1
00:00:00,120 --> 00:00:03,720
In the race to dominate 
generative AI, big tech firms 

2
00:00:03,880 --> 00:00:06,320
haven't just been building, 
they've been buying. 

3
00:00:06,560 --> 00:00:09,520
But there's something strange 
about most of the deals that 

4
00:00:09,520 --> 00:00:13,160
they've struck. 
Companies like Meta, Microsoft, 

5
00:00:13,160 --> 00:00:17,440
Amazon, Google and NVIDIA are 
embedding themselves deep within

6
00:00:17,440 --> 00:00:22,080
the AI ecosystem through 
strategic investments, exclusive

7
00:00:22,080 --> 00:00:26,040
partnerships, and talent 
acquisitions with deals that 

8
00:00:26,040 --> 00:00:28,880
stop just short of formal 
takeovers. 

9
00:00:29,160 --> 00:00:33,000
But the economic impact of these
deals, it turns out, is 

10
00:00:33,000 --> 00:00:35,560
indistinguishable from full 
control. 

11
00:00:36,000 --> 00:00:40,200
This new approach, which some 
analysts are describing as non 

12
00:00:40,200 --> 00:00:44,040
acquisition acquisitions, is 
designed to work around 

13
00:00:44,040 --> 00:00:47,800
antitrust law. 
The idea is to avoid drawing 

14
00:00:47,800 --> 00:00:52,320
regulatory scrutiny with deals 
that allow tech giants to secure

15
00:00:52,520 --> 00:00:56,560
privileged access to 
foundational by models, critical

16
00:00:56,560 --> 00:00:59,760
data pipelines, and elite 
research talent. 

17
00:01:00,200 --> 00:01:03,840
Despite all the talk of slashing
regulation, and despite the 

18
00:01:03,840 --> 00:01:07,760
support that Trump has drawn 
from Silicon Valley so far, the 

19
00:01:07,760 --> 00:01:11,960
new administration is showing a 
surprising degree of continuity 

20
00:01:11,960 --> 00:01:15,640
with the Biden era approach to 
antitrust enforcement, 

21
00:01:15,880 --> 00:01:20,080
particularly when it comes to 
Big Tech and AI related deals. 

22
00:01:20,600 --> 00:01:24,760
According to the law firm 
Goodwin Proctor, US enforcers 

23
00:01:24,760 --> 00:01:28,360
are still willing to challenge 
deals they view as anti 

24
00:01:28,360 --> 00:01:32,160
competitive under traditional 
horizontal theories of harm, 

25
00:01:32,400 --> 00:01:36,480
particularly mergers between 
direct competitors with leading 

26
00:01:36,480 --> 00:01:40,640
market positions. 
So far, there's been no sign of 

27
00:01:40,640 --> 00:01:46,120
reduced FTC enforcement. 
The tactics used by today's tech

28
00:01:46,120 --> 00:01:50,400
giants to quietly consolidate 
power in the AI sector are not 

29
00:01:50,400 --> 00:01:54,000
without precedent. 
More than a century ago, John D 

30
00:01:54,000 --> 00:01:58,200
Rockefeller pioneered similar 
strategies during the rise of 

31
00:01:58,200 --> 00:02:02,200
Standard Oil, where he often 
acquired companies without 

32
00:02:02,200 --> 00:02:06,400
formal legal documentation, 
using silent partnerships, 

33
00:02:06,600 --> 00:02:11,240
verbal agreements, and trusted 
intermediaries to quietly absorb

34
00:02:11,240 --> 00:02:14,560
competitors. 
According to his biographer, 

35
00:02:14,680 --> 00:02:18,880
Standard Oil executives worried 
that this secrecy could cause 

36
00:02:18,880 --> 00:02:23,280
problems for the firm, as if a 
seller died unexpectedly, his 

37
00:02:23,280 --> 00:02:27,320
heirs might mistakenly claim 
ownership of a refinery that 

38
00:02:27,320 --> 00:02:29,760
Standard Oil had bought and paid
for. 

39
00:02:30,560 --> 00:02:34,080
To further obscure control of 
the businesses he had purchased,

40
00:02:34,320 --> 00:02:39,480
Rockefeller created the Standard
Oil Trust in 1882, a legal 

41
00:02:39,480 --> 00:02:43,360
innovation that allowed him to 
centralized authority over 

42
00:02:43,360 --> 00:02:46,800
dozens of companies without 
technically owning them 

43
00:02:46,800 --> 00:02:50,640
outright. 
These tactics eventually led to 

44
00:02:50,640 --> 00:02:55,720
the 1911 Supreme Court decision 
that broke up Standard Oil under

45
00:02:55,720 --> 00:03:00,160
the Sherman Antitrust Act. 
Today, Big Tech's use of non 

46
00:03:00,160 --> 00:03:04,080
acquisition acquisitions, 
exclusive infrastructure deals 

47
00:03:04,240 --> 00:03:07,560
and strategic talent 
consolidation echoes 

48
00:03:07,600 --> 00:03:12,600
Rockefeller's playbook. 
Meadows recently announced $14.8

49
00:03:12,600 --> 00:03:16,960
billion investment in Scale AI, 
which involved hiring. 

