【C&K GO!】AI投资风险加剧！4个维度拆解市场递增逻辑帮你避险

1
00:00:00,220 --> 00:00:02,419
That we are in some form of a bubble

2
00:00:02,419 --> 00:00:05,459
Are we over spending does the math of all of this

3
00:00:05,459 --> 00:00:06,549
Make sense

4
00:00:06,549 --> 00:00:09,130
So this is really complicated

5
00:00:09,130 --> 00:00:12,609
And i want to separate out the technological side of it

6
00:00:12,609 --> 00:00:16,260
From the economic side of it on the technological side of it

7
00:00:16,260 --> 00:00:18,179
I feel really solid

8
00:00:18,179 --> 00:00:20,420
I think i'm one of the most bullish people around

9
00:00:20,420 --> 00:00:23,910
And i think it pencils out on the economic side

10
00:00:23,910 --> 00:00:25,170
You know i i

11
00:00:25,170 --> 00:00:28,719
I have my concerns where even if the technology is is

12
00:00:28,719 --> 00:00:31,899
You know really powerful and fulfills all its promises

13
00:00:31,899 --> 00:00:33,960
I think there may be players in the ecosystem

14
00:00:33,960 --> 00:00:35,979
Who if they just make a timing error

15
00:00:35,979 --> 00:00:37,899
If they just get it off by a little bit

16
00:00:37,899 --> 00:00:39,039
Bad things could happen

17
00:00:39,039 --> 00:00:39,340
Hello

18
00:00:39,340 --> 00:00:39,799
大家好

19
00:00:39,799 --> 00:00:41,640
欢迎来到c k go频道

20
00:00:41,640 --> 00:00:43,340
2026年2月

21
00:00:43,340 --> 00:00:46,179
谷歌做了一件让金融圈震惊的事

22
00:00:46,179 --> 00:00:48,729
他们发行了百年债券

23
00:00:48,729 --> 00:00:49,950
这意味着说

24
00:00:49,950 --> 00:00:52,250
如果你现在买了这张债券

25
00:00:52,250 --> 00:00:55,689
要等到2126年才能拿回本金

26
00:00:55,689 --> 00:00:58,070
为什么一家曾经轻资产

27
00:00:58,070 --> 00:00:59,710
靠着一行行代码

28
00:00:59,710 --> 00:01:02,670
就能躺着赚取广告钱的科技巨头

29
00:01:02,670 --> 00:01:06,250
现在却要借这跨越一个世纪的巨款呢

30
00:01:06,250 --> 00:01:08,978
缺钱呢还不仅仅是谷歌一家

31
00:01:08,978 --> 00:01:11,269
2026年1月16日

32
00:01:11,269 --> 00:01:15,230
OpenAI的CEOSAMODMAN发推文说

33
00:01:15,230 --> 00:01:19,299
我们正在开始在ChatGPT当中测试广告

34
00:01:19,299 --> 00:01:22,709
很显然很多人用AI却不愿意付钱

35
00:01:22,709 --> 00:01:24,670
为了让公司活下去

36
00:01:24,670 --> 00:01:27,719
广告成了OpenAI的最后手段

37
00:01:27,719 --> 00:01:29,640
这绝不是危言耸听

38
00:01:29,640 --> 00:01:33,280
而是sam omen早在2024年10月的时候

39
00:01:33,280 --> 00:01:35,150
自己亲口说的

40
00:01:35,150 --> 00:01:38,609
I kind of think of ads as like a last resort for us for a business model

41
00:01:38,609 --> 00:01:39,049
Um

42
00:01:39,049 --> 00:01:39,730
I would do it

43
00:01:39,730 --> 00:01:42,810
If it meant that was the only way to get everybody on the world

44
00:01:42,810 --> 00:01:43,349
In the world

45
00:01:43,349 --> 00:01:45,450
Like access to great services

46
00:01:45,450 --> 00:01:47,730
But if we can find something that doesn't do that

