【Ticker Symbol YOU】我正在买入这家AI巨头

1
00:00:00,060 --> 00:00:02,140
博通公司正在发生重大变化

2
00:00:04,799 --> 00:00:08,439
博通的AI芯片营收同比激增逾两倍

3
00:00:08,439 --> 00:00:11,699
他们正悄然与谷歌签订巨额协议

4
00:00:11,699 --> 00:00:12,359
Meta

5
00:00:12,359 --> 00:00:14,699
OpenAI和Anthropic

6
00:00:14,699 --> 00:00:17,309
博通能否成为AI新霸主

7
00:00:17,309 --> 00:00:20,629
这对英伟达股票未来意味着什么

8
00:00:20,629 --> 00:00:21,769
你的时间很宝贵

9
00:00:21,769 --> 00:00:23,460
让我们直接进入正题

10
00:00:23,460 --> 00:00:23,739
首先

11
00:00:23,739 --> 00:00:24,219
先说重点

12
00:00:24,219 --> 00:00:25,760
我不是来绑架你的

13
00:00:25,760 --> 00:00:26,579
这里全部内容

14
00:00:26,579 --> 00:00:27,879
我将在视频中讲解

15
00:00:27,879 --> 00:00:29,579
博通到底做什么业务

16
00:00:29,579 --> 00:00:31,079
他们如何盈利

17
00:00:31,079 --> 00:00:34,679
他们的AI软硬件如何与英伟达竞争

18
00:00:34,679 --> 00:00:38,509
最新财报透露AI革命的推进速度

19
00:00:38,509 --> 00:00:41,409
当然还有这对博通股票的影响

20
00:00:41,409 --> 00:00:42,329
因此

21
00:00:42,329 --> 00:00:45,049
博通分为两大核心业务

22
00:00:45,049 --> 00:00:48,810
首先是芯片业务，其次是芯片上的软件

23
00:00:48,810 --> 00:00:51,710
博通是顶尖半导体公司

24
00:00:51,710 --> 00:00:53,670
他们负责芯片设计

25
00:00:53,670 --> 00:00:58,530
然后将设计交给台积电等厂商制造

26
00:00:58,530 --> 00:01:00,229
博通的芯片应用于AI领域

27
00:01:00,229 --> 00:01:01,119
数据中心

28
00:01:01,119 --> 00:01:02,280
网络设备

29
00:01:02,280 --> 00:01:03,320
存储系统

30
00:01:03,320 --> 00:01:04,099
智能手机

31
00:01:04,099 --> 00:01:05,260
以及更多领域

32
00:01:05,260 --> 00:01:09,870
对投资者最重要的业务是AI和数据中心

33
00:01:09,870 --> 00:01:15,250
博通为行业巨头设计定制AI加速器

34
00:01:15,250 --> 00:01:17,370
包括谷歌的TPU项目

35
00:01:17,370 --> 00:01:19,359
Meta的自研AI芯片

36
00:01:19,359 --> 00:01:23,400
以及与OpenAI、Anthropic等新合作

37
00:01:23,400 --> 00:01:25,079
与英伟达GPU不同

38
00:01:25,079 --> 00:01:27,140
博通芯片按需定制

39
00:01:27,140 --> 00:01:31,280
针对每个客户的模型和基础设施单独调优

40
00:01:31,280 --> 00:01:32,319
此外

41
00:01:32,319 --> 00:01:37,668
博通销售连接数据中心AI芯片的高速神经系统

42
00:01:37,668 --> 00:01:40,528
具体是Tomahawk和Jericho芯片

43
00:01:40,528 --> 00:01:40,909
提供

44
00:01:40,909 --> 00:01:43,149
超高速以太网交换机和路由器

45
00:01:43,149 --> 00:01:46,319
在数千个AI加速器间传输数据

46
00:01:46,319 --> 00:01:50,519
如果你听说过百万GPU或千兆瓦级AI集群

47
00:01:50,519 --> 00:01:54,039
大量数据流量依赖博通网络芯片

48
00:01:54,039 --> 00:01:56,019
他们的第二引擎是软件业务

49
00:01:56,019 --> 00:01:57,858
主要得益于VMware

50
00:01:57,858 --> 00:01:58,858
Vmware的平台

51
