh, uh, ity, he did a great piece about Monte Carlo.

104
00:04:58,920 --> 00:05:01,770
I don't even know what Monte Carlo does, but he has a great

105
00:05:01,770 --> 00:05:04,620
piece about how a bunch of people were trying to get Monte Carlo.

106
00:05:04,620 --> 00:05:06,270
They were like, why am I talking about Monte Carlo?

107
00:05:06,450 --> 00:05:07,560
And no one knew what it was.

108
00:05:07,920 --> 00:05:11,790
This is the modern software era and we can't even make

109
00:05:11,790 --> 00:05:15,720
real money off of AI in this kind of specious software era.

110
00:05:15,720 --> 00:05:17,070
It's crazy to me.

111
00:05:17,070 --> 00:05:17,460
Man.

112
00:05:17,850 --> 00:05:19,680
It, the whole thing feels insane.

113
00:05:20,220 --> 00:05:22,620
Well, as you said, it's automation, and this is on some level from

114
00:05:22,620 --> 00:05:25,350
a coding assistant perspective, the purest form of automation of

115
00:05:25,350 --> 00:05:27,930
the workflow, which is copying and pasting outta a stack Overflow.

116
00:05:28,170 --> 00:05:29,310
IDs aren't new.

117
00:05:30,150 --> 00:05:34,200
Like, I mean, it's just, I'm not saying that this is, and one, one common

118
00:05:34,200 --> 00:05:37,020
mischaracterization of my work, and frankly, I've done myself no favors.

119
00:05:37,020 --> 00:05:40,620
In my earlier stuff, I was a little unilateral, but one

120
00:05:41,070 --> 00:05:44,790
common mischaracterization is that I believe there's no value.

121
00:05:44,850 --> 00:05:47,940
No, I think this is a $50 billion total addressable market.

122
00:05:48,000 --> 00:05:52,110
I think that we are nowhere near the top, but the top is much smaller,

123
00:05:52,470 --> 00:05:56,640
and I don't think there is going to be some magical breakthrough.

124
00:05:56,700 --> 00:06:01,440
Every breakthrough that's happened is just

125
00:06:01,440 --> 00:06:04,530
kind of an incremental one since about 2023.

126
00:06:04,830 --> 00:06:07,920
Even with the launch of reasoning, remember what everyone was

127
00:06:07,920 --> 00:06:12,480
beating off about strawberry and it comes out and nothing happens.

128
00:06:12,600 --> 00:06:13,050
Yes.

129
00:06:13,050 --> 00:06:15,150
The number of RS in strawberry, which then all

130
00:06:15,150 --> 00:06:17,610
the models special case that that'll solve it.

131
00:06:17,640 --> 00:06:19,770
We'll just play whack-a-mole a hundred million times.

132
00:06:19,770 --> 00:06:21,570
Every time someone finds something embarrassing.

133
00:06:21,840 --> 00:06:22,920
It's so cool.

134
00:06:22,980 --> 00:06:25,710
It, it's so well, the strawberry one was there for a while.

135
00:06:26,039 --> 00:06:28,080
I know because every single time a new model would

136
00:06:28,080 --> 00:06:29,580
come out, I'd be like, how do you, how many Rs?

137
00:06:29,580 --> 00:06:31,020
And just the first thing I did.

138
00:06:31,020 --> 00:06:34,739
Or it'd be like, give me 50 sta, give me all the states with R in them.

139
00:06:34,890 --> 00:06:36,390
And I'd sit there and be like, woo hoo.

140
00:06:36,450 --> 00:06:37,229
They did it again.

141
00:06:37,229 --> 00:06:39,419
And then you easy posts.

142
00:06:39,510 --> 00:06:41,340
The one I love is trying to cyber bully it

143
00:06:41,340 --> 00:06:43,799
into ranking the US presidents by absorbency.

144
00:06:43,859 --> 00:06:45,900
That that can go surprisingly well.

145
00:06:45,990 --> 00:06:51,210
It's so funny that we're like boiling lakes to do, but it, it's

146
00:06:51,210 --> 00:06:56,789
just, it's very frustrating because I'm, I don't feel genuine.

147
00:06:56,789 --> 00:06:59,429
I genuinely, nothing you hear is really anger.

148
00:06:59,609 --> 00:07:04,710
I'm frustrated because what I am saying is true and it will event

149
00:07:04,710 --> 00:07:09,299
gravity exists and everyone's going to feel like me eventually, I think.

150
00:07:09,900 --> 00:07:13,140
And it's just like, how long do you want do this to yourself?

151
00:07:13,140 --> 00:07:14,909
How long do you wanna to base yourself

152
00:07:15,060 --> 00:07:17,580
pretending that these children are gifted?

153
00:07:17,820 --> 00:07:18,150
Yes.

154
00:07:18,180 --> 00:07:20,700
Uh, we're, we're still trying to summon God via JSON and

155
00:07:20,700 --> 00:07:23,250
we're hoping for the best, but it's not going super well.

156
00:07:23,340 --> 00:07:25,349
Uh, you've been tracking this longer than I have.

157
00:07:25,349 --> 00:07:30,270
I, I didn't want to be tracking this, but AWS had a corporate

158
00:07:30,300 --> 00:07:33,840
inability to shut up about any, about anything touching

159
00:07:33,840 --> 00:07:36,750
ai and it was all that they were willing to talk about.

160
00:07:36,750 --> 00:07:38,520
So I was brought in kicking and screaming

161
00:07:38,580 --> 00:07:41,490
relatively late 'cause I resisted for a while.

162
00:07:41,940 --> 00:07:43,320
What have you seen economically?

163
00:07:43,320 --> 00:07:44,370
Let's talk about the numbers.

164
00:07:44,460 --> 00:07:48,659
So recently came out as in like a day before we spoke the

165
00:07:48,719 --> 00:07:52,740
micro, so Microsoft fun story about Microsoft and numbers.

166
00:07:53,280 --> 00:07:55,664
Two quarters ago Microsoft said they had $10

167
00:07:55,664 --> 00:07:59,789
billion of annualized recurring revenue from ai.

168
00:08:00,060 --> 00:08:04,530
Now Microsoft does not have an AI section of their earnings.

169
00:08:04,530 --> 00:08:07,920
They have intelligent cloud productivity and software and I think.

170
00:08:09,270 --> 00:08:12,599
Like something like it's something and more intelligent

171
00:08:12,599 --> 00:08:14,760
cloud, which also speaks to the existence of more on

172
00:08:14,760 --> 00:08:16,770
cloud somewhere, but they don't like to talk about it.

173
00:08:16,800 --> 00:08:17,070
Yeah.

174
00:08:17,130 --> 00:08:19,830
Yeah, that's, that's actually what large language models are.

175
00:08:20,099 --> 00:08:21,960
I, so they have these, they don't have these

176
00:08:21,960 --> 00:08:25,230
sections, but AI is basically all combined AI revenue.

177
00:08:25,349 --> 00:08:30,450
So last quarter, then quarterly earnings in 2024, they said $10 billion a RR.

178
00:08:30,450 --> 00:08:35,520
So month times 12, Hey, January rolls around, proud as can be.

179
00:08:35,610 --> 00:08:39,960
Microsoft's at $13 billion of a RR in ai.

180
00:08:39,990 --> 00:08:41,039
Exciting, right?

181
00:08:41,220 --> 00:08:42,870
Next quarter, earnings come round.

182
00:08:42,870 --> 00:08:46,145
They do not update, the number doesn't update.

183
00:08:46,525 --> 00:08:47,305
Why would they do that?

184
00:08:47,535 --> 00:08:50,310
Then the information came out a few weeks ago with a story that

185
00:08:50,310 --> 00:08:55,950
said, uh, Microsoft's projected revenue in AI was $13 billion.

186
00:08:56,040 --> 00:08:59,550
And then another story came out yesterday that said that

187
00:08:59,910 --> 00:09:02,790
10 billion, well, first of all, $10 billion of that was

188
00:09:02,790 --> 00:09:07,260
open AI's spend, but OpenAI is as yours biggest customer.

189
00:09:08,505 --> 00:09:12,855
And it's very bad that that's the case because Azure, they pay a discounted

190
00:09:12,855 --> 00:09:17,415
rate on Azure and there's a decent chance that a lot of that money isn't real.

191
00:09:17,415 --> 00:09:22,545
'cause Microsoft invested 10 billion ish in 2023 with a large portion of

192
00:09:22,545 --> 00:09:25,845
that, that we don't know how large in cloud credits SEMA four reported.

193
00:09:26,205 --> 00:09:29,685
And so we have the, we have one of the largest

194
00:09:29,685 --> 00:09:31,725
cloud providers feeding itself cardboard.

195
00:09:32,220 --> 00:09:34,320
Like it's what it, that is what's happening.

196
00:09:34,320 --> 00:09:35,820
I've spoken to multiple report.

197
00:09:36,030 --> 00:09:38,730
It's like Azure's old biggest customer to my understanding was Xbox.

198
00:09:38,730 --> 00:09:39,690
I'm like, okay, great.

199
00:09:39,690 --> 00:09:40,680
Congratulations.

200
00:09:40,710 --> 00:09:41,940
Uh, the Snake eats itself.

201
00:09:41,940 --> 00:09:42,990
But that makes it sense.

202
00:09:43,050 --> 00:09:43,830
That makes sense.

203
00:09:43,890 --> 00:09:47,850
If you're providing, I imagine 365 is probably a very, very large Azure

204
00:09:48,060 --> 00:09:52,860
customer just because, but it isn't aort as revenue because it is cost,

205
00:09:52,860 --> 00:09:55,290
which probably has tax preferential, blah, blah, blah, blah, blah.

206
00:09:55,350 --> 00:09:56,910
They've also stuffed AI into it.

207
00:09:56,910 --> 00:09:59,070
So how much do you want to bet that that is now being

208
00:09:59,070 --> 00:10:01,560
considered AI revenue, despite the fact that just

209
00:10:01,560 --> 00:10:03,870
people who want to do want to write documents for work?

210
00:10:03,930 --> 00:10:07,830
It is, but Microsoft's only making $3 billion of annualized revenue

211
00:10:07,830 --> 00:10:12,390
from all of their AI bullshit other than open AI's $10 billion.

212
00:10:13,170 --> 00:10:17,520
But what's crazy about the story is that apparently Microsoft invested

213
00:10:17,520 --> 00:10:21,870
in open AI belie with top executives believing that they would fail.

214
00:10:22,980 --> 00:10:24,360
This is this.

215
00:10:24,570 --> 00:10:25,860
I've talked to multiple reports today.

216
00:10:25,860 --> 00:10:26,880
I'm trying to be like, Hey.

217
00:10:27,420 --> 00:10:28,469
This is weird.

218
00:10:28,469 --> 00:10:31,050
Have you looked and everyone's like, yeah, yeah, maybe.

219
00:10:31,050 --> 00:10:32,520
You know, these things happen.

220
00:10:32,550 --> 00:10:34,410
You know what people do?

221
00:10:34,500 --> 00:10:35,699
No one does this.

222
00:10:35,819 --> 00:10:39,030
I have never heard in my entire history in Silicon Valley ever

223
00:10:39,300 --> 00:10:42,750
of someone investing in a company and thinking they would die.

224
00:10:43,050 --> 00:10:44,339
There's only one reason.

225
00:10:44,550 --> 00:10:47,459
These, these throw money at it and early stage ideas, but they throw

226
00:10:47,459 --> 00:10:50,280
a million here, 2 million there out of a hundred million dollars fund.

227
00:10:50,310 --> 00:10:52,140
But that's a slight difference.

228
00:10:52,140 --> 00:10:58,050
They are believing it might die, but kind of hedging their bets.

229
00:10:58,380 --> 00:11:00,510
Microsoft executives believe they would

230
00:11:00,510 --> 00:11:04,170
eventually fail, is a vastly different situation.

231
00:11:04,709 --> 00:11:08,730
And the reason in my opinion, is that Microsoft has access to all of

232
00:11:08,730 --> 00:11:13,860
open AI's research IP and they host all of their cloud infrastructure.

233
00:11:14,370 --> 00:11:16,589
Kind of feels like they can let OpenAI die and then

234
00:11:16,589 --> 00:11:19,829
take everything OpenAI has because they already own it.

235
00:11:20,220 --> 00:11:22,199
But nevertheless, putting all that aside.

236
00:11:23,820 --> 00:11:25,980
Microsoft's biggest as your customer is a

237
00:11:25,980 --> 00:11:30,570
company that op that loses the money That is bad.

238
00:11:30,570 --> 00:11:34,320
We are, and Microsoft is the company making the most from

239
00:11:34,320 --> 00:11:38,880
AI by handing money to a company to hand it back to them,

240
00:11:39,090 --> 00:11:42,660
Google is doing a deal to serve compute, to open ai.

241
00:11:43,440 --> 00:11:46,140
By which I mean they are hiring core weave to provide the

242
00:11:46,140 --> 00:11:48,630
compute to Google, to provide the compute, to open ai.

243
00:11:48,840 --> 00:11:50,670
This is fracking wash trading.

244
00:11:51,300 --> 00:11:53,580
It's wash trading in the cloud.

245
00:11:53,940 --> 00:11:56,460
We, this should terrify people.

246
00:11:56,520 --> 00:11:58,920
'cause when you remove open AI's, compute, Microsoft's

247
00:11:58,920 --> 00:12:03,510
making $3 billion a year from AI off of what?

248
00:12:03,570 --> 00:12:05,700
$80 billion of pledged CapEx.

249
00:12:06,030 --> 00:12:07,470
What the frack are we doing?

250
00:12:07,950 --> 00:12:10,680
Like this is, this is not even radical stuff.

251
00:12:10,680 --> 00:12:12,780
I'm speaking objective fact.

252
00:12:13,050 --> 00:12:15,090
Like this is stuff that is real.

253
00:12:15,270 --> 00:12:17,715
But people are like, ah, he's just, he's just a pessimist.

254
00:12:17,715 --> 00:12:18,630
He's just a cenic.

255
00:12:19,080 --> 00:12:20,700
You just have to believe harder at.

256
00:12:21,225 --> 00:12:24,045
I don't even know what I'm believing in at this point.

257
00:12:24,315 --> 00:12:28,245
No one can express what it is that I am not believing in beyond.

258
00:12:28,515 --> 00:12:32,475
They will work it out and model get better.

259
00:12:32,715 --> 00:12:33,885
How does model get better?

260
00:12:34,125 --> 00:12:35,025
Oh, I'm sorry.

261
00:12:35,025 --> 00:12:35,955
I can't answer that.

262
00:12:36,375 --> 00:12:37,905
The what?

263
00:12:38,535 --> 00:12:39,705
Oh, millions of people.

264
00:12:39,765 --> 00:12:40,665
A hundred million.

265
00:12:41,025 --> 00:12:43,545
It's actually a really good that we're on a cloud podcast.

266
00:12:44,415 --> 00:12:45,045
Sure.

267
00:12:45,795 --> 00:12:46,605
Here's a fun fact.

268
00:12:46,605 --> 00:12:50,565
Do you ever notice the, the OpenAI never announces monthly active users.

269
00:12:50,565 --> 00:12:51,585
They only do weekly.

270
00:12:52,245 --> 00:12:52,395
Mm-hmm.

271
00:12:52,785 --> 00:12:53,265
I did.

272
00:12:53,625 --> 00:12:54,075
Yeah.

273
00:12:54,195 --> 00:12:57,345
Do you, do you, do you know why they're not doing that?

274
00:12:57,345 --> 00:12:58,275
I have a theory.

275
00:12:58,905 --> 00:13:00,615
I'm, I don't have anything concrete.

276
00:13:00,615 --> 00:13:01,815
I'm curious to hear your theory.

277
00:13:02,175 --> 00:13:07,065
So the information reported earlier in the year that, uh, I think it was, they

278
00:13:07,065 --> 00:13:13,935
have 15.5 billion, sorry, 15.5 million million paying subscribers on chat GPT.

279
00:13:13,935 --> 00:13:14,355
Right.

280
00:13:15,045 --> 00:13:16,995
I reckon their monthly active users are much

281
00:13:16,995 --> 00:13:19,425
higher than 500 million, which would seem good.

282
00:13:19,425 --> 00:13:19,785
Right.

283
00:13:20,280 --> 00:13:23,760
Would seem good to have except, okay, so say they're 600,

284
00:13:23,760 --> 00:13:27,660
700 million and then you divide that by 15.5 million.

