BV1oCRKBUEgG · 浑水摸鱼清源1 00:00:00,100 --> 00:00:02,379 周三收盘后我们迎来了创收高峰 2 00:00:02,379 --> 00:00:04,379 这影响着每一位投资者 3 00:00:04,379 --> 00:00:06,059 无论你投资的是指数基金 4 00:00:06,059 --> 00:00:07,900 还是人工智能股票 5 00:00:07,900 --> 00:00:10,410 所以我们要讨论并总结关键要点 6 00:00:10,410 --> 00:00:10,990 从那天开始 7 00:00:10,990 --> 00:00:12,730 邀请Jose Naharro来做客 8 00:00:12,730 --> 00:00:16,010 我们将涵盖Alphabet、Meta、亚马逊和微软 9 00:00:16,010 --> 00:00:18,050 首先由Jose开始 10 00:00:18,050 --> 00:00:19,649 我认为当天最重要的消息 11 00:00:19,649 --> 00:00:20,809 就是Alphabet的表现 12 00:00:20,809 --> 00:00:22,109 我想先给出几个数据 13 00:00:22,109 --> 00:00:24,219 这些数据简直令人震惊 14 00:00:24,219 --> 00:00:25,480 搜索业务 15 00:00:25,480 --> 00:00:27,500 本应被人工智能颠覆的 16 00:00:27,500 --> 00:00:31,280 人工智能业务在本季度增长19% 17 00:00:31,280 --> 00:00:35,030 但这家公司仍面临诸多利好 18 00:00:35,030 --> 00:00:39,200 云业务增长63% 19 00:00:39,200 --> 00:00:41,979 他们还从过去十年的投资中获得了一笔资金 20 00:00:41,979 --> 00:00:44,649 这些投资带来的收益 21 00:00:44,649 --> 00:00:47,329 约300亿美元利润 22 00:00:47,329 --> 00:00:47,890 这些利润 23 00:00:47,890 --> 00:00:49,310 我们稍后会详细分析 24 00:00:49,310 --> 00:00:50,530 但当我们看这份报告时 25 00:00:50,530 --> 00:00:52,140 让我们聚焦云业务数据 26 00:00:52,140 --> 00:00:53,369 什么让你印象深刻 27 00:00:53,369 --> 00:00:53,710 是的 28 00:00:53,710 --> 00:00:55,969 有几个点让我印象深刻 29 00:00:55,969 --> 00:00:56,829 看到这个数字时 30 00:00:56,829 --> 00:00:57,250 嗯 31 00:00:57,250 --> 00:00:57,729 Travis先说 32 00:00:57,729 --> 00:01:00,310 正如你提到的63%同比增速 33 00:01:00,310 --> 00:01:02,340 这是云业务增长最快的 34 00:01:02,340 --> 00:01:04,439 显然在三大巨头中 35 00:01:04,439 --> 00:01:04,840 嗯 36 00:01:04,840 --> 00:01:06,109 这是增速最低的 37 00:01:06,109 --> 00:01:07,790 但我们可以确定的是 38 00:01:07,790 --> 00:01:09,549 他们拥有自主研发的AI芯片 39 00:01:09,549 --> 00:01:11,049 经过长期研发 40 00:01:11,049 --> 00:01:14,799 因此能比其他竞争对手更快扩展产能 41 00:01:14,799 --> 00:01:16,299 我相信这正在帮助他们 42 00:01:16,299 --> 00:01:16,980 可以看到这种 43 00:01:16,980 --> 00:01:17,700 云业务增长 44 00:01:17,700 --> 00:01:21,299 他们在计算资源领域非常激进 45 00:01:21,299 --> 00:01:22,569 在 46 00:01:22,569 --> 00:01:24,459 他们正在构建的计算能力 47 00:01:24,459 --> 00:01:26,680 甚至在财报会议上 48 00:01:26,680 --> 00:01:28,280 我们听到资本支出 49 00:01:28,280 --> 00:01:31,269 他们上调了上下限 50 00:01:31,269 --> 00:01:31,790 嗯 51 00:01:31,790 --> 00:01:32,670 我原谅了 52 00:01:32,670 --> 00:01:34,209 这相当于十亿或五亿美元 53 00:01:34,209 --> 00:01:34,810 无论如何 54 00:01:34,810 --> 00:01:36,769 仅增加了微不足道的舍入误差 55 00:01:36,769 --> 00:01:40,439 当谈到年度一百八十亿美元的资本支出时 56 00:01:40,439 --> 00:01:41,159 而且 57 00:01:41,159 --> 00:01:42,079 更重要的是 58 00:01:42,079 --> 00:01:43,420 特拉维斯他们给了我们一些 59 00:01:43,420 --> 00:01:44,739 这只是第一季度 60 00:01:44,739 --> 00:01:48,659 但他们已经发出信号，针对2027年 61 00:01:48,659 --> 00:01:52,558 他们还预计caics将显著高于2026年 62 00:01:52,558 --> 00:01:54,450 所以啊 63 00:01:54,450 --> 00:01:56,509 这两大点引起了我的注意 64 00:01:56,509 --> 00:02:00,929 两年前谷歌云仅占公司总收入的12% 65 00:02:00,929 --> 00:02:03,668 我刚才做了快速计算 66 00:02:03,668 --> 00:02:07,049 现在季度收入占比已达18.2% 67 00:02:07,049 --> 00:02:07,588 嗯 68 00:02:07,588 --> 00:02:09,169 总收入的18.2% 69 00:02:09,169 --> 00:02:09,709 嗯 70 00:02:09,709 --> 00:02:13,069 这家公司曾被视为纯广告业务 71 00:02:13,069 --> 00:02:15,110 当时还有些云计算业务嗯 72 00:02:15,110 --> 00:02:16,150 或许确实如此 73 00:02:16,150 --> 00:02:18,879 但近五分之一的收入来自云业务 74 00:02:18,879 --> 00:02:21,460 我认为我们正见证公司重大转变 75 00:02:21,460 --> 00:02:24,319 在广告领域出现大幅增长 76 00:02:24,319 --> 00:02:25,550 正如你提到的激增 77 00:02:25,550 --> 00:02:27,969 同样在云业务中更为重要 78 00:02:27,969 --> 00:02:28,509 是的 79 00:02:28,509 --> 00:02:29,870 其中一点 80 00:02:29,870 --> 00:02:30,729 有必要 81 00:02:30,729 --> 00:02:31,009 我觉得 82 00:02:31,009 --> 00:02:32,629 将这些数据置于语境中 83 00:02:32,629 --> 00:02:33,770 毕竟 84 00:02:33,770 --> 00:02:34,530 如果这是顺风 85 00:02:34,530 --> 00:02:35,860 则所有船只都将受益 86 00:02:35,860 --> 00:02:37,919 但让我们先看看这四大超大规模云服务商 87 00:02:37,919 --> 00:02:39,900 这些数据中有趣的部分是 88 00:02:39,900 --> 00:02:42,219 谷歌云同比增长63% 89 00:02:42,219 --> 00:02:45,120 微软Azure同比增长40% 90 00:02:45,120 --> 00:02:47,560 这是巨大数字 91 00:02:47,560 --> 00:02:48,680 但仍远低于63% 92 00:02:48,680 --> 00:02:50,879 而AWS同比增长28% 93 00:02:50,879 --> 00:02:53,159 投资者需注意的另一个数据是待执行订单 94 00:02:53,159 --> 00:02:56,689 这是一个惊人数字 95 00:02:56,689 --> 00:02:58,929 还记得我们曾获得 96 00:02:58,929 --> 00:02:59,808 嗯 97 00:02:59,808 --> 00:03:00,368 来自甲骨文的剩余履约义务 98 00:03:00,368 --> 00:03:03,359 三四季度前他们与OpenAI签订巨额合同时 99 00:03:03,359 --> 00:03:07,019 但现在迎来谷歌云 100 00:03:07,019 --> 00:03:08,758 这就是谷歌云 101 00:03:08,758 --> 00:03:12,509 我认为这可能是一个更加多元化的业务 102 00:03:12,509 --> 00:03:15,889 他们肯定有一些非常大的客户，比如Anthropic 103 00:03:15,889 --> 00:03:17,870 但比你们之前与Oracle的业务更加多元化 104 00:03:17,870 --> 00:03:18,349 你知道的 105 00:03:18,349 --> 00:03:19,810 Oracle签署了一项单一协议 106 00:03:19,810 --> 00:03:24,889 这相当于四百六十亿美元的待执行订单 107 00:03:24,889 --> 00:03:29,169 谷歌云具体有什么优势 108 00:03:29,169 --> 00:03:31,650 让它能够超越竞争对手 109 00:03:31,650 --> 00:03:35,349 因为我认为它们确实比一些大公司规模小 110 00:03:35,349 --> 00:03:37,479 但如果这些增长势头持续 111 00:03:37,479 --> 00:03:42,520 可能只需几年时间就能成为最大的云服务商 112 00:03:42,520 --> 00:03:44,219 看起来 113 00:03:44,219 --> 00:03:47,759 作为半导体和AI专家 114 00:03:47,759 --> 00:03:48,740 但当我观察这个情况 115 00:03:48,740 --> 00:03:52,050 这是真正垂直整合的公司 116 00:03:52,050 --> 00:03:53,270 他们自主研发模型 117 00:03:53,270 --> 00:03:54,289 他们制造自己的芯片 118 00:03:54,289 --> 00:03:55,229 你之前提到过 119 00:03:55,240 --> 00:03:59,370 因此他们可能能为客户提供差异化解决方案 120 00:03:59,370 --> 00:04:01,259 而像AWS 121 00:04:01,259 --> 00:04:01,659 你知道的 122 00:04:01,659 --> 00:04:02,419 微软 123 00:04:02,419 --> 00:04:05,329 他们追求标准化产品 124 00:04:05,329 --> 00:04:08,329 希望将模型商品化 125 00:04:08,329 --> 00:04:13,610 这是否说明谷歌的垂直整合正在见效 126 00:04:13,610 --> 00:04:17,120 还是另有原因，他们似乎抓住了计算需求 127 00:04:17,120 --> 00:04:18,199 并且他们已经具备了 128 00:04:18,199 --> 00:04:18,720 到底怎么回事 129 00:04:18,720 --> 00:04:19,600 我觉得你说得很对 130 00:04:19,600 --> 00:04:21,339 你完全抓住了关键点 131 00:04:21,339 --> 00:04:21,959 嗯，Travis 132 00:04:21,959 --> 00:04:25,168 我认为这得益于其完全整合的系统 133 00:04:25,168 --> 00:04:25,689 嗯 134 00:04:25,689 --> 00:04:28,629 我也认为亚马逊正在接近这一点 135 00:04:28,629 --> 00:04:30,869 最终他们没有完全自主的 136 00:04:30,869 --> 00:04:32,009 