47
00:01:47,730 --> 00:01:48,569
I'd prefer that

48
00:01:48,569 --> 00:01:52,989
以AI模型cloud闻名的ANTHROPIC创始人以及CEO

49
00:01:52,989 --> 00:01:54,539
近期的一次访谈

50
00:01:54,539 --> 00:01:56,618
可以说很好地道出了

51
00:01:56,618 --> 00:01:59,789
2026年AI投资当中的风险

52
00:01:59,789 --> 00:02:00,608
市场

53
00:02:00,608 --> 00:02:02,769
好像也对AI这个词脱敏了

54
00:02:02,769 --> 00:02:06,129
比起去年的提到AI股价就上涨

55
00:02:06,129 --> 00:02:07,510
现在却是财报

56
00:02:07,510 --> 00:02:11,818
提及AI市场反倒可能砸盘诸神之战

57
00:02:11,818 --> 00:02:16,169
这个视频我们用最直观的四个维度来拆解一下

58
00:02:16,169 --> 00:02:20,439
2026年AI投资必须知道的底层逻辑

59
00:02:20,439 --> 00:02:23,580
第一类是我们最熟悉的SAAS公司

60
00:02:23,580 --> 00:02:25,889
像是salesforce adobe

61
00:02:25,889 --> 00:02:27,449
那么告诉投资者

62
00:02:27,449 --> 00:02:29,870
AI让我们的产品更好用了

63
00:02:29,870 --> 00:02:33,849
员工的效率也因为使用AI工具更高了

64
00:02:33,849 --> 00:02:36,990
曾经的cs公司是华尔街的宠儿

65
00:02:36,990 --> 00:02:40,409
写一套代码成为的软件可以卖给百万人

66
00:02:40,409 --> 00:02:43,229
使用边际成本几乎为零

67
00:02:43,229 --> 00:02:45,930
2026年的软件及服务

68
00:02:45,930 --> 00:02:47,930
确实面临着巨大的挑战

69
00:02:47,930 --> 00:02:50,069
归根结底原因在于

70
00:02:50,069 --> 00:02:54,110
曾经的SAS是企业员工使用的辅助工具

71
00:02:54,110 --> 00:02:58,409
现在新增的AI运用大多是通过聊天窗口

72
00:02:58,409 --> 00:03:02,370
可能可以帮员工更有效率地总结分析数据

73
00:03:02,370 --> 00:03:04,139
写写邮件草稿

74
00:03:04,139 --> 00:03:05,580
但到了2026年

75
00:03:05,580 --> 00:03:08,360
这些企业的高层期待不一样了

76
00:03:08,360 --> 00:03:10,710
他们想要的软件即服务

77
00:03:10,710 --> 00:03:13,750
是可以帮他们直接减少人工开支的

78
00:03:13,750 --> 00:03:15,039
AI代理

79
00:03:15,280 --> 00:03:16,879
什么是AI代理呢

80
00:03:16,879 --> 00:03:19,800
就是它不能够仅仅是一个聊天窗口

81
00:03:19,800 --> 00:03:22,800
而是需要AI本身成为你的客服

82
00:03:22,800 --> 00:03:24,019
你的会计

83
00:03:24,019 --> 00:03:26,859
这些SAAS公司正在疯狂转型

84
00:03:26,859 --> 00:03:30,639
好让自己能够继续被客户企业看得上眼

85
00:03:30,639 --> 00:03:33,379
市场却已经开始抱有怀疑

86
00:03:33,379 --> 00:03:36,960
这就好比当电力刚刚被普及使用

87
00:03:36,960 --> 00:03:40,819
一家制冰公司可能还在说自己怎么在用电

88
00:03:40,819 --> 00:03:43,490
帮客户把冰制造的更快一些

89
00:03:43,490 --> 00:03:45,530
储存时间能够更久一些

90
00:03:46,569 --> 00:03:50,729
结果市场当中直接出现了电冰箱空调

91
00:03:50,849 --> 00:03:53,030
当然AI就目前为止

92
00:03:53,030 --> 00:03:56,270
对SAAS企业的威胁还是有区别的

93
00:03:56,270 --> 00:03:58,240
并不能一棍子打死

94
00:03:58,240 --> 00:04:00,780
CK觉得这之中可以分为两类

95
00:04:00,780 --> 00:04:03,849
第一类为结构性嵌入的公司

96
00:04:03,849 --> 00:04:07,930
第二类则主要是以提升效率为核心的公司

97
00:04:07,930 --> 00:04:09,530
AI目前的进展

98
00:04:09,530 --> 00:04:11,870
对第二类公司所构成的威胁

99