285
00:13:27,839 --> 00:13:29,610
That's a dog shit conversion rate.

286
00:13:29,729 --> 00:13:34,439
That's absolute Do-do that is some of the worst conversion rates in SaaS.

287
00:13:34,589 --> 00:13:37,079
And you have to treat open a, like a SA business.

288
00:13:37,079 --> 00:13:39,360
You cannot treat it like a consumer subscription.

289
00:13:39,540 --> 00:13:41,400
They lose money on every customer, but they

290
00:13:41,400 --> 00:13:44,160
want to be given SaaS style valuations.

291
00:13:44,400 --> 00:13:47,130
Okay, so you have a 2% conversion rate.

292
00:13:47,130 --> 00:13:49,439
The information came out yesterday, well, day before yesterday,

293
00:13:49,439 --> 00:13:52,410
I think with a whole thing about retention rates open.

294
00:13:52,410 --> 00:13:57,780
AI's got like a 73, 72% retention rate after six months,

295
00:13:57,990 --> 00:14:01,020
which is in line with like Spotify, which has 72%.

296
00:14:02,160 --> 00:14:05,849
That's not very good for a company that wants SaaS style valuations.

297
00:14:06,540 --> 00:14:08,910
That's more consumer product valuation than anything else.

298
00:14:09,060 --> 00:14:09,719
Exactly.

299
00:14:09,719 --> 00:14:12,449
And a consumer product that only loses money.

300
00:14:13,349 --> 00:14:14,834
That's not great.

301
00:14:15,555 --> 00:14:16,815
It's not, it's not brilliant.

302
00:14:16,875 --> 00:14:17,385
It's not.

303
00:14:17,475 --> 00:14:19,275
I juggle my consumer stuff all the time.

304
00:14:19,305 --> 00:14:22,155
The business stuff that I use that tends to stay put for years.

305
00:14:22,155 --> 00:14:25,215
That's why you see relatively small churn rates.

306
00:14:25,215 --> 00:14:28,395
That's why you see renewal rates being very sticky.

307
00:14:29,235 --> 00:14:31,365
It, there's not these big drop off at one

308
00:14:31,365 --> 00:14:32,835
year cliffs at most of these companies.

309
00:14:32,895 --> 00:14:36,015
And the thing is, it's because they're effectively consumer subscription.

310
00:14:36,075 --> 00:14:38,145
They, no one really, you can't really express.

311
00:14:38,265 --> 00:14:41,955
I realize that a lot of SaaS is sold in this kind of specious way.

312
00:14:41,955 --> 00:14:44,685
You, where you sell to A CTO or a CEO or a CIO

313
00:14:44,685 --> 00:14:46,575
that doesn't really touch the rest of the stack.

314
00:14:47,265 --> 00:14:50,865
You can't really do that with AI quite as good because

315
00:14:51,675 --> 00:14:53,985
there's not really an outcome that you can even look at.

316
00:14:54,135 --> 00:14:57,015
It's just kind of a thing you offer and it's so heavily commoditized.

317
00:14:57,015 --> 00:14:59,595
Sure, you could get cursor, sure, you could get chat

318
00:14:59,595 --> 00:15:03,225
GPT, but at the same time, okay, I've now got it.

319
00:15:03,495 --> 00:15:05,205
How are you making more money outta these companies?

320
00:15:05,205 --> 00:15:06,975
'cause that's how SaaS companies really cook.

321
00:15:07,215 --> 00:15:08,985
You can't really upsell someone with chat

322
00:15:08,985 --> 00:15:11,565
GPT because, oh, my model will be better.

323
00:15:11,865 --> 00:15:12,225
How?

324
00:15:12,690 --> 00:15:13,890
One company goes out with a good model.

325
00:15:13,890 --> 00:15:17,460
The other one winds up matching or exceeding it within the next two weeks.

326
00:15:17,490 --> 00:15:20,310
Uh, if you suddenly say, you're not allowed to use this one, I can

327
00:15:20,310 --> 00:15:23,340
switch to the other one for anything that I'm doing in less than

328
00:15:23,340 --> 00:15:25,890
an hour's worth of work, and we're right back where we started.

329
00:15:25,890 --> 00:15:26,580
There's no moat.

330
00:15:26,580 --> 00:15:27,450
They're they're fungible.

331
00:15:27,510 --> 00:15:28,230
Exactly.

332
00:15:28,230 --> 00:15:31,350
And on top of that, if you were trying to sell into the enterprise,

333
00:15:31,740 --> 00:15:34,710
I guess if you're doing enterprise chat, GPT subscriptions

334
00:15:35,100 --> 00:15:38,130
fine, but they're not getting the kind of revenue they want.

335
00:15:38,250 --> 00:15:39,510
They're not getting the kind of revenue that

336
00:15:39,510 --> 00:15:42,180
like a Salesforce would have out of that.

337
00:15:42,480 --> 00:15:42,930
They're not.

338
00:15:43,170 --> 00:15:46,260
They are, right now they are, let's see, let's see.

339
00:15:46,260 --> 00:15:51,120
Salesforce earnings bring 'em up live because I think that

340
00:15:51,120 --> 00:15:56,130
they want to, for chat GPT and open AI to make sense, they

341
00:15:56,130 --> 00:16:01,200
need the combined value of the consumer, sorry, the consumers.

342
00:16:01,200 --> 00:16:03,930
I think it's the smartphone market's like half a trillion dollars.

343
00:16:04,260 --> 00:16:05,640
And the enterprise SaaS market, which is

344
00:16:05,640 --> 00:16:08,220
about a quarter of a trillion, they need both.

345
00:16:08,715 --> 00:16:10,455
That's the only way any of this makes sense.

346
00:16:10,455 --> 00:16:14,145
And they are not even at the total revenue.

347
00:16:14,234 --> 00:16:16,305
They may be just a little bit above the total revenue.

348
00:16:16,604 --> 00:16:19,935
If you combine open ai, anthropic, Microsoft, Amazon,

349
00:16:20,084 --> 00:16:23,984
all of the projected revenues, they're like 38 billion.

350
00:16:25,124 --> 00:16:29,805
The global smartwatch revenue for this year is projected to be about 32 billion.

351
00:16:30,749 --> 00:16:32,339
So congratulations everyone.

352
00:16:32,339 --> 00:16:34,530
All the kings, horses, all the kingsmen came together

353
00:16:34,680 --> 00:16:36,989
and we've barely beaten the smartwatch industry.

354
00:16:37,349 --> 00:16:39,390
The register had an article come out this morning that

355
00:16:39,390 --> 00:16:42,810
says that, uh, was a quarter of all Gen ai Proof of

356
00:16:42,810 --> 00:16:45,269
concepts don't, are turned off by the end of the year.

357
00:16:45,659 --> 00:16:46,169
People are.

358
00:16:46,394 --> 00:16:49,334
It's people are experimenting with it, and I work on cloud bills.

359
00:16:49,334 --> 00:16:51,374
I see what the spend looks like.

360
00:16:51,435 --> 00:16:54,435
People are experimenting with these things, but I'm not seeing

361
00:16:54,435 --> 00:16:57,824
anyone saying, well, we're, we're spending $150 million a year

362
00:16:57,824 --> 00:17:01,154
with AWS, but we'd better make a 200 million on our commitment.

363
00:17:01,184 --> 00:17:05,174
'cause of all the Gen ai, they're doing $50,000 a month here and there.

364
00:17:05,174 --> 00:17:07,935
But on that basis of a spend, what's it matter?

365
00:17:07,935 --> 00:17:10,514
It's, if it succeeds, maybe they'll start rolling, rolling

366
00:17:10,514 --> 00:17:12,674
it out more wildly, and then care about optimizing it.

367
00:17:12,944 --> 00:17:16,364
But it's getting turned off if it doesn't hit certain, uh, certain metrics.

368
00:17:16,395 --> 00:17:19,065
And I'm not seeing the stories that show that this

369
00:17:19,065 --> 00:17:21,585
stuff is actually being transformative in any way.

370
00:17:21,614 --> 00:17:24,464
And if it were even slightly, they wouldn't shut up about it.

371
00:17:24,525 --> 00:17:25,545
Exactly.

372
00:17:25,634 --> 00:17:27,435
These motherfrackers love going to conferences

373
00:17:27,435 --> 00:17:28,995
and talking about how important stuff is.

374
00:17:28,995 --> 00:17:32,625
But I guess another Nick Suresh suresh piece, I'll have to link you to,

375
00:17:32,924 --> 00:17:36,284
but talking about why people get, uh, why people talk about Snowflake.

376
00:17:36,674 --> 00:17:37,815
Like it's a database company.

377
00:17:37,815 --> 00:17:39,915
It's like, it's the most boring thing, but guys like buying

378
00:17:39,915 --> 00:17:42,134
Snowflake so they can go and talk about using Snowflake.

379
00:17:42,465 --> 00:17:45,284
But with AI it's like, it's all experimental.

380
00:17:45,284 --> 00:17:48,074
I don't think people realize how bad that is because

381
00:17:48,105 --> 00:17:50,415
it would be one thing saying this a year ago, I was

382
00:17:50,415 --> 00:17:53,685
right then and I'm right now, but a year ago, fine.

383
00:17:54,255 --> 00:17:56,054
It's experimental fine.

384
00:17:56,384 --> 00:17:57,915
There's experimentations are taking an

385
00:17:57,915 --> 00:18:01,215
awfully long time by enterprise standards.

386
00:18:01,335 --> 00:18:01,844
Exactly.

387
00:18:01,844 --> 00:18:04,215
Like they would say this is a, this is a good thing.

388
00:18:04,215 --> 00:18:05,235
Like this is moving well.

389
00:18:05,235 --> 00:18:07,034
Also, I wanna correct myself, I brought it up.

390
00:18:07,034 --> 00:18:10,065
Salesforce's revenue is $9.83 billion in the

391
00:18:10,065 --> 00:18:14,685
last quarter, so only four x to go for open ai.

392
00:18:14,745 --> 00:18:19,965
Also, they made $1.5 billion in profit open AI is yet to make $1 in profit.

393
00:18:20,445 --> 00:18:23,835
I spent top dollar for Slack and I sort of work my way between

394
00:18:23,835 --> 00:18:26,985
all the AI stuff that they're shoehorning into it against my will.

395
00:18:27,195 --> 00:18:30,134
Oh, I, I am paying for Slack with a gun to my head.

396
00:18:30,134 --> 00:18:34,005
Like I do not wanna pay the idea of paying Mark Benioff sickens me.

397
00:18:34,065 --> 00:18:34,545
But.

398
00:18:35,384 --> 00:18:37,694
Salesforce, as I mentioned earlier, they say

399
00:18:37,694 --> 00:18:39,824
there's not gonna be any revenue growth from this.

400
00:18:39,944 --> 00:18:46,065
It's just so weird because all of these things have been said publicly.

401
00:18:46,335 --> 00:18:50,054
All of these things kept, keep being said publicly, but

402
00:18:50,054 --> 00:18:52,094
when you put 'em all together and you say, Hey, this

403
00:18:52,094 --> 00:18:54,705
doesn't look good, everyone goes,  you rude bastard.

404
00:18:55,125 --> 00:18:55,844
How dare you?

405
00:18:56,475 --> 00:18:58,364
How dare you insult my beautiful child?

406
00:18:58,364 --> 00:18:58,844
Ai?

407
00:18:59,054 --> 00:19:02,534
It's like, I dunno, because it sucks.

408
00:19:03,165 --> 00:19:06,554
Like even if it worked, even if it was flawless, if these

409
00:19:06,554 --> 00:19:09,165
were the revenues and the losses, it would still kind of suck.

410
00:19:09,344 --> 00:19:11,384
Like it's not, it's not good business and

411
00:19:11,384 --> 00:19:13,574
frankly I don't think it's great software either.

412
00:19:13,814 --> 00:19:15,750
. Um, it's.

413
00:19:16,229 --> 00:19:18,749
It's just so bizarre because sure, there are people who really

414
00:19:18,749 --> 00:19:21,360
like it and I think there's a philosophical angle here where

415
00:19:21,360 --> 00:19:23,850
there are people who are really attached to being attached to the

416
00:19:23,850 --> 00:19:27,449
future where they want, they wanna be the guy that said it first.

417
00:19:27,449 --> 00:19:28,379
I think you see it in journalism.

418
00:19:28,379 --> 00:19:31,019
I think you see it in software engineering and founders in particular.

419
00:19:31,019 --> 00:19:34,229
They wanna be the person that said, I was on this trend before everyone.

420
00:19:34,229 --> 00:19:35,279
Look how smart I am.

421
00:19:35,399 --> 00:19:37,409
I think that that's what Andy Jassy's doing.

422
00:19:37,829 --> 00:19:38,879
Management consultant.

423
00:19:38,939 --> 00:19:39,870
MBA motherfracker.

424
00:19:40,709 --> 00:19:41,519
He, okay.

425
00:19:41,519 --> 00:19:43,829
I don't think he was a management consultant is an MBA though.

426
00:19:44,100 --> 00:19:46,110
He does have at MBA, but he was a marketing

427
00:19:46,110 --> 00:19:49,110
manager at at Amazon before he started AWS.

428
00:19:49,410 --> 00:19:52,800
He was one of the 21 people that take credit for founding AWS.

429
00:19:53,010 --> 00:19:55,230
It's like a, there's like a bunch of people.

430
00:19:55,350 --> 00:19:56,430
You ever notice it's, yeah.

431
00:19:56,430 --> 00:19:58,860
Can you tell on LinkedIn whether project's successful or not by how many

432
00:19:58,860 --> 00:20:01,770
people claim credit for the thing versus distance the hell outta themselves?

433
00:20:01,830 --> 00:20:05,460
But I think the Jassy, I think it's an AWS,

434
00:20:06,090 --> 00:20:08,280
it's kind of a hard situation for AWS as well.

435
00:20:08,280 --> 00:20:09,780
'cause of course they have to jump on this.

436
00:20:09,990 --> 00:20:11,850
Like the, they couldn't, not the stock would

437
00:20:11,850 --> 00:20:16,200
get pummeled, but they are, they're doing this.

438
00:20:16,200 --> 00:20:18,570
Jassy's staking his future on this.

439
00:20:18,600 --> 00:20:21,720
He wants this to be, here's everything store.

440
00:20:21,780 --> 00:20:23,520
He wants this to be the thing that defines

441
00:20:23,520 --> 00:20:25,910
the future of Amazon, except it's dog shit.

442
00:20:26,040 --> 00:20:26,580
Amazon.

443
00:20:26,610 --> 00:20:30,930
The Amazon may be one of the most embarrassing AI companies.

444
00:20:30,930 --> 00:20:32,520
I think Apple truly is.

445
00:20:32,520 --> 00:20:34,290
I think Apple's doing the smartest thing

446
00:20:34,710 --> 00:20:38,100
in that they are kind of slow walking back.

447
00:20:38,130 --> 00:20:41,040
But Apple intelligence, it was like the most radicalizing thing.

448
00:20:41,190 --> 00:20:42,870
They shot their mouth off too early.

449
00:20:43,020 --> 00:20:43,830
I love it.

450
00:20:43,920 --> 00:20:44,400
I love it.

451
00:20:44,400 --> 00:20:46,530
So I love that they were like, what if.

452
00:20:47,025 --> 00:20:49,665
Hey, you know, autocorrect, what if it sucked?

453
00:20:50,685 --> 00:20:52,785
And okay, I, I don't like that idea.

454
00:20:52,845 --> 00:20:56,925
What if we randomly added buttons that loaded things you didn't want?

455
00:20:57,315 --> 00:20:58,515
No, I don't want that.

456
00:20:58,515 --> 00:20:58,965
Cool.

457
00:20:58,965 --> 00:21:00,765
This update will apply automatically.

458
00:21:01,185 --> 00:21:04,305
I think they radicalized millions of people against ai.

459
00:21:04,455 --> 00:21:07,275
And on the keynote too, it's what if your Siri had this

460
00:21:07,275 --> 00:21:09,525
magic knowledge of everything you did on your iPhone?

461
00:21:09,525 --> 00:21:11,415
It could tie it to, it's like, wow, that'd be kind of cool.

462
00:21:11,415 --> 00:21:12,135
Like, when do we get that?

463
00:21:12,140 --> 00:21:12,405
No, no, no.

