1
00:00:00,000 --> 00:00:04,357
Hello everybody. We are here
at Palantir's ninth AIP Con with the CEO

2
00:00:04,440 --> 00:00:06,237
and co-founder of Palantir, Alex Karp.

3
00:00:06,320 --> 00:00:09,237
Alex, thanks for being here.
>> Great to be with you.

4
00:00:09,320 --> 00:00:12,757
>> Over the past 6 months, a lot
of people have had a lot to say about your

5
00:00:12,840 --> 00:00:16,717
company. The media has stuff to say,
investors have stuff to say, short

6
00:00:16,800 --> 00:00:19,917
sellers have stuff to say. What do you
have to say about what's happened over

7
00:00:20,000 --> 00:00:22,157
the past 6 months with Palantir?

8
00:00:22,240 --> 00:00:26,197
>> Um well, you know, maybe we should
ask what the mullahs have to say about

9
00:00:26,280 --> 00:00:29,317
what's happened over the last 6 months
at Palantir. I think you should start

10
00:00:29,400 --> 00:00:31,637
there. You know, like let's go right to

11
00:00:31,720 --> 00:00:33,837
the people of the customers.

12
00:00:33,920 --> 00:00:36,397
Ask the customers. You know, it's

13
00:00:36,480 --> 00:00:40,957
uh yeah, I you know, um we spent

14
00:00:41,040 --> 00:00:45,797
uh we started 20 years ago saying
we're going to do value creation.

15
00:00:45,880 --> 00:00:49,597
Uh don't for you know, it's like we're
not going to you know, take your weird

16
00:00:49,680 --> 00:00:53,317
category you learned at business school
at Harvard probably, you know, about how

17
00:00:53,400 --> 00:00:57,677
to build software, build a parasitic
little trinket and install it because

18
00:00:57,760 --> 00:01:00,837
that's what venture capitalists wanted,
because that's what analysts wanted,

19
00:01:00,920 --> 00:01:04,157
because that's what banks
would underwrite, because that's

20
00:01:04,240 --> 00:01:07,277
how they gave you your underwriting fee

21
00:01:07,360 --> 00:01:09,557
and and pre-bought your equity before

22
00:01:09,640 --> 00:01:12,037
they sold it to retail after their hedge

23
00:01:12,120 --> 00:01:17,677
fund friends got in and you got [__]
and they got laid. Uh um and we said

24
00:01:17,760 --> 00:01:20,517
we're not going to do any of that.
We were going to build things

25
00:01:20,600 --> 00:01:23,397
uh that were super valuable and de facto

26
00:01:23,480 --> 00:01:25,677
get paid. I've been telling people for

27
00:01:25,760 --> 00:01:29,837
decades now we essentially get paid uh

28
00:01:29,920 --> 00:01:32,437
uh uh after we create value for you.

29
00:01:32,520 --> 00:01:36,477
>> Right. >> On your terms, according
to your institution. We did this with

30
00:01:36,560 --> 00:01:39,197
clandestine services, special forces,

31
00:01:39,280 --> 00:01:43,797
Maven, every institution in government,
every ally in the world, some of which

32
00:01:43,880 --> 00:01:47,797
love to admit they work with us, many
of which don't admit they work with us. We

33
00:01:47,880 --> 00:01:51,677
then went into commercial. Everyone
said we had no commercial product. But we

34
00:01:51,760 --> 00:01:54,277
grew commercial at a rate
no one's ever seen before.

35
00:01:54,360 --> 00:01:56,997
>> The last earnings were one of the
craziest earnings we've seen in the

36
00:01:57,080 --> 00:02:00,517
history of not just Palantir,
but in the entire stock market.

37
00:02:00,600 --> 00:02:04,757
Why are you guys growing
as fast as you guys are growing?

38
00:02:04,840 --> 00:02:08,917
>> Well, again, it's very much linked to
this. Everybody is still You build these

39
00:02:09,000 --> 00:02:12,917
organizations commercial,
military, institutional,

40
00:02:13,000 --> 00:02:14,837
clandestine services.

41
00:02:14,920 --> 00:02:17,637
You can't really They're not built to

42
00:02:17,720 --> 00:02:20,517
change. At best, they're built to adapt.