但他们确实有自己的 137 00:04:32,009 --> 00:04:34,620 但我不确定谁在使用Nova或其他模型 138 00:04:34,620 --> 00:04:35,699 他们宣布的 139 00:04:35,699 --> 00:04:36,319 但啊 140 00:04:36,319 --> 00:04:40,569 谷歌拥有自己的大语言模型 141 00:04:40,569 --> 00:04:43,910 拥有自主研发的AI芯片 142 00:04:43,910 --> 00:04:45,050 他们已经 143 00:04:45,050 --> 00:04:49,189 或现在已是第八代AI芯片与CPU 144 00:04:49,189 --> 00:04:50,730 更重要的是 145 00:04:50,730 --> 00:04:55,649 我认为Alphabet是唯一没有缩减资本支出的玩家 146 00:04:55,649 --> 00:04:57,459 在投资方面没有刹车 147 00:04:57,459 --> 00:05:00,160 回想当初 148 00:05:00,160 --> 00:05:01,798 当这一切开始时 149 00:05:01,798 --> 00:05:04,478 微软在资本支出上投入很大 150 00:05:04,478 --> 00:05:07,658 后来似乎微软踩了刹车 151 00:05:07,658 --> 00:05:11,670 现在看来微软似乎真的感受到了这种痛苦 152 00:05:11,670 --> 00:05:13,430 我们需要时间才能感受到这种痛苦 153 00:05:13,430 --> 00:05:14,300 但呃 154 00:05:14,300 --> 00:05:15,060 在Azure方面 155 00:05:15,060 --> 00:05:16,180 尽管它增长了40% 156 00:05:16,180 --> 00:05:19,620 过去几个季度的增长率在38%到40%之间 157 00:05:19,620 --> 00:05:21,220 它的增长空间已经有限了 158 00:05:21,220 --> 00:05:22,279 这取决于计算资源 159 00:05:22,279 --> 00:05:23,100 而谷歌 160 00:05:23,100 --> 00:05:24,819 我认为他们一直保持 161 00:05:24,819 --> 00:05:26,428 始终保持着强劲投入 162 00:05:26,428 --> 00:05:29,088 这让他们实现了快速增长 163 00:05:29,088 --> 00:05:29,468 嗯 164 00:05:29,468 --> 00:05:34,309 亚马逊最初与微软截然相反 165 00:05:34,309 --> 00:05:35,980 亚马逊显得非常从容 166 00:05:35,980 --> 00:05:36,519 相当放松 167 00:05:36,519 --> 00:05:38,470 在资本支出方面的投资 168 00:05:38,470 --> 00:05:40,949 现在我们看到亚马逊采取了激进策略 169 00:05:40,949 --> 00:05:43,649 同时看到AWS增长加速 170 00:05:43,649 --> 00:05:47,279 我认为这就是你提到的全栈解决方案 171 00:05:47,279 --> 00:05:49,610 更重要的是它从未 172 00:05:49,610 --> 00:05:51,230 它从未自我怀疑 173 00:05:51,230 --> 00:05:52,889 我认为其他公司曾有过自我怀疑 174 00:05:52,889 --> 00:05:53,209 他们就像 175 00:05:53,209 --> 00:05:55,949 哦也许也许感觉 176 00:05:55,949 --> 00:05:57,509 谷歌就像我们在坚持这样做 177 00:05:57,509 --> 00:05:58,370 我们坚持到底 178 00:05:58,370 --> 00:05:59,600 这对我们有效 179 00:05:59,600 --> 00:06:00,540 这是否是 180 00:06:00,540 --> 00:06:03,079 它几乎处于防御态势 181 00:06:03,079 --> 00:06:04,050 所以回顾过去 182 00:06:04,050 --> 00:06:06,230 我想我们迎来了三周年纪念 183 00:06:06,230 --> 00:06:06,649 呃 184 00:06:06,649 --> 00:06:11,550 可能是在这次采访中，萨提亚·纳德拉提到 185 00:06:11,550 --> 00:06:12,629 你知道吗 186 00:06:12,629 --> 00:06:14,420 我很高兴让谷歌不得不调整 187 00:06:14,420 --> 00:06:16,579 他对此充满信心 188 00:06:16,579 --> 00:06:18,120 他们与OpenAI建立了合作 189 00:06:18,120 --> 00:06:19,300 看起来他们正在推进 190 00:06:19,300 --> 00:06:23,089 这种垂直整合模式未能成功 191 00:06:23,089 --> 00:06:25,269 可能未达预期效果 192 00:06:25,269 --> 00:06:26,550 但事实是 193 00:06:26,550 --> 00:06:31,889 他们早期似乎采取了防御性策略 194 00:06:31,889 --> 00:06:32,730 我们得加大投入 195 00:06:32,730 --> 00:06:33,589 因为如果不投入 196 00:06:33,589 --> 00:06:34,790 你知道OpenAI真的可能颠覆我们 197 00:06:34,790 --> 00:06:35,750 开放 198 00:06:35,750 --> 00:06:38,329 AI确实可能严重威胁公司 199 00:06:38,329 --> 00:06:39,930 最终使我们 200 00:06:39,930 --> 00:06:42,430 如今他们几乎处于绝佳地位 201 00:06:42,430 --> 00:06:46,300 因为他们必须应对几年前发生的一些事情 202 00:06:46,300 --> 00:06:47,120 而恰好发生的是 203 00:06:47,120 --> 00:06:50,230 他们还有现金流可以大量涌入市场 204 00:06:50,230 --> 00:06:50,949 当你观察 205 00:06:50,949 --> 00:06:51,230 你知道的 206 00:06:51,230 --> 00:06:51,750 购买芯片 207 00:06:51,750 --> 00:06:52,509 建设数据中心 208 00:06:52,509 --> 00:06:53,430 他们拥有其他公司不具备的基础设施 209 00:06:53,430 --> 00:06:54,829 其他公司根本没有 210 00:06:54,829 --> 00:06:55,430 是啊 211 00:06:55,430 --> 00:06:56,370 我觉得是这样 212 00:06:56,370 --> 00:06:56,709 呃 213 00:06:56,709 --> 00:06:56,949 特拉维斯 214 00:06:56,949 --> 00:06:59,939 我认为微软逼他们不得不调整 215 00:06:59,939 --> 00:07:02,478 或者为了三十秒的歌曲 216 00:07:02,478 --> 00:07:03,858 我想最多可以说 217 00:07:03,858 --> 00:07:04,319 嗯 218 00:07:05,358 --> 00:07:06,120 他们 219 00:07:06,120 --> 00:07:07,600 这迫使他们更加 220 00:07:07,600 --> 00:07:08,199 嗯 221 00:07:08,199 --> 00:07:12,019 在发布AI模型时更高效 222 00:07:12,019 --> 00:07:14,709 当他们不断发布更多AI模型时 223 00:07:14,709 --> 00:07:16,649 他们开始意识到哦看来 224 00:07:16,649 --> 00:07:19,620 或许我们需要对OpenAI更防御 225 00:07:19,620 --> 00:07:21,600 但我们也要更进取 226 00:07:21,600 --> 00:07:24,779 因为我们现在确实有一些很酷的东西可以开发 227 00:07:24,779 --> 00:07:29,899 所以我认为这种防御策略既创造了防御也推动了进攻，谷歌因此受益 228 00:07:29,899 --> 00:07:31,920 现在他们正在收获成果 229 00:07:31,920 --> 00:07:32,759 嗯 230 00:07:32,759 --> 00:07:34,699 还有一个我们还没提到的点 231 00:07:34,699 --> 00:07:36,420 特拉维斯，他们现在在卖芯片 232 00:07:36,420 --> 00:07:37,860 没错这太疯狂了 233 00:07:37,860 --> 00:07:40,040 听说他们 234 00:07:40,040 --> 00:07:40,610 他们 235 00:07:40,610 --> 00:07:42,389 他们正在进入新阶段 236 00:07:42,389 --> 00:07:45,959 在芯片销售方面开辟新的收入来源 237 00:07:45,959 --> 00:07:47,879 因为这与Anthropic合作有关 238 00:07:47,879 --> 00:07:49,339 我想这就是你在暗示的 239 00:07:49,339 --> 00:07:50,949 Anthropic将会是 240 00:07:50,949 --> 00:07:53,110 不确定他们是否直接从他们采购芯片 241 00:07:53,110 --> 00:07:55,009 但他们将通过云合作伙伴采购 242 00:07:55,009 --> 00:07:55,990 他们的云合作伙伴 243 00:07:55,990 --> 00:07:58,529 当谈到Anthropic时，他们在多方布局 244 00:07:58,529 --> 00:08:00,209 他们现在是AI领域最炙手可热的公司 245 00:08:00,209 --> 00:08:02,649 但在多方进行布局 246 00:08:02,649 --> 00:08:03,949 我们使用哪个云服务 247 00:08:03,949 --> 00:08:05,709 我们基于哪些芯片开发 248 00:08:05,709 --> 00:08:07,910 他们不想绑定在英伟达 249 00:08:07,910 --> 00:08:09,430 他们不想绑定在TPU 250 00:08:09,430 --> 00:08:10,709 他们也不想绑定在训练Anium 251 00:08:10,709 --> 00:08:11,430 我觉得这些有点像 252 00:08:11,430 --> 00:08:15,158 他们目前正在使用的三大主要芯片 253 00:08:15,158 --> 00:08:18,819 但另一部分与Anthropic相关 254 00:08:18,819 --> 00:08:19,199 你知道的 255 00:08:19,199 --> 00:08:22,439 这部分积压来自Anthropic为谷歌云 256 00:08:22,439 --> 00:08:25,259 但这个数字绝对疯狂 257 00:08:25,259 --> 00:08:32,139 本季度他们所谓的非运营收入达377亿美元 258 00:08:32,139 --> 00:08:36,100 仅仅一个季度对谷歌或整个字母表集团 259 00:08:36,100 --> 00:08:39,899 主要来自他们在Anthropic的股份 260 00:08:39,899 --> 00:08:41,600 还有SpaceX 261 00:08:41,600 --> 00:08:44,669 所以为了使事情 262 00:08:44,669 --> 00:08:46,289 更加疯狂 263 00:08:46,289 --> 00:08:48,679 当今世界最热门的公司 264 00:08:48,679 --> 00:08:52,740 字母表恰好持有这些公司的大额股份 265 00:08:52,740 --> 00:08:55,399 这也是他们资产负债表上的另一利好 266 00:08:55,399 --> 00:08:55,639 是的 267 00:08:55,639 --> 00:08:55,840 definitely 268 00:08:55,840 --> 00:08:56,080 我说 269 00:08:56,080 --> 00:08:57,320 我觉得这太疯狂了 270 00:08:57,320 --> 00:08:58,679 我在看图表想确认 271 00:08:58,679 --> 00:08:59,179 大概在那里 272 00:08:59,179 --> 00:09:00,769 在下面的项目中我忘了叫什么 273 00:09:00,769 --> 00:09:01,129 嗯 274 00:09:01,129 --> 00:09:01,730 这些投资 275 00:09:01,730 --> 00:09:03,960 但与去年相比差距巨大 276 00:09:03,960 --> 00:09:07,460 我确定主要是基本业务和Anthropic推动了增长 277 00:09:07,460 --> 00:09:07,759 呃 278 00:09:07,759 --> 00:09:09,419 但快速回到芯片话题 279 00:09:09,419 --> 00:09:09,779 呃 280 00:09:09,779 --> 00:09:10,120 特拉维斯 281 00:09:10,120 --> 00:09:13,750 他们在财报会议上提到实际上 282 00:09:13,750 --> 00:09:16,289 尽管他们与博通合作开发这些芯片 283 00:09:16,289 --> 00:09:17,429 他们也将看到收入 284 00:09:17,429 --> 00:09:19,169 他们还提到将会看到 285 00:09:19,169 --> 00:09:19,690 呃 286 00:09:19,690 --> 00:09:21,509 这部分收入将波动不定 287 00:09:21,509 --> 00:09:23,330 取决于芯片何时交付 288 00:09:23,330 --> 00:09:23,730 呃 289 00:09:23,730 --> 00:09:26,549 但今年晚些时候将有收入 290 00:09:26,549 --> 00:09:31,490 并在2027年上半年也有收入 291 00:09:31,490 --> 00:09:35,129 他们提到正在向特定客户销售 292 00:09:35,129 --> 00:09:37,049 比如AI实验室 293 00:09:37,049 --> 00:09:37,789 我认为这是 294 00:09:37,789 --> 00:09:38,370 就是Anthropic 295 00:09:38,370 --> 00:09:38,809 正如你所说 296 00:09:38,809 --> 00:09:40,590 但他们还提到类似 297 00:09:40,590 --> 00:09:42,929 一些金融机构 298 00:09:42,929 --> 00:09:44,330 我认为有交易部门 299 00:09:44,330 --> 00:09:47,600 交易人员也会购买这些TPU 300 00:09:47,600 --> 00:09:50,159 嗯所以他们就是这样安排的 301 00:09:50,159 --> 00:09:50,909 是的 302 00:09:50,909 --> 00:09:51,750 这就是我看到的 303 00:09:51,750 --> 00:09:52,269 我听说过 304 00:09:52,269 --> 00:09:53,149 X 我无法确认 305 00:09:53,149 --> 00:09:54,110 但这是我从下一期听到的 306 00:09:54,110 --> 00:09:56,049 那就是正在购买的那个 307 00:09:56,049 --> 00:09:56,470 呃 308 00:09:56,470 --> 00:09:57,350 有些这些CPU 309 00:09:57,350 --> 00:09:57,629 是的 310 00:09:57,629 --> 00:09:58,629 这很有趣 311 00:09:58,629 --> 00:09:59,870 那个呃 312 00:09:59,870 --> 00:10:00,970 人工智能机器学习 313 00:10:00,970 --> 00:10:01,350 你知道的 314 00:10:01,350 --> 00:10:04,370 无论你想回顾人工智能的历史 315 00:10:04,370 --> 00:10:08,000 嗯，通常都始于金融业 316 00:10:08,000 --> 00:10:09,700 并解释如何获得这种优势 317 00:10:09,700 --> 00:10:10,159 你知道的 318 00:10:10,159 --> 00:10:13,830 市场中千分之一秒的优势就能利用 319 00:10:13,830 --> 00:10:15,509 如果你拥有 Citadel 的基础设施 320 00:10:15,509 --> 00:10:17,078 或这些大型交易公司 321 00:10:17,078 --> 00:10:18,698 这就是他们部署的地方 322 00:10:18,698 --> 00:10:20,399 所以这是一个有趣的用例 323 00:10:20,399 --> 00:10:22,359 这也是你知道的 324 00:10:22,359 --> 00:10:23,558 Deep Sea 的起源 325 00:10:23,558 --> 00:10:26,599 这家公司现在在做AI模型吗 326 00:10:26,599 --> 00:10:28,120 但最初是一家对冲基金 327 00:10:28,120 --> 00:10:29,039 确实如此 328 00:10:29,039 --> 00:10:29,440 呃 329 00:10:29,440 --> 00:10:29,679 就是 330 00:10:29,679 --> 00:10:30,019 是的 331 00:10:30,019 --> 00:10:31,720 这一直是前沿领域之一 332 00:10:31,720 --> 00:10:35,240 让我们将此与其它公司财报结合 333 00:10:35,240 --> 00:10:39,820 因为市场对Alphabet的反应特别强劲 334 00:10:39,820 --> 00:10:41,299 股价上涨约7% 335 00:10:41,299 --> 00:10:43,339 我们正在市场开盘后录制 336 00:10:43,339 --> 00:10:46,538 但之前提到的微软 337 00:10:46,538 --> 00:10:47,899 云业务增长强劲 338 00:10:47,899 --> 00:10:49,739 我认为云业务增长了40% 339 00:10:49,739 --> 00:10:53,839 今日早盘股价下跌5% 340 00:10:53,839 --> 00:10:56,958 这是不是过度反应 341 00:10:56,958 --> 00:10:57,558 你知道的 342 00:10:57,558 --> 00:10:59,759 我们总能看到细微差别 343 00:10:59,759 --> 00:11:00,339 这里或那里 344 00:11:00,339 --> 00:11:01,639 市场可能惩罚某只股票 345 00:11:01,639 --> 00:11:04,610 但似乎趋势向对的方向发展 346 00:11:04,610 --> 00:11:05,210 他们只是 347 00:11:05,210 --> 00:11:07,409 还是与Alphabet不同 348 00:11:07,409 --> 00:11:08,669 微软的情况如何 349 00:11:08,669 --> 00:11:10,210 我与微软的情况 350 00:11:10,210 --> 00:11:11,330 首先有几个要点 351 00:11:11,330 --> 00:11:15,529 你拥有类似萨斯球体的整体萨斯波伊普索 352 00:11:15,529 --> 00:11:17,820 这是削弱效果的浪费之一 353 00:11:17,820 --> 00:11:18,580 另一件事是 354 00:11:18,580 --> 00:11:23,960 其他两位玩家特拉维斯实际上在云增长方面出现了强劲加速 355 00:11:23,960 --> 00:11:24,500 好的 356 00:11:24,500 --> 00:11:25,240 所以呃 357 00:11:25,240 --> 00:11:26,799 AWS增长了28% 358 00:11:26,799 --> 00:11:31,019 我认为28到29左右，之前增长约20% 359 00:11:31,019 --> 00:11:32,659 低20%左右 360 00:11:32,659 --> 00:11:33,139 嗯 361 00:11:33,139 --> 00:11:33,860 即使上个季度 362 00:11:33,860 --> 00:11:36,080 我记得大概是两两或两三 363 00:11:36,080 --> 00:11:38,639 但似乎增长有所加速 364 00:11:38,639 --> 00:11:41,559 我们在谷歌身上看到了这种加速 365 00:11:41,559 --> 00:11:44,259 但微软没有这种加速 366 00:11:44,259 --> 00:11:44,480 对 367 00:11:44,480 --> 00:11:47,460 微软上季度增长39% 368 00:11:47,460 --> 00:11:48,740 之前那次是38% 369 00:11:48,740 --> 00:11:51,620 在40到39%之间波动 370 00:11:51,620 --> 00:11:53,639 过去三到四个季度 371 00:11:53,639 --> 00:11:55,340 我很好奇市场是否认为 372 00:11:55,340 --> 00:11:56,720 嗯嗯 373 00:11:56,720 --> 00:11:59,379 如果其他人都在加速增长 374 00:11:59,379 --> 00:12:00,539 你却保持不变 375 00:12:00,539 --> 00:12:02,500 难道你在失去市场份额吗 376 00:12:02,500 --> 00:12:05,519 难道我们该为此担忧吗 377 00:12:05,519 --> 00:12:07,879 谷歌为何能实现63%的增长 378 00:12:07,879 --> 00:12:10,259 所以我想这可能引发 379 00:12:10,259 --> 00:12:10,879 嗯 380 00:12:10,879 --> 00:12:12,139 我觉得更多是 381 00:12:12,139 --> 00:12:16,669 阿舍目前尚未达到其他公司的增长水平 382 00:12:16,669 --> 00:12:20,090 但依赖他们公布的资本支出数据 383 00:12:20,090 --> 00:12:23,590 我认为他们又开始在数据中心大力投资 384 00:12:23,590 --> 00:12:24,659 或许这会改变现状 385 00:12:24,659 --> 00:12:25,039 嗯 386 00:12:25,039 --> 00:12:27,559 市场已发展到这种程度 387 00:12:27,559 --> 00:12:29,600 而这在每个周期都会发生 388 00:12:29,600 --> 00:12:34,799 对，40%的增长率让投资者失望 389 00:12:34,799 --> 00:12:37,659 这说明情况与几年前不同 390 00:12:37,659 --> 00:12:38,659 比如英伟达 391 00:12:38,659 --> 00:12:38,940 他们 392 00:12:38,940 --> 00:12:42,720 曾因股价飙升创出惊人业绩 393 00:12:42,720 --> 00:12:43,860 而现在 394 00:12:43,860 --> 00:12:44,720 无论是英伟达 395 00:12:44,720 --> 00:12:45,799 无论是微软 396 00:12:45,799 --> 00:12:47,200 都取得优异成绩 397 00:12:47,200 --> 00:12:49,169 市场只是淡淡回应 398 00:12:49,169 --> 00:12:49,730 这没问题 399 00:12:49,730 --> 00:12:52,370 但增速比一年前放缓 400 00:12:52,370 --> 00:12:55,159 这就是加速与减速的体现 401 00:12:55,159 --> 00:12:57,860 这是投资者始终关注的重点 402 00:12:57,860 --> 00:13:01,250 因为从30%的增长到63%的增长 403 00:13:01,250 --> 00:13:03,029 你开始算算账吧兄弟 404 00:13:03,029 --> 00:13:04,769 如果你能持续几年 405 00:13:04,769 --> 00:13:05,990 数字可能会很惊人 406 00:13:05,990 --> 00:13:08,659 从GCP和特拉维斯出来后 407 00:13:08,659 --> 00:13:11,269 