Ai硬件暴跌泡沫破裂了吗

1
00:00:00,300 --> 00:00:04,679
这两天站在光里的兄弟应该跌的很痛苦啊

2
00:00:04,679 --> 00:00:07,839
呃实际上这种的还不算最惨的啊

3
00:00:07,839 --> 00:00:10,160
最惨的是223年啊

4
00:00:10,160 --> 00:00:11,730
中概打一折

5
00:00:11,730 --> 00:00:13,230
那个时候比现在还惨

6
00:00:13,230 --> 00:00:15,470
但是中概那个时候呢现金流很好

7
00:00:15,470 --> 00:00:16,550
业绩也很好

8
00:00:16,550 --> 00:00:19,289
那增速呢不比现在的那些所谓的光啊

9
00:00:19,289 --> 00:00:20,859
存储啊增速慢

10
00:00:20,859 --> 00:00:22,420
那互联网有个什么特点呢

11
00:00:22,420 --> 00:00:24,219
就是边际成本

12
00:00:24,219 --> 00:00:26,120
互联网的边际成本几乎为零啊

13
00:00:26,120 --> 00:00:27,420
你比如拼多多啊

14
00:00:27,420 --> 00:00:30,890
用1.9万人达到了七到8万亿的GMV

15
00:00:30,890 --> 00:00:34,039
那阿里呢啊也是不到10万人吧

16
00:00:34,039 --> 00:00:35,560
当时做了80000EGMV

17
00:00:35,560 --> 00:00:36,859
但阿里后面人多

18
00:00:36,859 --> 00:00:38,619
就是因为收购大润发啊

19
00:00:38,619 --> 00:00:40,128
收购红红星美凯龙啊

20
00:00:40,128 --> 00:00:41,168
如果他不收购的话

21
00:00:41,168 --> 00:00:42,249
其实效率很高的啊

22
00:00:42,249 --> 00:00:42,848
包括美团啊

23
00:00:42,848 --> 00:00:43,429
京东啊

24
00:00:43,429 --> 00:00:44,408
虽然人多啊

25
00:00:44,408 --> 00:00:46,509
但他实际上他这个做平台的人不多

26
00:00:46,509 --> 00:00:48,229
所以你当时看他们的人效啊

27
00:00:48,229 --> 00:00:49,909
其实非常恐怖的对吧

28
00:00:49,909 --> 00:00:52,810
你像拼多多1万多人赚1000多亿

29
00:00:52,810 --> 00:00:53,899
什么概念

30
00:00:55,039 --> 00:00:57,899
一人净利润的话就是1000万呀

31
00:00:57,899 --> 00:00:58,259
对不对

32
00:00:58,259 --> 00:00:59,479
将近1000万净利润

33
00:00:59,479 --> 00:01:01,479
这个人效在全球都是很高的

34
00:01:01,479 --> 00:01:03,200
所以互联网的本质上

35
00:01:03,200 --> 00:01:07,340
它当时啊比这个所谓的AI硬件啊猛多了

36
00:01:07,340 --> 00:01:09,099
那跌起来呢更快啊

37
00:01:09,099 --> 00:01:12,650
互联网大概用了两年的时间不到哈

38
00:01:12,650 --> 00:01:13,909
打了一折啊

39
00:01:13,909 --> 00:01:14,870
实际上AI硬件呢

40
00:01:14,870 --> 00:01:17,680
这一波呢主要还是这些纳税企业

41
00:01:17,680 --> 00:01:19,079
麦里面的几个巨头啊

42
00:01:19,079 --> 00:01:21,780
因为他们的支出呢从去年5500亿啊

43
00:01:21,780 --> 00:01:23,099
调到今年8000个亿

44
00:01:23,099 --> 00:01:26,200
而这些支出呢造就了这些啊

45
00:01:26,200 --> 00:01:27,219
无论是CPU啊

46
00:01:27,219 --> 00:01:28,000
GPU啊

47
00:01:28,000 --> 00:01:29,799