00:04:11,870 --> 00:04:14,669
远远超过第一类市场

100
00:04:14,669 --> 00:04:17,658
目前对于这一维度企业股价的打压

101
00:04:17,658 --> 00:04:20,519
正是因为担心第二维度的公司

102
00:04:20,519 --> 00:04:23,158
能够有朝一日发明出电冰箱

103
00:04:23,158 --> 00:04:24,619
空调

104
00:04:24,769 --> 00:04:27,029
这一维度是AI的大脑

105
00:04:27,029 --> 00:04:29,129
比如g b d five claw

106
00:04:29,129 --> 00:04:29,509
Four

107
00:04:29,509 --> 00:04:31,750
Gemini two等等的开发者

108
00:04:31,750 --> 00:04:35,970
这些开发者们因为拥有着底层的大语言模型

109
00:04:35,970 --> 00:04:39,029
开启了对应用方的毛利率侵蚀

110
00:04:39,029 --> 00:04:41,699
例如第一维度的SARS公司

111
00:04:41,699 --> 00:04:43,639
以前作为软件公司

112
00:04:43,639 --> 00:04:46,220
一套代码卖给百万人使用

113
00:04:46,220 --> 00:04:49,100
现在却是用户调用一次AI功能

114
00:04:49,100 --> 00:04:51,420
SARS公司就要付一笔钱

115
00:04:51,420 --> 00:04:54,160
给底层的大语言模型拥有者

116
00:04:54,160 --> 00:04:57,480
成为了仰人鼻息的中间人的存在

117
00:04:57,480 --> 00:05:02,300
但同时大语言模型拥有者玩的也是烧钱的游戏

118
00:05:02,300 --> 00:05:04,160
OpenAI的GB t five

119
00:05:04,160 --> 00:05:07,769
预计就可能烧了至少10亿来训练

120
00:05:07,769 --> 00:05:09,230
更糟糕的是

121
00:05:09,230 --> 00:05:12,389
即使你烧了大量资金推出的模型

122
00:05:12,389 --> 00:05:15,920
并不能保证你在市场当中的领导地位

123
00:05:15,920 --> 00:05:17,800
你得不停的升级

124
00:05:17,800 --> 00:05:22,160
并且需要购买足够的算力来确保模型的运行

125
00:05:22,160 --> 00:05:26,430
The idea they are building the data centers

126
00:05:26,430 --> 00:05:27,629
Has a long lag time

127
00:05:27,629 --> 00:05:29,009
It's like a year or two

128
00:05:29,009 --> 00:05:31,230
So i have to decide now literally now

129
00:05:31,230 --> 00:05:33,379
Or in some cases a few months ago

130
00:05:33,379 --> 00:05:37,110
How how how much compute i need to buy in

131
00:05:37,110 --> 00:05:37,930
You know early

132
00:05:37,930 --> 00:05:40,170
Twenty four to serve the models in early

133
00:05:40,170 --> 00:05:41,110
Twenty twenty seven

134
00:05:42,810 --> 00:05:44,009
And there's two couple

135
00:05:44,009 --> 00:05:44,670
Dangers

136
00:05:44,670 --> 00:05:45,589
One is that

137
00:05:45,589 --> 00:05:46,230
If i if

138
00:05:46,230 --> 00:05:47,730
I don't buy enough compute

139
00:05:47,730 --> 00:05:49,850
I won't be able to serve all the customers

140
00:05:49,850 --> 00:05:51,170
I want have to turn them away

141
00:05:51,170 --> 00:05:53,230
And send them to to to my competitors

142
00:05:53,230 --> 00:05:54,899
If i buy too much compute

143
00:05:54,899 --> 00:05:58,009
Of course i might not get enough for revenue to pay for that compute