50
00:03:16,960 --> 00:03:21,400
The data labeling startup CEO 
will test how the Trump 

51
00:03:21,400 --> 00:03:26,160
administration views Aqua hire 
deals, where a company buys a 

52
00:03:26,160 --> 00:03:30,800
startup as a means of hiring its
most important staff, which many

53
00:03:30,800 --> 00:03:35,080
have criticized as an attempt to
evade regulatory scrutiny. 

54
00:03:35,880 --> 00:03:42,000
Meta executed a $14.3 billion 
investment in Scale AIA leading 

55
00:03:42,000 --> 00:03:44,720
data labeling firm, earlier this
month. 

56
00:03:45,120 --> 00:03:50,920
On paper, Meta bought a 49% non 
voting stake in Scale, but in 

57
00:03:50,920 --> 00:03:55,680
practice they gained exclusive 
access to scale's Human in the 

58
00:03:55,680 --> 00:03:59,760
Loop data pipeline. 
The migration of Scale founder 

59
00:03:59,760 --> 00:04:04,480
Alexander Wang and key staff to 
Meta's new superintelligence 

60
00:04:04,480 --> 00:04:09,440
unit, and a web of contractual 
agreements that effectively 

61
00:04:09,440 --> 00:04:14,800
embedded Scales capabilities 
within Meta's AI infrastructure.

62
00:04:15,400 --> 00:04:18,839
The structure of this deal meant
that Meta could avoid the 

63
00:04:18,839 --> 00:04:22,960
formalities of a merger while 
achieving many of its strategic 

64
00:04:22,960 --> 00:04:25,560
benefits. 
In an excellent sub stack 

65
00:04:25,560 --> 00:04:28,840
article, which I'll link to in 
the description, Drayton de 

66
00:04:28,840 --> 00:04:32,440
Silva described the deal as 
being like the way medieval 

67
00:04:32,440 --> 00:04:36,440
Masons would move entire 
cathedrals from one side to 

68
00:04:36,480 --> 00:04:40,120
another, brick by brick, by 
numbering each brick, 

69
00:04:40,320 --> 00:04:44,280
transporting it to a new site, 
and then rebuilding the entire 

70
00:04:44,280 --> 00:04:49,000
cathedral exactly as before. 
The reconstructed cathedral, he 

71
00:04:49,000 --> 00:04:53,640
says, is in a new location, but 
in a spiritual and metaphysical 

72
00:04:53,640 --> 00:04:57,920
sense, He argues, it's the same 
cathedral scale. 

73
00:04:57,920 --> 00:05:00,920
AI, which was once an 
independent supplier to the 

74
00:05:00,920 --> 00:05:06,280
entire AI ecosystem, once this 
deal is done, now functions as a

75
00:05:06,280 --> 00:05:08,840
de facto internal division of 
Mehta. 

76
00:05:09,240 --> 00:05:13,160
De Silva argues that there are 
multiple elements in the deal 

77
00:05:13,160 --> 00:05:16,680
structure that make it 
vulnerable to FTC scrutiny. 

78
00:05:17,440 --> 00:05:21,880
From an antitrust perspective, 
the deal raises red flags across

79
00:05:21,880 --> 00:05:25,920
multiple dimensions. 
De facto control input 

80
00:05:25,920 --> 00:05:29,840
foreclosure, talent 
consolidation, coordination 

81
00:05:29,840 --> 00:05:32,440
risk, and dominance 
entrenchment. 

82
00:05:32,880 --> 00:05:37,040
Even if regulators ultimately 
blocked the deal, Mehta may have

83
00:05:37,040 --> 00:05:41,040
already achieved A strategic 
victory by disrupting rival 

84
00:05:41,040 --> 00:05:45,320
operations and gaining insight 
into scales operations. 

85
00:05:45,480 --> 00:05:49,200
In a funny piece about Mark 
Zuckerberg's new bro look in the

86
00:05:49,200 --> 00:05:54,680
FT, Hannah Murphy argues that 
Facebook feels desperately dated

87
00:05:54,680 --> 00:05:57,480
today. 
Instagram has been losing ground

88
00:05:57,480 --> 00:06:01,400
to TikTok for quite some time, 
and the metaverse, after which 

89
00:06:01,400 --> 00:06:06,080
Mark Zuckerberg renamed the 
entire company, is a ghost town,

90
00:06:06,240 --> 00:06:11,240
but without even ghosts. 
The FTC is already pursuing an 

91
00:06:11,240 --> 00:06:15,600
antitrust case against the firm,
saying that it's acquisitions of

92
00:06:15,600 --> 00:06:19,640
Instagram and WhatsApp were anti
competitive and aimed at 

93
00:06:19,640 --> 00:06:22,080
monopolizing the social media 
market. 