464
00:21:12,405 --> 00:21:13,755
We just said, what if, what if?

465
00:21:13,755 --> 00:21:14,445
Wouldn't it be neat?

466
00:21:14,445 --> 00:21:16,425
Like we don't know how to do it, but wouldn't it be neat?

467
00:21:16,455 --> 00:21:17,235
We are riffing.

468
00:21:17,295 --> 00:21:18,315
We are riffing right now.

469
00:21:18,615 --> 00:21:22,275
Well, uh, Google did the same thing at the last, uh, IO in 2024.

470
00:21:22,275 --> 00:21:26,505
They were like, imagine if you needed to return a shoe, you hit one button.

471
00:21:26,685 --> 00:21:29,325
An agent takes the shoe data, it does this, it does

472
00:21:29,325 --> 00:21:31,245
that, and then this happens, and then this happens.

473
00:21:31,845 --> 00:21:33,765
All of that is entirely theoretical.

474
00:21:33,765 --> 00:21:34,995
But wouldn't it be cool?

475
00:21:35,805 --> 00:21:38,565
I swear to God it's illegal for public companies to lie.

476
00:21:38,565 --> 00:21:41,235
I. Or like they, there should be, I thought there was

477
00:21:41,235 --> 00:21:43,395
something that stopped them from just making shit up.

478
00:21:43,395 --> 00:21:43,935
But there isn't.

479
00:21:43,935 --> 00:21:45,655
They can just do what they want.

480
00:21:46,255 --> 00:21:49,245
And then when you look behind the curtain, there's no money.

481
00:21:49,695 --> 00:21:52,245
There was an analyst that said, my Amazon is estimated

482
00:21:52,245 --> 00:21:55,575
to make $5 billion in generative AI this year.

483
00:21:55,755 --> 00:22:00,705
That's, that's, these are not just rookie numbers, these are alarming.

484
00:22:00,945 --> 00:22:03,765
One of their, uh, one of their job ads a while back

485
00:22:03,765 --> 00:22:07,065
listed S3 as an AI service, as one of the list of them.

486
00:22:07,065 --> 00:22:07,455
So, great.

487
00:22:07,515 --> 00:22:10,155
Where does that start And stop the, the broad buckets

488
00:22:10,155 --> 00:22:12,735
that no one will ever break down further for you.

489
00:22:12,765 --> 00:22:14,085
So here's the thing.

490
00:22:14,655 --> 00:22:16,245
S3 is storage, right?

491
00:22:16,245 --> 00:22:16,725
Forgive me,

492
00:22:16,725 --> 00:22:17,805
I'm, it is.

493
00:22:17,925 --> 00:22:20,205
I have heard that Anthropic is the biggest customer there.

494
00:22:20,700 --> 00:22:22,080
I have been told by someone.

495
00:22:22,680 --> 00:22:26,040
There are multiple exabyte scale customers on Esther.

496
00:22:26,040 --> 00:22:28,380
I don't, I don't have the details at this point.

497
00:22:28,425 --> 00:22:29,970
I, I would not be surprised.

498
00:22:30,150 --> 00:22:33,560
Oh, and this is just a, this is just doing the Joe Rogan "you know

499
00:22:33,560 --> 00:22:38,190
that guy told me that once", but it's still five billion dollars is

500
00:22:38,190 --> 00:22:41,280
not a lot of money on a hundred billion dollars of CapEx this year.

501
00:22:41,580 --> 00:22:43,965
Microsoft last year, I think only made $5

502
00:22:43,965 --> 00:22:46,830
billion in ai, and that included open AI spend.

503
00:22:47,370 --> 00:22:50,400
This is not an in, this isn't an industry.

504
00:22:50,580 --> 00:22:54,480
Like if you really look at the cost benefit analysis, this isn't real.

505
00:22:54,630 --> 00:22:57,270
Like that's it, it's a Fugazi.

506
00:22:57,420 --> 00:22:59,550
If all this growth is there, then why haven't we seen

507
00:22:59,550 --> 00:23:02,280
massive posted growth numbers from all of these companies?

508
00:23:02,280 --> 00:23:04,050
Rather than what looks a lot like typical

509
00:23:04,050 --> 00:23:06,780
organic, more of the same style growth.

510
00:23:06,990 --> 00:23:13,110
I mean, Microsoft has been amortizing $2.5 billion of, of

511
00:23:13,110 --> 00:23:16,800
revenue from open AI's cloud spend across intelligent cloud.

512
00:23:16,980 --> 00:23:19,020
Any growth within intelligent cloud.

513
00:23:19,409 --> 00:23:20,699
Is now suspicious.

514
00:23:21,330 --> 00:23:21,629
Yes.

515
00:23:21,750 --> 00:23:22,169
Yes.

516
00:23:22,379 --> 00:23:25,590
And they're the biggest investor in OpenAI, so they're effectively putting

517
00:23:25,590 --> 00:23:29,429
their own money into back into their own pocket by laundering it through OpenAI.

518
00:23:29,550 --> 00:23:32,310
And if it's cloud credits, is it even money?

519
00:23:32,669 --> 00:23:34,949
I, I've never trusted Azure growth numbers and Azure usage

520
00:23:34,949 --> 00:23:37,770
numbers, just because they're so inextricably linked to Office

521
00:23:37,770 --> 00:23:41,580
365 to enterprise agreements that big companies have for windows.

522
00:23:42,060 --> 00:23:43,290
Whoa, whoa, whoa, whoa.

523
00:23:43,770 --> 00:23:45,870
Do you think that they're charging, that they're

524
00:23:47,399 --> 00:23:51,000
amortizing revenue from their own spend on Azure?

525
00:23:51,090 --> 00:23:51,929
That can't be.

526
00:23:52,470 --> 00:23:55,439
I would not be entirely surprised.

527
00:23:55,439 --> 00:23:57,240
I'm sure it's by, I'm sure it's above board.

528
00:23:57,240 --> 00:24:00,870
I would not want to accuse them of doing anything other, other than that.

529
00:24:01,139 --> 00:24:04,379
But Amy Hood, their CFO has been an absolute

530
00:24:04,379 --> 00:24:07,620
master of financial engineering for over a decade.

531
00:24:07,919 --> 00:24:13,050
Uh, it's impossible to get much signal from the things that they say based

532
00:24:13,050 --> 00:24:16,110
upon the way that they very intentionally have structured and laid them out.

533
00:24:16,470 --> 00:24:18,899
Can I run my favorite.

534
00:24:19,500 --> 00:24:23,010
This is tin foil hat, like this probably won't

535
00:24:23,010 --> 00:24:25,290
happen, but I wanna put my flag in the ground.

536
00:24:25,530 --> 00:24:27,990
I think, I think Amy Hood takes over Microsoft.

537
00:24:29,040 --> 00:24:32,429
I think when whatever happens with OpenAI shits the bed.

538
00:24:32,520 --> 00:24:36,389
I think they put a, they put like boxer in Animal Farm.

539
00:24:36,389 --> 00:24:37,610
They send Satya off.

540
00:24:38,310 --> 00:24:41,250
Bye by Satya because, 'cause he is, so, I'm reading between

541
00:24:41,250 --> 00:24:43,770
the lines here, but the information story that came out over

542
00:24:43,770 --> 00:24:46,800
this tense negotiation between open AI and Microsoft, there

543
00:24:46,800 --> 00:24:50,970
was a thing where top executives thought open AI would fail.

544
00:24:51,090 --> 00:24:53,070
Wall Street Journal Berber Jin there were like

545
00:24:53,070 --> 00:24:54,600
four different bylines on the information.

546
00:24:54,600 --> 00:24:55,980
But Berber Jin wrote journal one.

547
00:24:56,189 --> 00:25:00,750
That one had a thing about how a top executives didn't like the clause

548
00:25:00,750 --> 00:25:04,860
about  AGI Basically, when  AGI is called open AI doesn't have to share

549
00:25:04,860 --> 00:25:09,300
IP with Microsoft and, but Satya Nadella apparently pushed it through.

550
00:25:09,360 --> 00:25:14,110
What I think happened was Amy Hood went, this company is going to  die.

551
00:25:14,879 --> 00:25:16,379
This company is gonna die.

552
00:25:17,040 --> 00:25:19,440
This is stupid, but they're gonna die.

553
00:25:19,500 --> 00:25:21,300
Let's buy them effectively.

554
00:25:21,300 --> 00:25:23,130
And then when they die, we take all their stuff.

555
00:25:23,400 --> 00:25:26,070
And then, Satya Nadella said, well, the  AGI thing.

556
00:25:26,070 --> 00:25:27,070
And everyone went, no.

557
00:25:27,070 --> 00:25:27,450
Satya.

558
00:25:27,620 --> 00:25:28,500
That's  stupid.

559
00:25:28,960 --> 00:25:29,530
Don't do that.

560
00:25:29,530 --> 00:25:30,760
So, and he agreed to it Anyway.

561
00:25:30,910 --> 00:25:33,250
It's now a major sticking point, I think.

562
00:25:33,520 --> 00:25:35,530
I think when it all falls down and it, well,

563
00:25:35,860 --> 00:25:38,920
this is the Azure thing is a freaking scandal.

564
00:25:39,130 --> 00:25:41,200
I think the tech press is doing it, and business

565
00:25:41,200 --> 00:25:42,940
press is doing a terrible job with this.

566
00:25:43,240 --> 00:25:46,360
The fact that your largest customer is a company that burns money

567
00:25:46,360 --> 00:25:50,680
and has incredibly shaky finances, that is a counterparty risk.

568
00:25:51,010 --> 00:25:52,030
That is a huge thing.

569
00:25:52,360 --> 00:25:55,420
So when this breaks, I wouldn't be surprised if Satya is the

570
00:25:55,420 --> 00:25:59,620
one that gets blamed and Amy Hood in a responsible leadership

571
00:25:59,620 --> 00:26:02,110
role comes in and just starts freaking savaging the company.

572
00:26:02,560 --> 00:26:06,370
Sort of an interim CEO story where they conduct a search, but

573
00:26:06,370 --> 00:26:09,880
the search drags on and they remove interim from her Title I.

574
00:26:10,030 --> 00:26:11,350
Or whoever she picks.

575
00:26:11,350 --> 00:26:12,370
It's another like.

576
00:26:13,405 --> 00:26:16,534
Another elder, God, she goes back, finds a cthulhu

577
00:26:16,555 --> 00:26:20,125
adjacent creature to take over Microsoft, because I have

578
00:26:20,125 --> 00:26:22,975
friends who think that this Microsoft thing is a plan.

579
00:26:23,215 --> 00:26:26,665
It was always the plan kind of because non 'cause of the nonprofit

580
00:26:26,665 --> 00:26:28,975
situation, I've kind of like poorly explained the situation.

581
00:26:28,975 --> 00:26:29,784
Do you want me to?

582
00:26:30,504 --> 00:26:31,645
It's entirely up to you.

583
00:26:31,645 --> 00:26:33,625
I, I feel like there's, there's a lot of information

584
00:26:33,625 --> 00:26:35,365
around this floating around out there, but.

585
00:26:35,784 --> 00:26:38,725
Basically OpenAI needs to become a for-profit entity by the

586
00:26:38,725 --> 00:26:41,305
end of the year or they lose $20 billion from their funding.

587
00:26:41,485 --> 00:26:43,014
Microsoft is the stopping block.

588
00:26:43,014 --> 00:26:45,145
'cause OpenAI wants a bunch of concessions

589
00:26:45,145 --> 00:26:46,675
and Microsoft doesn't want to give them.

590
00:26:47,095 --> 00:26:47,395
Yeah.

591
00:26:47,754 --> 00:26:51,865
And um, no one knows what's gonna happen and no one wants to say

592
00:26:52,045 --> 00:26:55,195
what I'm saying, which is what if Microsoft just lets them die?

593
00:26:55,225 --> 00:26:57,595
Because if Microsoft says no, we won't let you convert.

594
00:26:58,585 --> 00:26:59,485
OpenAI dies.

595
00:26:59,889 --> 00:27:01,060
They can't go public.

596
00:27:01,690 --> 00:27:04,510
And uh, Microsoft owns all their IP and research

597
00:27:04,510 --> 00:27:06,070
anyway and handles all their infrastructure.

598
00:27:06,100 --> 00:27:09,070
Well, what about all the CapEx they did explicitly for open ai?

599
00:27:09,100 --> 00:27:09,340
Yeah.

600
00:27:09,370 --> 00:27:11,050
'cause those things can't ever be repurposed

601
00:27:11,050 --> 00:27:12,820
for other customers doing the different things.

602
00:27:12,820 --> 00:27:15,910
They also didn't, they never said it was just for open ai.

603
00:27:15,910 --> 00:27:17,139
They just said for ai.

604
00:27:17,440 --> 00:27:19,870
Amazon has been saying that Project Rainier, all

605
00:27:19,870 --> 00:27:22,750
of these specific data centers are for anthropic.

606
00:27:22,960 --> 00:27:24,700
And there was a New York Times story about it.

607
00:27:24,700 --> 00:27:25,030
Yeah.

608
00:27:25,600 --> 00:27:28,210
Indeed, and I'm very curious as to how that plays out because they're

609
00:27:28,210 --> 00:27:31,530
talking a lot about how Oh, anthropic says that , they're running,

610
00:27:31,860 --> 00:27:36,330
uh, Claude Opus or Claude, the Claude four on Trainium, which is

611
00:27:36,330 --> 00:27:39,660
interesting because I have looked far and wide and I can't find

612
00:27:39,945 --> 00:27:41,875
any Trainium or Inference customers.

613
00:27:42,055 --> 00:27:47,185
The two that I'm aware of from keynotes were Apple and Anthropic Anthropic.

614
00:27:47,185 --> 00:27:50,005
They've invested  four billion dollars in, and Apple has

615
00:27:50,065 --> 00:27:53,155
been known to be a large AWS customer for quite some time.

616
00:27:53,665 --> 00:27:57,565
When we see contracts being negotiated between these things, who says

617
00:27:57,565 --> 00:28:01,285
what on stage at whose conference is always negotiated into that?

618
00:28:01,465 --> 00:28:03,895
Apple never says shit about other companies'

619
00:28:03,895 --> 00:28:06,385
products, but they did at reinvent last year.

620
00:28:06,385 --> 00:28:09,715
Gee, I'm sure there was no contractual stipulation there.

621
00:28:10,405 --> 00:28:12,865
Also, no one has said that they're running their

622
00:28:12,865 --> 00:28:15,455
Claude four models exclusively on  Trainium.

623
00:28:15,625 --> 00:28:16,855
So what did they do?

624
00:28:16,855 --> 00:28:19,495
Did they try running a quick like, and here's the thing that

625
00:28:19,495 --> 00:28:22,465
ties the last bow on it and just this one side bit we can

626
00:28:22,465 --> 00:28:24,835
run on  Trainium and the rest, we using video like grownups.

627
00:28:25,285 --> 00:28:27,985
I don't know, but they're being cagey and I don't trust them.

628
00:28:28,195 --> 00:28:30,805
Well, they also, one of the reasons they wouldn't have said they

629
00:28:30,805 --> 00:28:34,315
exclusively run on  Trainium is that both Amazon and Google have.

630
00:28:34,755 --> 00:28:37,935
Separately claim they're the exclusive cloud provider of anthropic.

631
00:28:38,025 --> 00:28:38,685
It rocks.

632
00:28:38,685 --> 00:28:39,315
I love that.

633
00:28:39,315 --> 00:28:42,075
The just lying just lies.

634
00:28:42,195 --> 00:28:45,195
It's an amazing competition though between Amazon and Google

635
00:28:45,195 --> 00:28:47,595
Over who can hit Anthropic harder with the money stick.

636
00:28:47,595 --> 00:28:50,445
I, I wanna be the, the centerpiece in that competition.

637
00:28:50,450 --> 00:28:53,660
I, I would love to get like $100 million.

638
00:28:54,000 --> 00:28:55,610
That sounds so sick.

639
00:28:55,850 --> 00:28:57,030
4 billion each.

640
00:28:57,075 --> 00:29:00,465
Uh, then, then Amazon made it eight and I'm sure it's all AWS credit.

641
00:29:00,465 --> 00:29:02,205
So out of one pocket into the other.

642
00:29:02,325 --> 00:29:03,255
Oh, no, no, no, no.