144
00:05:58,009 --> 00:05:59,670
And and and you know

145
00:05:59,670 --> 00:06:01,310
And and you know in the extreme case

146
00:06:01,310 --> 00:06:03,579
There's kind of the risk of going bankrupt

147
00:06:03,579 --> 00:06:06,100
正如sam oven正在经历的

148
00:06:06,100 --> 00:06:10,870
光靠大语言模型本身很难带来持续的营收盈利

149
00:06:10,870 --> 00:06:12,490
这在ck看来

150
00:06:12,490 --> 00:06:16,379
纯粹的AI模型目前几乎没有网络效应

151
00:06:16,379 --> 00:06:18,579
回想在互联网时代

152
00:06:18,579 --> 00:06:21,209
社交媒体开始热门的时候

153
00:06:21,209 --> 00:06:24,689
你使用FACEBOOK不是因为它的代码写得好

154
00:06:24,689 --> 00:06:29,019
而是因为你的朋友们都在那里迁移到别的app

155
00:06:29,019 --> 00:06:31,139
意味着失去你的网络社交

156
00:06:31,139 --> 00:06:34,180
这就是所谓的直接网络效应

157
00:06:34,180 --> 00:06:35,860
现在的AI时代呢

158
00:06:35,860 --> 00:06:39,620
模型本身的价值却是在于模型的智力

159
00:06:39,620 --> 00:06:41,060
任何竞争对手

160
00:06:41,060 --> 00:06:44,889
理论上都可以在明天就追平或者是超越

161
00:06:44,889 --> 00:06:48,470
如果明天新一代的cloud模型问世

162
00:06:48,470 --> 00:06:51,750
并且比现在的GEMINI聪明10%

163
00:06:51,750 --> 00:06:53,769
或者是便宜五美元

164
00:06:53,769 --> 00:06:58,019
你只需要将指令迁移到新的模型即可

165
00:06:58,019 --> 00:07:00,459
几个人一起做group assignment

166
00:07:00,459 --> 00:07:03,339
可以各自使用着不同的AI模型辅助

167
00:07:03,339 --> 00:07:06,009
但对合作没有任何的障碍

168
00:07:06,009 --> 00:07:07,589
没有任何的社群

169
00:07:07,589 --> 00:07:11,069
能够把你锁定在特定的聊天机器人上

170
00:07:11,069 --> 00:07:13,709
但如果模型想要生存下去

171
00:07:13,709 --> 00:07:15,670
确实需要烧更多的资金

172
00:07:15,670 --> 00:07:18,850
从而让自己的模型在智力排行榜上

173
00:07:18,850 --> 00:07:20,509
拥有一席之地

174
00:07:20,829 --> 00:07:24,949
这之中的高成本的支出和低粘性的营收之间

175
00:07:24,949 --> 00:07:26,189
巨大的缺口

176
00:07:26,189 --> 00:07:26,990
将意味着

177
00:07:26,990 --> 00:07:31,519
这之中的赢家往往不会是数学最强的那一个

178
00:07:31,519 --> 00:07:35,779
而是能够将数学隐藏在用户每天都必须使用的

179
00:07:35,779 --> 00:07:38,180
生态系统当中的那个模型

180
00:07:38,180 --> 00:07:40,160
在购买昂贵的数据

181
00:07:40,160 --> 00:07:42,139
高薪招揽顶级科学家

182
00:07:42,139 --> 00:07:44,399
来训练下一个顶级模型之外

183
00:07:44,399 --> 00:07:47,459
还有着算力这一个昂贵的支出

184
00:07:47,459 --> 00:07:52,310
这也让制造算力成为了AI投资的第三维度

185
00:07:52,709 --> 00:07:54,069
在AI时代

186
00:07:54,069 --> 00:07:55,649
算力就是石油

187
00:07:55,649 --> 00:07:57,269
谁拥有的数据中心

188
00:07:57,269 --> 00:08:01,100
多数据中心建造以及运营的效率更高

189
00:08:01,100 --> 00:08:04,139
谁就拥有了石油开采的能力

190
00:08:04,139 --> 00:08:05,980
现在的美股巨头

191
00:08:05,980 --> 00:08:07,288
除去苹果

192
00:08:07,288 --> 00:08:11,209
都在尽最大可能利用迄今为止的盈利模式

193
00:08:11,209 --> 00:08:13,108
带来的庞大的现金流

194
00:08:13,108 --> 00:08:16,300
不单单是建立了自己的大元模型

195
00:08:16,300 --> 00:08:20,870
还投入了大量的CAPEX去买地买芯片

196
00:08:20,870 --> 00:08:23,589
建造数据中心来制造算力

197
00:08:23,589 --> 00:08:25,970
甚至是签署了核电厂

198
00:08:25,970 --> 00:08:28,209
让他们不单单可以自给自足

199
00:08:28,209 --> 00:08:30,829
自己的模型所需要的算力之外

200
00:08:30,829 --> 00:08:32,730
还能够出售算力

201
00:08:32,730 --> 00:08:35,269
If you want to buy a gigawatt of compute

202
00:08:35,269 --> 00:08:38,529
Right that buying the chips and building the chips