00:01:58,858 --> 00:02:03,739
让企业能在同一物理服务器上运行多个独立虚拟机

52
00:02:03,739 --> 00:02:08,139
并在自有数据中心及混合云环境中统一管理大型工作负载

53
00:02:08,139 --> 00:02:10,199
以一致的方式实现

54
00:02:10,199 --> 00:02:11,658
自收购以来

55
00:02:11,658 --> 00:02:15,479
Vmware出现在博通的基础设施软件板块

56
00:02:15,479 --> 00:02:19,169
产生大量高利润率的持续性收入

57
00:02:19,169 --> 00:02:24,180
这种软硬件结合使博通不同于其他芯片股

58
00:02:24,180 --> 00:02:27,860
为数据中心的物理空气提供核心工具

59
00:02:27,860 --> 00:02:29,319
及其控制层

60
00:02:29,319 --> 00:02:34,349
每次AI基础设施支出增加博通都能多渠道获益

61
00:02:34,349 --> 00:02:38,899
说到这一点，我们来谈谈博通在AI栈中的具体位置

62
00:02:38,899 --> 00:02:42,058
以及其与英伟达的对比

63
00:02:42,058 --> 00:02:48,259
英伟达销售通用GPU和全规模服务器系统

64
00:02:48,259 --> 00:02:49,259
另一方面

65
00:02:49,259 --> 00:02:51,840
博通为少数大客户定制AI加速器

66
00:02:51,840 --> 00:02:55,300
并提供专用网络解决方案

67
00:02:55,300 --> 00:02:59,479
英伟达是即用型AI计算的默认选择

68
00:02:59,479 --> 00:03:03,139
博通则是需要定制芯片的企业首选

69
00:03:03,139 --> 00:03:04,620
专属网络方案

70
00:03:04,620 --> 00:03:07,569
并获得更多成本与性能控制权

71
00:03:07,569 --> 00:03:08,330
上一季度

72
00:03:08,330 --> 00:03:11,969
博通的AI营收达到84亿美元

73
00:03:11,969 --> 00:03:15,849
同比增长106%

74
00:03:15,849 --> 00:03:18,370
其中约三分之一来自AI网络

75
00:03:18,370 --> 00:03:21,150
其余三分之二来自定制计算

76
00:03:21,150 --> 00:03:24,490
英伟达GPU是AI训练的黄金标准

77
00:03:24,490 --> 00:03:27,830
但博通正在快速扩展并行芯片

78
00:03:27,830 --> 00:03:32,579
可针对客户定制并融入长期路线图

79
00:03:32,579 --> 00:03:33,438
例如

80
00:03:33,438 --> 00:03:37,758
谷歌将博通作为TPU的长期联合设计方

81
00:03:37,758 --> 00:03:40,550
用于训练和运行Gemini 3等模型

82
00:03:40,550 --> 00:03:44,469
而最新Meta的训练与推理系统均由博通开发

83
00:03:44,469 --> 00:03:48,469
随着Meta试图减少对英伟达GPU的依赖

84
00:03:48,469 --> 00:03:50,709
Anthropic更进一步

85
00:03:50,709 --> 00:03:51,829
宣布一项两

86
00:03:51,829 --> 00:03:52,908
十亿美元

87
00:03:52,908 --> 00:03:59,750
多年期协议，采购近百万TPU及博通全规模AI系统

88
00:03:59,750 --> 00:04:04,020
使其成为Anthropic的主要定制计算伙伴之一

89
00:04:04,020 --> 00:04:05,699
此外，OpenAI

90
00:04:05,699 --> 00:04:10,500
与博通宣布部署十吉瓦定制加速器计划

91
00:04:10,500 --> 00:04:15,889
确认OpenAI为博通AI业务的旗舰客户

92
00:04:15,889 --> 00:04:19,649
因此博通并非仅提供英伟达系统的周边部件

93
00:04:19,649 --> 00:04:22,370
而是直接替代GPU需求

94
00:04:22,370 --> 00:04:26,769
为超大规模客户提供专属AI芯片

95
00:04:26,769 --> 00:04:32,569
每部署一枚博通芯片即减少对英伟达GPU的工作负载

96
00:04:32,569 --> 00:04:36,370
英伟达占据数据中心GPU市场约90%

97
00:04:36,370 --> 00:04:40,459
成为通用AI计算的首选方案

98