643
00:29:03,255 --> 00:29:03,675
It did.

644
00:29:03,885 --> 00:29:09,375
Amazon then converted whatever it was into some sort of new vehicle.

645
00:29:09,435 --> 00:29:11,775
So they were out actually able to book that as

646
00:29:11,775 --> 00:29:14,565
profit, like their stock in Anthropic as profit.

647
00:29:15,435 --> 00:29:16,215
Corporate bulls.

648
00:29:16,275 --> 00:29:17,805
They're better at this than Microsoft.

649
00:29:17,805 --> 00:29:18,465
Microsoft is.

650
00:29:18,675 --> 00:29:22,335
Microsoft made such a bad, they really thought OpenAI would die faster.

651
00:29:22,485 --> 00:29:25,975
But looking at the original $4 billion investment from November

652
00:29:25,975 --> 00:29:29,955
22nd, 2024, Amazon and anthropic, deepen strategic collaboration.

653
00:29:29,955 --> 00:29:34,119
Amazon names AWS , it's primary trading partner and will use

654
00:29:34,149 --> 00:29:38,019
AWS Trainium to train and deploy its largest foundation models.

655
00:29:38,289 --> 00:29:39,429
Yeah, use them too.

656
00:29:40,029 --> 00:29:42,609
Great that that does an awful lot of heavy lifting.

657
00:29:42,759 --> 00:29:43,419
Use them.

658
00:29:43,419 --> 00:29:44,019
How?

659
00:29:44,049 --> 00:29:47,439
How many of them is Amazon still buying GPUs from Nvidia?

660
00:29:47,439 --> 00:29:49,119
I didn't hear them cancel their order.

661
00:29:49,824 --> 00:29:52,044
Yeah, I, I'm no expert in this space.

662
00:29:52,044 --> 00:29:53,514
I want to be very clear on that.

663
00:29:53,514 --> 00:29:56,394
For a long time it sounded like Amazon execs would get on stage and

664
00:29:56,394 --> 00:29:59,124
recite Star Trek techno babble when it came to machine learning.

665
00:29:59,574 --> 00:30:02,214
And now, like I, I tend to recognize patterns.

666
00:30:02,334 --> 00:30:05,454
I know that when people are telling me to get in on something now while

667
00:30:05,454 --> 00:30:08,934
it's on the ground floor, they're trying to sell me something every time.

668
00:30:09,144 --> 00:30:12,474
And I also know that when someone is selling something this

669
00:30:12,474 --> 00:30:16,824
hard, it doesn't necessarily live up to a lot of those things.

670
00:30:16,854 --> 00:30:19,614
My 4-year-old daughter loves ice cream, and I don't

671
00:30:19,614 --> 00:30:22,014
find myself having to shove it down her throat.

672
00:30:22,134 --> 00:30:24,054
It's a pull rather than a push.

673
00:30:24,384 --> 00:30:27,354
Everyone's pushing this so hard, I don't

674
00:30:27,354 --> 00:30:29,424
see the outcome that they're talking about.

675
00:30:29,664 --> 00:30:34,014
What if, just, what if the, the power of AI is more useful to folks

676
00:30:34,014 --> 00:30:37,134
on an individual level than it is an enterprise corporate one?

677
00:30:37,284 --> 00:30:41,874
I think your idea that this is a $50 billion tam is directionally correct.

678
00:30:42,084 --> 00:30:42,414
Yeah.

679
00:30:42,534 --> 00:30:44,544
I'll spend 20, 30 bucks a month for this stuff.

680
00:30:44,544 --> 00:30:45,114
No problem.

681
00:30:45,114 --> 00:30:46,164
Would I spend five grand?

682
00:30:46,164 --> 00:30:46,674
No.

683
00:30:46,884 --> 00:30:47,634
So.

684
00:30:48,009 --> 00:30:50,589
There is an inherent limit to all of this until

685
00:30:50,589 --> 00:30:53,469
they un they cover some massively awesome use case.

686
00:30:53,619 --> 00:30:55,299
But they've been looking for that for three

687
00:30:55,299 --> 00:30:57,879
years now and haven't much to show for it beyond.

688
00:30:58,059 --> 00:31:01,389
It's really good at writing code badly, just like our engineers.

689
00:31:01,779 --> 00:31:04,419
And the thing is, I like to talk about the iPhone

690
00:31:05,139 --> 00:31:07,059
and I picked up the iPhone singular wireless.

691
00:31:07,059 --> 00:31:09,459
It was at visiting Penn State where it went to college.

692
00:31:10,029 --> 00:31:12,129
And um, I remember getting it and I remember

693
00:31:12,129 --> 00:31:14,709
handing it to, it was dating a dating girl.

694
00:31:14,709 --> 00:31:18,399
It was like a social studies, like I had her dad worked as a local accountant

695
00:31:18,429 --> 00:31:22,359
in central pa. So really good representation of just like normal folk.

696
00:31:22,569 --> 00:31:23,949
Every single person who touched it.

697
00:31:23,949 --> 00:31:24,309
Got it.

698
00:31:24,969 --> 00:31:26,559
They were like, holy shit, there's like a

699
00:31:26,559 --> 00:31:28,149
little computer and you can just touch it.

700
00:31:28,989 --> 00:31:29,469
Wow.

701
00:31:29,589 --> 00:31:30,939
Like I can take photos on there.

702
00:31:31,149 --> 00:31:32,649
I, my music is on here too.

703
00:31:32,769 --> 00:31:33,099
Wow.

704
00:31:33,129 --> 00:31:34,509
Your email as well.

705
00:31:34,914 --> 00:31:36,834
Do you even like, they kind of immediately, like

706
00:31:36,984 --> 00:31:39,924
the channels started forming in their head with AWS.

707
00:31:39,924 --> 00:31:41,904
People loved, oh, my fracking, God, I want

708
00:31:41,904 --> 00:31:44,934
to strangle every person AWS lost many.

709
00:31:44,934 --> 00:31:48,924
First AWS did that big fracking deal.

710
00:31:49,404 --> 00:31:50,934
There was a path to profitability.

711
00:31:51,024 --> 00:31:53,964
There was an obvious way it which this would become profitable.

712
00:31:53,964 --> 00:31:57,444
On top of that, you didn't have to explain to people why AWS was important.

713
00:31:58,284 --> 00:31:59,424
It made sense.

714
00:31:59,544 --> 00:32:00,174
It made sense.

715
00:32:00,174 --> 00:32:00,744
Immediately.

716
00:32:00,804 --> 00:32:03,564
Suddenly you could provision stuff instantly, overnight.

717
00:32:03,564 --> 00:32:06,684
You could run an experiment in your dorm room and realize, oh, this is stupid.

718
00:32:06,684 --> 00:32:09,024
No one's gonna buy it and you're out 27 cents.

719
00:32:09,024 --> 00:32:12,414
As opposed to having to buy and rack servers and wait 16 weeks because

720
00:32:12,414 --> 00:32:15,234
Dell can't get its act together and sign leases and data centers.

721
00:32:15,414 --> 00:32:16,464
That was transformative.

722
00:32:16,734 --> 00:32:18,834
And it was obviously transformative.

723
00:32:19,614 --> 00:32:23,184
It was obviously like you didn't have to, like you had to sell people.

724
00:32:23,184 --> 00:32:24,894
It got, Amazon loves to sell it, but.

725
00:32:25,464 --> 00:32:27,564
You didn't have to like con people.

726
00:32:27,564 --> 00:32:28,584
There were no people going around it.

727
00:32:28,584 --> 00:32:31,404
If you don't get, I imagine there was someone saying, if you don't get in on

728
00:32:31,404 --> 00:32:35,304
the cloud early, there's always someone, but it's, there was an obvious point.

729
00:32:35,304 --> 00:32:38,004
You didn't have to like talk like a wizard.

730
00:32:38,664 --> 00:32:43,734
I got seriously into cloud in 2015, 2016  era, very late by a lot of standards.

731
00:32:43,734 --> 00:32:45,114
I seemed to have done okay with it.

732
00:32:45,899 --> 00:32:48,474
I, I got into enterprise tech in like 2013.

733
00:32:48,594 --> 00:32:52,164
That was when I started doing PR clients and that, and it's like, guess what?

734
00:32:52,224 --> 00:32:55,134
It's like it was still cooking because there were obvious things.

735
00:32:55,134 --> 00:32:57,114
You have the mixture of on-prem and off, off-prem.

736
00:32:57,114 --> 00:32:58,794
You have all sorts of solutions within it.

737
00:32:58,884 --> 00:33:01,194
There were obvious things and obvious ways it could proliferate.

738
00:33:01,464 --> 00:33:03,474
I am a gizmos and gadgets guy.

739
00:33:04,134 --> 00:33:06,174
I work in a, I run a PR firm.

740
00:33:06,264 --> 00:33:07,914
I have to find reasons.

741
00:33:07,914 --> 00:33:09,354
Things are interesting all the time.

742
00:33:09,564 --> 00:33:11,064
I know what to look for.

743
00:33:11,484 --> 00:33:12,594
I genuinely can't hear.

744
00:33:12,669 --> 00:33:17,769
I genu, I, I have tried even the coding I have on the show, I had Carl

745
00:33:17,769 --> 00:33:20,949
Brown from Internet of Bugs who I adore, and he really explained it where

746
00:33:20,949 --> 00:33:24,669
it's like, yeah, it can do some code, but mediocre code can be dangerous.

747
00:33:24,789 --> 00:33:26,409
If you know what you're looking for, it's great.

748
00:33:26,409 --> 00:33:28,059
But do you always know what you're looking for?

749
00:33:28,059 --> 00:33:29,649
How much of this are you handing off?

750
00:33:30,159 --> 00:33:31,329
And it's like, yeah.

751
00:33:31,329 --> 00:33:34,359
It sounds like pretty much every cloud automation solution

752
00:33:34,359 --> 00:33:36,519
I've ever freaking heard of this is really useful.

753
00:33:36,519 --> 00:33:37,929
It can speed up your workflow.

754
00:33:38,139 --> 00:33:40,569
Something that would take you 10 minutes might take you two.

755
00:33:40,749 --> 00:33:41,979
You need to watch it.

756
00:33:42,369 --> 00:33:44,049
You can't trust it on its own.

757
00:33:44,139 --> 00:33:46,479
But this can automate and turn a 10 minute job

758
00:33:46,479 --> 00:33:48,279
into a two minute job, which compounds over time.

759
00:33:48,399 --> 00:33:48,909
Awesome.

760
00:33:48,969 --> 00:33:50,049
$50 billion.

761
00:33:50,079 --> 00:33:50,859
Tam Market.

762
00:33:50,859 --> 00:33:51,579
Absolutely.

763
00:33:51,849 --> 00:33:54,129
Rock and roll requires stealing from everyone.

764
00:33:54,129 --> 00:33:55,869
It's horribly horrible from the environment.

765
00:33:56,244 --> 00:33:59,783
But I even think that once this bubble pops, all that shit still exists,

766
00:33:59,963 --> 00:34:02,664
but you're not gonna see the scale and you're probably gonna see more

767
00:34:02,664 --> 00:34:08,094
money invested ever in finding ways to make inference as like nothing.

768
00:34:08,243 --> 00:34:10,314
I don't think they are even close.

769
00:34:10,314 --> 00:34:14,153
And I think because if they were, my evidence here is real simple.

770
00:34:14,394 --> 00:34:18,833
If they were close, if they were there, they would say, we've made influence.

771
00:34:18,894 --> 00:34:19,554
Influence.

772
00:34:19,554 --> 00:34:19,974
Nothing.

773
00:34:20,214 --> 00:34:21,024
Influence doesn't cut.

774
00:34:21,054 --> 00:34:22,073
They would just say it.

775
00:34:22,434 --> 00:34:24,533
They would say, we have made this profitable.

776
00:34:24,743 --> 00:34:27,024
It would be the easiest thing in the world.

777
00:34:27,024 --> 00:34:31,283
You would just say, you would just say like, it's profitable, it's good.

778
00:34:31,314 --> 00:34:35,243
It's good When they have these, when with the growth of Azure, the

779
00:34:35,243 --> 00:34:38,214
classic growth of Azure, the growth of AWS, the growth of any software

780
00:34:38,214 --> 00:34:42,474
as a service solution, you, they trumpeted in very simple terms.

781
00:34:42,474 --> 00:34:45,175
They say, Mark Benioff saunters out, slimly.

782
00:34:45,180 --> 00:34:48,054
And it's like, ah, number has gone up so many times.

783
00:34:48,649 --> 00:34:51,399
And he says the word  agentic and the industry shits itself for the next

784
00:34:51,399 --> 00:34:55,179
18 months trying to fall over themselves to get in on that buzzword now.

785
00:34:55,329 --> 00:34:57,969
So last year I sat down and I went through a bunch of

786
00:34:57,969 --> 00:35:01,479
Salesforce, um, Dreamforce conference things, the transcripts.

787
00:35:01,689 --> 00:35:06,069
He said they were in AI like five times since 2014.

788
00:35:06,069 --> 00:35:09,609
I think he has said Einstein AI was happening so many times.

789
00:35:09,788 --> 00:35:13,479
It is remarkable how bad the business press is, and it's remarkable

790
00:35:13,479 --> 00:35:19,419
how disconnected the valuations are because this is foolhardy stuff.

791
00:35:19,449 --> 00:35:21,009
These people are not smart.

792
00:35:21,639 --> 00:35:26,739
But when you look at these numbers as well, this is just,

793
00:35:26,739 --> 00:35:28,989
this is the hubris point that's always been coming for them.

794
00:35:29,229 --> 00:35:32,739
When you have just had businesses that it's kind of incredible

795
00:35:32,739 --> 00:35:35,469
how much they've grown, how big these businesses have got.

796
00:35:35,469 --> 00:35:37,869
It's kind of impressive, good or bad people that they

797
00:35:37,869 --> 00:35:40,089
are, they've been able to make remarkable businesses.

798
00:35:40,269 --> 00:35:42,129
I just think that they got it by accident.

799
00:35:42,129 --> 00:35:44,559
I don't think that they were super intentional

800
00:35:44,559 --> 00:35:46,239
with everything because if they were.

801
00:35:46,899 --> 00:35:48,489
Well, they would be doing it again.

802
00:35:49,118 --> 00:35:52,449
They would've, someone would've found, and Microsoft

803
00:35:52,449 --> 00:35:55,229
was the one that seemed to have worked it out.

804
00:35:55,349 --> 00:35:57,019
They had more even if it wasn't profit,

805
00:35:57,038 --> 00:35:58,659
they had more revenue than everyone else.

806
00:35:59,469 --> 00:36:03,069
And then you see, oh, it's, they're meeting cardboard.

807
00:36:03,159 --> 00:36:08,019
They fed themselves their own money, and there should be shareholder lawsuit.

808
00:36:08,019 --> 00:36:10,659
I'm just, it's unbelievable.

809
00:36:10,839 --> 00:36:14,319
This episode is sponsored by my own company, the

810
00:36:14,319 --> 00:36:18,069
Duck Bill Group, having trouble with your AWS bill.

811
00:36:18,159 --> 00:36:20,949
Perhaps it's time to renegotiate a contract with them.

812
00:36:21,159 --> 00:36:23,259
Maybe you're just wondering how to predict

813
00:36:23,259 --> 00:36:26,319
what's going on in the wide world of AWS.

814
00:36:26,499 --> 00:36:29,439
Well, that's where the Duck Bill group comes in to help.

815
00:36:29,859 --> 00:36:32,109
Remember, you can't duck the duck bill.

816
00:36:32,109 --> 00:36:34,209
Bill, which I am reliably informed by my

817
00:36:34,209 --> 00:36:36,909
business partner is absolutely not our motto.

818
00:36:37,449 --> 00:36:40,209
One question I do have for you on the inference economics, though,

819
00:36:40,209 --> 00:36:43,179
I, I hear what you're saying, but one of the arguments I would

820
00:36:43,179 --> 00:36:46,328
make against that would be the fact that Google has been shoving

821
00:36:46,328 --> 00:36:51,009
their dumb AI nonsense into every search result for a while now.

822
00:36:51,339 --> 00:36:54,368
And I, I have a hard time believing that they're losing money on every

823
00:36:54,368 --> 00:36:59,139
search by shoving that in that that would, that would imp well on some

824
00:36:59,139 --> 00:37:02,589
level that would show from orbit at the scale that they tend to operate at.