43
00:02:20,600 --> 00:02:23,037
Our organization is maximally adaptive

44
00:02:23,120 --> 00:02:26,037
on the back of a couple I principles and

45
00:02:26,120 --> 00:02:28,797
has maximal range of motion. And that

46
00:02:28,880 --> 00:02:30,917
and they are built for right now,

47
00:02:31,000 --> 00:02:35,837
meaning people want value. And like,
you know, now people who are trying to

48
00:02:35,920 --> 00:02:39,837
understand where the value is work with
broad categories, large language models

49
00:02:39,920 --> 00:02:42,197
versus software versus hardware versus

50
00:02:42,280 --> 00:02:44,877
like the client, whether it is the US

51
00:02:44,960 --> 00:02:49,317
military now doing in my view, you know,
it is all of our jobs to protect the

52
00:02:49,400 --> 00:02:53,397
warfighter. Even if you don't agree
with any other part of government or this

53
00:02:53,480 --> 00:02:57,237
administration, once the warfighter
is putting their life in danger, it's our

54
00:02:57,320 --> 00:03:01,037
responsibility, in my view, to support
them, especially business leaders who

55
00:03:01,120 --> 00:03:05,797
are making doing very, very well in this
country. But it's like, you know, at the

56
00:03:05,880 --> 00:03:09,397
end of the day. And then the other thing,
by the way, is one of the biggest

57
00:03:09,480 --> 00:03:14,277
critiques of Palantir still is, but what
is it? Is it software, hardware, LLM? I

58
00:03:14,360 --> 00:03:18,557
mean, read read the somewhat innate
often I don't even think they're I think

59
00:03:18,640 --> 00:03:21,877
at this point they're probably just
hired help when they're writing these

60
00:03:21,960 --> 00:03:25,837
crazy short things about us. But it's
like, but but the legitimate thing is we

61
00:03:25,920 --> 00:03:27,717
don't fit into any of these categories.

62
00:03:27,800 --> 00:03:29,477
>> Yeah. >> But that's
why we're valuable. That's

63
00:03:29,560 --> 00:03:33,957
why the client All of our clients need
to move from categories that make no

64
00:03:34,040 --> 00:03:37,477
sense for them to the category that does
make sense to them, which I'm the best

65
00:03:37,560 --> 00:03:41,517
in the world, whether it's the US
CENTCOM or whether it's commercial

66
00:03:41,600 --> 00:03:44,677
clients. And by the way, it's the same
thing internally. Like we're just trying

67
00:03:44,760 --> 00:03:46,957
to slot people into a position where we

68
00:03:47,040 --> 00:03:51,957
can get the most value for themselves
and then therefore for Palantir.

69
00:03:52,040 --> 00:03:54,877
>> Software companies have gone through
a fundamental re-rating in the broader

70
00:03:54,960 --> 00:03:58,197
market. A lot of people are wondering
is AI eating software? The interesting

71
00:03:58,280 --> 00:04:01,237
thing about Palantir is I don't think
you ever characterize yourself as just

72
00:04:01,320 --> 00:04:02,717
an AI company, just a software company.

73
00:04:02,800 --> 00:04:06,677
It was we are a value company. We get
paid when we deliver value. How do you

74
00:04:06,760 --> 00:04:09,277
think you survive in this new
world of AI disruption?

75
00:04:09,360 --> 00:04:12,317
>> Well, no, the thing is they they they
of course they need broad categories and

76
00:04:12,400 --> 00:04:15,917
the broad category they may have
of course but again, in the beginning the

77
00:04:16,000 --> 00:04:19,277
broad category that they had of us is we
weren't a software company because we

78
00:04:19,360 --> 00:04:22,917
were a services company. By the way, one
of the most important indicators in the

79
00:04:23,000 --> 00:04:25,357
market is everybody wants to be us.

80
00:04:25,440 --> 00:04:27,877
>> Yeah. >> Every two days >> Well,
four deployed engineers is now

81
00:04:27,960 --> 00:04:29,997
being used everywhere.
>> like you're going to see all the

82
00:04:30,080 --> 00:04:34,157
research labs are all everybody's going
to have like a Palantir because it's the

83
00:04:34,240 --> 00:04:37,117
only way to create money for the
customer which is the only way to get

84
00:04:37,200 --> 00:04:38,557
paid. Everything is too amorphous.

85
00:04:38,640 --> 00:04:41,637
Everything else is too amorphous and not
quantifiable and you're going to be

86
00:04:41,720 --> 00:04:44,877
competing at the end of the day with
somebody who's going to buy Palantir and

87
00:04:44,960 --> 00:04:48,557
how the [__] you're going to explain
to them that your your hype trinket is

88
00:04:48,640 --> 00:04:49,957
going to create value when it doesn't.