00:04:40,459 --> 00:04:45,079
但博通掌握定制AI加速器市场约70%

99
00:04:45,079 --> 00:04:49,329
及数据中心以太网交换芯片市场约80%

100
00:04:49,329 --> 00:04:50,089
事实上

101
00:04:50,089 --> 00:04:52,769
网络领域对比更加明显

102
00:04:52,769 --> 00:04:56,639
英伟达拥有自己的InfiniBand和Spectrum X以太网堆栈

103
00:04:56,639 --> 00:05:00,959
博通是高端以太网交换芯片的主要供应商

104
00:05:00,959 --> 00:05:03,220
通过他们的Tomahawk和Jericho系列

105
00:05:03,220 --> 00:05:05,519
Tomahawk是高带宽交换芯片

106
00:05:05,519 --> 00:05:09,439
连接数据中心机架内外的GPU和ASIC

107
00:05:09,439 --> 00:05:13,540
而Jericho则是将大型AI集群连接起来的路由器

108
00:05:13,540 --> 00:05:17,730
甚至将整个数据中心整合为一个超大规模计算系统

109
00:05:17,730 --> 00:05:19,329
像Arista Networks这样的厂商

110
00:05:19,329 --> 00:05:23,658
Juniper和部分思科交换机均基于这些芯片构建

111
00:05:23,658 --> 00:05:25,158
因此博通获利丰厚

112
00:05:25,158 --> 00:05:27,930
即使数据中心不直接采购其产品

113
00:05:27,930 --> 00:05:31,798
尽管英伟达目前主导AI基础设施市场

114
00:05:31,798 --> 00:05:37,449
大公司为特定应用寻求每瓦更高性能时会转向博通

115
00:05:37,449 --> 00:05:40,298
并希望更掌控整体AI架构

116
00:05:40,298 --> 00:05:44,459
这就是博通AI业务同比增长106%

117
00:05:44,459 --> 00:05:47,418
同比激增的原因，现在了解博通的业务

118
00:05:47,418 --> 00:05:49,199
及其在AI栈中的定位

119
00:05:49,199 --> 00:05:52,250
与英伟达相比，让我们分析财报

120
00:05:52,250 --> 00:05:53,310
说到财报

121
00:05:53,310 --> 00:05:54,529
新研究显示，

122
00:05:54,529 --> 00:05:59,430
每天使用AI的美国员工日薪比不用者高40%

123
00:05:59,430 --> 00:06:01,629
这意味着AI不可或缺

124
00:06:01,629 --> 00:06:05,668
要么拥有优势，要么被他人超越

125
00:06:05,668 --> 00:06:07,488
这就是Outskill的作用

126
00:06:07,488 --> 00:06:09,028
本视频的赞助商

127
00:06:09,028 --> 00:06:11,149
Outskill正在举办为期两天的

128
00:06:11,149 --> 00:06:13,069
AI大师班本周末举行

129
00:06:13,069 --> 00:06:14,370
十六小时培训

130
00:06:14,370 --> 00:06:20,100
汇聚来自英伟达、微软等百位专家知识

131
00:06:20,100 --> 00:06:23,060
助你自信应用AI

132
00:06:23,060 --> 00:06:27,500
工具快速迭代中保持领先并将技能转化为高价值工作

133
00:06:27,500 --> 00:06:28,600
获得更高薪酬

134
00:06:28,600 --> 00:06:32,850
前一千名通过我的链接注册者获免费C

135
00:06:32,850 --> 00:06:36,290
无论你在科技、销售管理或营销领域

136
00:06:36,290 --> 00:06:38,300
将学习使用AI代理

137
00:06:38,300 --> 00:06:40,060
创建自动化流程

138
00:06:40,060 --> 00:06:44,029
并与日常使用的软件和电子表格连接

139
00:06:44,029 --> 00:06:46,689
这是提升AI知识的绝佳途径

140
00:06:46,689 --> 00:06:48,610
获得真实竞争优势

141
00:06:48,610 --> 00:06:55,139
已有超过1000万人参与，全球观众

142
00:06:55,139 --> 00:06:58,360
本次名额正以更快速度被填满

143
00:06:58,360 --> 00:07:00,779
务必通过下方链接注册免费席位