我知道你总在寻找 408 00:13:11,269 --> 00:13:13,789 或许这个CAC故事的漏洞 409 00:13:13,789 --> 00:13:16,409 你居然没提到这个我有点意外 410 00:13:16,409 --> 00:13:18,590 那我告诉你这个 411 00:13:18,590 --> 00:13:19,549 嗯还有啊 412 00:13:19,549 --> 00:13:21,350 你可以随意使用 413 00:13:21,350 --> 00:13:25,149 但事实上我们看到本季度资本支出增加 414 00:13:25,149 --> 00:13:27,350 但我其实并不兴奋 415 00:13:27,350 --> 00:13:29,870 我不兴奋的主要原因 416 00:13:29,870 --> 00:13:32,578 本季度大部分资本支出增长 417 00:13:32,578 --> 00:13:36,759 似乎完全由内存价格驱动 418 00:13:36,759 --> 00:13:39,129 并不是他们生产了更多芯片 419 00:13:39,129 --> 00:13:39,769 所以这 420 00:13:39,769 --> 00:13:40,809 他们仍然预期 421 00:13:40,809 --> 00:13:43,389 大概和上季度相同数量的芯片 422 00:13:43,389 --> 00:13:48,200 只是内存价格飙升 423 00:13:48,200 --> 00:13:48,539 呃 424 00:13:48,539 --> 00:13:50,460 我们看到三星电子发布财报 425 00:13:50,460 --> 00:13:50,740 呃 426 00:13:50,740 --> 00:13:50,980 哦 427 00:13:50,980 --> 00:13:54,470 两周前或一周前他们实现60%营收增长 428 00:13:54,470 --> 00:13:58,950 因为本季度平均售价环比上涨60% 429 00:13:58,950 --> 00:14:00,029 环比增长60% 430 00:14:00,029 --> 00:14:01,429 而这不仅限于内存 431 00:14:01,429 --> 00:14:02,490 这不仅仅是内存的问题 432 00:14:02,490 --> 00:14:04,059 看看布罗德能源 433 00:14:04,059 --> 00:14:04,679 呃 434 00:14:04,679 --> 00:14:07,399 看看GE新能源的情况 435 00:14:07,399 --> 00:14:10,129 整个供应链中的所有人 436 00:14:10,129 --> 00:14:10,850 你知道的 437 00:14:10,850 --> 00:14:12,470 这些是上市公司 438 00:14:12,470 --> 00:14:13,850 这些是中小企业 439 00:14:13,850 --> 00:14:16,039 这些是在屋顶工作的人 440 00:14:16,039 --> 00:14:16,840 这些是正在 441 00:14:16,840 --> 00:14:17,220 你知道的 442 00:14:17,220 --> 00:14:19,740 购买土地铺沥青的人 443 00:14:19,740 --> 00:14:22,889 大家都在说需求激增 444 00:14:22,889 --> 00:14:24,389 这些公司资金充裕 445 00:14:24,389 --> 00:14:27,090 劳动力和材料供应不足 446 00:14:27,090 --> 00:14:29,429 我们可以随意定价 447 00:14:29,429 --> 00:14:32,120 我觉得这一点值得指出 448 00:14:32,120 --> 00:14:32,879 是的 449 00:14:32,879 --> 00:14:34,759 资本支出在增加 450 00:14:34,759 --> 00:14:38,600 但投入的资本可能并未同步增长 451 00:14:38,600 --> 00:14:40,399 看看接下来会发生什么 452 00:14:40,399 --> 00:14:43,000 这对你来说是个巨大的助力 453 00:14:43,000 --> 00:14:44,440 看看标普的热力图 454 00:14:44,440 --> 00:14:46,049 今年以来已上涨500点 455 00:14:46,049 --> 00:14:46,409 你知道的 456 00:14:46,409 --> 00:14:47,769 软件股表现疲软 457 00:14:47,769 --> 00:14:48,970 许多成长型股票 458 00:14:48,970 --> 00:14:50,330 科技股表现不佳 459 00:14:50,330 --> 00:14:51,649 真正没受影响的是 460 00:14:51,649 --> 00:14:52,149 你知道的 461 00:14:52,149 --> 00:14:53,870 全球的毛毛虫企业 462 00:14:53,870 --> 00:14:54,509 呃 463 00:14:54,509 --> 00:14:56,090 全球的公用事业公司 464 00:14:56,090 --> 00:15:00,399 因为他们看到了下游的AI支出影响 465 00:15:00,399 --> 00:15:03,379 这就是我们需要考虑的动态 466 00:15:03,379 --> 00:15:05,099 无论你如何划分市场 467 00:15:05,099 --> 00:15:07,499 是的，我认为对于所有AI和半导体 468 00:15:07,499 --> 00:15:08,458 所有类型的投资者 469 00:15:08,458 --> 00:15:09,918 必须记住这一点 470 00:15:09,918 --> 00:15:10,639 CAC故事 471 00:15:10,639 --> 00:15:11,099 嗯 472 00:15:11,099 --> 00:15:13,360 可能芯片领域会有所增长 473 00:15:13,360 --> 00:15:16,360 但似乎大部分由成本上涨驱动 474 00:15:16,360 --> 00:15:19,120 由于许多组件成本通胀 475 00:15:19,120 --> 00:15:19,720 呃 476 00:15:19,720 --> 00:15:21,149 所以这是一个 477 00:15:21,149 --> 00:15:22,509 我不知道谁来做 478 00:15:22,509 --> 00:15:23,350 也许我来做 479 00:15:23,350 --> 00:15:25,909 但最终会看到结构分解 480 00:15:25,909 --> 00:15:27,360 如何拆解资本支出 481 00:15:27,360 --> 00:15:28,379 内存占比多少 482 00:15:28,379 --> 00:15:31,379 是否存在资本支出下降的世界 483 00:15:31,379 --> 00:15:33,360 但获得更多芯片 484 00:15:33,360 --> 00:15:35,120 因为内存价格下降 485 00:15:35,120 --> 00:15:35,559 呃 486 00:15:35,559 --> 00:15:38,059 因此许多动态将由此产生 487 00:15:38,059 --> 00:15:38,809 嗯 488 00:15:38,809 --> 00:15:41,210 在如此疯狂的定价环境中 489 00:15:41,210 --> 00:15:41,450 好吧 490 00:15:41,450 --> 00:15:42,610 这就是你开始看到的 491 00:15:42,610 --> 00:15:44,490 公司也会做出调整对吧 492 00:15:44,490 --> 00:15:45,049 好的 493 00:15:45,049 --> 00:15:47,279 这是价值链的瓶颈 494 00:15:47,279 --> 00:15:48,980 现在我们需要突破这个点 495 00:15:48,980 --> 00:15:52,610 无论是提升模型效率还是调整运行方式 496 00:15:52,610 --> 00:15:55,389 或者可能重新设计数据中心 497 00:15:55,389 --> 00:15:56,549 通常不会长期存在 498 00:15:56,549 --> 00:16:00,759 这些行业通常不会长期有瓶颈 499 00:16:00,759 --> 00:16:02,210 我们将观察结果如何发展 500 00:16:02,210 --> 00:16:03,350 嗯 你打算接手 501 00:16:03,350 --> 00:16:07,049 你接下来想讲AWS还是Meta那个野生的？ 502 00:16:07,049 --> 00:16:08,090 也许我们可以留到最后再说 503 00:16:08,090 --> 00:16:09,490 这会消耗掉10%的资源 504 00:16:09,490 --> 00:16:11,320 这看起来有点疯狂 505 00:16:11,320 --> 00:16:12,240 但AWS 506 00:16:12,240 --> 00:16:13,360 只是感觉他们 507 00:16:13,360 --> 00:16:16,399 过去几周他们宣布了一些有趣的合作伙伴关系 508 00:16:16,399 --> 00:16:16,799 嗯 509 00:16:16,799 --> 00:16:17,120 你知道的 510 00:16:17,120 --> 00:16:19,759 我觉得这周OpenAI 511 00:16:19,759 --> 00:16:20,480 呃 512 00:16:20,480 --> 00:16:21,539 有点自毁了 513 00:16:21,539 --> 00:16:23,250 脱离了微软 514 00:16:23,250 --> 00:16:24,289 微软表示 515 00:16:24,289 --> 00:16:24,730 好吧 516 00:16:24,730 --> 00:16:28,440 也许我们通过 517 00:16:28,440 --> 00:16:31,019 要求他们使用微软边缘服务 518 00:16:31,019 --> 00:16:34,340 顺便说一下我们持有OpenAI约三分之一的股份 519 00:16:34,340 --> 00:16:35,759 所以我们不希望 520 00:16:35,759 --> 00:16:36,480 你知道的 521 00:16:36,480 --> 00:16:38,419 失去这项资产的价值 522 00:16:38,419 --> 00:16:41,139 所以我们允许你们脱离我们的云服务 523 00:16:41,139 --> 00:16:43,440 他们随即与AWS签约 524 00:16:43,440 --> 00:16:45,799 这对AWS来说是个好交易 525 00:16:45,799 --> 00:16:47,440 目前数据还没体现 526 00:16:47,440 --> 00:16:48,080 但我们确实看到 527 00:16:48,080 --> 00:16:50,460 正如你所说加速效应开始显现 528 00:16:50,460 --> 00:16:54,659 我其实看到Travis在线说很多人在使用 529 00:16:54,659 --> 00:16:56,200 类似Anthropic的模型 530 00:16:56,200 --> 00:16:58,159 当你使用AWS时 531 00:16:58,159 --> 00:17:00,500 只能选择有限的模型 532 00:17:00,500 --> 00:17:02,139 有些人不愿切换 533 00:17:02,139 --> 00:17:03,440 我已经在AWS了 534 00:17:03,440 --> 00:17:05,259 所有系统都在AWS 535 00:17:05,259 --> 00:17:05,940 AWS 536 00:17:05,940 --> 00:17:07,839 AWS是最大的云服务商 537 00:17:07,839 --> 00:17:08,599 他们不会 538 00:17:08,599 --> 00:17:10,250 如果我的系统都在那里 539 00:17:10,250 --> 00:17:13,190 全部迁移到Azure会很困难 540 00:17:13,190 --> 00:17:14,670 仅仅为了一个AI模型 541 00:17:14,670 --> 00:17:17,900 Anthropic有时领先 542 00:17:17,900 --> 00:17:20,259 有时落后于OpenAI 543 00:17:20,259 --> 00:17:22,160 但若竞争激烈 544 00:17:22,160 --> 00:17:22,819 链条 545 