203
00:08:38,529 --> 00:08:40,658
Building everything that costs

204
00:08:40,658 --> 00:08:41,678
You know roughly

205
00:08:41,678 --> 00:08:42,158
Let's say

206
00:08:42,158 --> 00:08:44,418
Fifty billion dollars of of capital

207
00:08:44,418 --> 00:08:46,879
Expense to to kind of to kind of fund that

208
00:08:46,879 --> 00:08:49,639
And you can think of that as useful lifetime

209
00:08:49,639 --> 00:08:50,639
As people argue about it

210
00:08:50,639 --> 00:08:51,759
But maybe it's five years

211
00:08:51,759 --> 00:08:54,899
So that's ten ten billion dollars for basically

212
00:08:54,899 --> 00:08:56,210
Basically for five years

213
00:08:56,210 --> 00:08:57,690
就以谷歌为例

214
00:08:57,690 --> 00:08:58,649
2026年

215
00:08:58,649 --> 00:09:02,399
谷歌计划投入1850亿的cap acks

216
00:09:02,399 --> 00:09:05,820
为2025年914亿的翻倍

217
00:09:05,820 --> 00:09:08,799
主要就是用来建造数据中心等等的

218
00:09:08,799 --> 00:09:11,610
AI基建和OpenAI

219
00:09:11,610 --> 00:09:14,250
通过预计未来的营收来决定

220
00:09:14,250 --> 00:09:16,318
现在的资本投入不同

221
00:09:16,318 --> 00:09:20,619
谷歌当前有着2400亿积压着的未完成订单

222
00:09:20,619 --> 00:09:24,179
代表着这1850亿的CPACKS投入

223
00:09:24,179 --> 00:09:28,220
基本有着很好的以签署的算力需求作为保障

224
00:09:28,220 --> 00:09:30,679
确保了建成后的营收

225
00:09:30,679 --> 00:09:35,559
但这仍然是在用现在的现金来换取未来的营收

226
00:09:35,559 --> 00:09:39,740
这未来的营收还可能是分摊到5年的时间流入

227
00:09:39,740 --> 00:09:41,179
即使是谷歌

228
00:09:41,179 --> 00:09:44,799
也没有办法轻易承担那么大的现金流压力

229
00:09:44,799 --> 00:09:47,340
截止到2025年第四季度

230
00:09:47,340 --> 00:09:50,679
谷歌的现金储备为1268亿

231
00:09:50,679 --> 00:09:54,379
CAB bx的投入确实需要1850亿

232
00:09:54,379 --> 00:09:58,179
所以能够借到足够便宜够长期的钱

233
00:09:58,179 --> 00:10:00,679
比如现在的100年的债券

234
00:10:00,679 --> 00:10:03,419
来减轻当下的现金流压力

235
00:10:03,419 --> 00:10:06,230
是谷歌的解决方式之一

236
00:10:06,230 --> 00:10:07,909
但这也意味着

237
00:10:07,909 --> 00:10:09,389
这些软件公司

238
00:10:09,389 --> 00:10:13,029
已经脱离过去的软件类的轻资产模式

239
00:10:13,029 --> 00:10:14,309
垂直整合

240
00:10:14,309 --> 00:10:17,889
进入了典型的工业巨头的运作模式

241
00:10:18,210 --> 00:10:19,509
第四维度

242
00:10:19,509 --> 00:10:22,210
卖铲子以及制作铲子的人

243
00:10:22,210 --> 00:10:23,610
像是NVIDIA

244
00:10:23,610 --> 00:10:27,000
AMD台积电ASML

245
00:10:27,000 --> 00:10:29,000
在AI的淘金热里

246
00:10:29,000 --> 00:10:30,840
无论谁挖到金子

247
00:10:30,840 --> 00:10:33,730
卖铲子的人永远稳赚不赔

248
00:10:33,730 --> 00:10:36,190
但这个稳赚不赔的前提是

249
00:10:36,190 --> 00:10:40,230
他们需要保证更多的人保持挖金子的热情

250
00:10:40,230 --> 00:10:43,230
NVIDIACEOJENSENHUANG表示

251
00:10:43,230 --> 00:10:46,389
市场认为AI对软件公司有威胁

252
00:10:46,389 --> 00:10:48,409
这事是判断错误

253
00:10:48,409 --> 00:10:52,549
另外当想挖金子的人却没有钱买铲子时

254
00:10:52,549 --> 00:10:54,840
卖铲子的人也会大方的说

255
00:10:54,840 --> 00:10:55,620
没事

256
00:10:55,620 --> 00:10:56,679
我先给你钱

257
00:10:56,679 --> 00:10:58,120
买铲子以后

258
00:10:58,120 --> 00:11:00,080
等你挖到金子了再还给我

259
00:11:00,080 --> 00:11:00,799
Make sense

260
00:11:00,799 --> 00:11:03,519