144
00:07:00,779 --> 00:07:02,529
今日立即行动

145
00:07:02,529 --> 00:07:03,170
好的

146
00:07:03,170 --> 00:07:05,149
让我们深入分析博通财报

147
00:07:05,149 --> 00:07:11,029
因为最新季度数据最清晰展现其AI业务走向

148
00:07:11,029 --> 00:07:14,509
博通营收达193亿美元

149
00:07:14,509 --> 00:07:15,529
同比增长约2%

150
00:07:15,529 --> 00:07:20,620
同比上涨9%，略高于华尔街预计的192亿美元

151
00:07:20,620 --> 00:07:24,600
每股收益达到1.50美元

152
00:07:24,600 --> 00:07:28,119
相比去年同期的1.14美元纸面增长

153
00:07:28,119 --> 00:07:29,738
这看起来是一个正常的业绩超预期

154
00:07:29,738 --> 00:07:32,598
但表面之下他们正经历大幅增长

155
00:07:32,598 --> 00:07:35,970
得益于AI业务，193亿美元中

156
00:07:35,970 --> 00:07:37,449
84%来自

157
00:07:37,449 --> 00:07:39,060
正如我之前所说

158
00:07:39,060 --> 00:07:41,939
同比激增106%

159
00:07:41,939 --> 00:07:45,500
但值得注意的是，AI现在约占

160
00:07:45,500 --> 00:07:48,980
在Broadcom总收入中占比约4%

161
00:07:48,980 --> 00:07:52,759
AI已成为Broadcom业务占比超过AMD的

162
00:07:52,759 --> 00:07:57,600
尽管多数投资者仍认为AMD是NVIDIA在

163
00:07:57,600 --> 00:08:03,329
AI领域Broadcom业务增长更快

164
00:08:03,329 --> 00:08:07,149
半导体解决方案营收约125亿美元

165
00:08:07,149 --> 00:08:08,069
增长了5

166
00:08:08,069 --> 00:08:09,620
2个百分点

167
00:08:09,620 --> 00:08:14,620
现在占Broadcom总收入的65%

168
00:08:14,620 --> 00:08:17,269
软件业务贡献了其余68亿美元

169
00:08:17,269 --> 00:08:21,170
主要得益于VMware，但仅同比增长1%

170
00:08:21,170 --> 00:08:25,779
软件部门毛利率达到93%

171
00:08:25,779 --> 00:08:28,860
运营毛利率78%

172
00:08:28,860 --> 00:08:33,889
增长来自AI芯片，而软件提供极高的利润率

173
00:08:33,889 --> 00:08:37,960
这使Broadcom整体利润率也非常强劲

174
00:08:37,960 --> 00:08:41,360
他们报告毛利率68%

175
00:08:41,360 --> 00:08:44,730
或调整后毛利率77%

176
00:08:44,730 --> 00:08:47,779
比AMD高15个百分点

177
00:08:47,779 --> 00:08:50,519
几乎与NVIDIA在同一水平

178
00:08:50,519 --> 00:08:53,659
Broadcom运营利润率66.4%

179
00:08:53,659 --> 00:08:56,200
同比略升1%

180
00:08:56,200 --> 00:08:59,080
这意味着他们正在扩大盈利能力

181
00:08:59,080 --> 00:09:01,818
同时加速AI产品布局

182
00:09:01,818 --> 00:09:06,119
Broadcom本季度自由现金流达80亿美元

183
00:09:06,119 --> 00:09:07,458
或占总收入的40

184
00:09:07,458 --> 00:09:10,090
约41%的总收入

185
00:09:10,090 --> 00:09:16,090
NVIDIA上季度自由现金流占比51%，AMD约23%

186
00:09:16,090 --> 00:09:16,919
3%

187
00:09:16,919 --> 00:09:21,169
但Broadcom指引才是其AI故事的关键

188
00:09:21,169 --> 00:09:22,870
Broadcom预计下一季度营收约220亿美元

189
00:09:22,870 --> 00:09:25,289
同比增长47%

190
00:09:25,289 --> 00:09:26,570
并将具体指出AI半导体营收达107亿美元