00:17:22,819 --> 00:17:25,269 我迁移的动力不存在 546 00:17:25,269 --> 00:17:28,490 现在OpenAI能直接部署 547 00:17:28,490 --> 00:17:31,950 GPT 5.5和5.4到AWS 548 00:17:31,950 --> 00:17:34,390 我认为这对所有人都是巨大胜利 549 00:17:34,390 --> 00:17:35,329 对所有参与者 550 00:17:35,329 --> 00:17:35,549 没错 551 00:17:35,549 --> 00:17:36,430 将会有一些公司 552 00:17:36,430 --> 00:17:36,710 哦看 553 00:17:36,710 --> 00:17:39,339 我其实想和GPT-5.5合作 554 00:17:39,339 --> 00:17:40,819 也许我可以回到Bedrock 555 00:17:40,819 --> 00:17:41,619 嗯为了做这件事 556 00:17:41,619 --> 00:17:43,099 就像你提到的微软 557 00:17:43,099 --> 00:17:45,809 因为他们拥有OpenAI的大部分股份 558 00:17:45,809 --> 00:17:46,910 扩大规模 559 00:17:46,910 --> 00:17:49,888 OpenAI的投入对他们来说可能更值得 560 00:17:49,888 --> 00:17:50,308 嗯 561 00:17:50,308 --> 00:17:52,409 AWS展示的其他亮眼数据 562 00:17:52,409 --> 00:17:52,729 是 563 00:17:52,729 --> 00:17:55,709 他们提到内部芯片解决方案 564 00:17:55,709 --> 00:17:56,880 嗯 565 00:17:56,880 --> 00:17:59,920 投入2250亿用于训练 566 00:17:59,920 --> 00:18:00,200 Anium 567 00:18:00,200 --> 00:18:01,000 这是他们的AI 568 00:18:01,000 --> 00:18:03,019 芯片主要涉及Anthropic和OpenAI 569 00:18:03,019 --> 00:18:05,119 但他们刚发布的芯片 570 00:18:05,119 --> 00:18:08,339 Tranium 3已全面预订，Tranium 4 571 00:18:08,339 --> 00:18:12,470 预计15-18个月后推出 572 00:18:12,470 --> 00:18:15,970 再次提前大量预订 573 00:18:15,970 --> 00:18:17,069 我交给你了 574 00:18:17,069 --> 00:18:17,549 Travis 575 00:18:17,549 --> 00:18:17,970 嗯 576 00:18:17,970 --> 00:18:21,250 我们看到谷歌在销售自家AI芯片 577 00:18:21,250 --> 00:18:23,130 他们正在销售TPU 578 00:18:23,130 --> 00:18:27,390 亚马逊CEO开始以独立业务视角 579 00:18:27,390 --> 00:18:28,670 如果我们 580 00:18:28,670 --> 00:18:30,509 如果我们出售芯片 581 00:18:30,509 --> 00:18:32,970 实际会有约50亿美元 582 00:18:32,970 --> 00:18:33,529 嗯 583 00:18:33,529 --> 00:18:35,069 芯片业务的营收规模 584 00:18:35,069 --> 00:18:37,029 不知道他们是否在暗示 585 00:18:37,029 --> 00:18:40,490 让我们预览未来几个季度的趋势 586 00:18:40,490 --> 00:18:41,349 嗯是的 587 00:18:41,349 --> 00:18:41,730 嗯 588 00:18:41,730 --> 00:18:45,109 但他们的用词很有趣 589 00:18:45,109 --> 00:18:46,650 如果我们是独立芯片公司 590 00:18:46,650 --> 00:18:49,309 这就是我们的营收规模 591 00:18:49,309 --> 00:18:50,509 你怎么看这个 592 00:18:50,509 --> 00:18:54,420 因为这三家公司处境特殊 593 00:18:54,420 --> 00:18:57,160 嗯我们之前讨论过其中一个 594 00:18:57,160 --> 00:18:59,529 你一直是英伟达的坚定支持者 595 00:18:59,529 --> 00:19:01,369 但英伟达的风险在于 596 00:19:01,369 --> 00:19:03,140 客户集中度过高 597 00:19:03,140 --> 00:19:05,339 我认为在之前的季度 598 00:19:05,339 --> 00:19:09,679 我们看到四大客户贡献了约60%营收 599 00:19:09,679 --> 00:19:11,838 现在这些客户包括 600 00:19:11,838 --> 00:19:16,338 特别是Alphabet的GCP和亚马逊 601 00:19:16,338 --> 00:19:17,278 凭借他们的tranium 602 00:19:17,278 --> 00:19:18,409 打招呼吧 603 00:19:18,409 --> 00:19:20,429 我们正在尝试要做的事 604 00:19:20,429 --> 00:19:23,348 客户根本看不到芯片对吧 605 00:19:23,348 --> 00:19:24,788 他们正在地基上构建 606 00:19:24,788 --> 00:19:27,449 在地表之下 607 00:19:27,769 --> 00:19:30,229 当然如果英伟达芯片最高效 608 00:19:30,229 --> 00:19:31,068 我们会使用它 609 00:19:31,068 --> 00:19:32,068 但我们想打造 610 00:19:32,068 --> 00:19:34,319 我们想尝试 commoditize 这一层 611 00:19:34,319 --> 00:19:36,900 而他们在云领域做到了这一点 612 00:19:36,900 --> 00:19:39,150 没错你最初用的是英特尔芯片 613 00:19:39,150 --> 00:19:41,069 然后转向AMD芯片 614 00:19:41,069 --> 00:19:41,549 嗯 615 00:19:41,549 --> 00:19:41,950 你知道的 616 00:19:41,950 --> 00:19:44,589 或者因为更便宜一些 617 00:19:44,589 --> 00:19:45,029 它们 618 00:19:45,029 --> 00:19:45,529 它们 619 00:19:45,529 --> 00:19:45,970 你知道的 620 00:19:45,970 --> 00:19:47,390 你能获得更好的定价权 621 00:19:47,390 --> 00:19:49,619 或者可以在这两者之间博弈 622 00:19:49,619 --> 00:19:51,519 正如你看到这些交易 623 00:19:51,519 --> 00:19:52,779 我同意你说的 624 00:19:52,779 --> 00:19:53,640 TPU胜出 625 00:19:53,640 --> 00:19:56,680 Tranium的胜利对这些公司有利 626 00:19:56,680 --> 00:20:00,089 但这如何影响芯片领域的另一巨头 627 00:20:00,089 --> 00:20:01,210 那就是英伟达 628 00:20:01,210 --> 00:20:01,869 是的 629 00:20:01,869 --> 00:20:02,630 我认为 630 00:20:02,630 --> 00:20:03,190 嗯 631 00:20:03,190 --> 00:20:07,160 这在当前周期确实值得注意 632 00:20:07,160 --> 00:20:07,500 而且 633 00:20:07,500 --> 00:20:10,380 我们看到的AI市场很多内容 634 00:20:10,380 --> 00:20:13,460 更依赖于周期持续对吧 635 00:20:13,460 --> 00:20:15,220 在当前周期 636 00:20:15,220 --> 00:20:18,940 我们听到每个人其实都 637 00:20:18,940 --> 00:20:23,980 我刚做了些研究，所有四份记录都说同一件事 638 00:20:23,980 --> 00:20:24,579 无论哪种方式 639 00:20:24,579 --> 00:20:25,750 他们面临计算资源限制 640 00:20:25,750 --> 00:20:26,470 谷歌曾表示 641 00:20:26,470 --> 00:20:27,269 如果获得更多芯片 642 00:20:27,269 --> 00:20:29,409 就能在此次市场中做到 643 00:20:29,409 --> 00:20:31,308 我不太担心英伟达 644 00:20:31,308 --> 00:20:32,868 如果最终周期下滑 645 00:20:32,868 --> 00:20:35,489 显然客户会削减资本支出 646 00:20:35,489 --> 00:20:36,828 不仅削减资本支出 647 00:20:36,828 --> 00:20:39,039 还会推进自身 648 00:20:39,039 --> 00:20:39,960 嗯 649 00:20:39,960 --> 00:20:41,240 投入自研芯片 650 00:20:41,240 --> 00:20:43,960 可能进一步降低成本 651 00:20:43,960 --> 00:20:44,920 当前的行业格局 652 00:20:44,920 --> 00:20:46,880 只要人工智能创新持续进行 653 00:20:46,880 --> 00:20:47,480 我们正在见证这一点 654 00:20:47,480 --> 00:20:48,200 我们正深入其中 655 00:20:48,200 --> 00:20:49,099 人工智能代理 656 00:20:49,099 --> 00:20:51,299 人工智能代理不可能是最终答案 657 00:20:51,299 --> 00:20:51,500 人工智能 658 00:20:51,500 --> 00:20:52,579 必须存在更深层次的东西 659 00:20:52,579 --> 00:20:53,539 超越当前的层面 660 00:20:53,539 --> 00:20:55,059 或是对现有方案的改进 661 00:20:55,059 --> 00:20:59,980 嗯，我仍然认为人工智能代理足以成为该行业的需求缺口 662 00:20:59,980 --> 00:21:02,119 另一个问题是供应链 663 00:21:02,119 --> 00:21:04,210 供应链短缺确实存在 664 00:21:04,210 --> 00:21:05,670 我们缺乏足够的内存 665 00:21:05,670 --> 00:21:07,279 我们严重短缺 666 00:21:07,279 --> 00:21:08,240 台积电 667 00:21:08,240 --> 00:21:09,119 台积电曾提到 668 00:21:09,119 --> 00:21:11,259 他们产能不足无法扩大生产 669 00:21:11,259 --> 00:21:12,200 芯片制造 670 00:21:12,200 --> 00:21:16,269 其他企业都在实现可观的自由现金流 671 00:21:16,269 --> 00:21:18,930 或强劲的运营现金流 672 00:21:18,930 --> 00:21:20,670 必须明智使用资金 673 00:21:20,670 --> 00:21:21,609 他们将资金投入何处 674 00:21:21,609 --> 00:21:24,559 用于建设数据中心和采购服务器 675 00:21:24,559 --> 00:21:27,700 英伟达也拥有强大的现金流 676 00:21:27,700 --> 00:21:30,240 但英伟达如何运用这笔资金 677 00:21:30,240 --> 00:21:32,219 直接投入供应链 678 00:21:32,219 --> 00:21:36,259 因此可以抢占约60-70%的 679 00:21:36,259 --> 00:21:36,699 假设 680 00:21:36,699 --> 00:21:37,778 理论上我在举例 681 00:21:37,778 --> 