So if you want to buy a gigawatt of compute

261
00:11:03,519 --> 00:11:06,788
Right that buying the chips and building the chips

262
00:11:06,788 --> 00:11:08,899
Building everything that costs

263
00:11:08,899 --> 00:11:09,919
You know roughly

264
00:11:09,919 --> 00:11:10,419
Let's say

265
00:11:10,419 --> 00:11:12,659
Fifty billion dollars of of capital

266
00:11:12,659 --> 00:11:15,120
Expense to to kind of to kind of fund that

267
00:11:15,120 --> 00:11:17,899
And you can think of that as useful lifetime

268
00:11:17,899 --> 00:11:18,879
As people argue about it

269
00:11:18,879 --> 00:11:20,039
But maybe it's five years

270
00:11:20,039 --> 00:11:23,149
So that's ten ten billion dollars for basically

271
00:11:23,149 --> 00:11:24,460
Basically for five years

272
00:11:24,460 --> 00:11:26,139
And and so if you're a company

273
00:11:26,139 --> 00:11:27,279
You know your your

274
00:11:27,279 --> 00:11:27,659
You know

275
00:11:27,659 --> 00:11:30,299
You're a company is making you know a ten billion dollars of revenue

276
00:11:30,299 --> 00:11:31,399
You think that's growing

277
00:11:31,399 --> 00:11:31,659
You

278
00:11:31,659 --> 00:11:32,340
Don't know how fast

279
00:11:32,340 --> 00:11:34,220
It's growing you have to make the decision

280
00:11:34,220 --> 00:11:34,539
Right now

281
00:11:34,539 --> 00:11:36,019
And you don't have fifty billion dollars

282
00:11:36,019 --> 00:11:37,100
Finance for a year

283
00:11:37,100 --> 00:11:38,740
So you're you're you know you're basically saying

284
00:11:38,740 --> 00:11:41,220
I need to get ten billion dollars of of of revenue

285
00:11:41,220 --> 00:11:42,960
Per year two years from now

286
00:11:42,960 --> 00:11:43,779
Right um

287
00:11:43,779 --> 00:11:45,679
So i don't think there's anything wrong with that

288
00:11:45,679 --> 00:11:48,828
One player has capital and has an interest

289
00:11:48,828 --> 00:11:50,089
Because they're selling the

290
00:11:50,089 --> 00:11:51,129
Because they're selling the

291
00:11:51,129 --> 00:11:52,480
You know they're selling the chips

292
00:11:52,480 --> 00:11:55,279
And the other player is pretty confident

293
00:11:55,279 --> 00:11:56,919
They'll have they'll have the revenue at the right time

294
00:11:56,919 --> 00:11:59,059
But they don't have fifty billion dollars at hand

295
00:11:59,059 --> 00:12:00,509
So i don't think there's anything

296
00:12:00,509 --> 00:12:02,230
Inappropriate about that in principle

297
00:12:02,230 --> 00:12:02,509
Now

298
00:12:02,509 --> 00:12:03,929
If you start stacking these

299
00:12:03,929 --> 00:12:05,960
Where they get to huge amounts of money

300
00:12:05,960 --> 00:12:06,759
And you're saying

301
00:12:06,759 --> 00:12:07,639
You know by by twenty

302
00:12:07,639 --> 00:12:08,399
Twenty seven or twenty

303
00:12:08,399 --> 00:12:08,759
Twenty eight

304
00:12:08,759 --> 00:12:10,899
I need to make two hundred billion dollars a year