191
00:09:26,570 --> 00:09:28,669
同比增长140%

192
00:09:28,669 --> 00:09:33,450
换句话说

193
00:09:33,450 --> 00:09:34,669
70亿美元

194
00:09:34,669 --> 00:09:39,460
同比增长约140%

195
00:09:39,460 --> 00:09:40,600
换言之

196
00:09:40,600 --> 00:09:44,299
Broadcom的AI营收预计环比增长27%

197
00:09:44,299 --> 00:09:49,590
占总收入约50%

198
00:09:49,590 --> 00:09:52,789
若能达成这些数字对投资者而言

199
00:09:52,789 --> 00:09:55,669
Broadcom财报证实三大关键点

200
00:09:55,669 --> 00:09:59,259
首先，人工智能已成为他们业务的主要增长驱动力

201
00:09:59,259 --> 00:09:59,779
其次

202
00:09:59,779 --> 00:10:01,460
他们正在扩大人工智能收入规模

203
00:10:01,460 --> 00:10:05,840
而未牺牲利润率，这正是大多数硬件公司梦寐以求的

204
00:10:05,840 --> 00:10:06,700
第三

205
00:10:06,700 --> 00:10:08,879
他们的AI业务不仅在增长

206
00:10:08,879 --> 00:10:10,480
而且增速正在加快

207
00:10:10,480 --> 00:10:13,919
这对整个AI革命具有重大影响

208
00:10:13,919 --> 00:10:14,899
例如

209
00:10:14,899 --> 00:10:16,320
博通首席执行官霍克

210
00:10:16,320 --> 00:10:18,769
谭先生向分析师表示他已清晰看到

211
00:10:18,769 --> 00:10:23,580
到2027年AI芯片收入将超过1000亿美元

212
00:10:23,580 --> 00:10:26,240
他明确指出这仅指芯片

213
00:10:26,240 --> 00:10:27,700
而非软件

214
00:10:27,700 --> 00:10:28,990
也不是服务

215
00:10:28,990 --> 00:10:30,509
仅仅是半导体

216
00:10:30,509 --> 00:10:35,840
这意味着博通的AI芯片收入到2027年将再次翻倍以上

217
00:10:35,840 --> 00:10:39,639
即使从当前年400亿美元的基数大幅增长

218
00:10:39,639 --> 00:10:43,120
他还表示博通已锁定供应链

219
00:10:43,120 --> 00:10:43,919
晶圆

220
00:10:43,919 --> 00:10:45,259
先进封装

221
00:10:45,259 --> 00:10:48,179
以及支持该目标的高带宽内存

222
00:10:48,179 --> 00:10:53,279
此外，博通的总订单 backlog 超过1600亿美元

223
00:10:53,279 --> 00:10:55,960
其中包括730亿美元的 backlog

224
00:10:55,960 --> 00:11:00,960
直接关联来自超大规模数据中心和AI实验室的定制加速器订单

225
00:11:00,960 --> 00:11:02,860
和AI网络产品

226
00:11:02,860 --> 00:11:07,480
因此他们预估的AI收入增长已有很大部分已签约

227
00:11:07,480 --> 00:11:10,529
霍克坦也提到谷歌的需求

228
00:11:10,529 --> 00:11:11,070
Meta

229
00:11:11,070 --> 00:11:15,649
Anthropic和OpenAI等大客户的实际需求正在加速

230
00:11:15,649 --> 00:11:20,490
他预计AI将成为半导体业务增长的主要驱动力

231
00:11:20,490 --> 00:11:22,490
未来多年都将如此

232
00:11:22,490 --> 00:11:23,490
如果你是投资者

233
00:11:23,490 --> 00:11:27,360
试图判断我们是否接近AI支出周期的顶部

234
00:11:27,360 --> 00:11:30,100
还是仍处于AI革命的早期阶段

235
00:11:30,100 --> 00:11:33,659
博通的订单 backlog 表明我们仍处于非常早期阶段

236
00:11:33,659 --> 00:11:36,659
这也降低了近期执行风险

237
00:11:36,659 --> 00:11:42,279
因为大量AI需求已通过合同锁定，供应链也已确定

238