94
00:06:22,680 --> 00:06:26,800
The case is currently in trial, 
with arguments focused on 

95
00:06:26,800 --> 00:06:31,120
whether Meta's acquisitions were
part of a broader strategy to 

96
00:06:31,120 --> 00:06:34,480
eliminate competition and 
maintain its dominance. 

97
00:06:35,000 --> 00:06:39,080
Meta has struggled from the very
beginning to innovate internally

98
00:06:39,240 --> 00:06:42,800
and has mostly had to rely on 
acquisitions for growth. 

99
00:06:43,160 --> 00:06:47,000
If Meta loses this antitrust 
case, the company could be 

100
00:06:47,000 --> 00:06:49,720
forced to sell off Instagram and
WhatsApp. 

101
00:06:50,000 --> 00:06:52,920
They would most likely still be 
allowed to hold on to the 

102
00:06:52,920 --> 00:06:56,280
metaverse. 
And those creepy AI glasses that

103
00:06:56,280 --> 00:07:00,480
Zuckerberg wears with the built 
in cameras just because nobody 

104
00:07:00,480 --> 00:07:04,920
else actually uses them? 
Meta has been very aggressive in

105
00:07:04,920 --> 00:07:07,720
its attempts to grow in 
generative AI. 

106
00:07:08,000 --> 00:07:11,360
And the new investments that 
Meta and almost all of the other

107
00:07:11,360 --> 00:07:15,880
big tech firms have been making 
in AI startups are not just 

108
00:07:15,880 --> 00:07:19,640
financial investments. 
They're strategic and designed 

109
00:07:19,640 --> 00:07:24,040
to secure early access to 
cutting edge models, proprietary

110
00:07:24,040 --> 00:07:29,120
data, and elite research talent.
These deals, it could be argued,

111
00:07:29,280 --> 00:07:33,440
are driven by a belief amongst 
Big Tech CE OS that the next 

112
00:07:33,440 --> 00:07:38,000
wave of AI breakthroughs will be
shaped not just by algorithms, 

113
00:07:38,200 --> 00:07:41,880
but by control over the 
infrastructure and inputs that 

114
00:07:41,880 --> 00:07:45,360
power them. 
Data labeling pipelines, cloud 

115
00:07:45,360 --> 00:07:49,000
compute access, and model 
training workflows have become 

116
00:07:49,000 --> 00:07:52,960
the new battlegrounds. 
Microsoft's relationship with 

117
00:07:52,960 --> 00:07:57,640
Open AI is a master class in 
contractual leverage, but the 

118
00:07:57,640 --> 00:08:02,680
deal still has its flaws. 
Since 2019, Microsoft has 

119
00:08:02,680 --> 00:08:08,760
invested over $13 billion into 
Open A is for profit subsidiary,

120
00:08:08,920 --> 00:08:13,000
securing exclusive commercial 
rights to its models and 

121
00:08:13,000 --> 00:08:17,400
integrating them deeply into its
Azure cloud platform and 

122
00:08:17,400 --> 00:08:21,440
enterprise software products. 
Yet despite the scale of this 

123
00:08:21,440 --> 00:08:26,520
investment, Microsoft holds no 
formal equity stake in Open AIS 

124
00:08:26,520 --> 00:08:30,680
nonprofit parent and no voting 
control over its board. 

125
00:08:31,280 --> 00:08:35,440
This structure was carefully 
designed to avoid the appearance

126
00:08:35,480 --> 00:08:38,280
of a merger. 
Microsoft didn't report the 

127
00:08:38,280 --> 00:08:42,480
transaction to the FTC at the 
time because the investment 

128
00:08:42,480 --> 00:08:44,880
didn't amount to control of the 
company. 

129
00:08:45,040 --> 00:08:50,720
Under U.S. law, Open AI remains 
an independent entity, at least 

130
00:08:50,720 --> 00:08:53,960
on paper. 
But in practice, Microsoft has a

131
00:08:53,960 --> 00:08:58,240
level of influence that rivals 
are exceeds that of a majority 

132
00:08:58,240 --> 00:09:01,200
shareholder. 
The FTC is now investigating 

133
00:09:01,840 --> 00:09:05,120
whether this partnership 
constitutes a de facto 

134
00:09:05,160 --> 00:09:08,600
acquisition. 
A year and a half ago, when Open

135
00:09:08,680 --> 00:09:13,000
A IS board of directors ousted 
Sam Altman over concerns about 

136
00:09:13,040 --> 00:09:17,680
AI safety, Microsoft received 
just minutes notice of what was 

137
00:09:17,680 --> 00:09:22,000
going on and its executives were
not consulted in the decision. 