存储啊啊链接呀

48
00:01:29,799 --> 00:01:31,000
还有这些啊

49
00:01:31,000 --> 00:01:33,120
生产这个存储设备的

50
00:01:33,120 --> 00:01:35,319
这些公司的一个繁荣景象啊

51
00:01:35,319 --> 00:01:39,849
但是呢他这些8000个亿呢实际上确实是规划

52
00:01:39,849 --> 00:01:43,180
但是呢昨天扎克伯格呢说哎呀

53
00:01:43,180 --> 00:01:44,959
他要出租剩余算力啊

54
00:01:44,959 --> 00:01:46,448
把整个市场吓坏了

55
00:01:46,448 --> 00:01:48,769
实际上你说现在算力有没有过剩呢

56
00:01:48,769 --> 00:01:49,868
肯定没有过剩

57
00:01:49,868 --> 00:01:52,620
那meta呢其实来讲

58
00:01:52,620 --> 00:01:55,480
他只是把那些老旧的H100H200啊

59
00:01:55,480 --> 00:01:56,819
然后出动完之后呢

60
00:01:56,819 --> 00:01:59,420
然后降低它的啊成本

61
00:01:59,420 --> 00:02:01,519
因为折旧的话摊销下来很吓人嘛

62
00:02:01,519 --> 00:02:03,959
这样可以延长它的产品的使命使用周期

63
00:02:03,959 --> 00:02:07,790
所以本质上它不是说这个meta就缩小了开支啊

64
00:02:07,790 --> 00:02:11,109
他没有他那些顶尖的那些CPU还是自用的嘛

65
00:02:11,109 --> 00:02:13,908
包括他还找谷歌呢去买算力嘛

66
00:02:13,908 --> 00:02:16,389
实际上算力其实是需求还是很大的

67
00:02:16,389 --> 00:02:19,509
所以这个市场呢肯定是反反复复反反复复

68
00:02:19,509 --> 00:02:22,569
但是呢你要说AI这个商业模式啊

69
00:02:22,569 --> 00:02:24,150
和互联网商业模式哪个好

70
00:02:24,150 --> 00:02:26,599
实际上我觉得AI硬件的商业模式

71
00:02:26,599 --> 00:02:28,259
还不一定比得上互联网

72
00:02:28,259 --> 00:02:28,879
那当然了

73
00:02:28,879 --> 00:02:30,969
你要对比中国的这些啊

74
00:02:30,969 --> 00:02:31,610
谷歌啊

75
00:02:31,610 --> 00:02:34,330
不中国的这个阿里巴巴呀之类的确实很差啊

76
00:02:34,330 --> 00:02:37,359
但是你要看一下这个美国的这些互联网公司啊

77
00:02:37,359 --> 00:02:39,579
你像亚马逊啊啊吧

78
00:02:39,579 --> 00:02:42,800
你像这个谷歌呀哈其实苹果也算互联网公司

79
00:02:42,800 --> 00:02:44,960
他们的人效是非常非常高的对

80
00:02:44,960 --> 00:02:47,360
但是呢如果这些互联网公司出钱的

81
00:02:47,360 --> 00:02:48,879
它的估值呢只有20多倍啊

82
00:02:48,879 --> 00:02:50,709
你像那个啊英伟达啊

83
00:02:50,709 --> 00:02:53,568
但现在大概是25倍的动态市盈率啊

84
00:02:53,568 --> 00:02:55,748
然后谷歌也是30~35倍啊

85
00:02:55,748 --> 00:02:58,620
然后亚马逊啊也是25~30倍对吧

86
00:02:58,620 --> 00:03:00,709
然后苹果的话贵1.35倍左右吧

87
00:03:00,709 --> 00:03:02,468
所以你要看一下这些公司的市盈率

88
00:03:02,468 --> 00:03:04,549
你会发现这AI硬件就高估了

89
00:03:04,549 --> 00:03:05,370
那为什么呢

90
00:03:05,370 --> 00:03:07,889
实际上如果你是个买家啊

91
00:03:07,889 --> 00:03:10,030
你能接受你的净利润20%多