89
00:04:50,040 --> 00:04:51,677
Especially where the delta value

90
00:04:51,760 --> 00:04:54,317
creation on timelines, margins and our

91
00:04:54,400 --> 00:04:57,997
pure alpha is so large. Um, but

92
00:04:58,080 --> 00:05:01,717
to your point, no, we were never ever

93
00:05:01,800 --> 00:05:04,117
viewed ourselves as like we are X plus

94
00:05:04,200 --> 00:05:06,957
minus one. Right. We are a sui generis

95
00:05:07,040 --> 00:05:09,037
unique way of doing things which does

96
00:05:09,120 --> 00:05:11,197
include like a

97
00:05:11,280 --> 00:05:14,237
actual human labor augmenting the labor

98
00:05:14,320 --> 00:05:18,917
now with AI, uh encoding it both
the knowledge of your institution meaning

99
00:05:19,000 --> 00:05:22,157
the data but also more importantly your
assumptions around the data, especially

100
00:05:22,240 --> 00:05:25,637
assumptions that are impossible
to capture like basically your tribal

101
00:05:25,720 --> 00:05:30,317
knowledge and melding that into
something that is unique. And and part

102
00:05:30,400 --> 00:05:32,517
of the reason people don't want
to understand is you know this famous

103
00:05:32,600 --> 00:05:36,597
thing, you know, you sit you sit you
stand where you sit. It's very hard to

104
00:05:36,680 --> 00:05:41,037
acknowledge these things because this
is a radically have have not world and a

105
00:05:41,120 --> 00:05:45,837
lot of people including the software
thing, if you can't plausibly create

106
00:05:45,920 --> 00:05:48,997
value rapidly, you're going to disappear.

107
00:05:49,080 --> 00:05:52,837
Like and like but by the way, it's like,
you know, white-collar workers, certain

108
00:05:52,920 --> 00:05:57,077
weak software companies, it's like steak
dinners and a widget, that's not going

109
00:05:57,160 --> 00:06:00,397
to survive. It doesn't matter. You could
do weird financial machinations. None of

110
00:06:00,480 --> 00:06:05,357
it is going to work. Uh and if you're
just doing pure, you know, pure pure

111
00:06:05,440 --> 00:06:08,437
pure LLMs, well, that's very hard
to make that work in enterprise. You see

112
00:06:08,520 --> 00:06:10,717
this again, without going into all the

113
00:06:10,800 --> 00:06:14,797
Pentagon drama, it's like, "Yeah, but

114
00:06:14,880 --> 00:06:18,357
you know, it's like that's
nice and we're replacing you."

115
00:06:18,440 --> 00:06:22,397
>> Well, 2 years ago you said at AIP Con
number four that the world will split up

116
00:06:22,480 --> 00:06:26,037
into chips and ontology. And quite
frankly, you said it in a very simple to

117
00:06:26,120 --> 00:06:28,837
understand way that to this day I hear
from people saying, "Oh yeah, Karp told

118
00:06:28,920 --> 00:06:32,637
us 2 years ago the LLMs will be useful,
but they will not be where the value

119
00:06:32,720 --> 00:06:35,197
comes from." Do you think
the world has figured that out yet?

120
00:06:35,280 --> 00:06:39,197
>> No, what the world has figured out,
I don't know. It's very hard. I think it's

121
00:06:39,280 --> 00:06:43,677
it's super hard for people like No one's
figured out this world completely. And

122
00:06:43,760 --> 00:06:45,397
we have never seen anything like this.

123
00:06:45,480 --> 00:06:50,517
We have like technology has in the modern
world has been very aligned with

124
00:06:50,600 --> 00:06:54,397
the society itself. >> Yes.
>> But this technology is not purely

125
00:06:54,480 --> 00:06:57,277
aligned with the society the way it
originally was. You have like people

126
00:06:57,360 --> 00:07:01,317
who've historically had the best jobs
in the world, the most protected, most

127
00:07:01,400 --> 00:07:05,397
respected, most interesting. Yeah,
but those jobs are going to disappear and

128
00:07:05,480 --> 00:07:09,757
very rapidly. And those people are not
used to having like, you know, the way

129
00:07:09,840 --> 00:07:11,557
vocational labors are treated, right?