138
00:09:22,520 --> 00:09:26,560
At the time, Microsoft offered 
Altman and the core research 

139
00:09:26,560 --> 00:09:30,360
team roles within Microsoft, 
which would have likely drawn 

140
00:09:30,360 --> 00:09:34,280
regulatory scrutiny. 
The bizarre structure of Open 

141
00:09:34,400 --> 00:09:38,640
AI, which includes A nonprofit 
entity and a capped profit 

142
00:09:38,640 --> 00:09:41,960
subsidiary, has caused all sorts
of conflicts. 

143
00:09:42,240 --> 00:09:45,600
Now that the company is trying 
to convert into a for profit 

144
00:09:45,600 --> 00:09:50,360
company, the FT reports that 
Microsoft has considered halting

145
00:09:50,360 --> 00:09:54,640
discussions if the two sides 
can't agree on critical issues 

146
00:09:54,840 --> 00:09:59,000
like the size of Microsoft's 
future stake in Open AI. 

147
00:09:59,800 --> 00:10:03,600
If Microsoft was to walk away 
from these discussions, it could

148
00:10:03,600 --> 00:10:07,440
still rely on the existing 
commercial contract to retain 

149
00:10:07,440 --> 00:10:11,440
access to Open AI's technology 
until 2030. 

150
00:10:12,160 --> 00:10:16,520
The Wall Street Journal recently
reported that Open AI executives

151
00:10:16,640 --> 00:10:20,320
have discussed leveling 
antitrust accusations against 

152
00:10:20,320 --> 00:10:25,360
Microsoft as a last ditch option
for loosening the firm's control

153
00:10:25,360 --> 00:10:28,520
under their deal. 
So while Microsoft's deal may 

154
00:10:28,520 --> 00:10:32,240
have been structured to avoid 
FTC scrutiny, the fact that they

155
00:10:32,240 --> 00:10:36,520
don't have full control is still
causing them plenty of problems.

156
00:10:37,280 --> 00:10:41,160
Amazon and Google have each 
invested billions of dollars in 

157
00:10:41,160 --> 00:10:45,720
Anthropic, the AI startup behind
the Claude family of models. 

158
00:10:46,000 --> 00:10:51,640
In return, Claude is hosted on 
AWS and Google Cloud, raising 

159
00:10:51,640 --> 00:10:56,080
concerns about cloud dependency 
and preferential access. 

160
00:10:56,440 --> 00:11:01,120
NVIDIA, meanwhile, has taken 
equity stakes in dozens of AI 

161
00:11:01,120 --> 00:11:05,280
startups, bundling those 
investments with early access to

162
00:11:05,280 --> 00:11:09,720
hardware developer support and 
Co marketing opportunities. 

163
00:11:10,160 --> 00:11:14,640
In one of the stranger deals, 
NVIDIA invested in Core Weave, a

164
00:11:14,640 --> 00:11:18,920
business that buys and rents out
access to NVIDIA graphics 

165
00:11:18,920 --> 00:11:23,040
processing units for training 
artificial intelligence models. 

166
00:11:23,640 --> 00:11:28,520
Nvidia's investment strategy 
creates a self reinforcing loop.

167
00:11:28,760 --> 00:11:32,680
It funds the companies most 
likely to buy its products and 

168
00:11:32,680 --> 00:11:36,280
in doing so, ensures continued 
demand for its hardware. 

169
00:11:36,640 --> 00:11:41,120
This dual role raises questions 
about conflicts of interest and 

170
00:11:41,120 --> 00:11:45,080
fair market practices. 
Preferential treatment can 

171
00:11:45,080 --> 00:11:49,280
distort competition, not because
NVIDIA is denying access 

172
00:11:49,320 --> 00:11:53,360
outright to certain competitors,
but because it's subtly shaping 

173
00:11:53,360 --> 00:11:56,360
the playing field through 
capital and supply chain 

174
00:11:56,360 --> 00:12:00,320
influence. 
The term Aqua hire is used in 

175
00:12:00,320 --> 00:12:03,280
finance to describe the 
acquisition of a company 

176
00:12:03,520 --> 00:12:07,920
primarily to hire its employees 
rather than for the operating 

177
00:12:07,920 --> 00:12:10,760
business. 
It's a strategy used by larger 

178
00:12:10,760 --> 00:12:15,520
companies to quickly acquire 
specific talent, particularly in

179
00:12:15,520 --> 00:12:18,920
industries where skill teams can
be hard to find through 

180
00:12:18,920 --> 00:12:23,240
traditional recruitment. 
The Information reported last 