00:21:41,150 约占HBM总产能的60-70% 682 00:21:41,150 --> 00:21:43,109 因此占据优势地位 683 00:21:43,109 --> 00:21:43,369 思考 684 00:21:43,369 --> 00:21:44,549 得益于充足的自由现金流 685 00:21:44,549 --> 00:21:46,809 无需自建数据中心 686 00:21:46,809 --> 00:21:48,529 无需担忧其他投资 687 00:21:48,529 --> 00:21:50,650 可将现金流用于巩固 688 00:21:50,650 --> 00:21:52,589 供应链领域 689 00:21:52,589 --> 00:21:54,529 当然可以自行研发芯片 690 00:21:54,529 --> 00:21:56,509 但产能有限 691 00:21:56,509 --> 00:21:57,980 其余部分由我们承担 692 00:21:57,980 --> 00:21:59,539 最终这种模式将不可持续 693 00:21:59,539 --> 00:22:00,519 但在当前周期 694 00:22:00,519 --> 00:22:02,680 我认为这为他们提供了增长机会 695 00:22:02,680 --> 00:22:02,880 是的 696 00:22:02,880 --> 00:22:03,319 甚至更多 697 00:22:03,319 --> 00:22:04,759 这正是让我担忧的 698 00:22:04,759 --> 00:22:05,559 整个局势 699 00:22:05,559 --> 00:22:06,880 我们绝不能忽视 700 00:22:06,880 --> 00:22:07,039 这一点 701 00:22:07,039 --> 00:22:09,779 我们昨天得到的数据简直令人难以置信 702 00:22:09,779 --> 00:22:11,200 尤其是我简直惊呆了 703 00:22:11,200 --> 00:22:13,160 看着云端字母公司的数据 704 00:22:13,160 --> 00:22:17,089 但这一切背后还有一个层面 705 00:22:17,089 --> 00:22:18,109 直到它不再运转 706 00:22:18,109 --> 00:22:22,109 而我足够老去记得互联网泡沫破裂 707 00:22:22,109 --> 00:22:22,490 呃 708 00:22:22,490 --> 00:22:23,849 我们之前讨论过这个 709 00:22:23,849 --> 00:22:24,859 你知道的 710 00:22:24,859 --> 00:22:26,859 真正的挑战通常是这些事物 711 00:22:26,859 --> 00:22:27,759 当它们破裂时 712 00:22:27,759 --> 00:22:28,599 如果可以这样说的话 713 00:22:28,599 --> 00:22:30,779 并非因为投资停止 714 00:22:30,779 --> 00:22:32,839 如果你看看电信基建 715 00:22:32,839 --> 00:22:34,920 电信建设呈指数增长曲线 716 00:22:34,920 --> 00:22:36,420 就像我们现在看到的AI建设 717 00:22:36,420 --> 00:22:38,150 然后就会趋于平缓 718 00:22:38,150 --> 00:22:39,650 长达二十五年 719 00:22:39,650 --> 00:22:41,589 我说这在2021年达到顶峰 720 00:22:41,589 --> 00:22:42,950 而现在我们刚刚突破 721 00:22:42,950 --> 00:22:43,490 呃 722 00:22:43,490 --> 00:22:46,118 去年光纤领域的投资 723 00:22:46,118 --> 00:22:46,919 呃 724 00:22:46,919 --> 00:22:47,298 你知道的 725 00:22:47,298 --> 00:22:47,699 大致估算 726 00:22:47,699 --> 00:22:48,519 这差不多对 727 00:22:48,519 --> 00:22:49,680 所以它其实 728 00:22:49,680 --> 00:22:51,500 这将非常有趣 729 00:22:51,500 --> 00:22:52,500 你知道我们仍处于那个阶段 730 00:22:52,500 --> 00:22:52,700 当 731 00:22:52,700 --> 00:22:53,279 正如你所说 732 00:22:53,279 --> 00:22:54,019 字母公司说 733 00:22:54,019 --> 00:22:55,359 明年我们将继续加大投入 734 00:22:55,359 --> 00:22:58,730 顺便说一下字母公司还有现金流持续投入 735 00:22:58,730 --> 00:23:01,950 亚马逊现在到了花光所有营收的地步 736 00:23:01,950 --> 00:23:03,500 并承担更多债务 737 00:23:03,500 --> 00:23:04,420 你知道的 738 00:23:04,420 --> 00:23:05,940 同领域的甲骨文公司 739 00:23:05,940 --> 00:23:08,180 现在面临完全不同的风险特征 740 00:23:08,180 --> 00:23:10,069 再加上涉及的金额 741 00:23:10,069 --> 00:23:11,230 如果利润率变化 742 00:23:11,230 --> 00:23:13,309 如果需求变化 743 00:23:13,309 --> 00:23:15,210 风险特征会有所不同 744 00:23:15,210 --> 00:23:17,569 具体取决于所分析的公司 745 00:23:17,569 --> 00:23:19,430 让我们谈谈Meta平台 746 00:23:19,430 --> 00:23:20,750 因为Meta是 747 00:23:20,750 --> 00:23:24,069 我认为这里真正令人震惊的反应 748 00:23:24,069 --> 00:23:25,569 股价下跌了10% 749 00:23:25,569 --> 00:23:26,430 在我们录制时 750 00:23:26,430 --> 00:23:29,470 这与其它公司业务完全不同 751 00:23:29,470 --> 00:23:30,279 他们不是 752 00:23:30,279 --> 00:23:31,480 你知道超大规模云服务商 753 00:23:31,480 --> 00:23:31,880 从某种意义上说 754 00:23:31,880 --> 00:23:32,400 他们正在 755 00:23:32,400 --> 00:23:34,750 他们为其他公司提供模型服务 756 00:23:34,750 --> 00:23:36,509 但他们确实想参与这个游戏 757 00:23:36,509 --> 00:23:37,589 他们正在构建自有模型 758 00:23:37,589 --> 00:23:38,289 打造自己的 759 00:23:38,289 --> 00:23:39,369 你知道AI团队 760 00:23:39,369 --> 00:23:43,950 他们在过去六个月进行了全面升级，所以当你看这个 761 00:23:43,950 --> 00:23:44,940 是的 762 00:23:44,940 --> 00:23:48,539 Meta为何大幅削减投入背后的故事是什么 763 00:23:48,539 --> 00:23:50,039 因为我认为这家公司规模在扩张 764 00:23:50,039 --> 00:23:52,529 我估计本季度增长了3.3% 765 00:23:52,529 --> 00:23:54,549 表面看起来非常出色 766 00:23:54,549 --> 00:23:55,029 是的 767 00:23:55,029 --> 00:23:55,470 我是说 768 00:23:55,470 --> 00:23:58,230 我认为这对Meta来说很不幸 769 00:23:58,230 --> 00:24:00,190 它没有云业务作为支撑可以说 770 00:24:00,190 --> 00:24:02,269 这才是实际营收增长所在 771 00:24:02,269 --> 00:24:03,710 其他竞争者都能这么说 772 00:24:03,710 --> 00:24:04,569 嗯 773 00:24:04,569 --> 00:24:05,769 我我 774 00:24:05,769 --> 00:24:08,069 我相信他们在 775 00:24:08,069 --> 00:24:10,730 AI对其推荐系统至关重要 776 00:24:10,730 --> 00:24:11,549 他们曾提及 777 00:24:11,549 --> 00:24:11,910 过去 778 00:24:11,910 --> 00:24:13,349 我忘了具体模型名称 779 00:24:13,349 --> 00:24:14,519 他们开发的名为Gem 780 00:24:14,519 --> 00:24:16,400 或类似新启动的 781 00:24:16,400 --> 00:24:16,880 嗯 782 00:24:16,880 --> 00:24:21,039 他们提到这个推荐系统正像LLM一样成长 783 00:24:21,039 --> 00:24:22,920 投入的计算资源越多 784 00:24:22,920 --> 00:24:23,940 效果越好 785 00:24:23,940 --> 00:24:24,420 嗯 786 00:24:24,420 --> 00:24:26,000 这对核心业务至关重要 787 00:24:26,000 --> 00:24:26,220 没错 788 00:24:26,220 --> 00:24:27,339 如果你有更多算力 789 00:24:27,339 --> 00:24:27,660 呃 790 00:24:27,660 --> 00:24:28,460 我认为我们看到 791 00:24:28,460 --> 00:24:34,339 广告展示量和广告成本同比双双上升 792 00:24:34,339 --> 00:24:36,929 这在我看来是 793 00:24:36,929 --> 00:24:38,949 若广告效率提升 794 00:24:38,949 --> 00:24:40,929 就能提高广告收费 795 00:24:40,929 --> 00:24:43,068 这说明这个推荐系统 796 00:24:43,068 --> 00:24:44,949 AI投入对他们有效 797 00:24:44,949 --> 00:24:46,929 并同时助力其营收 798 00:24:46,929 --> 00:24:47,249 嗯 799 00:24:47,249 --> 00:24:48,528 正在提升主营业务收入 800 00:24:48,528 --> 00:24:52,660 另一方面我们也看到负面变化，这是首次 801 00:24:52,660 --> 00:24:55,180 我们看到每日活跃用户数量下降 802 00:24:55,180 --> 00:25:00,319 他们提到了网络中断以及冲突中的一些紧张局势 803 00:25:00,319 --> 00:25:01,470 中东地区局势动荡 804 00:25:01,470 --> 00:25:05,799 这正在导致某些地区在平台上的活跃度降低 805 00:25:05,799 --> 00:25:06,319 所以 806 00:25:06,319 --> 00:25:07,059 嗯 807 00:25:07,059 --> 00:25:10,059 看看这个故事有多严重会很有趣 808 00:25:10,059 --> 00:25:11,079 仅针对本季度 809 00:25:11,079 --> 00:25:12,950 这个故事将持续多久 810 00:25:12,950 --> 00:25:16,730 嗯，我们的投资者要么在买入，要么相信这是事实 811 00:25:16,730 --> 00:25:19,289 我认为这可能引发了一些担忧 812 00:25:19,289 --> 00:25:21,329 我认为对于投资者来说 813 00:25:21,329 --> 00:25:22,930 这是一个完全不同的商业模式 814 00:25:22,930 --> 00:25:25,549 尽管他们在技术层面做了很多相同的事情 815 00:25:25,549 --> 00:25:27,000 这本质上是一个注意力经济 816 00:25:27,000 --> 00:25:27,839 从这个角度来看 817 00:25:27,839 --> 00:25:30,299 可能更类似于YouTube的情况 818 00:25:30,299 --> 