305
00:12:10,899 --> 00:12:11,179
Then

306
00:12:11,179 --> 00:12:11,460
Yeah

307
00:12:11,460 --> 00:12:12,200
You can you can

308
00:12:12,200 --> 00:12:13,759
You can can over extend yourself

309
00:12:13,759 --> 00:12:16,690
2026年卖铲子还面临的危机

310
00:12:16,690 --> 00:12:18,769
那就是他们要确保自己

311
00:12:18,769 --> 00:12:21,769
仍然供应着市面上最好的铲子

312
00:12:21,769 --> 00:12:23,389
为了控制成本

313
00:12:23,389 --> 00:12:25,859
不让NVIDIA赚走所有的利润

314
00:12:25,859 --> 00:12:28,078
处在第三维度的巨头们

315
00:12:28,078 --> 00:12:29,038
像是微软

316
00:12:29,038 --> 00:12:29,979
谷歌等等

317
00:12:29,979 --> 00:12:34,169
都在跨足第四维度研发自己的芯片

318
00:12:34,169 --> 00:12:36,350
这场自研芯片大战呢

319
00:12:36,350 --> 00:12:39,990
将决定谁未来能够拥有更高的利润率

320
00:12:39,990 --> 00:12:41,370
值得一提的是

321
00:12:41,370 --> 00:12:43,750
当所有巨头都在烧钱

322
00:12:43,750 --> 00:12:46,230
有一个巨头却显得异常冷静

323
00:12:46,230 --> 00:12:49,929
这就是苹果一度因为没有大语言模型

324
00:12:49,929 --> 00:12:50,970
看似掉队了的

325
00:12:50,970 --> 00:12:52,539
他现在看来

326
00:12:52,539 --> 00:12:54,580
却是将自己定位在了一个

327
00:12:54,580 --> 00:12:57,399
脱离这四个维度之外的位置

328
00:12:57,399 --> 00:13:00,100
掌握终端的超级护城河

329
00:13:00,100 --> 00:13:04,779
苹果本身呢其实就可以被看作是最大的SAAS公司

330
00:13:04,779 --> 00:13:07,889
它不需要开发出世界上最强的AI

331
00:13:07,889 --> 00:13:11,389
只选择将apple intelligence作为那个空壳

332
00:13:11,389 --> 00:13:14,889
可以将市面上的大语言模型作为插件

333
00:13:14,889 --> 00:13:17,600
坐等各大模型竞争的赢家

334
00:13:17,600 --> 00:13:21,720
苹果只需要专注开发出那个最好用的界面

335
00:13:21,720 --> 00:13:26,169
同时苹果也是第四维度芯片商的隐形王者

336
00:13:26,169 --> 00:13:29,830
自主研发了apple intelligence所需要的芯片

337
00:13:29,830 --> 00:13:34,289
尽管说苹果AI的进程没有大家预想的那么顺利

338
00:13:34,289 --> 00:13:38,309
但却成功避开了现在越发拥挤的大语言模型

339
00:13:38,309 --> 00:13:41,629
以及数据中心赛道一旦成功

340
00:13:41,629 --> 00:13:46,029
这意味着苹果不需要借百年债券去盖数据中心

341
00:13:46,029 --> 00:13:49,490
它可以让全球25亿的用户自己买单

342
00:13:49,490 --> 00:13:52,100
帮他分担算力成本

343
00:13:52,379 --> 00:13:55,620
那2016年的AI竞争已进入白热化

344
00:13:55,620 --> 00:13:57,700
意味着这之中的赢家

345
00:13:57,700 --> 00:14:00,549
谁能够拥有更充足的现金流

346
00:14:00,549 --> 00:14:03,210
谁能够将科技隐藏在用户

347
00:14:03,210 --> 00:14:05,669
每天都使用的生态系统当中

348
00:14:05,669 --> 00:14:09,120
谁才最有可能成为那最终的赢家

349
00:14:09,120 --> 00:14:11,480
那今天的视频就到这里啦

350
00:14:11,480 --> 00:14:13,580
希望对大家有所帮助

351
00:14:13,580 --> 00:14:15,179
喜欢这个视频的话

352
00:14:15,179 --> 00:14:17,029
记得订阅点赞哦

353
00:14:17,029 --> 00:14:18,269
我是CK

354
00:14:18,269 --> 00:14:19,909
我们下期再见

评论采集失败

采集完成，该视频无弹幕