130
00:07:11,640 --> 00:07:15,717
So, it's like that's going to be We
we have on the military front, all of a

131
00:07:15,800 --> 00:07:19,997
sudden America's doing things that, you
know, like 2 3 years ago a lot of people

132
00:07:20,080 --> 00:07:22,677
were looking at America as like, "Well,
we could definitely do that, especially

133
00:07:22,760 --> 00:07:26,277
if we didn't spend half the money,
right?" No one thinks that now.

134
00:07:26,360 --> 00:07:30,077
And and it's like on every front.
So, this is a We're in We're in unchar-

135
00:07:30,160 --> 00:07:34,477
charted territory. And the best thing
that can happen to you is you And again,

136
00:07:34,560 --> 00:07:38,717
this is where people could learn from
retail. You first of all, you have to be

137
00:07:38,800 --> 00:07:43,277
on the front line yourself. You cannot
outsource what you're doing to an expert

138
00:07:43,360 --> 00:07:47,597
unless that expert then conversely the
definition of an expert nowadays has to

139
00:07:47,680 --> 00:07:51,637
be you were right and therefore I think
you might be right tomorrow not you have

140
00:07:51,720 --> 00:07:54,797
some weird degree or something or someone
says you're an expert or you you

141
00:07:54,880 --> 00:07:58,957
know and and then you really have
to investigate these things yourself and

142
00:07:59,040 --> 00:08:03,557
you have to be open to yeah things
are changing so rapidly that yeah what

143
00:08:03,640 --> 00:08:06,557
existed yesterday is not
going to exist tomorrow.

144
00:08:06,640 --> 00:08:09,317
>> There's a war happening right now. I'm
sure there's a lot of people that have a

145
00:08:09,400 --> 00:08:12,277
lot of thoughts about it. What is
Palantir's message about what's happening?

146
00:08:12,360 --> 00:08:15,357
>> Well, everybody who's watching this

147
00:08:15,440 --> 00:08:19,477
should just reflect no matter
where you are on war or this

148
00:08:19,560 --> 00:08:22,117
administration or anything else this

149
00:08:22,200 --> 00:08:25,397
country has an ability to define itself

150
00:08:25,480 --> 00:08:28,397
meaning the our constitutional order and

151
00:08:28,480 --> 00:08:30,197
the spirit around it

152
00:08:30,280 --> 00:08:33,397
and our economic success because it is

153
00:08:33,480 --> 00:08:36,237
defended by courageous men and women who

154
00:08:36,320 --> 00:08:40,597
are willing to put
their life on the line and

155
00:08:40,680 --> 00:08:44,517
by the way life on the line they also
have many most have like children at

156
00:08:44,600 --> 00:08:47,477
home spouses at home like these are

157
00:08:47,560 --> 00:08:52,237
people taking on enormous risk that have
a lot to lose and at Palantir we're you

158
00:08:52,320 --> 00:08:53,997
know as has been reported on the front

159
00:08:54,080 --> 00:08:56,477
line of a lot of these things and we

160
00:08:56,560 --> 00:08:59,517
obviously believe once once once the

161
00:08:59,600 --> 00:09:04,397
proverbial bullets start flying we have
one side we're on that side we're going

162
00:09:04,480 --> 00:09:08,157
to support it all the way through
beginning during and after we're not

163
00:09:08,240 --> 00:09:10,597
particularly attenuated to popularity or

164
00:09:10,680 --> 00:09:13,677
polls the warfighter is fighting for us

165
00:09:13,760 --> 00:09:15,837
and we are going to fight for them.

166
00:09:15,920 --> 00:09:19,357
>> A lot of people have debated the value
of what makes an AI company an AI

167
00:09:19,440 --> 00:09:22,317
company and you know you have what's
happening in the world right now with

168
00:09:22,400 --> 00:09:25,117
different models competing for the
infrastructure to be able to provide

169
00:09:25,200 --> 00:09:29,077
value some models not being liked
by the government. Is Palantir in a strong

170
00:09:29,160 --> 00:09:31,997
position over the coming years
to continue supporting the Department of

171
00:09:32,080 --> 00:09:33,877
War regardless of what AI models are in

172
00:09:33,960 --> 00:09:35,957
power or in popularity?

173
00:09:36,040 --> 00:09:40,637
>> Well, I mean,
it's uh we provide the the

174
00:09:40,720 --> 00:09:42,717
the air the layer that makes these

175
00:09:42,800 --> 00:09:45,517
models value valuable and it it it

176
00:09:45,600 --> 00:09:48,477
orchestrates where where and how they're

177
00:09:48,560 --> 00:09:50,797
used and and a lot of things that can't

178
00:09:50,880 --> 00:09:55,837
be discussed in public, but um I look,
I'm very focused on we will do

179
00:09:55,920 --> 00:09:57,717
whatever's best for the warfighter.