00:25:31,619 或者像 819 00:25:31,619 --> 00:25:32,079 呃 820 00:25:32,079 --> 00:25:33,739 像Netflix一样 821 00:25:33,739 --> 00:25:36,759 现在把这四个因素综合起来 822 00:25:36,759 --> 00:25:41,099 或许我们可以各自对这四只股票进行排名 823 00:25:41,099 --> 00:25:42,740 所以你有字母表公司 824 00:25:42,740 --> 00:25:43,400 亚马逊 825 00:25:43,400 --> 00:25:45,299 微软和Meta 826 00:25:45,299 --> 00:25:46,859 在今日的走势之后 827 00:25:46,859 --> 00:25:48,680 你最看好的是哪一家 828 00:25:48,680 --> 00:25:50,740 我可以选最看好的吗 829 00:25:50,740 --> 00:25:53,579 长期来看并基于当前估值 830 00:25:53,579 --> 00:25:54,059 是的 831 00:25:54,059 --> 00:25:54,740 让我们开始 832 00:25:54,740 --> 00:25:56,019 让我们设定三年时间框架 833 00:25:56,019 --> 00:25:56,700 三年或更久 834 00:25:56,700 --> 00:25:57,000 好的 835 00:25:57,000 --> 00:25:57,779 你最看好的是哪个 836 00:25:57,779 --> 00:25:58,500 我只能说这四家都行 837 00:25:58,500 --> 00:25:59,759 但如果必须选一个 838 00:25:59,759 --> 00:26:00,259 我认为 839 00:26:00,259 --> 00:26:01,099 应该是谷歌 840 00:26:01,099 --> 00:26:01,380 我的意思是 841 00:26:01,380 --> 00:26:03,589 谷歌已经构建了完整的生态系统 842 00:26:03,589 --> 00:26:05,329 他们在 843 00:26:05,329 --> 00:26:06,400 AI芯片领域表现优异 844 00:26:06,400 --> 00:26:07,900 拥有不错的模型 845 00:26:07,900 --> 00:26:09,160 我个人并不使用 846 00:26:09,160 --> 00:26:10,740 但我知道很多人在用 847 00:26:10,740 --> 00:26:11,279 嗯 848 00:26:11,279 --> 00:26:13,619 然后是广告业务 849 00:26:13,619 --> 00:26:14,359 YouTube 850 00:26:14,359 --> 00:26:17,480 他们的所有业务都是紧密相连的 851 00:26:17,480 --> 00:26:18,559 他们是非常优秀的投资者 852 00:26:18,559 --> 00:26:19,160 看起来确实如此 853 00:26:19,160 --> 00:26:21,959 嗯，显然 cuckoo 在那里排名第一 854 00:26:21,959 --> 00:26:23,349 长期来看 855 00:26:23,349 --> 00:26:24,789 我知道我曾经拥有过一个 856 00:26:24,789 --> 00:26:24,990 是的 857 00:26:24,990 --> 00:26:26,250 我完全有同感 858 00:26:26,250 --> 00:26:26,789 呃 859 00:26:26,789 --> 00:26:28,490 你知道这是我的最大持仓 860 00:26:28,490 --> 00:26:32,660 而且我称它为人工智能的便捷按钮 861 00:26:32,660 --> 00:26:34,299 因为他们没有模型 862 00:26:34,299 --> 00:26:35,559 他们拥有最多的计算资源 863 00:26:35,559 --> 00:26:36,569 他们有自己的芯片 864 00:26:36,569 --> 00:26:37,730 他们掌握着分发渠道 865 00:26:37,730 --> 00:26:38,329 我们还没谈到 866 00:26:38,329 --> 00:26:39,809 我们没太多讨论分发渠道 867 00:26:39,809 --> 00:26:41,720 但在搜索领域 868 00:26:41,720 --> 00:26:45,480 过去几年我们一直在讨论聊天机器人 869 00:26:45,480 --> 00:26:48,839 普通人仍然主要使用搜索栏 870 00:26:48,839 --> 00:26:51,019 你知道他们只是 871 00:26:51,019 --> 00:26:51,480 哦 872 00:26:51,480 --> 00:26:52,099 看看这个 873 00:26:52,099 --> 00:26:54,000 这个AI模式挺酷的 874 00:26:54,000 --> 00:26:55,079 直接给我答案 875 00:26:55,079 --> 00:26:56,079 我甚至不用离开当前位置 876 00:26:56,079 --> 00:26:59,299 它已经融入了我的日常流程 877 00:26:59,299 --> 00:27:02,279 这对他们来说是绝佳的市场地位 878 00:27:02,279 --> 00:27:03,539 接下来你选谁 879 00:27:03,539 --> 00:27:07,579 接下来我将根据评估给出我的首选 880 00:27:07,579 --> 00:27:10,299 嗯，我认为应该是微软 881 00:27:10,299 --> 00:27:12,019 我认为市场明显 882 00:27:12,019 --> 00:27:14,980 因云业务增长而过度惩罚微软 883 00:27:14,980 --> 00:27:18,920 他们认为微软在失去市场份额 884 00:27:18,920 --> 00:27:21,960 我认为他们之前几个季度投入不足 885 00:27:21,960 --> 00:27:23,859 他们已彻底改变策略 886 00:27:23,859 --> 00:27:25,980 我们将看到他们的加速增长 887 00:27:25,980 --> 00:27:27,740 所以这就是我的两个第一 888 00:27:27,740 --> 00:27:31,650 我认为谷歌有潜力成为第一公司 889 00:27:31,650 --> 00:27:33,809 不得不承认他们是强劲竞争对手 890 00:27:33,809 --> 00:27:35,170 与英伟达竞争 891 00:27:35,170 --> 00:27:39,119 但或许谷歌会成为最大公司 892 00:27:39,119 --> 00:27:41,400 我认为微软被高估了 893 00:27:41,400 --> 00:27:43,230 所以这是我的第二选择 894 00:27:43,230 --> 00:27:45,390 因为增长正在加速 895 00:27:45,390 --> 00:27:45,809 是的 896 00:27:45,809 --> 00:27:46,230 第二名 897 00:27:46,230 --> 00:27:48,710 我必须选择Meta 898 00:27:48,710 --> 00:27:51,549 这股票我从未持有，完全错过Meta 899 00:27:51,549 --> 00:27:52,699 这是我的失误 900 00:27:52,699 --> 00:27:54,519 但最终我们会遇到其中之一 901 00:27:54,519 --> 00:27:54,818 你知道吗 902 00:27:54,818 --> 00:27:55,699 二零二二年的关键时刻 903 00:27:55,699 --> 00:27:58,119 那里价格低廉而且人们 904 00:27:58,119 --> 00:28:00,539 他们使用的很多产品都在Instagram上 905 00:28:00,539 --> 00:28:02,579 你知道Facebook可能可能更 906 00:28:02,579 --> 00:28:04,949 我那一代人在千禧年初 907 00:28:08,940 --> 00:28:11,259 然后看看他们未来的可能性 908 00:28:11,259 --> 00:28:12,000 他们为未来提供的选择 909 00:28:12,000 --> 00:28:13,319 因此获得不错的估值 910 00:28:13,319 --> 00:28:13,819 嗯 911 00:28:13,819 --> 00:28:15,859 获得表现优异的业务 912 00:28:15,859 --> 00:28:16,420 收入增长 913 00:28:16,420 --> 00:28:18,659 以及我们提到的利润率扩张 914 00:28:18,659 --> 00:28:20,598 但再想想未来的可能性 915 00:28:20,598 --> 00:28:22,919 如果有人能搞清楚 916 00:28:22,919 --> 00:28:24,858 AI驱动的眼镜会是什么样子 917 00:28:24,858 --> 00:28:27,479 我会押注Meta而非苹果 918 00:28:27,479 --> 00:28:28,219 例如 919 00:28:28,219 --> 00:28:29,118 所以啊 920 00:28:29,118 --> 00:28:31,298 那里确实存在增长机会 921 00:28:31,298 --> 00:28:32,900 尽管他们曾遇到困难 922 00:28:32,900 --> 00:28:34,920 实际上在这些方面盈利 923 00:28:34,920 --> 00:28:35,579 嗯 好吧 924 00:28:35,579 --> 00:28:36,480 第三名是谁 925 00:28:36,480 --> 00:28:37,200 第三名 926 00:28:37,200 --> 00:28:37,619 哦 天啊 927 00:28:37,619 --> 00:28:38,339 这个有点难 928 00:28:38,339 --> 00:28:40,519 我认为只是因为增长放缓 929 00:28:40,519 --> 00:28:42,119 我选择第三名给Meta 930 00:28:42,119 --> 00:28:42,700 对的 931 00:28:42,700 --> 00:28:44,799 如果Meta没出现这次下跌 932 00:28:44,799 --> 00:28:46,059 或许我会选第一 933 00:28:46,059 --> 00:28:48,759 但正如你所说它有未来潜力 934 00:28:48,759 --> 00:28:50,200 这就是我排第三的原因 935 00:28:50,200 --> 00:28:52,180 因为这种潜力既是 936 00:28:52,180 --> 00:28:56,160 嗯 既是增长机会也可能未来有风险 937 00:28:56,160 --> 00:28:58,500 我觉得微软更稳健 938 00:28:58,500 --> 00:29:00,019 但但Meta这边 939 00:29:00,019 --> 00:29:01,220 极佳的估值 940 00:29:01,220 --> 00:29:02,880 扎克伯格有远见 941 00:29:02,880 --> 00:29:04,339 我认同这个愿景 942 00:29:04,339 --> 00:29:07,140 但市场有时会质疑 943 00:29:07,140 --> 00:29:09,579 我们将看到类似的价格反应 944 00:29:09,579 --> 00:29:10,000 嗯 945 00:29:10,000 --> 00:29:10,920 未来几个季度 946 00:29:10,920 --> 00:29:11,660 就像这样 947 00:29:11,660 --> 00:29:13,299 他们会打问号 948 00:29:13,299 --> 00:29:16,099 你是否正确组建了超级实验室团队 949 00:29:16,099 --> 00:29:18,579 你是否正确进行这类投资 950 00:29:18,579 --> 00:29:21,559 不过从长远来看，我认为这确实是个正确的决定 951 00:29:21,559 --> 00:29:22,039 好的 952 00:29:22,039 --> 00:29:24,769 我会支持亚马逊这边 953 00:29:24,769 --> 00:29:25,890 给他们一点支持 954 00:29:25,890 --> 00:29:30,210 其实我之前没续过Prime会员，现在重新续了 955 00:29:30,210 --> 