00:11:42,279 --> 00:11:45,909
但投资者仍需了解其他潜在风险

239
00:11:45,909 --> 00:11:47,970
如果你觉得这个视频有价值

240
00:11:47,970 --> 00:11:52,370
请点赞并订阅频道以支持我

241
00:11:52,370 --> 00:11:55,099
这能让我知道制作更多此类内容

242
00:11:55,099 --> 00:11:56,099
谢谢现在

243
00:11:56,099 --> 00:11:59,038
让我们讨论可能破坏博通业务的因素

244
00:11:59,038 --> 00:12:00,490
如果这些因素出错

245
00:12:00,490 --> 00:12:03,129
首要重大风险是客户集中度

246
00:12:03,129 --> 00:12:07,850
博通的AI收入严重依赖极少数超大规模数据中心

247
00:12:07,850 --> 00:12:09,529
以及像谷歌这样的AI实验室

248
00:12:09,529 --> 00:12:10,129
Meta

249
00:12:10,129 --> 00:12:10,490
Open

250
00:12:10,490 --> 00:12:10,830
AI

251
00:12:10,830 --> 00:12:11,990
以及人类中心主义

252
00:12:11,990 --> 00:12:17,720
分析人士估计，博通已掌控定制AI芯片市场超过60%的份额

253
00:12:17,720 --> 00:12:24,960
其三大主要客户到2027年可能带来超过600亿美元的年度AI芯片营收

254
00:12:24,960 --> 00:12:28,059
如果当前路线图保持不变，这听起来很棒

255
00:12:28,059 --> 00:12:30,600
但即使其中一个大客户放缓步伐

256
00:12:30,600 --> 00:12:35,049
他们的支出可能转向部署或更多工作负载回迁至英伟达

257
00:12:35,049 --> 00:12:36,909
一旦大规模推出Vera Rubin芯片

258
00:12:36,909 --> 00:12:41,139
博通的增长可能开始下滑，股价也将随之下跌

259
00:12:41,139 --> 00:12:45,149
第二大风险来自超大规模云服务商的利润压力

260
00:12:45,149 --> 00:12:50,210
定制芯片利润率较低，因为它们是为单一客户设计多年

261
00:12:50,210 --> 00:12:56,000
前期研发成本极高，通常摊销到单一客户身上

262
00:12:56,000 --> 00:12:58,190
而非更广泛的安装基础

263
00:12:58,190 --> 00:12:59,250
与此同时

264
00:12:59,250 --> 00:13:01,450
超大规模云服务商拥有巨大议价权

265
00:13:01,450 --> 00:13:05,019
这限制了博通可实现的利润率上限

266
00:13:05,019 --> 00:13:07,580
在客户转向英伟达之前

267
00:13:07,580 --> 00:13:12,039
博通和AMD实际上是在销售与英伟达相比的成本节省

268
00:13:12,039 --> 00:13:14,120
而非自身独特平台

269
00:13:14,120 --> 00:13:19,419
如CUDA和博通的全机架系统可能进一步压低利润率

270
00:13:19,419 --> 00:13:21,960
当博通销售完整AI机架时

271
00:13:21,960 --> 00:13:26,840
他们将芯片与网络解决方案及大量第三方组件（如内存）捆绑

272
00:13:26,840 --> 00:13:28,440
其他种类处理器

273
00:13:28,440 --> 00:13:30,450
甚至有时包括GPU

274
00:13:30,450 --> 00:13:34,500
博通几乎以成本价将这些转嫁给客户

275
00:13:34,500 --> 00:13:37,480
这意味着存在大量低利润率的系统营收

276
00:13:37,480 --> 00:13:40,450
叠加在其高利润率芯片营收之上

277
00:13:40,450 --> 00:13:42,370
英伟达则没有这些问题

278
00:13:42,370 --> 00:13:46,529
因为其GPU是标准化产品，面向众多客户销售

279
00:13:46,529 --> 00:13:50,610
这意味着研发成本能在更广泛的客户群中摊销

280
00:13:50,610 --> 00:13:52,590
他们还拥有强大定价权

281
00:13:52,590 --> 00:13:54,269
得益于CUDA生态系统

282
00:13:54,269 --> 00:13:56,818
客户无法在别处获得

283
00:13:56,818 --> 00:13:59,139
当英伟达销售完整系统时