92
00:03:10,030 --> 00:03:12,830
但是呢你的上游供应商的利润80%吗

93
00:03:12,830 --> 00:03:13,509
如果是我的话

94
00:03:13,509 --> 00:03:14,509
我肯定不接受呀

95
00:03:14,789 --> 00:03:16,310
因为你这个太暴力了

96
00:03:16,310 --> 00:03:19,819
所以呢这个AI的爹呢我觉得很正常

97
00:03:19,819 --> 00:03:23,099
他应该是回到一个合理的利润率

98
00:03:23,099 --> 00:03:24,860
就这个市场的量上来了

99
00:03:24,860 --> 00:03:27,629
但你不可能说我采购的量也上来了

100
00:03:27,629 --> 00:03:28,669
你的利润也上来了

101
00:03:28,669 --> 00:03:30,009
这个是不可取的

102
00:03:30,009 --> 00:03:33,430
因为任何的时候这个这个啊这个是贸易理赔

103
00:03:33,430 --> 00:03:33,789
应该这样

104
00:03:33,789 --> 00:03:34,430
比如你量大

105
00:03:34,430 --> 00:03:35,550
你就便宜点啊

106
00:03:35,550 --> 00:03:36,669
啊你要是量小的

107
00:03:36,669 --> 00:03:37,469
你就贵一点

108
00:03:37,469 --> 00:03:39,289
你不能说采购的量越大

109
00:03:39,289 --> 00:03:40,110
价格越高啊

110
00:03:40,110 --> 00:03:41,789
这绝对不符合正常的生意逻辑

111
00:03:41,789 --> 00:03:45,800
所以最终呢你这些买家不可能是全部买单的

112
00:03:45,800 --> 00:03:47,240
所以卖塔只是个导火索啊

113
00:03:47,240 --> 00:03:48,340
说这个啊

114
00:03:48,340 --> 00:03:49,989
因为出租剩余算力啊

115
00:03:49,989 --> 00:03:52,068
本质上还是自由现金流不足了对吧

116
00:03:52,068 --> 00:03:52,868
然后以战养战

117
00:03:52,868 --> 00:03:54,908
那接下来呢我觉得业务指引上

118
00:03:54,908 --> 00:03:57,360
无论是微软呀啊无论是谷歌呀

119
00:03:57,360 --> 00:03:58,800
他们也会降速

120
00:03:58,800 --> 00:03:59,860
很简单的道理嘛

121
00:03:59,860 --> 00:04:00,939
就他还会投

122
00:04:00,939 --> 00:04:03,219
但它会拉长时间会慢下来对吧

123
00:04:03,219 --> 00:04:04,909
因为大家肯定撑不住嘛

124
00:04:04,909 --> 00:04:05,870
慢下来之后呢

125
00:04:05,870 --> 00:04:08,349
带来的结果是就有时间压价了嘛是吧

126
00:04:08,349 --> 00:04:10,430
他不可能说让硬件赚这么多钱走啊

127
00:04:10,430 --> 00:04:12,189
如果说我们采购我们也不允许是吧

128
00:04:12,189 --> 00:04:12,710
特别是苹果

129
00:04:12,710 --> 00:04:14,069
你看一下苹果

130
00:04:14,069 --> 00:04:16,589
实际上就是靠库克这几年做供应链啊

131
00:04:16,589 --> 00:04:17,730
赚走了利润大头

132
00:04:17,730 --> 00:04:18,509
但话说回来

133
00:04:18,509 --> 00:04:21,088
因为他有C端的市场能力对吧

134
00:04:21,088 --> 00:04:23,509
C端是完全是基于消费者的选择

135
00:04:23,509 --> 00:04:25,348
B端不是B端的话

136
00:04:25,348 --> 00:04:26,069
我可以踩

137
00:04:26,069 --> 00:04:26,629
我可以不踩

138