00:29:31,769 大概五六年没用了 956 00:29:31,769 --> 00:29:33,529 能够做到这一点简直太疯狂了 957 00:29:33,529 --> 00:29:33,890 你知道的 958 00:29:33,890 --> 00:29:35,450 在十点左右下单 959 00:29:35,450 --> 00:29:37,869 然后早上四点就能送到门口 960 00:29:37,869 --> 00:29:38,470 嗯 961 00:29:38,470 --> 00:29:39,930 但另一个问题是 962 00:29:39,930 --> 00:29:45,339 我认为有人会扮演大宗商品云服务提供商的角色 963 00:29:45,339 --> 00:29:48,579 我觉得亚马逊在这方面比任何人都做得好 964 00:29:48,579 --> 00:29:50,119 他们正是用AWS做到了这一点 965 00:29:50,119 --> 00:29:51,799 他们没做什么疯狂的事 966 00:29:51,799 --> 00:29:53,180 只是顺势而为 967 00:29:53,180 --> 00:29:53,559 他们 968 00:29:53,559 --> 00:29:55,739 让产品更便宜更高效 969 00:29:55,739 --> 00:29:57,858 所以我认为他们将在未来云领域扮演重要角色 970 00:29:57,858 --> 00:30:00,278 未来云领域将发挥重要作用 971 00:30:00,278 --> 00:30:02,839 最终可能需要拆分这些业务 972 00:30:02,839 --> 00:30:07,420 因为AWS业务规模巨大，可能占公司价值的绝大部分 973 00:30:07,420 --> 00:30:10,140 今天可能占公司价值的100% 974 00:30:10,140 --> 00:30:12,240 来看我们的第四点 975 00:30:12,240 --> 00:30:15,250 微软在市场获得了信任 976 00:30:15,250 --> 00:30:16,910 当OpenAI出现时 977 00:30:16,910 --> 00:30:18,369 他们之前就有合作 978 00:30:18,369 --> 00:30:19,990 当时没人真正了解OpenAI 979 00:30:19,990 --> 00:30:21,890 他们投资了十亿美元 980 00:30:21,890 --> 00:30:22,630 之后 981 00:30:22,630 --> 00:30:24,529 这将是巨大的收益 982 00:30:24,529 --> 00:30:28,279 我不认为人们会认为微软的Copilot产品 983 00:30:28,279 --> 00:30:32,089 在人工智能领域有优秀的产品 984 00:30:32,089 --> 00:30:33,769 只是说 985 00:30:33,769 --> 00:30:35,630 公司是微软的 986 00:30:35,630 --> 00:30:38,170 所以我们必须用微软的东西 987 00:30:38,170 --> 00:30:40,710 我不认为这是未来的增长动力 988 00:30:40,710 --> 00:30:43,799 如果AI要改变人们的工作方式 989 00:30:43,799 --> 00:30:47,119 微软可能不会拥有最佳界面 990 00:30:47,119 --> 00:30:48,380 最终胜出的将是那个 991 00:30:48,380 --> 00:30:50,599 所以我不确定这就是我选择他们的原因 992 00:30:50,599 --> 00:30:52,779 为什么选择亚马逊呢 993 00:30:52,779 --> 00:30:53,779 第四位 994 00:30:53,779 --> 00:30:55,740 只是不错的上涨 995 00:30:55,740 --> 00:30:57,700 不过我同意 996 00:30:57,700 --> 00:30:59,500 他们的AWS业务将加速增长 997 00:30:59,500 --> 00:31:02,680 让我再说第一点还很遥远 998 00:31:02,680 --> 00:31:03,888 第二第三第四 999 00:31:03,888 --> 00:31:05,269 这取决于周的情况 1000 00:31:05,269 --> 00:31:06,328 特拉维斯，是你让我来的 1001 00:31:06,328 --> 00:31:06,528 是的 1002 00:31:06,528 --> 00:31:08,368 嗯，能不能直接关掉啊 1003 00:31:08,368 --> 00:31:09,489 那些完全可以关掉 1004 00:31:09,489 --> 00:31:10,608 但呃亚马逊 1005 00:31:10,608 --> 00:31:11,138 我是说 1006 00:31:11,138 --> 00:31:13,369 作为最大的云服务提供商 1007 00:31:13,369 --> 00:31:14,769 没错，现在他们开放了 1008 00:31:14,769 --> 00:31:18,309 仅此一项就为他们的云业务打开了巨大机遇 1009 00:31:18,309 --> 00:31:18,970 仅此一项 1010 00:31:18,970 --> 00:31:20,369 他们刚推出新模型 1011 00:31:20,369 --> 00:31:21,269 就像 1012 00:31:21,269 --> 00:31:21,789 好的 1013 00:31:21,789 --> 00:31:22,630 你们就待在这里 1014 00:31:22,630 --> 00:31:23,829 你们要花更多代币 1015 00:31:23,829 --> 00:31:24,349 所以对我来说 1016 00:31:24,349 --> 00:31:25,269 这个AI基础设施 1017 00:31:25,269 --> 00:31:26,250 作为最大的玩家让他们能够增长 1018 00:31:26,250 --> 00:31:28,680 嗯 1019 00:31:28,680 --> 00:31:29,099 我不会撒谎 1020 00:31:29,099 --> 00:31:29,940 我持有亚马逊股票 1021 00:31:29,940 --> 00:31:30,900 我从不关注他们的电商财务数据 1022 00:31:30,900 --> 00:31:34,140 我只是非常担心AWS业务 1023 00:31:34,140 --> 00:31:36,099 我看他们的广告业务也算不错 1024 00:31:36,099 --> 00:31:38,920 算是订阅业务的一个亮点 1025 00:31:38,920 --> 00:31:40,779 不错的试水 1026 00:31:40,779 --> 00:31:45,119 显然因为像你这样的人在七年没用后又重新订阅了 1027 00:31:45,119 --> 00:31:45,819 但呃 1028 00:31:46,500 --> 00:31:48,000 我很少关注这些 1029 00:31:48,000 --> 00:31:48,240 哦 1030 00:31:48,240 --> 00:31:54,019 国际销售在下滑，分销也不行 1031 00:31:54,019 --> 00:31:56,059 我通常跳过这些 1032 00:31:56,059 --> 00:31:56,680 正常来说 1033 00:31:56,680 --> 00:31:57,519 好的 1034 00:31:57,519 --> 00:31:59,190 最后一个问题 1035 00:31:59,190 --> 00:32:01,619 我知道这会让你心痛 1036 00:32:01,619 --> 00:32:05,440 英伟达今天的市值刚过五万亿 1037 00:32:05,440 --> 00:32:09,549 比如谷歌母公司市值四点四或四点 1038 00:32:10,650 --> 00:32:11,349 抱歉 1039 00:32:11,349 --> 00:32:12,170 呃 1040 00:32:12,170 --> 00:32:13,950 谁先达到十万亿 1041 00:32:13,950 --> 00:32:15,829 谁先达到十万亿 1042 00:32:15,829 --> 00:32:16,430 嗯 1043 00:32:16,430 --> 00:32:17,829 我还是看好英伟达 1044 00:32:17,829 --> 00:32:18,069 呃 1045 00:32:18,069 --> 00:32:18,769 主要原因在于 1046 00:32:18,769 --> 00:32:19,089 呃 1047 00:32:19,089 --> 00:32:22,650 尽管现在AI芯片存在竞争 1048 00:32:22,650 --> 00:32:23,190 嗯 1049 00:32:23,190 --> 00:32:25,690 英伟达无疑正在以更快的速度前进 1050 00:32:25,690 --> 00:32:26,289 我们看到 1051 00:32:26,289 --> 00:32:26,650 呃 1052 00:32:26,650 --> 00:32:31,209 他们最近宣布的Ruben搭配新Pro Aqui产品 1053 00:32:31,209 --> 00:32:33,169 我认为TCO仍将保持 1054 00:32:33,169 --> 00:32:33,568 呃 1055 00:32:33,568 --> 00:32:34,769 这对他们来说非常强劲 1056 00:32:34,769 --> 00:32:37,089 这将使他们成为强劲的竞争对手 1057 00:32:37,089 --> 00:32:39,150 呃，他们正在确保这一AI 1058 00:32:39,150 --> 00:32:40,009 我认为英伟达 1059 00:32:40,009 --> 00:32:43,409 他们正在确保这一领域保持创新 1060 00:32:43,409 --> 00:32:45,608 因为如果创新更多 1061 00:32:45,608 --> 00:32:46,449 芯片数量也会增加 1062 00:32:46,449 --> 00:32:46,949 这个男人 1063 00:32:46,949 --> 00:32:47,449 呃 1064 00:32:47,449 --> 00:32:52,119 我认为英伟达拥有最佳思维和最佳机会 1065 00:32:52,119 --> 00:32:52,700 呃 1066 00:32:52,700 --> 00:32:54,980 至少在这个周期内要达到十万亿 1067 00:32:54,980 --> 00:32:58,380 如果你想了解何塞为何错误 1068 00:32:58,380 --> 00:33:00,680 以及为何谷歌将成为首个 1069 00:33:00,680 --> 00:33:01,059 顺便说一句 1070 00:33:01,059 --> 00:33:03,410 谷歌要么第一要么第二 1071 00:33:03,410 --> 00:33:05,119 是英伟达最大客户 1072 00:33:05,119 --> 00:33:08,460 如果他们真的在投资TPU并扩大TPU 1073 00:33:08,460 --> 00:33:09,559 这才是对英伟达的真实威胁 1074 00:33:09,559 --> 00:33:10,440 但呃 1075 00:33:10,440 --> 00:33:10,980 肯定有事情 1076 00:33:10,980 --> 00:33:11,319 我说 1077 00:33:11,319 --> 00:33:11,660 你知道的 1078 00:33:11,660 --> 00:33:12,759 世界上最好的公司之一 1079 00:33:12,759 --> 00:33:14,029 惊人的数据 1080 00:33:14,029 --> 00:33:15,869 我们还有很多内容要讲 1081 00:33:15,869 --> 00:33:16,390 所有人 1082 00:33:16,390 --> 00:33:18,589 别忘了在Motley Fool订阅 1083 00:33:18,589 --> 00:33:20,740 我们将带来更多内容 1084 00:33:20,740 --> 00:33:22,160 更多你喜爱的股票 1085 00:33:22,160 --> 00:33:23,440 尤其是在财报季 1086 00:33:23,440 --> 00:33:25,180 接下来几周有很多事情发生 1087 00:33:25,180 --> 00:33:26,299 所以请订阅 1088 00:33:26,299 --> 00:33:27,819 感谢何塞和特拉维斯·霍利姆 1089 00:33:27,819 --> 00:33:28,819 我们下次再见
评论采集失败
采集完成，该视频无弹幕