181
00:12:23,240 --> 00:12:28,080
week that Meta is in advance 
talks to hire the prominent AI 

182
00:12:28,080 --> 00:12:32,640
investor, not Friedman and 
Daniel Gross, to help lead it's 

183
00:12:32,720 --> 00:12:36,680
AI efforts, and that as part of 
those talks, Meta might 

184
00:12:36,680 --> 00:12:41,200
partially buy out Friedman and 
Gross's venture capital fund, 

185
00:12:41,440 --> 00:12:46,040
which holds stakes in top AI 
startups and is worth billions 

186
00:12:46,040 --> 00:12:49,680
of dollars on paper. 
If the talks are successful, 

187
00:12:49,920 --> 00:12:53,760
Gross would leave Safe Super 
Intelligence, which he Co 

188
00:12:53,760 --> 00:12:58,120
founded with the former Open AI 
chief scientist last year, and 

189
00:12:58,120 --> 00:13:01,120
work mostly on AI products at 
Meta. 

190
00:13:01,760 --> 00:13:06,360
As part of the deal, Meta could 
end up owning minority stakes in

191
00:13:06,360 --> 00:13:09,600
the start-ups that the venture 
capital firm has invested in, 

192
00:13:09,880 --> 00:13:14,000
but allegedly won't get 
information about and control 

193
00:13:14,200 --> 00:13:17,320
over these start-ups, according 
to the article. 

194
00:13:17,920 --> 00:13:22,160
Drayton da Silva's Sub Stack 
article shows how Meta broke its

195
00:13:22,160 --> 00:13:27,120
non acquisition of Scale AI into
4 main parts, each of which 

196
00:13:27,120 --> 00:13:30,440
could raise red flags under US 
antitrust law. 

197
00:13:30,840 --> 00:13:35,760
Meta bought 49% of Scale's 
shares, but with no voting 

198
00:13:35,760 --> 00:13:40,080
rights, they got exclusive 
access to Scale's data labeling 

199
00:13:40,080 --> 00:13:43,800
pipeline, a key resource for 
training AI models. 

200
00:13:44,040 --> 00:13:48,800
They hired Scale's CEO and key 
staff, meaning that competitors 

201
00:13:48,800 --> 00:13:53,800
could no longer access them, and
they kept Scale CEO on Scales 

202
00:13:53,800 --> 00:13:57,200
board while he leads Metas AI 
Lab. 

203
00:13:57,600 --> 00:14:01,080
Each of these moves on its own 
might seem harmless, but 

204
00:14:01,080 --> 00:14:05,200
together they give Meta almost 
complete control without calling

205
00:14:05,200 --> 00:14:08,680
the deal a merger. 
Drayton points out the 

206
00:14:08,680 --> 00:14:11,960
traditional antitrust 
enforcement was built around 

207
00:14:11,960 --> 00:14:15,920
clear ownership and control, but
in the world of technology, 

208
00:14:16,120 --> 00:14:20,160
influence can be exerted through
contracts, infrastructure and 

209
00:14:20,160 --> 00:14:24,120
informal arrangements. 
The FTC and the Department of 

210
00:14:24,120 --> 00:14:28,760
Justice have recently started to
adopt A substance over form 

211
00:14:28,760 --> 00:14:33,280
approach, focusing on the 
economic reality of deals rather

212
00:14:33,280 --> 00:14:37,880
than their legal structure. 
The FTC has published 11 merger 

213
00:14:37,880 --> 00:14:44,360
guidelines, and Drayton focuses 
on Guidelines 3456 and 11 in his

214
00:14:44,360 --> 00:14:47,760
piece, which are most relevant 
to this structured deal. 

215
00:14:48,440 --> 00:14:52,600
Guideline 11 is about how 
minority stakes can still give 

216
00:14:52,600 --> 00:14:58,040
control, and under this deal, 
Meta owns 49% of scale and might

217
00:14:58,040 --> 00:15:01,240
be able to influence decisions 
with that leverage. 

218
00:15:01,560 --> 00:15:05,240
Guideline 5 is about how 
blocking rivals from access to 

219
00:15:05,240 --> 00:15:09,880
key resources is bad, and under 
this deal Meta gets exclusive 

220
00:15:09,880 --> 00:15:14,480
access to scales data. 
Guideline 4 is about how hiring 

221
00:15:14,480 --> 00:15:18,840
away key people can kill 
competition, and in this deal 

222
00:15:19,000 --> 00:15:22,960
Meta is hiring away scale CEO 
and other top talent. 

223
00:15:23,320 --> 00:15:27,200
Guideline 3 is about how shared 
board members can lead to 

224
00:15:27,200 --> 00:15:31,640
collusion, and in this deal 
scale CEO is on both sides. 