284
00:13:59,139 --> 00:14:01,960
大部分价值来自其自有硬件和软件

285
00:14:01,960 --> 00:14:07,009
第三方组件转售给客户的低利润率更少

286
00:14:07,009 --> 00:14:08,250
归根结底

287
00:14:08,250 --> 00:14:12,568
与英伟达竞争将是任何AI公司的最大风险

288
00:14:12,568 --> 00:14:16,908
但博通管理层在最新财报电话会上显得相当自信

289
00:14:16,908 --> 00:14:20,919
并指引调整后毛利率达77%

290
00:14:20,919 --> 00:14:23,860
但随着产品组合转向定制AI

291
00:14:23,860 --> 00:14:27,440
我更关注其毛利率是否会随时间下降

292
00:14:27,440 --> 00:14:28,179
好的

293
00:14:28,179 --> 00:14:29,440
现在整合所有信息

294
00:14:29,440 --> 00:14:33,720
看看博通股票是否值得与英伟达长期并列

295
00:14:33,720 --> 00:14:36,620
正如我多年来一直强调的AI投资组合

296
00:14:36,620 --> 00:14:39,759
博通是英伟达唯一的真正竞争对手

297
00:14:39,759 --> 00:14:42,440
因为他们瞄准技术栈的不同部分

298
00:14:42,440 --> 00:14:43,000
定制

299
00:14:43,000 --> 00:14:45,759
为少数科技巨头开发的AI处理器

300
00:14:45,759 --> 00:14:49,899
连接数千加速器的以太网交换芯片

301
00:14:49,899 --> 00:14:53,740
英伟达占据了数据中心GPU市场约90%份额

302
00:14:53,740 --> 00:14:58,600
博通则掌控了约70%的定制AI加速器市场

303
00:14:58,600 --> 00:15:02,039
几乎80%的以太网交换芯片均由其掌握

304
00:15:02,039 --> 00:15:05,259
这就是数据中心真正的竞争格局

305
00:15:05,259 --> 00:15:08,620
但博通无需成为英伟达就能赢得市场

306
00:15:08,620 --> 00:15:12,419
只需赢得那些不愿完全依赖英伟达的企业

307
00:15:12,419 --> 00:15:17,419
如果超大规模云服务商每年在AI基建上投入数百亿美元

308
00:15:17,419 --> 00:15:23,230
博通160亿美元的订单 backlog将持续与之同步增长

309
00:15:23,230 --> 00:15:25,309
AI业务已占博通总收入的四成

310
00:15:28,639 --> 00:15:32,359
他们预计2027年这一比例将再翻一番

311
00:15:32,359 --> 00:15:35,259
那么博通股票是否优于英伟达

312
00:15:35,259 --> 00:15:37,269
我认为这不是正确的问题

313
00:15:37,269 --> 00:15:42,309
博通业务更多元化，为我的投资组合提供了AI不同侧面的布局

314
00:15:42,309 --> 00:15:45,669
AI支出占六成，而非GPU领域

315
00:15:45,669 --> 00:15:46,429
以太网技术

316
00:15:46,429 --> 00:15:49,110
对比无限扩展与定制化基建

317
00:15:49,110 --> 00:15:50,970
对比现成平台解决方案

318
00:15:50,970 --> 00:15:52,070
这样我就能赢

319
00:15:52,070 --> 00:15:54,360
无论AI市场如何演变

320
00:15:54,360 --> 00:15:58,480
但若你相信我们仍处于多阶段AI建设的早期阶段

321
00:15:58,480 --> 00:16:02,409
则博通是与英伟达并持的优质股票

322
00:16:02,409 --> 00:16:03,789
而非替代选择

323
00:16:03,789 --> 00:16:07,389
这正是稳健致富的绝佳方式

324
00:16:07,389 --> 00:16:11,068
如果你想了解我如何通过稳健策略致富

325
00:16:11,068 --> 00:16:13,600
请观看下期视频

326
00:16:13,600 --> 00:16:14,299
感谢观看

327
00:16:14,299 --> 00:16:17,139
下次再见，这里是股票代码你

328
00:16:17,139 --> 00:16:18,240
我是亚历克斯

329
00:16:18,240 --> 00:16:22,500
提醒您最好的投资就是投资自己

评论采集失败

采集完成，该视频无弹幕