BV1CBGQ6HEVT · 请不要叫我梁同学1 00:00:00,040 --> 00:00:01,560 在纽可乐的淘金热中 2 00:00:01,560 --> 00:00:04,400 90%的投资者会犯的一个致命错误是 3 00:00:04,400 --> 00:00:06,320 只盯着爆火的明星公司 4 00:00:06,320 --> 00:00:08,119 new可乐的最初兴起的时候 5 00:00:08,119 --> 00:00:09,740 大家只顾着投资co威夫 6 00:00:09,740 --> 00:00:11,929 却错过了翻了四倍的奈贝尔斯 7 00:00:11,929 --> 00:00:14,949 紧接着同样逻辑的IRN和CIFR 8 00:00:14,949 --> 00:00:16,230 顶着市场的质疑 9 00:00:16,230 --> 00:00:18,100 再度完成了五倍的拉升 10 00:00:18,100 --> 00:00:20,920 许多投资者每天看着他们狂奔的市值 11 00:00:20,920 --> 00:00:21,800 望洋兴叹 12 00:00:21,800 --> 00:00:24,129 后悔自己没有早一点发现他们 13 00:00:24,129 --> 00:00:26,769 但如果你是我们ruby投资笔记的老朋友 14 00:00:26,769 --> 00:00:27,530 就会发现 15 00:00:27,530 --> 00:00:30,899 我们早在十几美元的时候就分享过这两家公司 16 00:00:30,899 --> 00:00:32,719 要是你没有吃到这波红利 17 00:00:32,719 --> 00:00:33,390 别急 18 00:00:33,390 --> 00:00:34,390 这段时间呢 19 00:00:34,390 --> 00:00:37,090 我们团队又在大家视线一步之遥的地方 20 00:00:37,090 --> 00:00:37,950 发现了一家 21 00:00:37,950 --> 00:00:41,509 正在悄悄构筑自己的AI军火库的算力云厂商 22 00:00:41,509 --> 00:00:43,710 同样是把英伟达的顶级GPU 23 00:00:43,710 --> 00:00:45,479 一车车的搬进数据中心 24 00:00:45,479 --> 00:00:49,140 同样是为大模型训练和推理提供GPU租赁服务 25 00:00:49,140 --> 00:00:51,600 但他教出来的财务和业务数据 26 00:00:51,600 --> 00:00:54,170 却远超现在的纽克劳德领头羊 27 00:00:54,170 --> 00:00:57,210 这家公司过去12个月的毛利率高达8% 28 00:00:57,210 --> 00:00:58,149 17.26 29 00:00:58,149 --> 00:00:59,170 更重要的是 30 00:00:59,170 --> 00:01:02,060 他手里还有着超过10亿美元的订单落地 31 00:01:02,060 --> 00:01:05,019 而他目前的总市值却只有12亿美元 32 00:01:05,019 --> 00:01:07,439 他会是下一个翻身的NE贝尔斯吗 33 00:01:07,439 --> 00:01:08,680 那今天这期视频 34 00:01:08,680 --> 00:01:12,480 我们就来全面拆解这位新兴的new cloud黑马 35 00:01:14,280 --> 00:01:15,200 hi朋友们 36 00:01:15,200 --> 00:01:16,459 欢迎来到ruby投资笔记 37 00:01:16,459 --> 00:01:18,109 我是你们的老朋友ruby 38 00:01:18,109 --> 00:01:20,950 今天咱们这期视频的主角就是wild fiber 39 00:01:20,950 --> 00:01:22,590 简称WYFI 40 00:01:22,590 --> 00:01:24,870 在深入拆解WYFI之前 41 00:01:24,870 --> 00:01:27,170 为了帮助大家更好地理解这家公司 42 00:01:27,170 --> 00:01:29,609 我们来看一下什么是new cloud 43 00:01:29,609 --> 00:01:31,090 不知道大家还记不记得 44 00:01:31,090 --> 00:01:32,689 2025年8月份的时候 45 00:01:32,689 --> 00:01:35,510 整个硅谷讨论最火的一个词就是new克劳德 46 00:01:35,510 --> 00:01:37,620 也就是新一代AI云厂商 47 00:01:37,620 --> 00:01:39,319 当时很多人不太当回事 48 00:01:39,319 --> 00:01:42,180 觉得这可能又是华尔街包装出来的新概念 49 00:01:42,180 --> 00:01:43,840 吵一阵子就结束了 50 00:01:43,840 --> 00:01:45,599 但如果你站在2026年 51 00:01:45,599 --> 00:01:47,379 今天这个时间点再回头看 52 00:01:47,379 --> 00:01:49,870 会发现事情根本没有这么简单 53 00:01:49,870 --> 00:01:53,079 new coud是一场真正的云计算范式转移 54 00:01:53,079 --> 00:01:54,260 估计直到现在 55 00:01:54,260 --> 00:01:56,780 很多人脑子里可能还是同样一个疑惑 56 00:01:56,780 --> 00:01:59,280 云计算这门生意不是早就被亚马逊 57 00:01:59,280 --> 00:01:59,640 微软 58 00:01:59,640 --> 00:02:01,978 谷歌三家公司彻底垄断了吗 59 00:02:01,978 --> 00:02:04,259 人家一年几千亿美元的资本开支 60 00:02:04,259 --> 00:02:05,698 全球到处扫GPU 61 00:02:05,698 --> 00:02:09,438 像WYFI这种市值才十几亿美元的小公司 62 00:02:09,438 --> 00:02:12,319 凭什么去跟这些万亿巨头抢市场呢 63 00:02:12,319 --> 00:02:14,599 听起来确实像是鸡蛋碰石头 64 00:02:14,599 --> 00:02:15,800 但问题就在于 65 00:02:15,800 --> 00:02:18,479 AI时代的算力需求已经彻底变了 66 00:02:18,479 --> 00:02:21,020 过去大厂的云更像是一个超级杂货铺 67 00:02:21,020 --> 00:02:23,740 它的底层架构是为了服务互联网世界 68 00:02:23,740 --> 00:02:24,800 各种各样的需求 69 00:02:24,800 --> 00:02:25,840 是个通用云 70 00:02:25,840 --> 00:02:27,969 这在互联网时代里是优势 71 00:02:27,969 --> 00:02:31,699 但在现在的大模型时代却是限制它发展的劣势 72 00:02:31,699 --> 00:02:33,960 因为大模型训练需要的是几万张 73 00:02:33,960 --> 00:02:35,560 甚至是几10万张GPU 74 00:02:35,560 --> 00:02:37,400 在极短的时间内同步协作 75 00:02:37,400 --> 00:02:39,879 通用云根本达不到AI训练的需求 76 00:02:39,879 --> 00:02:43,039 于是以co威夫和奈贝尔斯为代表的专有云 77 00:02:43,039 --> 00:02:44,180 开始真正崛起 78 00:02:44,180 --> 00:02:46,360 也就是我们所说的new cloud 79 00:02:46,360 --> 00:02:48,639 那new cloud的玩法非常直接 80 00:02:48,639 --> 00:02:51,919 先是利用资本渠道去抢最先进的英伟达芯片 81 00:02:51,919 --> 00:02:53,379 塞进专用机房里 82 00:02:53,379 --> 00:02:56,500 然后用最直接的裸金属架构把算力拉满 83 00:02:56,500 --> 00:02:59,990 转手再租给G消迅模型的AI巨头和独角兽 84 00:02:59,990 --> 00:03:03,430 而这套在2025年跑通的轻资产的垂直架构 85 00:03:03,430 --> 00:03:06,870 直接给他们带来了两个延续至今的颠覆性优势 86 00:03:06,870 --> 00:03:08,710 一个是计算效率极高 87 00:03:08,710 --> 00:03:09,669 没有中间商 88 00:03:09,669 --> 00:03:11,909 也没有冗余的软件层去赚差价 89 00:03:11,909 --> 00:03:14,409 算力能够百分之百的用在刀刃上 90 00:03:14,409 --> 00:03:16,569 第二个也是最让大厂头疼的就是 91 00:03:16,569 --> 00:03:19,449 他们能用比传统云计算大厂低30% 92 00:03:19,449 --> 00:03:20,919 到50%的价格 93 00:03:20,919 --> 00:03:24,329 提供同等甚至是更高性能的算力租赁服务 94 00:03:24,329 --> 00:03:26,788 站在2026年今天这个时间点上 95 00:03:26,788 --> 00:03:30,120 全球AI算力荒的问题其实还远远没有结束 96 00:03:30,120 --> 00:03:32,800 谁能在这个阶段持续提供更便宜 97 00:03:32,800 --> 00:03:34,099 更高效的AI算力 98 00:03:34,099 --> 00:03:36,300 谁就等于掌握了新的印钞机 99 00:03:36,300 --> 00:03:38,240 这也是为什么WYFI 100 00:03:38,240 --> 00:03:40,349 能在几家巨头的夹缝之间 101 00:03:40,349 --> 00:03:42,509 硬生生撕开一块市场的原因 102 00:03:42,509 --> 00:03:43,669 那顺便说一下 103 00:03:43,669 --> 00:03:45,469 我们上期视频给大家准备的 104 00:03:45,469 --> 00:03:47,229 new克劳德赛道的投资途径 105 00:03:47,229 --> 00:03:48,930 已经进行了全面的更新 106 00:03:48,930 --> 00:03:50,530 里面除了WYFI 107 00:03:50,530 --> 00:03:53,430 还深入拆解了另外五家new cloud明星公司 108 00:03:53,430 --> 00:03:55,650 可以直接点击评论区的置顶链接 109 00:03:55,650 --> 00:03:57,169 获取好视线 110 00:03:57,169 --> 00:03:59,020 拉回来 111 00:04:00,280 --> 00:04:01,280 那讲到这里 112 00:04:01,280 --> 00:04:03,500 可能有些朋友们心里会有一个疑问 113 00:04:03,500 --> 00:04:05,860 既然new克劳德这个模式这么赚钱 114 00:04:05,860 --> 00:04:09,219 岂不是谁有钱谁就能去买GPU改数据中心 115 00:04:09,219 --> 00:04:12,719 那W和FI一个后来者凭什么能给call wave 116 00:04:12,719 --> 00:04:16,000 甚至是AWS这些传统大厂正面竞争呢 117 00:04:16,000 --> 00:04:18,420 他真正的护城河到底是什么呢 118 00:04:18,420 --> 00:04:20,060 其实答案并不是芯片 119 00:04:20,060 --> 00:04:22,579 因为买到GPU只是拿到了入场券 120 00:04:22,579 --> 00:04:24,220 真正决定胜负的东西 121 00:04:24,220 --> 00:04:27,439 就藏在大多数人根本不会注意的网络架构里 122 00:04:27,439 --> 00:04:28,899 不知道大家还记不记得 123 00:04:28,899 --> 00:04:31,180 今年的光通信为什么会突然爆发 124 00:04:31,180 --> 00:04:33,959 原因就在于它能够提升数据的传输效率 125 00:04:33,959 --> 00:04:35,819 从而带动AI的运行速度 126 00:04:35,819 --> 00:04:38,259 而在AI集群里面也是一样的 127 00:04:38,259 --> 00:04:41,819 真正限制AI集群效率的并不是单张GPU的性能 128 00:04:41,819 --> 00:04:44,079 而是几万张GPU连在一起后 129 00:04:44,079 --> 00:04:45,800 它们之间的通行效率 130 00:04:45,800 --> 00:04:49,769 当成千上万张GPU同时计算一个超大模型的时候 131 00:04:49,769 --> 00:04:52,649 数据需要在芯片之间不断的同步交换 132 00:04:52,649 --> 00:04:54,350 一旦网络设计不够好 133 00:04:54,350 --> 00:04:56,029 整个集群就会开始堵车 134 00:04:56,029 --> 00:04:58,300 业内呢叫做集群效率损耗 135 00:04:58,300 --> 00:05:01,100 打个比方就是你买了1万张的GPU 136 00:05:01,100 --> 00:05:03,529 但是因为底层网络通信效率太差 137 00:05:03,529 --> 00:05:06,610 最后真正能发挥出来的效果可能还不如别人 138 00:05:06,610 --> 00:05:08,670 6000张优化后的GPU集群 139 00:05:08,670 --> 00:05:12,269 而WFI真正有前瞻性的地方就在这里 140 00:05:12,269 --> 00:05:15,009 他很早以前就在自己的AI数据中心里 141 00:05:15,009 --> 00:05:15,930 大规模部署 142 00:05:15,930 --> 00:05:19,279 drive nets驱动网络的AI以太网矩阵架构 143 00:05:19,279 --> 00:05:21,620 我在AVP2那一期视频里面说过 144 00:05:21,620 --> 00:05:23,360 过去超大规模的AI集群 145 00:05:23,360 --> 00:05:24,980 基本都依赖昂贵封闭 146 00:05:24,980 --> 00:05:28,230 且面临严重供应链垄断的infinite banner网络 147 00:05:28,230 --> 00:05:31,750 而WYFI现在推荐的这套全新的以太网 148 00:05:31,750 --> 00:05:32,470 AI架构 149 00:05:32,470 --> 00:05:34,870 就是在打破INFBANNER的垄断 150 00:05:34,870 --> 00:05:36,009 他用更开放 151 00:05:36,009 --> 00:05:38,209 更容易扩展成本更低的以太网 152 00:05:38,209 --> 00:05:40,750 实现了超大规模GPU集群运行 153 00:05:40,750 --> 00:05:44,170 AI工作负载时的零丢包和最大化吞吐 154 00:05:44,170 --> 00:05:47,230 用大白话来解释就是别人家的AI数据中心 155 00:05:47,230 --> 00:05:49,740 虽然也买了1万张最顶级的GPU 156 00:05:49,740 --> 00:05:51,560 但是因为网络经常堵车 157 00:05:51,560 --> 00:05:53,699 最后整个集群可能发挥出来的效率 158 00:05:53,699 --> 00:05:55,019 只有70%左右 159 00:05:55,019 --> 00:05:56,600 那剩下的30% 160 00:05:56,600 --> 00:05:58,970 全浪费在等待和通信损耗里 161 00:05:58,970 --> 00:06:02,449 但是WOFI靠着这套drive nice网络矩阵 162 00:06:02,449 --> 00:06:04,769 能让每一张英伟达芯片的综合性能 163 00:06:04,769 --> 00:06:06,459 发挥到95%以上 164 00:06:06,459 --> 00:06:08,579 对于那些每天训练大模型 165 00:06:08,579 --> 00:06:10,860 一天烧几百万美元的AI公司来说 166 00:06:10,860 --> 00:06:12,220 就是实打实的效率 167 00:06:12,220 --> 00:06:13,189 就是省钱 168 00:06:13,189 --> 00:06:16,470 这也是为什么WYFI虽然是后来者 169 00:06:16,470 --> 00:06:19,750 却依然能慢慢吸引越来越多的高端AI客户 170 00:06:19,750 --> 00:06:21,980 因为到了AI基础设施这个阶段 171 00:06:21,980 --> 00:06:25,300 客户真正买的其实已经不只是GPU的数量了 172 00:06:25,300 --> 00:06:27,939 而是谁能够把这些GPU组织的更高效 173 00:06:27,939 --> 00:06:28,579 更稳定 174 00:06:28,579 --> 00:06:30,019 更便宜 175 00:06:32,180 --> 00:06:35,379 当然任何公司都要面临一个很现实的事情 176 00:06:35,379 --> 00:06:36,660 技术讲的再厉害 177 00:06:36,660 --> 00:06:37,959 如果最后没人买单 178 00:06:37,959 --> 00:06:40,439 那在机构眼里也只是PPT公司 179 00:06:40,439 --> 00:06:44,319 但WYFI最让我觉得有潜力的地方就在于它 180 00:06:44,319 --> 00:06:47,420 现在很多算力服务器甚至都还没有运进机房 181 00:06:47,420 --> 00:06:50,129 插上电源就已经被AI公司抢完了 182 00:06:50,129 --> 00:06:51,290 未来几年的产能 183 00:06:51,290 --> 00:06:53,259 我们直接看数据说哈 184 00:06:53,259 --> 00:06:55,860 根据WYFI最新披露的财报 185 00:06:55,860 --> 00:06:57,600 截至2026年的3月底 186 00:06:57,600 --> 00:06:59,620 公司账上的剩余履约业务 187 00:06:59,620 --> 00:07:02,730 也就是RPO已经达到了9.21亿美元 188 00:07:02,730 --> 00:07:04,810 熟悉美股的朋友们应该都知道 189 00:07:04,810 --> 00:07:06,709 RPO这个指标很关键 190 00:07:06,709 --> 00:07:08,029 它代表的是客户 191 00:07:08,029 --> 00:07:10,800 