225
00:15:31,880 --> 00:15:36,280
Guideline 6 is about how deals 
that make a dominant firm even 

226
00:15:36,280 --> 00:15:40,720
stronger are risky, and Meta 
weakens its AI rivals by 

227
00:15:40,720 --> 00:15:43,000
disrupting their access to 
scale. 

228
00:15:43,600 --> 00:15:48,040
Drayton argues that even if the 
FTC or the courts blocked the 

229
00:15:48,040 --> 00:15:50,960
deal, Zuckerberg already has a 
win. 

230
00:15:51,280 --> 00:15:55,160
Open AIXAI. 
Google and Microsoft have 

231
00:15:55,160 --> 00:16:00,200
already scaled back or paused 
their work with Scale AI, citing

232
00:16:00,400 --> 00:16:03,960
data privacy risks and conflict 
of interest issues. 

233
00:16:04,280 --> 00:16:08,800
Anthropic has not yet announced 
anything publicly, but it also 

234
00:16:08,800 --> 00:16:13,000
works with scale. 
Together, these five labs plan 

235
00:16:13,000 --> 00:16:19,520
to spend over $250 billion on AI
related CapEx in 2025. 

236
00:16:19,800 --> 00:16:24,240
Meta has already induced months 
long delays and disruptions to 

237
00:16:24,240 --> 00:16:27,960
their critical workflows, likely
costing them billions. 

238
00:16:28,280 --> 00:16:32,640
Drayton estimates that setting 
up and announcing this deal cost

239
00:16:32,640 --> 00:16:37,400
Meta approximately $40 million 
across external advisors and 

240
00:16:37,400 --> 00:16:41,160
internal resources over six 
months, which is a very 

241
00:16:41,160 --> 00:16:44,400
effective asymmetric attack on 
their rivals. 

242
00:16:45,120 --> 00:16:49,280
He finishes up his piece with an
amusing playbook for non 

243
00:16:49,280 --> 00:16:53,280
acquisitions for Tech CE OS 
Hoping to avoid regulatory 

244
00:16:53,280 --> 00:16:55,880
scrutiny. 
I'll leave a link to a sub stack

245
00:16:55,880 --> 00:16:59,720
in the video description. 
It's worth checking out the 

246
00:16:59,720 --> 00:17:03,880
deals being struck by Meta. 
Microsoft's Open AI partnership 

247
00:17:04,079 --> 00:17:08,800
and Nvidia's customer investment
loop all challenge regulators to

248
00:17:08,800 --> 00:17:12,800
look beyond formal mergers in 
assessing competition within an 

249
00:17:12,800 --> 00:17:15,079
industry. 
While the deals that are being 

250
00:17:15,079 --> 00:17:18,640
struck might not trigger 
traditional review thresholds, 

251
00:17:18,880 --> 00:17:22,680
their cumulative effect can 
without a doubt be expected to 

252
00:17:22,680 --> 00:17:25,359
reshape competition within an 
industry. 

253
00:17:26,160 --> 00:17:30,600
Big Tech CE OS don't seem to 
just expect AI to be a 

254
00:17:30,600 --> 00:17:34,280
technological shift, but a 
structural shift too. 

255
00:17:34,640 --> 00:17:37,840
They believe that as 
foundational models become the 

256
00:17:37,840 --> 00:17:41,800
engines of the next digital 
economy, the companies that 

257
00:17:41,800 --> 00:17:45,400
control their development, 
deployment and distribution 

258
00:17:45,640 --> 00:17:48,960
could end up wielding 
extraordinary influence. 

259
00:17:49,240 --> 00:17:52,920
Unlike past industrial 
consolidations, this one is 

260
00:17:52,920 --> 00:17:56,600
being shaped less through 
traditional mergers and more 

261
00:17:56,600 --> 00:18:01,240
through strategic entanglements.
While regulators may not have 

262
00:18:01,240 --> 00:18:05,120
noticed this in the past, the 
sheer volume of these deals has 

263
00:18:05,120 --> 00:18:07,240
brought this issue to their 
attention. 

264
00:18:07,880 --> 00:18:11,320
At present, it would appear that
there is plenty of competition 

265
00:18:11,320 --> 00:18:13,840
in the AI space. 
There are a number of 

266
00:18:13,840 --> 00:18:17,240
competitors and they're all 
racing to develop better and 

267
00:18:17,240 --> 00:18:20,320
better models. 
Some of the huge investment in 

268
00:18:20,320 --> 00:18:24,840
the space has quite likely been 
driven by a belief that AI will 

269
00:18:24,840 --> 00:18:29,200
have winner take all economics 
like Internet search did, where 

270
00:18:29,200 --> 00:18:32,880
Google built the best search 
engine many years ago and then 

271
00:18:32,880 --> 00:18:35,880
dominated that sector for the 
next few decades. 