180
00:09:57,800 --> 00:10:01,637
We're That's that's the priority,
maximum priority for us.

181
00:10:01,720 --> 00:10:04,517
Uh I think we're doing
pretty well as a company and

182
00:10:04,600 --> 00:10:08,237
you know, people will We're just
going to be a very exciting ride.

183
00:10:08,320 --> 00:10:12,237
>> The operating margins of Palantir also
have grown pretty exponentially. I mean,

184
00:10:12,320 --> 00:10:15,757
a year ago they were at 17%. We're now
pushing 50%. We have not seen anything

185
00:10:15,840 --> 00:10:19,317
like this in software. What is >>
Well, what about our rule of 127?

186
00:10:19,400 --> 00:10:23,917
>> Well, it's it's a new rule
every 90 days >> Hey, size matters.

187
00:10:24,000 --> 00:10:25,757
It matters. People want to say it

188
00:10:25,840 --> 00:10:28,597
doesn't matter.

189
00:10:28,680 --> 00:10:33,237
>> Um what is the magic behind putting
up the most insane numbers we've seen in

190
00:10:33,320 --> 00:10:36,837
software history?
Is it internal culture? Is it

191
00:10:36,920 --> 00:10:39,557
>> it's it's the sum is the parts are very

192
00:10:39,640 --> 00:10:41,877
strong and the sum is much stronger than

193
00:10:41,960 --> 00:10:46,797
the parts. I mean, we are an artistic
Look, there there there are very

194
00:10:46,880 --> 00:10:51,117
Nietzsche once said there were two
artists of the German language and

195
00:10:51,200 --> 00:10:55,637
somewhat arrogantly, a kind of arrogance
I relate to, said, "Uh myself and

196
00:10:55,720 --> 00:10:59,037
Heine." There are very
few artists in business.

197
00:10:59,120 --> 00:11:00,957
I am one of them. Palantir is an

198
00:11:01,040 --> 00:11:04,277
artistic project and everybody else is

199
00:11:04,360 --> 00:11:07,997
trying to build things
in like a non-artistic,

200
00:11:08,080 --> 00:11:11,117
very much based on what they has been

201
00:11:11,200 --> 00:11:15,437
understood in the past kind of way when
the there is no road map, there's no

202
00:11:15,520 --> 00:11:17,877
playbook, and it's it really is the

203
00:11:17,960 --> 00:11:21,357
neuro divergence Yes. of Palantir as a

204
00:11:21,440 --> 00:11:26,317
company and our ability
to manage highly in their own way

205
00:11:26,400 --> 00:11:29,437
differentiated and often in their
own way without even realizing it

206
00:11:29,520 --> 00:11:34,397
neurodivergent people against a task
that is more important than us. And then

207
00:11:34,480 --> 00:11:35,877
that seeps into the quality of the

208
00:11:35,960 --> 00:11:39,397
products and and and the way we execute.

209
00:11:39,480 --> 00:11:43,797
And you know, it's that concatenation
um and then you could just judge us by

210
00:11:43,880 --> 00:11:48,557
our enemies. Look at them.
Judges by our enemies.

211
00:11:48,640 --> 00:11:51,477
Yeah. >> Got it. Okay.
Alex Karp, CEO, thank you so much.

212
00:11:51,560 --> 00:11:56,517
>> Yeah. Thank you. Yes.
>> That was great. Hopefully nice meeting

213
00:11:56,600 --> 00:11:58,877
you for you. Good stuff. >> Yeah,
I know. I always like seeing you.

214
00:11:58,960 --> 00:12:01,637
It's one of my highlights. >> Yeah,
you know, I like seeing you, too.

215
00:12:01,720 --> 00:12:03,637
It's fun. >> Alex Karp, what do
you think about the brick work here?

216
00:12:03,720 --> 00:12:06,597
>> I like the whole thing. >> You know,
my background at my home studio is bricks.

217
00:12:06,680 --> 00:12:08,397
>> Yeah, I know.
>> I know Iliano picked out the

218
00:12:08,480 --> 00:12:12,077
>> It's fake bricks. So just the ongoing
joke online is like oh, the bricks are

219
00:12:12,160 --> 00:12:13,837
fake, they're real.
>> Yeah, everyone's debating.

220
00:12:13,920 --> 00:12:17,800
>> Bricks and mortar. Hey.