825
00:37:03,158 --> 00:37:04,479
And their engineers who I, some of them

826
00:37:04,479 --> 00:37:06,459
whom I trust have said that they are not.

827
00:37:06,459 --> 00:37:08,649
But I, I have a hard time seeing how that gets there

828
00:37:08,649 --> 00:37:11,259
except for the fact that TPUs sound like what Trainium and

829
00:37:11,259 --> 00:37:14,559
in wish they were as far as actually supporting things.

830
00:37:14,769 --> 00:37:17,589
I don't have any insight into the economics of it.

831
00:37:17,589 --> 00:37:21,969
I just can go by what I see and they have a stupendous amount of

832
00:37:21,969 --> 00:37:25,299
inference running all the time now for these things that no one wants.

833
00:37:26,139 --> 00:37:27,699
And it's fast inference too, because people

834
00:37:27,699 --> 00:37:29,589
aren't gonna wait five minutes for a page to load.

835
00:37:29,589 --> 00:37:30,519
A few things.

836
00:37:30,759 --> 00:37:31,269
One.

837
00:37:32,004 --> 00:37:36,114
They're not just using TPUs, they're absolutely using, uh, Nvidia chips.

838
00:37:36,114 --> 00:37:37,674
They buy absolute shit.

839
00:37:38,004 --> 00:37:40,134
They have, they did a recent thing.

840
00:37:40,134 --> 00:37:40,344
Wow.

841
00:37:40,344 --> 00:37:42,774
We can run one model on one h, 100

842
00:37:45,444 --> 00:37:46,314
well done.

843
00:37:46,703 --> 00:37:49,614
Um, doesn't mean it's, doesn't mean they're not losing money.

844
00:37:49,764 --> 00:37:53,094
And now engineer, they may have some insight,

845
00:37:53,094 --> 00:37:54,864
but how much financial insight do they have?

846
00:37:54,894 --> 00:37:58,134
If Google was, and also real simple, if Google was not

847
00:37:58,134 --> 00:37:59,964
losing money, they would say they weren't losing money.

848
00:38:00,114 --> 00:38:02,453
How would we see that they're losing money?

849
00:38:02,453 --> 00:38:04,314
Where would that be on their earnings?

850
00:38:04,374 --> 00:38:06,834
We wouldn't, 'cause they bury every, these companies.

851
00:38:07,283 --> 00:38:09,864
Amy Hood is not the only finance wizard out there.

852
00:38:09,864 --> 00:38:13,194
I forget what the CFO at Google is probably intentional there.

853
00:38:13,464 --> 00:38:15,294
But Sundar is a McKinsey guy.

854
00:38:15,924 --> 00:38:17,214
These people are evil.

855
00:38:17,964 --> 00:38:20,694
Uh, McKinsey, flesh eating bacteria of business.

856
00:38:21,894 --> 00:38:22,913
They're not gonna say it.

857
00:38:22,913 --> 00:38:25,434
But also they would just say, in instance, free.

858
00:38:25,434 --> 00:38:27,413
And would they burn billions of dollars?

859
00:38:27,684 --> 00:38:28,644
Are you asking.

860
00:38:29,109 --> 00:38:33,249
Whether, just wanna be clear, a hyperscaler would burn

861
00:38:33,249 --> 00:38:38,169
billions of dollars chasing AI because they already have.

862
00:38:38,828 --> 00:38:40,029
They already are.

863
00:38:40,089 --> 00:38:41,288
They would burn billions.

864
00:38:41,349 --> 00:38:44,859
Maybe they're even wrapping some of that loss into their capital expenditures.

865
00:38:44,919 --> 00:38:45,879
Who freaking knows?

866
00:38:45,879 --> 00:38:47,019
They don't tell us.

867
00:38:47,078 --> 00:38:48,788
But there is plenty of evidence that these

868
00:38:48,788 --> 00:38:50,618
companies are willing to lose billions.

869
00:38:50,709 --> 00:38:56,299
Google hired, they bought character AI kind of for $2.7 billion just to get Noam

870
00:38:56,299 --> 00:38:59,949
Shazeer back, who's one of the original attention is all you need paper writers.

871
00:38:59,949 --> 00:39:01,328
The transformer paper that kicked off this

872
00:39:01,328 --> 00:39:05,049
generation, Google does dumb shit all the time.

873
00:39:05,259 --> 00:39:07,149
Everyone hates their integrations.

874
00:39:07,449 --> 00:39:09,099
Yeah, I think they are losing money.

875
00:39:09,099 --> 00:39:11,559
I think they're willingly doing so and I think they'll do it again.

876
00:39:11,739 --> 00:39:16,839
I think that they have their nasty growth pig search and all of

877
00:39:16,839 --> 00:39:20,709
their monopolies allow them to finance many money losing operations.

878
00:39:21,038 --> 00:39:23,919
Do you think I, I would actually really need to look into the economics

879
00:39:23,919 --> 00:39:26,529
of Chrome, 'cause I can't imagine that that has direct revenue.

880
00:39:26,949 --> 00:39:29,349
But it's kind of like meta meta's.

881
00:39:29,349 --> 00:39:30,788
Absolutely losing billions.

882
00:39:30,788 --> 00:39:33,249
100%. And do you not think that

883
00:39:34,179 --> 00:39:37,029
Meta's very clearly thrashing and desperately trying to find the next thing.

884
00:39:37,239 --> 00:39:38,979
I love, I love what they're doing.

885
00:39:39,129 --> 00:39:39,788
I love it.

886
00:39:39,849 --> 00:39:42,699
I think it's the funniest shit in the world I've ever seen.

887
00:39:42,849 --> 00:39:45,609
It's like watching Zuckerberg's midlife crisis play out before us.

888
00:39:45,939 --> 00:39:46,839
It really is.

889
00:39:46,839 --> 00:39:49,029
He is having like a hell of a midlife crisis.

890
00:39:49,029 --> 00:39:53,199
I thought Elon Musk was having any uh uh, like a midlife crisis.

891
00:39:53,199 --> 00:39:56,649
But someone said something to me on Blue Sky about this earlier.

892
00:39:56,649 --> 00:39:56,919
Yeah.

893
00:39:57,279 --> 00:40:00,339
That meta is being the LIV golf of AI research.

894
00:40:00,459 --> 00:40:02,769
I think that that is one of the funniest ways to put it.

895
00:40:02,769 --> 00:40:06,578
'cause it really is just like this giant evil company that everyone hates being

896
00:40:06,578 --> 00:40:12,519
like, do you want more money than realistically a person can conceptualize to

897
00:40:12,519 --> 00:40:17,078
come and work with me on large language models becoming super intelligence.

898
00:40:17,259 --> 00:40:20,809
By the way, here's my chief AI scientist, his name's Yann LeCun.

899
00:40:21,038 --> 00:40:22,449
He disagrees with me.

900
00:40:22,449 --> 00:40:23,859
He does not think this is possible.

901
00:40:23,979 --> 00:40:25,449
This is your boss.

902
00:40:25,809 --> 00:40:28,809
Oh, and, and Altman was saying that Meta was offering a hundred million

903
00:40:28,809 --> 00:40:33,129
dollars signing bonuses that people were not taking for bullshit.

904
00:40:33,189 --> 00:40:35,709
That is generational wealth money to come work

905
00:40:35,709 --> 00:40:37,569
with some douche bag for a couple of years.

906
00:40:38,198 --> 00:40:38,948
Who were they offered to?

907
00:40:38,948 --> 00:40:39,609
Sam Altman?

908
00:40:39,819 --> 00:40:40,658
Two things.

909
00:40:41,049 --> 00:40:44,529
One, they have hired some people, including a guy

910
00:40:44,529 --> 00:40:46,929
with like one of the most insane names I've ever said.

911
00:40:46,929 --> 00:40:49,569
I just, I've been looking forward to saying this out loud all day.

912
00:40:49,569 --> 00:40:50,679
Lemme just get to this.

913
00:40:51,249 --> 00:40:53,229
Oh God, this name is so good too.

914
00:40:54,209 --> 00:40:57,129
His name is Trapit Bansal.

915
00:40:57,749 --> 00:40:58,689
That is quite the name.

916
00:40:58,959 --> 00:41:00,759
That's a incredible name.

917
00:41:00,879 --> 00:41:02,019
I'm called Ed Zitron.

918
00:41:02,019 --> 00:41:06,609
I have a phenomenal name, but  Trapit    Bansal might have to marry him.

919
00:41:06,999 --> 00:41:11,679
Uh, but is it, is it  Trapit   Bansal  Zitron is actually pretty anyway.

920
00:41:11,889 --> 00:41:15,489
Uh, girlfriend's gonna hate that, but there we are.

921
00:41:15,788 --> 00:41:16,509
She'll understand.

922
00:41:16,509 --> 00:41:19,569
It's for the content, but where was it?

923
00:41:19,569 --> 00:41:21,279
Yeah, these a hundred million dollar offers.

924
00:41:21,279 --> 00:41:23,828
I reckon Mr.  Bansal, I probably took one.

925
00:41:24,953 --> 00:41:27,233
Well, as Altman said, he offered to their best employees.

926
00:41:27,233 --> 00:41:30,294
So definitionally, anyone who takes it is not one of their best.

927
00:41:30,624 --> 00:41:31,134
Oh, of.

928
00:41:31,493 --> 00:41:35,063
And of course he's gonna say that, but I think it's great

929
00:41:35,063 --> 00:41:38,033
because like I said, Yann LeCun, chief AI scientist says that

930
00:41:38,033 --> 00:41:40,344
large language models will not become super intelligence.

931
00:41:40,464 --> 00:41:41,964
It's like, this is your boss.

932
00:41:42,084 --> 00:41:46,523
I've hired you to do a job that your boss doesn't like, but I am also your boss.

933
00:41:46,523 --> 00:41:48,294
And his boss have fun.

934
00:41:48,563 --> 00:41:48,924
Oh good.

935
00:41:48,924 --> 00:41:50,304
Mommy and daddy are fighting again!

936
00:41:50,393 --> 00:41:54,714
And constantly like it's like being born into a divorce.

937
00:41:54,834 --> 00:41:56,153
Like insane stuff.

938
00:41:56,153 --> 00:41:59,153
But on top of that, what are you freaking doing there?

939
00:41:59,214 --> 00:42:02,663
You, the head of your team is Alexandr Wang of scale ai.

940
00:42:02,754 --> 00:42:03,384
Not for nothing.

941
00:42:03,384 --> 00:42:05,453
If they offer me a hundred million dollars signing bonus,

942
00:42:05,453 --> 00:42:07,344
I'm doing whatever the hell they ask me to do until that

943
00:42:07,344 --> 00:42:09,504
earnout is done and then you'll never hear from me again.

944
00:42:09,594 --> 00:42:11,754
You want me to hit, hit myself in the nuts every morning?

945
00:42:11,754 --> 00:42:12,294
It'll happen.

946
00:42:12,294 --> 00:42:13,823
I can buy, I can buy new ones.

947
00:42:14,453 --> 00:42:14,844
That's right.

948
00:42:14,844 --> 00:42:15,563
I, I would learn.

949
00:42:15,624 --> 00:42:15,983
You're right.

950
00:42:15,983 --> 00:42:16,764
Hit myself in the nuts.

951
00:42:16,764 --> 00:42:18,384
Every morning I would learn node for that.

952
00:42:18,443 --> 00:42:18,684
Yeah.

953
00:42:18,684 --> 00:42:20,754
I mean, I will a hundred million dollars.

954
00:42:20,783 --> 00:42:21,384
Yeah, sure.

955
00:42:21,653 --> 00:42:22,523
But on top of that.

956
00:42:23,154 --> 00:42:25,703
The other thing, there's two other things that I think have pushed people back.

957
00:42:25,703 --> 00:42:28,104
The information had some good reporting on this as well, where

958
00:42:28,104 --> 00:42:31,314
it was, first of all, people were like the Yann LeCun thing.

959
00:42:31,314 --> 00:42:34,283
They were like, Hey, what are you doing here?

960
00:42:34,283 --> 00:42:35,663
I don't think you really understand.

961
00:42:35,694 --> 00:42:38,064
'cause Mark Zuckerberg, it's actually kind of a compelling offer.

962
00:42:38,304 --> 00:42:42,024
He was saying, I will give you a shit ton of money and also you won't

963
00:42:42,024 --> 00:42:45,953
have to worry about compute resources because OpenAI classically would,

964
00:42:46,374 --> 00:42:50,364
Sam Altman is very vengeful and he will restrict access on compute and

965
00:42:50,364 --> 00:42:53,993
resources at OpenAI if he doesn't like you being reported repeatedly.

966
00:42:54,474 --> 00:42:56,964
I think it was in Karen Howe's Empires of AI as well

967
00:42:56,964 --> 00:42:59,154
and Washington Post and Natasha Tku had some stuff.

968
00:42:59,484 --> 00:43:01,344
I think, pardon me if I got that wrong.

969
00:43:01,554 --> 00:43:04,254
No one has ever offered me all the compute resources

970
00:43:04,254 --> 00:43:06,743
I can use because they know my kind of horse shit.

971
00:43:06,834 --> 00:43:09,234
I mean, I would be working out ways to burn it.

972
00:43:09,234 --> 00:43:12,323
I would, I would just go and there'd be like a hundred million dollars.

973
00:43:12,323 --> 00:43:15,384
This is a, this is time to be the most annoying gentleman.

974
00:43:15,624 --> 00:43:16,734
But the other thing is.

975
00:43:17,124 --> 00:43:20,243
Is, I wonder if that's not a hundred million upfront.

976
00:43:20,243 --> 00:43:22,674
Maybe it's more like still $5 million.

977
00:43:22,733 --> 00:43:24,264
I would probably still do the nuts thing.

978
00:43:24,894 --> 00:43:25,134
Yeah.

979
00:43:25,163 --> 00:43:25,943
20 year earnout.

980
00:43:26,004 --> 00:43:26,214
Yeah.

981
00:43:26,453 --> 00:43:26,874
Yeah.

982
00:43:26,964 --> 00:43:29,334
But it's probably structured in a weird way.

983
00:43:29,964 --> 00:43:32,334
Oh, there's no way they're cutting checks for that.

984
00:43:32,334 --> 00:43:33,294
There's, there's just no way.

985
00:43:33,323 --> 00:43:35,453
It's probably got weird cliffs to it.

986
00:43:36,203 --> 00:43:37,193
Probably milestones.

987
00:43:37,193 --> 00:43:42,983
And on top of it even you are, and on top of that, the actual

988
00:43:42,983 --> 00:43:45,684
job sounds like it would be fracking miserable, because do you

989
00:43:45,684 --> 00:43:48,504
think Mark Zuckerberg is gonna put more or less pressure on you?

990
00:43:48,624 --> 00:43:49,044
No.

991
00:43:49,163 --> 00:43:53,544
You are the, you are the person that he's going to blame when this doesn't work.

992
00:43:53,544 --> 00:43:56,424
That is what you, I would take a hundred million dollars to do that.

993
00:43:56,663 --> 00:43:57,174
Oh, sure.

994
00:43:57,174 --> 00:43:57,953
I, I'll do it.

995
00:43:57,953 --> 00:43:59,424
And I'll just expect that I'm gonna need

996
00:43:59,424 --> 00:44:01,163
to cry to my therapist after every meeting.

997
00:44:01,163 --> 00:44:01,584
Fine.

998
00:44:01,674 --> 00:44:03,983
Oh, I, I would, I wouldn't be crying at all.

999
00:44:03,983 --> 00:44:05,214
I'd have a hundred million dollars.

1000
00:44:05,214 --> 00:44:07,943
I would go into work every day dressing.

1001
00:44:07,943 --> 00:44:09,943
However, Mark Zuckerberg last dressed.

1002
00:44:10,464 --> 00:44:12,069
And then just claim I'm copying him.

1003
00:44:12,069 --> 00:44:13,959
I would just create chaos like I would

1004
00:44:13,959 --> 00:44:16,658
jokify the entirety of Meta's AI department.

1005
00:44:16,868 --> 00:44:17,739
It would be amazing.

1006
00:44:17,979 --> 00:44:20,199
Well, you don't need a hundred million dollars to do that.

1007
00:44:20,288 --> 00:44:22,779
No, but I would, if I got that money, I would really, I

1008
00:44:22,779 --> 00:44:25,359
would have someone, my joker makeup guy every morning.