272
00:18:36,400 --> 00:18:40,120
Since the DeepSeek breakthrough 
earlier this year, there's been 

273
00:18:40,120 --> 00:18:44,040
growing concern amongst 
investors that while it might be

274
00:18:44,040 --> 00:18:48,440
extremely expensive to achieve 
AI breakthroughs, the new models

275
00:18:48,440 --> 00:18:51,760
once built might be quite easy 
to replicate. 

276
00:18:52,000 --> 00:18:56,280
Meaning that there may not be a 
dominant AI company in the near 

277
00:18:56,280 --> 00:19:00,320
future and that the industry 
could instead remain extremely 

278
00:19:00,320 --> 00:19:03,480
competitive with very little 
pricing power. 

279
00:19:04,040 --> 00:19:08,600
Meta, Microsoft, Amazon, Google 
and NVIDIA are without a doubt 

280
00:19:08,600 --> 00:19:13,640
trying to build moats around AI1
contract at a time, but 

281
00:19:13,640 --> 00:19:18,800
regulators do appear to already 
be on to what's going on and 

282
00:19:18,800 --> 00:19:21,800
could shut these deals down if 
they appear to be harming 

283
00:19:21,800 --> 00:19:25,520
competition in the industry. 
We've all seen the sudden 

284
00:19:25,520 --> 00:19:28,920
transformation of Mark 
Zuckerberg from the left-leaning

285
00:19:28,920 --> 00:19:32,280
tech nerd who wrote an open 
letter to his child about his 

286
00:19:32,280 --> 00:19:36,000
goal of building inclusive and 
welcoming communities while 

287
00:19:36,000 --> 00:19:39,920
pouring millions into social 
justice causes into the gold 

288
00:19:39,920 --> 00:19:44,760
chain wearing MMA bro who told 
Joe Rogan that corporate America

289
00:19:44,760 --> 00:19:49,400
had been culturally neutered and
workplaces needed more masculine

290
00:19:49,400 --> 00:19:52,480
energy. 
This all happened right around 

291
00:19:52,480 --> 00:19:56,600
the time of Trump's election 
victory, and this metamorphosis 

292
00:19:56,880 --> 00:20:00,040
could be explained as much by 
Zuckerberg's need to align 

293
00:20:00,040 --> 00:20:03,400
himself with the new 
administration as an actual 

294
00:20:03,400 --> 00:20:06,440
change. 
In his beliefs, should any of 

295
00:20:06,440 --> 00:20:10,320
these deals be deemed anti 
competitive, being aligned with 

296
00:20:10,320 --> 00:20:13,760
the current administration, 
along with the purchase of the 

297
00:20:13,760 --> 00:20:17,680
correct meme coins could solve a
lot of Mark Zuckerberg's 

298
00:20:17,680 --> 00:20:21,320
problems. 
As an example, the US Supreme 

299
00:20:21,320 --> 00:20:25,920
Court unanimously upheld a 
federal law requiring TikTok to 

300
00:20:25,920 --> 00:20:29,960
either be sold by its Chinese 
parent company or be banned in 

301
00:20:29,960 --> 00:20:32,960
the United States more than six 
months ago. 

302
00:20:33,200 --> 00:20:36,800
But enforcement and has been 
repeatedly delayed by executive 

303
00:20:36,800 --> 00:20:41,120
orders from the US president, 
who even invited its CEO to his 

304
00:20:41,120 --> 00:20:44,680
inauguration. 
It's not obvious that any of the

305
00:20:44,680 --> 00:20:48,360
big tech firms have anything 
like a monopoly on AI. 

306
00:20:48,360 --> 00:20:52,840
At present, it appears to be an 
extremely competitive space. 

307
00:20:53,120 --> 00:20:56,840
While the technology is driving 
massive investment and strategic

308
00:20:56,840 --> 00:21:01,600
growth, profitability remains 
elusive for most AI companies 

309
00:21:01,840 --> 00:21:04,800
other than NVIDIA. 
According to the Wall Street 

310
00:21:04,800 --> 00:21:10,880
Journal, Microsoft loses around 
$20 per user per month on GitHub

311
00:21:10,880 --> 00:21:14,400
Copilot. 
The expectation for these 

312
00:21:14,400 --> 00:21:18,920
companies, whose core businesses
are very profitable, appears to 

313
00:21:18,920 --> 00:21:23,120
be that long term gains will 
come as AI becomes more 

314
00:21:23,120 --> 00:21:26,920
efficient, cheaper to run, and 
more deeply integrated into 

315
00:21:26,920 --> 00:21:30,600
enterprise workflows. 
At present they have a lot of 

316
00:21:30,600 --> 00:21:34,920
money and are spending some of 
it on generative AI to make sure

317
00:21:34,920 --> 00:21:37,440
that they are still competitive 
in the future. 