已经跟公司签了白纸黑字的法律合同 192 00:07:10,800 --> 00:07:12,019 甚至交了押金 193 00:07:12,019 --> 00:07:13,560 只是在会计层面上 194 00:07:13,560 --> 00:07:15,420 公司还没有完全交付算力 195 00:07:15,420 --> 00:07:18,579 暂时还没有确认成收入的在手订单额 196 00:07:18,579 --> 00:07:21,879 这笔钱呢主要是来自NCE长期托管协议 197 00:07:21,879 --> 00:07:22,699 也就是说 198 00:07:22,699 --> 00:07:26,298 WOFI未来几年的收入又有着一个确定性 199 00:07:26,298 --> 00:07:28,259 但更夸张的还在后面 200 00:07:28,259 --> 00:07:29,759 就在今年的5月份 201 00:07:29,759 --> 00:07:32,610 WYFI又扔出了一颗重磅炸弹 202 00:07:32,610 --> 00:07:35,930 他们宣布已经和一家拥有投资级信用评级的 203 00:07:35,930 --> 00:07:36,949 顶级科技公司 204 00:07:36,949 --> 00:07:40,259 签署了一份新的长期AI算力服务协议 205 00:07:40,259 --> 00:07:41,480 大家注意这个词 206 00:07:41,480 --> 00:07:42,319 投资级 207 00:07:42,319 --> 00:07:44,600 这就意味着对方绝对不差钱 208 00:07:44,600 --> 00:07:45,360 信用极佳 209 00:07:45,360 --> 00:07:46,809 在全球排得上号 210 00:07:46,809 --> 00:07:50,129 而这单笔新合同的价值就超过了1.6亿美元 211 00:07:50,129 --> 00:07:52,560 是他25年全年收入的两倍 212 00:07:52,560 --> 00:07:55,699 如果把这笔合同再加上之前的RPU总额 213 00:07:55,699 --> 00:07:58,680 WYFI现在手里已经锁定了确定性 214 00:07:58,680 --> 00:08:01,649 合同总额直接来到了10.81亿美元 215 00:08:01,649 --> 00:08:04,009 而他现在整个公司的总市值呢 216 00:08:04,009 --> 00:08:05,610 也就12亿美元出头 217 00:08:05,610 --> 00:08:07,790 这其实是一个非常夸张的对比 218 00:08:07,790 --> 00:08:10,069 相当于他现在整个公司的估值 219 00:08:13,379 --> 00:08:16,819 也就是说WYFI的市值和它的订单挂钩 220 00:08:16,819 --> 00:08:19,009 更多的订单带来更大的市值 221 00:08:19,009 --> 00:08:22,259 这种情况在整个美股AI板块里并不常见 222 00:08:22,259 --> 00:08:24,540 所以在华尔街顶级机构的眼中 223 00:08:24,540 --> 00:08:26,620 评估WYFI的真实价值 224 00:08:26,620 --> 00:08:27,949 主要就看两个方向 225 00:08:27,949 --> 00:08:30,509 一是公司手里锁定了多少长期合同 226 00:08:30,509 --> 00:08:31,548 也就是IPO 227 00:08:31,548 --> 00:08:33,869 二是他未来还能扩展出多少 228 00:08:33,869 --> 00:08:35,408 可以交付的算力资产 229 00:08:35,408 --> 00:08:38,708 因为这两件事才能真正决定他未来几年的收入 230 00:08:38,708 --> 00:08:39,989 天花板 231 00:08:41,820 --> 00:08:42,659 那老规矩 232 00:08:42,659 --> 00:08:44,179 咱们不靠情绪喊目标价 233 00:08:44,179 --> 00:08:45,340 也不随便画大饼 234 00:08:45,340 --> 00:08:47,980 直接把前面拆出来的数据订单情况 235 00:08:47,980 --> 00:08:48,759 行业逻辑 236 00:08:48,759 --> 00:08:50,399 还有他现在的估值位置 237 00:08:50,399 --> 00:08:52,639 一起放进我们的多场景模型里面 238 00:08:52,639 --> 00:08:54,360 做一个公开透明的推演 239 00:08:54,360 --> 00:08:56,039 我会给大家看三个版本 240 00:08:56,039 --> 