1009
00:44:25,929 --> 00:44:27,158
Just, just a guy who comes in.

1010
00:44:27,699 --> 00:44:28,029
Yeah.

1011
00:44:28,029 --> 00:44:30,158
He makes $250,000 a year.

1012
00:44:30,399 --> 00:44:32,828
He's also my driver and he also dresses like the Joker.

1013
00:44:33,129 --> 00:44:35,439
'cause Mark Zuckerberg fracking hates me by the end of it.

1014
00:44:35,618 --> 00:44:37,419
Just like he's doing the Joker.

1015
00:44:37,629 --> 00:44:37,779
Yeah.

1016
00:44:37,779 --> 00:44:40,419
I try to be just like Mark proving that I too can hire a clown.

1017
00:44:41,634 --> 00:44:43,704
He does look like a clown without his makeup on.

1018
00:44:43,824 --> 00:44:44,424
He really does.

1019
00:44:44,424 --> 00:44:47,764
It's the frizzy hair and the weird, like, like weird tan.

1020
00:44:47,844 --> 00:44:50,064
We saw a few years of business mark where he,

1021
00:44:50,064 --> 00:44:51,474
he honestly, I will give him credit, we're due.

1022
00:44:51,594 --> 00:44:55,704
He has managed to reinvent himself in a way that few people necessarily can.

1023
00:44:55,734 --> 00:44:58,524
He became, he went from the hoodie dude to wearing a suit

1024
00:44:58,524 --> 00:45:01,464
all the time, dude, to people confused him for a robot dude.

1025
00:45:01,644 --> 00:45:04,254
And now he's just weird because I, I have to imagine at

1026
00:45:04,254 --> 00:45:06,294
that tier, you don't have to have anyone around you who

1027
00:45:06,294 --> 00:45:08,484
doesn't empower you to do whatever the hell you want to do.

1028
00:45:08,604 --> 00:45:10,913
You surround yourself by nature with yes men,

1029
00:45:11,154 --> 00:45:13,464
and that is not good psychologically for anyone.

1030
00:45:13,554 --> 00:45:18,564
You realize though, that like Mark Zuckerberg can't be fired.

1031
00:45:19,703 --> 00:45:19,913
Yeah.

1032
00:45:19,974 --> 00:45:23,154
Like he can't be, he, he owns more board seats than anyone.

1033
00:45:23,154 --> 00:45:23,964
There's nothing they can do.

1034
00:45:23,964 --> 00:45:26,243
They couldn't even sue him out, so he would

1035
00:45:26,243 --> 00:45:28,224
drive this bad boy into the, into the ground.

1036
00:45:28,224 --> 00:45:34,374
But top of all of that, his last rebrand worked only for a minute, and because

1037
00:45:34,374 --> 00:45:38,033
it worked on the media, which he thought would be enough, no, he needed

1038
00:45:38,033 --> 00:45:41,274
it to work on Donald Trump, except Donald Trump doesn't like new money.

1039
00:45:41,964 --> 00:45:44,724
The actual trick for Mark Zuckerberg would've been to align with

1040
00:45:44,724 --> 00:45:50,033
some sort of elder god creature, some sort of aged finance freak.

1041
00:45:50,033 --> 00:45:52,404
Like if Sheldon Adelson was still alive.

1042
00:45:52,899 --> 00:45:54,609
Our rest in piss.

1043
00:45:54,819 --> 00:45:55,449
Oh, yes.

1044
00:45:55,509 --> 00:45:57,759
My, my personal favorite Zuckerberg trivia on this is apparently

1045
00:45:57,759 --> 00:46:00,759
he's five foot eight or so and hates it to the point where

1046
00:46:00,788 --> 00:46:04,538
he has people stage his Instagram photos, so he looks taller.

1047
00:46:04,749 --> 00:46:07,089
Uh, in Congress, someone took a picture, there was a cushion

1048
00:46:07,269 --> 00:46:10,029
on the chair, so he didn't look like he was drowning in it.

1049
00:46:10,209 --> 00:46:11,169
Now, I don't give a shit.

1050
00:46:11,199 --> 00:46:12,279
People are gonna be over tall.

1051
00:46:12,279 --> 00:46:14,889
They're going to be, but the fact that he's so wildly

1052
00:46:14,889 --> 00:46:17,679
sensitive about him for all his money, he can't buy himself

1053
00:46:17,679 --> 00:46:20,679
four inches of height is nothing short of hilarious to me.

1054
00:46:20,859 --> 00:46:24,519
I'm, I'm five nine, neither short nor tall, and now that I

1055
00:46:24,519 --> 00:46:26,828
know I'm taller than Mark Zuckerberg, I feel so powerful.

1056
00:46:27,038 --> 00:46:29,529
Also, who cares if anyone insulted by height?

1057
00:46:29,674 --> 00:46:30,024
Sorry.

1058
00:46:30,069 --> 00:46:34,209
He's five, seven and a half, apparently, which whenever its people

1059
00:46:34,209 --> 00:46:37,299
start saying a half, that's how you know they're sensitive about it.

1060
00:46:37,299 --> 00:46:39,399
Oh, I hear Sam Altman's the same.

1061
00:46:39,819 --> 00:46:42,069
I, I've heard, I don't know his true height.

1062
00:46:42,158 --> 00:46:45,189
If Sam Altman's like five foot one, that would be that.

1063
00:46:45,189 --> 00:46:46,658
That's what I'm gonna start saying.

1064
00:46:47,078 --> 00:46:49,899
I realize we've kind of got off of topic, but the

1065
00:46:49,899 --> 00:46:53,889
fact is this does tie back, which is we're discussing

1066
00:46:53,889 --> 00:46:56,288
people just spending billions of dollars for no reason.

1067
00:46:56,948 --> 00:46:57,988
It's vanity projects

1068
00:46:58,389 --> 00:47:01,989
But it's vanity projects that have absorbed the entirety of the economy.

1069
00:47:02,259 --> 00:47:05,649
The thing that I scare people with, no one likes hearing this, so enjoy.

1070
00:47:06,249 --> 00:47:11,919
So the magnificent seven stocks, they are 35% of the economy, right?

1071
00:47:11,948 --> 00:47:15,339
The US economy, 19% of those stocks is Nvidia

1072
00:47:15,429 --> 00:47:17,649
NVIDIA's continued growth is the only thing.

1073
00:47:17,799 --> 00:47:20,529
Continuing NVIDIA's valuations.

1074
00:47:20,618 --> 00:47:24,639
How does Nvidia grow by the rest of the magnificent seven buying GPUs from them?

1075
00:47:25,749 --> 00:47:27,399
What happens if that changes?

1076
00:47:28,528 --> 00:47:29,453
Well think about this.

1077
00:47:29,453 --> 00:47:32,064
I think that the same people, you ever notice how a lot of

1078
00:47:32,064 --> 00:47:35,754
the AI folks who are the big, the most bullish folks you'll

1079
00:47:35,754 --> 00:47:39,113
meet, were also deep in the wool crypto bros. It's like, what?

1080
00:47:39,113 --> 00:47:39,384
What?

1081
00:47:39,384 --> 00:47:40,764
Why do these people all pivot?

1082
00:47:41,064 --> 00:47:43,854
Because they're Nvidia Street team all, as long

1083
00:47:43,854 --> 00:47:46,344
as they're using the GPUs, that's what it is.

1084
00:47:46,794 --> 00:47:49,224
And if this falls through, we'll go back to using 'em for gaming.

1085
00:47:49,493 --> 00:47:51,684
I, it's not the same like that.

1086
00:47:51,743 --> 00:47:53,004
I will push back on that one.

1087
00:47:53,453 --> 00:47:55,344
You're 50 to 80%, right?

1088
00:47:55,344 --> 00:47:55,794
And the, yeah.

1089
00:47:55,794 --> 00:47:58,044
A ton of people who are like, I'm in crypto, I'm

1090
00:47:58,044 --> 00:47:59,964
in AI Just 'cause they were never in anything.

1091
00:48:00,504 --> 00:48:01,434
They're just hype chasing.

1092
00:48:01,434 --> 00:48:01,554
Yeah.

1093
00:48:01,554 --> 00:48:04,733
It turns out that by being an influencer with no actual expertise

1094
00:48:04,733 --> 00:48:07,613
in a thing means you can retool and pivot super quickly.

1095
00:48:07,644 --> 00:48:08,903
'Cause you don't have to spend that time

1096
00:48:08,903 --> 00:48:11,033
gathering this pesky thing called expertise.

1097
00:48:11,064 --> 00:48:12,264
It's also sold in the same way.

1098
00:48:12,384 --> 00:48:13,974
It's, I have a specious idea.

1099
00:48:13,974 --> 00:48:16,524
I've connected nascent tech to You don't understand to it.

1100
00:48:16,524 --> 00:48:17,033
Wow.

1101
00:48:17,214 --> 00:48:18,653
And getting quick before it's too late.

1102
00:48:18,743 --> 00:48:19,374
Exactly.

1103
00:48:19,403 --> 00:48:20,363
Same kind of pressure.

1104
00:48:20,724 --> 00:48:23,754
The only thing is, is that the GPU side is just wholly different.

1105
00:48:23,754 --> 00:48:28,999
And I mean the rack mounted insanity of a, oh god, what is it?

1106
00:48:29,064 --> 00:48:31,193
GB 200 or whatever the Blackwell.

1107
00:48:31,704 --> 00:48:35,514
Things are, they're all these rack mounted, 3000 pound monstrosities.

1108
00:48:35,514 --> 00:48:36,294
They're horrifying.

1109
00:48:36,444 --> 00:48:37,314
They are the rack.

1110
00:48:37,524 --> 00:48:41,124
They, they really, and they have cooling issues because science

1111
00:48:41,184 --> 00:48:45,984
exists, but they're, these people are doing it because they're

1112
00:48:45,984 --> 00:48:48,894
too stupid to realize that what they're in is kind of a scam.

1113
00:48:49,044 --> 00:48:51,504
I think that they're scamming in the sense that they're selling something

1114
00:48:51,504 --> 00:48:54,354
they don't really understand, but they're not like, I think this will die.

1115
00:48:54,594 --> 00:48:56,454
I think that they think this will go forever.

1116
00:48:56,514 --> 00:48:57,654
Kind of like crypto.

1117
00:48:57,834 --> 00:49:00,413
And when I, this is the other thing, people with my

1118
00:49:00,413 --> 00:49:02,244
work, they're like, oh, you started like a year ago.

1119
00:49:03,204 --> 00:49:04,314
I was writing in 2020.

1120
00:49:04,404 --> 00:49:06,684
I was on crypto before everyone.

1121
00:49:07,104 --> 00:49:07,464
Yeah.

1122
00:49:07,524 --> 00:49:10,163
I, I found that crypto has been looking for a business

1123
00:49:10,163 --> 00:49:14,784
model that is not fraud or fraud adjacent, uh, for 15 years.

1124
00:49:14,784 --> 00:49:16,314
At this point, it doesn't seem to have one.

1125
00:49:16,464 --> 00:49:18,384
It's, uh, liquidity for venture capital.

1126
00:49:18,413 --> 00:49:19,764
That's what it actually is.

1127
00:49:19,884 --> 00:49:21,174
At, at least ai.

1128
00:49:21,174 --> 00:49:22,044
I could point at that.

1129
00:49:22,044 --> 00:49:24,654
I could make a robot say something funny very quickly.

1130
00:49:25,148 --> 00:49:25,688
Okay.

1131
00:49:25,688 --> 00:49:26,648
Is there value to that?

1132
00:49:26,648 --> 00:49:27,219
I don't know.

1133
00:49:27,219 --> 00:49:29,108
Not a trillion dollars worth, but yeah.

1134
00:49:29,108 --> 00:49:29,379
Okay.

1135
00:49:29,379 --> 00:49:30,759
It amuses me for 10 minutes

1136
00:49:30,908 --> 00:49:34,118
and I have the grand business idiot theory, which is that the people

1137
00:49:34,118 --> 00:49:38,648
in power who have the money and the power and the hands of the media,

1138
00:49:38,799 --> 00:49:42,009
they don't understand much of what's going on they never did before.

1139
00:49:42,249 --> 00:49:46,118
And AI is this kind of mystical thing where it kind of has value, it kind

1140
00:49:46,118 --> 00:49:49,479
of has something you can point to and said outcomes you can point to and

1141
00:49:49,479 --> 00:49:54,009
go, well, to extrapolate from what this does now, it will do this next year.

1142
00:49:54,009 --> 00:49:54,608
Why?

1143
00:49:55,538 --> 00:49:56,438
I don't know.

1144
00:49:56,469 --> 00:49:58,868
But these people have a lot of money and this

1145
00:49:58,868 --> 00:50:01,148
much money can't be wrong this much money.

1146
00:50:01,148 --> 00:50:03,339
Ha ha ha.

1147
00:50:03,639 --> 00:50:03,999
Damn.

1148
00:50:03,999 --> 00:50:05,078
Can it be freaking wrong?

1149
00:50:05,408 --> 00:50:05,528
Yeah.

1150
00:50:05,528 --> 00:50:07,749
But that's bet's Howard's big enough that when

1151
00:50:07,749 --> 00:50:09,759
it falls, it takes a lot of other people out.

1152
00:50:09,999 --> 00:50:13,209
And I mean, if, what happens when people don't buy GPUs?

1153
00:50:13,209 --> 00:50:14,139
'cause here's the other thing.

1154
00:50:14,528 --> 00:50:18,339
Whether or not you agree with me on ai, maybe you want to think about

1155
00:50:18,339 --> 00:50:22,749
something which is NVIDIA's, uh, data center revenue in the last quarter.

1156
00:50:22,749 --> 00:50:24,219
So the one where GPUs live.

1157
00:50:24,249 --> 00:50:27,069
It was, uh, $39.1 billion.

1158
00:50:27,069 --> 00:50:28,059
Pretty big, right?

1159
00:50:28,299 --> 00:50:31,389
Except they missed estimates by a couple hundred million,

1160
00:50:31,538 --> 00:50:35,229
which, you know, that's one I, it happened to me yesterday.

1161
00:50:35,379 --> 00:50:36,009
I knew a guy.

1162
00:50:36,459 --> 00:50:38,349
Um, here's the problem.

1163
00:50:38,948 --> 00:50:43,538
That number is the single most economic important economic indicator right now.

1164
00:50:44,559 --> 00:50:48,669
That number is, I think, 80 something percent of NVIDIA's revenue.

1165
00:50:48,759 --> 00:50:51,309
It is the reason that Nvidia grows every quarter.

1166
00:50:51,549 --> 00:50:55,899
If that number does not continue growing, because for me to be wrong,

1167
00:50:56,078 --> 00:50:59,559
for this to keep going, everyone needs to keep buying Nvidia GPUs.

1168
00:50:59,559 --> 00:51:02,679
Nvidia will need to, like in a, in a year or so, they'll need

1169
00:51:02,679 --> 00:51:06,729
to sell 60 to a hundred billion dollars of GPUs a quarter.

1170
00:51:06,879 --> 00:51:10,599
And then after that, they will need to sell 80 to $150 billion a

1171
00:51:10,599 --> 00:51:13,868
quarter, and then they will need to sell a hundred to 180 billion.

1172
00:51:14,229 --> 00:51:16,569
They will need to keep growing exponentially.

1173
00:51:16,569 --> 00:51:21,219
The rate that the market is requiring Nvidia to grow is too high.

1174
00:51:21,894 --> 00:51:24,983
At some point that growth has to run out because we

1175
00:51:24,983 --> 00:51:28,344
don't have the physical space for all of these GPUs.

1176
00:51:28,554 --> 00:51:30,743
And at some point, if we don't have the returns

1177
00:51:30,743 --> 00:51:32,934
coming out of them, in fact, we're losing money.

1178
00:51:33,474 --> 00:51:35,304
Maybe it's time to not buy them.

1179
00:51:35,573 --> 00:51:37,403
What if Google's TPUs help?

1180
00:51:38,153 --> 00:51:40,794
What if go Google actually can move to TPUs?

1181
00:51:40,913 --> 00:51:41,934
What would happen then?

1182
00:51:42,354 --> 00:51:44,754
What if we find out that the training paradigm we're

1183
00:51:44,754 --> 00:51:47,363
under does not work, which we're already finding out?