00:04:26,629 --> 00:04:27,199
对不对

139
00:04:27,199 --> 00:04:28,959
所以B端的这个风险最大的

140
00:04:28,959 --> 00:04:30,160
但是增长最快的

141
00:04:30,160 --> 00:04:31,519
所以你接下来会看到一个问题

142
00:04:31,519 --> 00:04:35,120
就是到底是哎掌握C端用户的这些公司厉害呢

143
00:04:35,120 --> 00:04:37,199
还是说啊在B端呢

144
00:04:37,199 --> 00:04:40,019
因为啊因为这个需求井喷带来的

145
00:04:40,019 --> 00:04:41,220
国家报道的公司厉害呢

146
00:04:41,220 --> 00:04:42,180
我觉得还是掌握谁更

147
00:04:42,180 --> 00:04:43,660
因为C端是最稳定的对

148
00:04:43,660 --> 00:04:44,540
因为你掌握了用户

149
00:04:44,540 --> 00:04:45,899
用户的心智不会轻易改变

150
00:04:45,899 --> 00:04:46,639
比如说苹果啊

151
00:04:46,639 --> 00:04:50,370
他现在啊这个笔笔记本大大幅涨价

152
00:04:50,370 --> 00:04:51,930
他这个大货涨价带来的结果是什么呢

153
00:04:51,930 --> 00:04:53,769
不是它涨价是全行业都涨

154
00:04:53,769 --> 00:04:54,589
大家都涨

155
00:04:54,589 --> 00:04:56,329
实际上和都不涨没有太多区别

156
00:04:56,329 --> 00:04:57,709
因为它还是占有率第一名

157
00:04:57,709 --> 00:04:58,889
但他手机没涨呀

158
00:04:58,889 --> 00:04:59,290
为什么

159
00:04:59,290 --> 00:05:01,310
因为他手机的软件收费利润高呀

160
00:05:01,310 --> 00:05:03,060
然后他慢慢的去对吧

161
00:05:03,060 --> 00:05:04,300
降低这个啊

162
00:05:04,300 --> 00:05:05,699
就是通过行业的涨价之后

163
00:05:05,699 --> 00:05:07,339
整个行业的需求减少之后呢

164
00:05:07,339 --> 00:05:08,420
它慢慢的去压价

165
00:05:08,420 --> 00:05:11,189
这个这个概率肯定会肯定会形成一个现实

166
00:05:11,189 --> 00:05:12,290
比如再过一两年

167
00:05:12,290 --> 00:05:14,550
那么这些硬件的利润就从80%

168
00:05:14,550 --> 00:05:14,990
净利润啊

169
00:05:14,990 --> 00:05:16,389
可能到了30%到40

170
00:05:16,389 --> 00:05:17,459
这是一个合理区间

171
00:05:17,459 --> 00:05:19,339
那如果他的利润到30%到40的时候

172
00:05:19,339 --> 00:05:21,660
你的投资预算那就会带来问题

173
00:05:21,660 --> 00:05:24,019
比如说你现在投啊一个经验上几千个亿

174
00:05:24,019 --> 00:05:26,509
你接下来你发现你在呃计算省的湖北周期呢

175
00:05:26,509 --> 00:05:27,430
就会拉长啊

176
00:05:27,430 --> 00:05:28,730
因为这是最终的一个结果啊

177
00:05:28,730 --> 00:05:29,949
没有一届市场说

178
00:05:29,949 --> 00:05:34,589
那那卖家啊比买家利润高4~5倍的对吧

179
00:05:34,589 --> 00:05:35,569
持平可以是吧

180
00:05:35,569 --> 00:05:36,910
都保持20~30的净利润

181
00:05:36,910 --> 00:05:39,959
但是如果说还是这种50%到80精人

182
00:05:39,959 --> 00:05:41,779
这个事儿肯定不可持续

共 7 条，按点赞数排序。

采集完成，该视频无弹幕