318
00:21:37,720 --> 00:21:41,920
They're also spending this money
as a moon shot investment where 

319
00:21:41,920 --> 00:21:45,880
maybe they'll develop a dominant
product which will matter as 

320
00:21:45,880 --> 00:21:50,240
much to the future Internet as 
search did to the Internet 25 

321
00:21:50,240 --> 00:21:53,400
years ago. 
The Economist made a very good 

322
00:21:53,400 --> 00:21:57,640
argument in a recent article 
that the FTC case calling for 

323
00:21:57,640 --> 00:22:00,480
the breakup of Mehta is without 
merit. 

324
00:22:00,840 --> 00:22:05,320
They say that the FTC's argument
against Mehta wasn't compelling 

325
00:22:05,320 --> 00:22:09,160
when it was launched in 2020, 
and in the years since the 

326
00:22:09,160 --> 00:22:13,920
complaint was filed, it's only 
become less persuasive. the FT 

327
00:22:13,920 --> 00:22:17,720
CS claim in the case is that 
network effects create 

328
00:22:17,720 --> 00:22:20,240
insurmountable barriers to 
entry. 

329
00:22:20,520 --> 00:22:23,880
But The Economist points out 
that the last few years have 

330
00:22:23,880 --> 00:22:27,600
seen a rush of new successful 
social media start-ups. 

331
00:22:27,880 --> 00:22:30,760
Some have fizzled, like 
Clubhouse and Be Real. 

332
00:22:31,040 --> 00:22:34,480
Others remain small, like Blue 
Sky and Truth Social. 

333
00:22:34,760 --> 00:22:39,440
But the arrival of TikTok made a
mockery of the idea that the 

334
00:22:39,440 --> 00:22:41,720
market was closed to new 
entrants. 

335
00:22:42,240 --> 00:22:46,680
They point out that partially 
because of TikTok, social media 

336
00:22:46,680 --> 00:22:51,000
apps now revolve around short 
form video, which has put metas,

337
00:22:51,000 --> 00:22:55,200
Instagram and Facebook in 
competition with YouTube, Shorts

338
00:22:55,200 --> 00:22:58,920
and Twitter, who are all 
pursuing the same short form 

339
00:22:58,920 --> 00:23:01,560
video audience that TikTok drew 
in. 

340
00:23:02,200 --> 00:23:06,000
They highlight that Spotify, 
which was originally a music 

341
00:23:06,000 --> 00:23:08,600
service, is pushing into video 
too. 

342
00:23:09,000 --> 00:23:13,320
Roblox, a gaming platform, acts 
as a social network for many 

343
00:23:13,320 --> 00:23:16,360
teenagers. 
They say that as competition 

344
00:23:16,360 --> 00:23:20,680
spreads across previously 
separate categories, the FT CS 

345
00:23:20,720 --> 00:23:24,120
narrow market definition looks 
ever sillier. 

346
00:23:24,800 --> 00:23:27,960
There's no doubt that all of 
these AI deals are being 

347
00:23:27,960 --> 00:23:31,680
specifically designed to work 
around the FTC rulebook. 

348
00:23:31,920 --> 00:23:35,400
And it does appear that some of 
the deals are anti competitive 

349
00:23:35,400 --> 00:23:39,880
in nature in that they're being 
done to block access to key AI 

350
00:23:39,880 --> 00:23:42,600
services, slowing down 
competitors. 

351
00:23:42,880 --> 00:23:46,800
But this scramble to invest and 
build better and better models 

352
00:23:46,960 --> 00:23:51,440
while keeping prices low doesn't
appear to be harming consumers 

353
00:23:51,640 --> 00:23:54,720
and reflects how competitive the
space is at present. 

354
00:23:55,680 --> 00:23:58,920
While these deals are worth 
keeping an eye on, it's not 

355
00:23:58,920 --> 00:24:02,200
obvious to me that they'll be 
stopped anytime soon. 

356
00:24:02,520 --> 00:24:04,800
Thanks for tuning into this 
week's podcast. 

357
00:24:04,920 --> 00:24:07,760
If you found it interesting, 
don't forget to subscribe and 

358
00:24:07,760 --> 00:24:11,360
hit the bell icon so that you're
alerted to new episodes. 

359
00:24:11,680 --> 00:24:14,640
A special thanks to my 
supporters on Patreon whose 

360
00:24:14,640 --> 00:24:16,840
support makes this podcast 
happen. 

361
00:24:17,120 --> 00:24:20,080
If you want to support the 
podcast, there's a link in the 

362
00:24:20,080 --> 00:24:22,600
show notes. 
Have a great week and talk to 

363
00:24:22,600 --> 00:24:24,000
you again soon. 
Bye.