1184
00:51:47,604 --> 00:51:50,663
Okay, that means we don't need to buy all of the GPUs because the GPUs

1185
00:51:50,663 --> 00:51:54,863
are really most useful for having a lot of them allows you to train.

1186
00:51:55,854 --> 00:51:58,794
If training isn't helpful, why would you need that?

1187
00:51:58,794 --> 00:52:02,094
Because inference is not as demanding on a GPU.

1188
00:52:03,113 --> 00:52:06,924
So it's like what do we, what do we do here?

1189
00:52:06,953 --> 00:52:08,033
What are we doing?

1190
00:52:08,033 --> 00:52:08,874
There's not much money.

1191
00:52:08,874 --> 00:52:12,294
There's a lot of expenses and I think the.

1192
00:52:13,284 --> 00:52:15,699
It is funny that I'm on this podcast 'cause I think it's probably

1193
00:52:15,699 --> 00:52:20,288
the most germane to my art, the actual way for AI to be sticky.

1194
00:52:20,288 --> 00:52:23,529
The actual way for AI to have made it was the enterprise.

1195
00:52:23,589 --> 00:52:24,759
It was never consumer.

1196
00:52:25,364 --> 00:52:27,639
It, it was going to be API access.

1197
00:52:27,729 --> 00:52:28,599
That's where the money is.

1198
00:52:28,719 --> 00:52:33,578
It is it it, but it was gonna be selling big seat SaaS

1199
00:52:33,578 --> 00:52:38,859
contracts, basically combining the way it should have worked.

1200
00:52:38,859 --> 00:52:41,199
If OpenAI would've been a functional business, would've been, it

1201
00:52:41,199 --> 00:52:45,639
would've looked a little like Microsoft 365 plus a, uh, Azure.

1202
00:52:45,939 --> 00:52:48,399
It would've been, we make a shit ton of money from selling

1203
00:52:48,489 --> 00:52:51,369
enterprise subscriptions, not consumer, but enterprise.

1204
00:52:51,699 --> 00:52:54,099
And we will make a shit ton of money on API.

1205
00:52:54,519 --> 00:52:57,578
They make a decent amount of money on both, because everyone

1206
00:52:57,578 --> 00:53:00,639
talks about AI and everyone mentions open AI every time

1207
00:53:00,879 --> 00:53:02,379
They won the branding marketing war.

1208
00:53:02,619 --> 00:53:03,339
Exactly.

1209
00:53:03,983 --> 00:53:07,733
And on top of that though, they're having, there's multiple stories, journal

1210
00:53:07,733 --> 00:53:10,644
and the information this week that say they're having trouble, they're having

1211
00:53:10,644 --> 00:53:14,033
to hit, actually you'll know this one, no one else would agree with me on this.

1212
00:53:14,033 --> 00:53:16,434
They're like, yeah, they're changing from a

1213
00:53:16,554 --> 00:53:19,344
subscription based model to a usage based model.

1214
00:53:19,584 --> 00:53:21,533
And I'm like, that's a bad sign.

1215
00:53:21,594 --> 00:53:24,684
That's a, if you know anything about SaaS, that's a bad sign.

1216
00:53:25,044 --> 00:53:27,983
That means that they're, they do not have product market fit.

1217
00:53:28,644 --> 00:53:30,894
OpenAI, the biggest name in the fracking game

1218
00:53:30,953 --> 00:53:33,294
does not have enterprise product market fit.

1219
00:53:34,044 --> 00:53:35,063
They don't have it.

1220
00:53:35,273 --> 00:53:36,474
That is doom.

1221
00:53:36,533 --> 00:53:38,514
That is certain doom.

1222
00:53:38,993 --> 00:53:42,413
If you, do you think a usage based model is gonna No, it'll probably

1223
00:53:42,413 --> 00:53:46,733
bring down, it will hopefully allow you to sell to scale more, but

1224
00:53:46,733 --> 00:53:49,644
it's not like it's gonna increase the amount that people are spending.

1225
00:53:49,794 --> 00:53:51,353
'cause if people wanted to increase the amount

1226
00:53:51,353 --> 00:53:53,334
of the spend, they use the subscription.

1227
00:53:54,189 --> 00:53:55,359
They'd use the subscription.

1228
00:53:55,899 --> 00:53:56,318
Right.

1229
00:53:57,129 --> 00:53:58,959
Uh, the problem too is you have a psychological issue,

1230
00:53:58,988 --> 00:54:01,059
especially in the consumer land when you do that, where people

1231
00:54:01,059 --> 00:54:03,399
feel that when everything you do is metered, then every time

1232
00:54:03,399 --> 00:54:05,408
you use it for anything, it feels like an investment decision.

1233
00:54:05,589 --> 00:54:07,719
Well, cons, they're not doing it for consumers,

1234
00:54:08,198 --> 00:54:09,639
they're just doing it for the enterprise.

1235
00:54:09,729 --> 00:54:10,389
Mm.

1236
00:54:10,568 --> 00:54:12,219
And they're having to, they're also having to

1237
00:54:12,219 --> 00:54:14,679
already deep discount to compete with Microsoft.

1238
00:54:16,359 --> 00:54:20,408
It's like these are on a SaaS and enterprise sales level.

1239
00:54:20,499 --> 00:54:25,988
These are terrifying stories, but most people, it's just really bad.

1240
00:54:26,379 --> 00:54:31,658
I built a dumb blue sky bot, uh, the, uh, AWS snark bot where it effectively

1241
00:54:31,658 --> 00:54:34,389
just tracks the, the news that it comes outta their, what's new feed,

1242
00:54:34,419 --> 00:54:37,929
summarizes it with inference and then puts it out there and that's it.

1243
00:54:38,198 --> 00:54:40,778
It's, I, it's, I haven't yet made it to anything funny on it, but

1244
00:54:40,778 --> 00:54:44,019
that costs me something like a dollar a month to run that thing

1245
00:54:44,019 --> 00:54:46,929
for the many hundreds of things that come across that RSS feed.

1246
00:54:47,379 --> 00:54:49,028
And it's okay.

1247
00:54:49,028 --> 00:54:51,549
That's an interesting toy, but it costs me nothing.

1248
00:54:51,939 --> 00:54:55,029
And if it started costing me something, I probably wouldn't do it.

1249
00:54:55,209 --> 00:54:56,974
And that is, it.

1250
00:54:56,979 --> 00:55:00,099
It, it fuels a bunch of fun, interesting toys like that.

1251
00:55:00,399 --> 00:55:02,559
But this does, I don't see that I could build a business

1252
00:55:02,559 --> 00:55:04,839
around that and I wouldn't be bold enough to try.

1253
00:55:04,989 --> 00:55:07,929
And even then, right now we're at the height of the mania.

1254
00:55:08,529 --> 00:55:10,989
If this was going to be the world's stickiest

1255
00:55:10,989 --> 00:55:14,229
business, there would be metrics that showed that.

1256
00:55:14,754 --> 00:55:16,119
It is very simple.

1257
00:55:16,119 --> 00:55:18,009
However anyone feels about ai.

1258
00:55:18,578 --> 00:55:20,769
Where is the actual business?

1259
00:55:20,769 --> 00:55:27,099
Because Perplexity, for example, uh, rag Search, their whole thing is they

1260
00:55:27,129 --> 00:55:31,389
re, I think they gave away $35 million last year, like discounts and refunds.

1261
00:55:31,719 --> 00:55:33,489
They're a massively lossy business.

1262
00:55:33,489 --> 00:55:34,059
It's just.

1263
00:55:34,403 --> 00:55:37,254
Instead of a three hour keynote where Randy Chassis gets up there and

1264
00:55:37,254 --> 00:55:40,014
talks about how great AI is, he would basically get on the earnings

1265
00:55:40,014 --> 00:55:44,903
call and like, okay, this past quarter we've posted 300% annual growth.

1266
00:55:45,084 --> 00:55:47,903
Uh, I will now be taking no questions.

1267
00:55:47,934 --> 00:55:48,983
Have a nice day.

1268
00:55:49,104 --> 00:55:51,384
Like that's what success story looks like.

1269
00:55:51,533 --> 00:55:53,004
And they would be kissing his ass.

1270
00:55:53,033 --> 00:55:55,554
They would be saying, Chassis, they, any of these ones,

1271
00:55:55,554 --> 00:55:57,863
they would be like, Satya Nadella has changed the world.

1272
00:55:57,863 --> 00:56:00,413
If this w this is the greatest investment ever.

1273
00:56:00,413 --> 00:56:04,613
Because he would say, AI revenue, blah, blah, profit, blah, blah.

1274
00:56:04,674 --> 00:56:05,519
He'd be like, beep, beep.

1275
00:56:05,618 --> 00:56:08,964
It would be, I don't need like, it's like Jensen Wong, Jensen Wong.

1276
00:56:09,424 --> 00:56:12,484
Goes up there and just like, sounds like he has a stroke.

1277
00:56:12,514 --> 00:56:14,554
He just goes and is like, yeah, we will have a computer.

1278
00:56:14,554 --> 00:56:17,163
They'll control AI with a computer that goes like this.

1279
00:56:17,613 --> 00:56:20,014
He will scream at someone that works for him and then he leaves

1280
00:56:20,014 --> 00:56:24,004
and everyone's like, yay, I love you Jensen sign my baby.

1281
00:56:24,064 --> 00:56:26,163
Well, for a year he was the one, he was

1282
00:56:26,163 --> 00:56:28,354
with doled out who got the GPU allocations.

1283
00:56:28,354 --> 00:56:31,024
He just made an entire year's study out of just, I'm

1284
00:56:31,024 --> 00:56:33,453
gonna go and speak at other people's, uh, keynotes.

1285
00:56:33,604 --> 00:56:34,504
Uh, why?

1286
00:56:34,533 --> 00:56:35,974
Because I choose to do so.

1287
00:56:36,004 --> 00:56:36,904
What are you gonna talk about?

1288
00:56:36,964 --> 00:56:38,554
Whatever I damn well, please.

1289
00:56:38,554 --> 00:56:39,573
Because what are they gonna do?

1290
00:56:39,573 --> 00:56:40,564
Risk of setting me?

1291
00:56:40,774 --> 00:56:42,424
And that's because he had a good business.

1292
00:56:42,424 --> 00:56:44,044
I think he sounds like a huge asshole.

1293
00:56:44,134 --> 00:56:46,624
Oh, but you see a gold rush sell a pickaxe.

1294
00:56:46,624 --> 00:56:47,194
My God.

1295
00:56:47,254 --> 00:56:50,944
But also, but also he, he performed scoreboard.

1296
00:56:51,573 --> 00:56:53,044
Like that's the thing, scoreboard.

1297
00:56:53,194 --> 00:56:56,584
Really like whether or not you like Nvidia, they got the fracking numbers.

1298
00:56:56,674 --> 00:56:57,124
Yeah.

1299
00:56:57,334 --> 00:56:58,863
Why doesn't anyone else?

1300
00:56:58,984 --> 00:57:01,203
In fact, here's a real weird question.

1301
00:57:01,474 --> 00:57:03,964
Why is it that everyone who buys GPUs from NVIDIA

1302
00:57:03,964 --> 00:57:06,184
installs them and then immediately starts losing money?

1303
00:57:07,419 --> 00:57:10,029
Is that not a li This is another thing that worries me.

1304
00:57:10,029 --> 00:57:14,439
It's like, not only is our economy built on buying GPUs, when you install

1305
00:57:14,439 --> 00:57:18,879
them, you start losing money like immediately because the costs are too high.

1306
00:57:19,059 --> 00:57:22,658
This feels like a corollary of something that I said the client might observed

1307
00:57:22,658 --> 00:57:25,779
years ago, and they're right, which is your AWS bill is not a function of

1308
00:57:25,779 --> 00:57:28,658
how many customers you have, but rather how many engineers you've hired.

1309
00:57:30,069 --> 00:57:30,549
Okay.

1310
00:57:30,759 --> 00:57:32,319
They're not entirely wrong.

1311
00:57:32,589 --> 00:57:34,899
Like once, once you have a very expensive hammer,

1312
00:57:35,139 --> 00:57:37,328
every problem starts to look like your thumb.

1313
00:57:37,569 --> 00:57:39,399
And that's the thing, the automation thing.

1314
00:57:39,399 --> 00:57:42,279
Software engineers are always gonna try and automate stuff.

1315
00:57:42,309 --> 00:57:44,559
Like it's just, that's what, that's what software.

1316
00:57:44,559 --> 00:57:45,639
They love that shit.

1317
00:57:45,819 --> 00:57:49,479
Like the containerization fest where everyone's like, we got containers now.

1318
00:57:49,479 --> 00:57:49,809
Why?

1319
00:57:49,868 --> 00:57:50,019
Ah.

1320
00:57:51,114 --> 00:57:55,374
I, as someone screamed Docker at me on the street and I'm just freaking out.

1321
00:57:55,493 --> 00:57:56,724
I need to install Docker.

1322
00:57:56,724 --> 00:57:57,234
What is it?

1323
00:57:57,234 --> 00:57:58,284
Do you work that out?

1324
00:57:58,434 --> 00:57:59,453
I need Docker now.

1325
00:57:59,663 --> 00:58:02,574
But even that did not work out so great for Docker because

1326
00:58:02,574 --> 00:58:04,794
it turns out that when people don't really know why they're

1327
00:58:04,794 --> 00:58:07,074
buying your shit, eventually, they go, why am I buying this?

1328
00:58:07,434 --> 00:58:09,084
They, they sold their enterprise business off.

1329
00:58:09,084 --> 00:58:13,044
My comment, the Morant, and my comment was to congratulate them on,

1330
00:58:13,044 --> 00:58:15,413
uh, getting rid of that pesky part of their business that made money.

1331
00:58:15,474 --> 00:58:15,924
Yes.

1332
00:58:15,924 --> 00:58:17,453
Uh, good work, everyone.

1333
00:58:18,084 --> 00:58:20,993
It's just, it's such a fantastical time.

1334
00:58:20,993 --> 00:58:24,924
And what's f what I'm finding is the conversations I have are more

1335
00:58:24,924 --> 00:58:28,614
like this one where people are just like, yeah, that is pretty weird.

1336
00:58:29,304 --> 00:58:30,114
What are we gonna do?

1337
00:58:30,203 --> 00:58:31,074
I don't know.

1338
00:58:31,134 --> 00:58:32,334
We'll watch and learn.

1339
00:58:32,484 --> 00:58:34,284
Oh, there's no next step for me.

1340
00:58:34,284 --> 00:58:37,734
I, I keep poking at the bear, but it turns out no one asked my

1341
00:58:37,734 --> 00:58:40,794
opinion before investing a hundred billion dollars into bullshit.

1342
00:58:40,974 --> 00:58:45,413
Well, that's the, my favorite thing here though, is no one, I mean,

1343
00:58:45,413 --> 00:58:48,984
people will mention it, but no one really wants to talk about the open ai.

1344
00:58:49,479 --> 00:58:50,169
Revenues.

1345
00:58:50,288 --> 00:58:51,879
No one wants to talk about the costs.

1346
00:58:52,029 --> 00:58:54,549
They will glance at 'em and be like, oh, they lost

1347
00:58:54,549 --> 00:58:57,729
$5 billion in a larger paragraph about some staff.

1348
00:58:57,819 --> 00:59:01,419
Well, I don't believe they have any formal reporting requirements, do they?

1349
00:59:01,749 --> 00:59:02,169
No.

1350
00:59:02,169 --> 00:59:03,279
They're private company.

1351
00:59:03,309 --> 00:59:04,538
I'm just saying the,

1352
00:59:04,779 --> 00:59:07,719
I'm trying to remember what the nonprofit reporting, uh, rules were around them.

1353
00:59:07,868 --> 00:59:08,439
Uh, no.

1354
00:59:08,439 --> 00:59:10,509
There's, they, no, they don't have to because

1355
00:59:10,509 --> 00:59:12,368
the for-profit entity is a separate wing.

1356
00:59:12,519 --> 00:59:13,689
Oh, that's, that's right.

1357
00:59:13,689 --> 00:59:19,149
But I mean, journalists, I mean, analysts, they don't want to touch this.

1358
00:59:19,328 --> 00:59:21,368
I, there are other people that have talked about these

1359
00:59:21,368 --> 00:59:23,679
numbers in passing, but no one else has sat down and gone,

1360
00:59:24,038 --> 00:59:26,649
Hey, look, I'm writing something properly like today.

1361
00:59:26,649 --> 00:59:29,049
That's called, why did Microsoft Invest in Open ai?

1362
00:59:29,559 --> 00:59:32,859
I've written open AI is a systemic risk to the tech industry.

1363
00:59:33,009 --> 00:59:34,448
Open AI is a bad business.

1364
00:59:34,448 --> 00:59:36,459
And what, how does OpenAI survive?

1365
00:59:36,549 --> 00:59:39,219
That headline feels like something Satya Nadella as saying to himself,

1366
00:59:39,219 --> 00:59:42,669
his head in his hand at his desks, like, why did Microsoft invest in this?

1367
00:59:42,908 --> 00:59:44,259
I'm asking it because.

1368
00:59:44,634 --> 00:59:46,794
The deal suggests many different things could be the

1369
00:59:46,794 --> 00:59:49,374
reason, such as maybe they invested hoping they'd die.

1370
00:59:49,884 --> 00:59:51,864
But no one wants to talk about this stuff.

1371
00:59:52,163 --> 00:59:56,514
And if I'm right and I believe I am, I very much, very much do.

1372
00:59:57,743 --> 00:59:59,993
I think this is one of the greatest failings

1373
00:59:59,993 --> 01:00:02,304
of financial journalism of all time.

1374
01:00:02,394 --> 01:00:05,934
I think it's only beaten by the great financial crisis.

1375
01:00:05,934 --> 01:00:07,674
I think the great financial crisis would be

1376
01:00:07,674 --> 01:00:09,654
much worse than what what's gonna happen here.

1377
01:00:10,134 --> 01:00:11,214
But it is.

1378
01:00:11,364 --> 01:00:13,854
And even then, I think had the great financial crisis happened

1379
01:00:13,854 --> 01:00:16,884
today, people would've noticed because you would, you had people

1380
01:00:16,884 --> 01:00:18,594
with the internet who could go and look at these things like

1381
01:00:18,594 --> 01:00:21,144
Michael Borough and is like, they had privileged information.

1382
01:00:21,324 --> 01:00:26,634
The proliferation of available internet services was large but not as large.

1383
01:00:26,634 --> 01:00:28,884
And you didn't have the proliferation of social

1384
01:00:28,884 --> 01:00:30,774
media at the scale that you didn't have Twitter.

1385
01:00:30,804 --> 01:00:33,714
We had Twitter was out, but it wasn't Point I'm making is

1386
01:00:35,064 --> 01:00:39,354
you have the biggest company in the ai, open ai, the most

1387
01:00:39,354 --> 01:00:43,074
well-known one, funded by the next biggest company in ai.

1388
01:00:43,449 --> 01:00:46,599
Microsoft, Microsoft's biggest customer is open

1389
01:00:46,599 --> 01:00:50,769
ai and the core finances of open AI are bad.

1390
01:00:50,769 --> 01:00:51,159
Bad.

1391
01:00:51,339 --> 01:00:52,599
They are so bad.

1392
01:00:52,659 --> 01:00:53,859
They are so awful.

1393
01:00:53,979 --> 01:00:58,149
This company burned, I think they spent $9 billion to lose $5 billion last year.

1394
01:00:58,509 --> 01:01:01,359
And this is something that modern journalists do not want to talk about.

1395
01:01:01,359 --> 01:01:03,339
They don't wanna write about it, they don't wanna read about

1396
01:01:03,339 --> 01:01:06,159
it, don't wanna hear about it in passing conversations.

1397
01:01:06,159 --> 01:01:09,369
They all say the same thing, which is, yeah, I just don't wanna think about it.

1398
01:01:09,489 --> 01:01:10,869
Yeah, that's not, that's uncomfortable if

1399
01:01:10,869 --> 01:01:12,279
I have to actually deal with any of it.

1400
01:01:12,459 --> 01:01:16,659
But it's, the problem is, is that, and I understand why they

1401
01:01:16,659 --> 01:01:19,569
don't wanna do it, because it starts justifying you in real time.

1402
01:01:20,109 --> 01:01:24,819
You start thinking, okay, well this is the biggest company in the industry

1403
01:01:24,819 --> 01:01:29,649
and they're burning billions of dollars and Microsoft needs them to

1404
01:01:29,649 --> 01:01:32,859
live, but Microsoft owns all their ip and this company can't go public

1405
01:01:32,859 --> 01:01:36,009
unless they become a, not a for-profit, which is in Microsoft's hands.

1406
01:01:36,249 --> 01:01:39,759
Even if they do though SoftBank has to keep funding them forever.

1407
01:01:40,299 --> 01:01:44,078
This company says they're gonna burn, what, 40, $50 billion?

1408
01:01:44,078 --> 01:01:46,509
By 2030, I think it's gonna be more.

1409
01:01:46,959 --> 01:01:47,889
Where are they getting that money?

1410
01:01:47,889 --> 01:01:48,879
'cause they lose money.

1411
01:01:48,879 --> 01:01:49,959
They lose money all the time.

1412
01:01:49,959 --> 01:01:51,009
They're constantly losing money.

1413
01:01:51,339 --> 01:01:52,689
And then when you think all those things you

1414
01:01:52,689 --> 01:01:57,069
go, is the entire AI industry built on sand?

1415
01:01:57,189 --> 01:01:58,839
And the answer is no.

1416
01:01:59,169 --> 01:02:01,389
The actual AI industry is built on dreams.

1417
01:02:02,319 --> 01:02:04,149
It's, it's on unicorn farts.

1418
01:02:04,479 --> 01:02:05,049
It is.

1419
01:02:05,229 --> 01:02:07,719
The only way for you to rationally look at this and

1420
01:02:07,719 --> 01:02:10,959
believe that it will keep going is fantasy Fanta?

1421
01:02:10,959 --> 01:02:12,009
I was gonna say fantastical.

1422
01:02:12,009 --> 01:02:12,459
It's fan.

1423
01:02:12,639 --> 01:02:13,719
Fantastical thinking.

1424
01:02:13,719 --> 01:02:14,379
Magical thinking.

1425
01:02:14,379 --> 01:02:14,889
There we go.

1426
01:02:15,609 --> 01:02:16,839
Could have delivered that point better.

1427
01:02:16,839 --> 01:02:20,528
But nevertheless, less you have to start engaging in fantasy.

1428
01:02:21,189 --> 01:02:21,459
You do.

1429
01:02:21,698 --> 01:02:24,908
That's the only, it's the only, it is an irrational approach to

1430
01:02:24,908 --> 01:02:28,269
this, to approach this from the perspective that OpenAI will survive.

1431
01:02:28,314 --> 01:02:33,444
Yeah, because SoftBank had to borrow $15 billion in a one year loan, a

1432
01:02:33,444 --> 01:02:39,324
bridge loan from, and they took 21 banks, 21 to get that $15 billion.

1433
01:02:39,654 --> 01:02:43,344
Note that SoftBank has also promised $19 billion to the Stargate Project.

1434
01:02:43,584 --> 01:02:46,224
The Stargate project is yet to incorporate, but Oracle is

1435
01:02:46,224 --> 01:02:49,044
already agreeing to put $40 billion of chips inside it.

1436
01:02:49,614 --> 01:02:51,533
They are being built by a company called Cruso

1437
01:02:51,533 --> 01:02:53,184
that has never built a data center before.

1438
01:02:53,844 --> 01:02:55,134
How hard could it possibly be?

1439
01:02:55,224 --> 01:02:55,913
But that's the thing.

1440
01:02:55,913 --> 01:03:02,724
It's like, it, it reminds me of the, one of the best threats of all time.

1441
01:03:02,724 --> 01:03:06,144
The IRA sent, uh, to Margaret Thatcher and

1442
01:03:06,144 --> 01:03:08,663
they said, you have to be lucky every time.

1443
01:03:08,663 --> 01:03:13,584
We only have to be lucky once it's for OpenAI to survive.

1444
01:03:13,764 --> 01:03:17,634
They will have to raise more money than anyone has ever raised

1445
01:03:17,754 --> 01:03:21,804
before, then raise it again, then likely raise it again.

1446
01:03:21,804 --> 01:03:24,624
They need to raise another $17 billion in 2027.

1447
01:03:24,624 --> 01:03:25,464
They fracking said it.

1448
01:03:25,464 --> 01:03:25,524
It.

1449
01:03:26,274 --> 01:03:29,214
They will then need to build a massive data

1450
01:03:29,214 --> 01:03:31,913
center that is moving along decently well.

1451
01:03:31,913 --> 01:03:34,254
But I've heard reports that it's having some issues.

1452
01:03:34,584 --> 01:03:37,974
They will also have to keep being able to

1453
01:03:37,974 --> 01:03:41,064
broker deals with various companies to pay them.

1454
01:03:41,064 --> 01:03:42,894
'Cause OpenAI is a Banana Republic.

1455
01:03:43,374 --> 01:03:45,474
They don't make enough money to run them.

1456
01:03:46,014 --> 01:03:47,663
All of their money has to come from outside.

1457
01:03:48,534 --> 01:03:51,774
SoftBank also has to keep getting that money and SoftBank does not have it.

1458
01:03:51,774 --> 01:03:54,504
They had to sell four and a half billion of T-Mobile stock.

1459
01:03:55,254 --> 01:03:57,953
And this is just to get a little bit, they need to give

1460
01:03:57,953 --> 01:04:01,524
open AI unless they don't convert another $30 billion.

1461
01:04:01,854 --> 01:04:03,564
Now they could syndicate $10 billion of

1462
01:04:03,564 --> 01:04:05,244
that, but still that's another $20 billion.

1463
01:04:05,244 --> 01:04:07,134
They do not have, they don't have the cash.

1464
01:04:07,314 --> 01:04:12,684
The 15 that they just borrowed, uh, six to seven of that went to amper.

1465
01:04:12,744 --> 01:04:16,524
Another company they're buying, the numbers are crazy, man.

1466
01:04:16,703 --> 01:04:19,913
When you sit and look at the numbers, they, they are crazy.

1467
01:04:20,244 --> 01:04:25,404
And I don't know what's wrong with the tech media, but they are missing this.

1468
01:04:25,779 --> 01:04:28,569
And when this blows up, I'm gonna look cool.

1469
01:04:28,599 --> 01:04:30,249
I'm gonna look cool as fracking shit.

1470
01:04:30,489 --> 01:04:31,899
This is gonna be very fun for me.

1471
01:04:31,899 --> 01:04:32,288
I'm

1472
01:04:32,408 --> 01:04:33,099
And everyone else is.

1473
01:04:33,099 --> 01:04:33,368
Very good.

1474
01:04:33,368 --> 01:04:34,929
Well, how could we have possibly known?

1475
01:04:35,139 --> 01:04:39,459
Oh, and I am not going to let anyone say that a moment anyone says that.

1476
01:04:39,609 --> 01:04:40,419
I will have link.

1477
01:04:40,449 --> 01:04:41,828
I have links ready to go.

1478
01:04:41,828 --> 01:04:42,969
I'm gonna be like, quick shot.

1479
01:04:43,809 --> 01:04:46,059
Oh, oh, oh, you, how could you have known?

1480
01:04:46,059 --> 01:04:47,739
You follow me, you follow me, cupcake.

1481
01:04:47,949 --> 01:04:48,759
Talk to me baby.

1482
01:04:48,759 --> 01:04:49,538
You have my cell.

1483
01:04:50,019 --> 01:04:51,519
I am going to be annoying.

1484
01:04:52,328 --> 01:04:53,469
I'm gonna be trotting around.

1485
01:04:53,469 --> 01:04:55,149
I'm gonna find the exact level of annoying

1486
01:04:55,149 --> 01:04:56,919
I can get before I start losing credit.

1487
01:04:57,489 --> 01:05:01,479
And what I think you're gonna see is you're gonna see a major outlet.

1488
01:05:01,479 --> 01:05:02,513
You're kind of already seeing it.

1489
01:05:02,513 --> 01:05:07,989
Deirdre Boer of CNN did a video to CNBC, even did a video today, uh,

1490
01:05:08,169 --> 01:05:11,049
talking about how the AI trade is in trouble 'cause the lack of revenue.

1491
01:05:11,409 --> 01:05:16,539
CNN's Alison Morrow has been on this ship for oh years now, just saying like,

1492
01:05:16,539 --> 01:05:22,058
Hey, this is rotten like this, this is a marketing scheme, but if I am wrong.

1493
01:05:22,989 --> 01:05:25,058
Something insane is gonna have to happen.

1494
01:05:25,058 --> 01:05:26,799
Something just completely unprecedented.

1495
01:05:26,979 --> 01:05:28,119
And I'll admit I'm wrong.

1496
01:05:28,209 --> 01:05:29,409
I'm not though.

1497
01:05:29,409 --> 01:05:32,319
It feels like Cassandra has never appreciated in her own time.

1498
01:05:32,558 --> 01:05:35,049
Ed, I, I really wanna thank you for taking the time to speak with me.

1499
01:05:35,289 --> 01:05:35,919
My pleasure.

1500
01:05:35,919 --> 01:05:36,699
Thank you for having me.

1501
01:05:36,939 --> 01:05:39,009
People wanna learn more about what you, uh,

1502
01:05:39,039 --> 01:05:41,109
what you espouse and I argue that they should.

1503
01:05:41,259 --> 01:05:42,969
Where's the best place for them to find you?

1504
01:05:43,089 --> 01:05:44,619
Go to be offline.com.

1505
01:05:45,129 --> 01:05:49,089
Just B-E-T-T-E-R-O-F-F-L-I-N-E.

1506
01:05:49,209 --> 01:05:50,379
I nearly didn't spell it.

1507
01:05:50,619 --> 01:05:51,279
Dot com.

1508
01:05:51,759 --> 01:05:52,719
Uh, it has everything.

1509
01:05:52,719 --> 01:05:53,679
It has links to everything.

1510
01:05:53,679 --> 01:05:55,569
Newsletter, podcast, PR firm.

1511
01:05:56,289 --> 01:05:57,668
Socials, all the good stuff.

1512
01:05:57,729 --> 01:06:00,008
And we'll of course put a link to that in the show notes as well.

1513
01:06:00,069 --> 01:06:01,058
And your snark bot.

1514
01:06:01,058 --> 01:06:02,048
I wanna see this.

1515
01:06:02,138 --> 01:06:04,508
I will include a link to that in the show notes as well.

1516
01:06:04,569 --> 01:06:08,918
Because the funny thing is, is like, I'm like a PR person, I'm a writer.

1517
01:06:09,098 --> 01:06:11,798
People don't know me like I'm a SaaS freak baby.

1518
01:06:11,798 --> 01:06:16,088
I, I know these companies, I know these economics and I feel like

1519
01:06:16,088 --> 01:06:19,029
maybe I'm just the only monster that could look at this stuff and make

1520
01:06:19,029 --> 01:06:23,319
sense of it because of the hours of SaaS related content I've consumed.

1521
01:06:23,499 --> 01:06:25,928
Oh, the, we like, um, Roger  Auer at the

1522
01:06:25,928 --> 01:06:28,088
end of Bladerunner talking about webinars.

1523
01:06:28,388 --> 01:06:30,008
I've seen things you people wouldn't believe.

1524
01:06:31,418 --> 01:06:34,058
It's just, it's three hour long reinvent keynotes,

1525
01:06:35,258 --> 01:06:37,539
burning, burning me in Las Vegas, Nevada.

1526
01:06:37,928 --> 01:06:39,459
Ed, thank you again for your time.

1527
01:06:39,459 --> 01:06:40,539
I deeply appreciate it.

1528
01:06:40,749 --> 01:06:44,499
Ed Zitron, host of better Offline, writer of Where's Your

1529
01:06:44,499 --> 01:06:47,678
Ed at Newsletter, and I'm cloud economist, Cory Quinn.

1530
01:06:47,919 --> 01:06:50,258
If you've enjoyed this podcast, please leave a five

1531
01:06:50,258 --> 01:06:52,419
star review on your podcast platform of choice.

1532
01:06:52,419 --> 01:06:55,838
Whereas if you've hated this podcast, please leave a five star review on

1533
01:06:55,838 --> 01:06:59,468
your podcast platform of choice along with an angry, insulting comment.

1534
01:06:59,619 --> 01:07:02,229
Uh, be sure to let me know in that comment,

1535
01:07:02,258 --> 01:07:04,869
which pony you've bet your AI horse on.