00:08:56,840 一是保守版 241 00:08:56,840 --> 00:08:57,659 二是乐观版 242 00:08:57,659 --> 00:08:59,360 还有就是极度乐观版 243 00:08:59,360 --> 00:09:01,240 第一种情况呢是保守场景 244 00:09:01,240 --> 00:09:02,639 先看最近进的情况 245 00:09:02,639 --> 00:09:03,799 假设未来3年 246 00:09:03,799 --> 00:09:06,159 全球AI算力供应开始慢慢变多 247 00:09:06,159 --> 00:09:09,259 AWS l zoo这些大厂继续去打价格战 248 00:09:09,259 --> 00:09:10,919 导致整个算力租赁行业 249 00:09:10,919 --> 00:09:12,828 的利润率出现一定的压缩 250 00:09:12,828 --> 00:09:14,548 那同时我们再保守一点 251 00:09:14,548 --> 00:09:17,000 假设WYFI后面没有再拿到 252 00:09:17,000 --> 00:09:18,360 特别惊艳的新订单 253 00:09:18,360 --> 00:09:20,159 只是把现在已经签下来的 254 00:09:20,159 --> 00:09:23,210 这10.8亿美元的长期合同稳稳兑现 255 00:09:23,210 --> 00:09:24,450 那在这个情况下 256 00:09:24,450 --> 00:09:25,769 他还能不能赚钱呢 257 00:09:25,769 --> 00:09:26,769 答案是能 258 00:09:26,769 --> 00:09:28,610 因为他现在最核心的优势 259 00:09:28,610 --> 00:09:30,529 就是那个非常夸张的毛利率 260 00:09:30,529 --> 00:09:31,370 8% 261 00:09:31,370 --> 00:09:34,539 17.26的毛利率比奈贝尔斯高出15个点 262 00:09:34,539 --> 00:09:36,799 意味着即便行业竞争变激烈 263 00:09:36,799 --> 00:09:39,669 它的底层现金流依然会有不错的韧性 264 00:09:39,669 --> 00:09:41,149 所以在这个场景里面 265 00:09:41,149 --> 00:09:42,960 我们不给他太激进的估值 266 00:09:42,960 --> 00:09:45,919 只要按照成长型云计算公司比较偏低的位置 267 00:09:45,919 --> 00:09:47,460 给他4~6倍的PS 268 00:09:47,460 --> 00:09:49,980 在这种相对保守的估值框架下 269 00:09:49,980 --> 00:09:50,779 3年之后 270 00:09:50,779 --> 00:09:52,299 W沃FI的合理市值 271 00:09:52,299 --> 00:09:54,940 大概依然有机会来到20亿美元左右 272 00:09:54,940 --> 00:09:55,820 换句话说 273 00:09:55,820 --> 00:09:58,340 即便按照最普通最克制的路径去推演 274 00:09:58,340 --> 00:09:59,019 未来几年 275 00:09:59,019 --> 00:10:01,830 它依然可能存在70%左右的回报空间 276 00:10:01,830 --> 00:10:05,210 这其实已经算是一个风险收益比不错的位置了 277 00:10:05,210 --> 00:10:06,889 第二种呢是乐观场景 278 00:10:06,889 --> 00:10:09,809 这个其实也是目前很多机构更愿意去交易的 279 00:10:09,809 --> 00:10:10,769 主线逻辑 280 00:10:10,769 --> 00:10:14,169 随着WYFI那套高效率的AI以太网矩阵 281 00:10:14,169 --> 00:10:15,870 开始被更多的客户认可 282 00:10:15,870 --> 00:10:17,269 尤其是在今年5月 283 00:10:17,269 --> 00:10:19,429 签下那家投资级科技巨头之后 284 00:10:19,429 --> 00:10:21,500 市场的示范效应开始出现 285 00:10:21,500 --> 00:10:25,379 越来越多中大型AI企业可能会主动寻找WOFI 286 00:10:25,379 --> 00:10:27,480 这种高效率的new cloud方案 287 00:10:27,480 --> 00:10:29,519 如果事情往这个方向发展 288 00:10:29,519 --> 00:10:31,659 它未来几年的GPU集群规模 289 00:10:31,659 --> 00:10:35,279 就很可能进入到一个两到三倍级别的扩张周期 290 00:10:35,279 --> 00:10:38,139 而订单也会开始像滚雪球一样增长 291 00:10:38,139 --> 00:10:39,240 到了这个阶段 292 00:10:39,240 --> 00:10:42,200 市场给他的估值逻辑会更接近AI基础设施 293 00:10:42,200 --> 00:10:43,230 核心资产 294 00:10:43,230 --> 00:10:46,090 对于这种具备高速成长能力的星云平台 295 00:10:46,090 --> 00:10:49,750 华尔街通常会愿意给8~10倍的PS估值空间 296 00:10:49,750 --> 00:10:51,629 如果这个逻辑真正兑现 297 00:10:51,629 --> 00:10:52,649 那么未来3年 298 00:10:52,649 --> 00:10:54,350 WYFI的总市值 299 00:10:54,350 --> 00:10:57,509 理论上是有机会可以冲到48亿美元左右 300 00:10:57,509 --> 00:11:00,590 对于现在的股价差不多是三倍的上涨空间 301 00:11:00,590 --> 00:11:01,629 这也是为什么 302 00:11:01,629 --> 00:11:04,309 最近会有越来越多的资金开始关注它 303 00:11:04,309 --> 00:11:06,629 因为市场真正想交易的其实是他 304 00:11:06,629 --> 00:11:10,350 从小型算力平台成长为行业核心玩家的过程 305 00:11:10,350 --> 00:11:13,370 那第三种情况呢就是极度乐观的场景 306 00:11:13,370 --> 00:11:17,049 这个场景呢就属于AI军备竞赛彻底失控的版本 307 00:11:17,049 --> 00:11:19,269 假设未来几年大语言模型 308 00:11:19,269 --> 00:11:19,529 物理 309 00:11:19,529 --> 00:11:20,490 AI机器人 310 00:11:20,490 --> 00:11:23,929 这些方向的爆发速度再次远远超过市场的预期 311 00:11:23,929 --> 00:11:27,370 那全球AI的算力需求就会持续失控式的增长 312 00:11:27,370 --> 00:11:29,529 整个行业会处于一个GPU不够 313 00:11:29,529 --> 00:11:30,090 电力不够 314 00:11:30,090 --> 00:11:31,610 数据中心不够的状态 315 00:11:31,610 --> 00:11:34,070 谁能提供稳定高效的AI基础设施 316 00:11:34,070 --> 00:11:35,519 谁就拥有定价权 317 00:11:35,519 --> 00:11:37,919 而WYFI在这种情况下 318 00:11:37,919 --> 00:11:40,019 凭借它的价格优势和网络效率 319 00:11:40,019 --> 00:11:43,299 顺利完成向超大规模AI托管平台的升级 320 00:11:43,299 --> 00:11:45,620 市场开始把它视作成为COVIE服 321 00:11:45,620 --> 00:11:48,679 奈维尔斯同级别的核心AI技术设施玩家 322 00:11:48,679 --> 00:11:51,779 那他拿到的就不只是普通成长股的估值 323 00:11:51,779 --> 00:11:54,080 而是顶级AI基础设施的溢价 324 00:11:54,080 --> 00:11:55,700 届时它的远期PS 325 00:11:55,700 --> 00:11:58,179 甚至有可能被推升到12倍以上 326 00:11:58,179 --> 00:12:01,370 对应的市值空间大概会来到100亿美元附近 327 00:12:01,370 --> 00:12:03,850 如果这个极端乐观路径真的出现 328 00:12:03,850 --> 00:12:05,169 那对于现在的位置 329 00:12:05,169 --> 00:12:08,068 理论上涨空间是现在的五倍或者是更多 330 00:12:08,068 --> 00:12:09,528 那当然我还是那句话 331 00:12:09,528 --> 00:12:11,788 任何估值推演都只是一个概率模型 332 00:12:11,788 --> 00:12:13,490 不是百分之百的水晶球 333 00:12:13,490 --> 00:12:16,210 真正的估值重构是当一家公司的订单 334 00:12:16,210 --> 00:12:17,330 现金流确定性 335 00:12:17,330 --> 00:12:21,649 行业位置和技术壁垒开始同时出现共振的时候 336 00:12:23,259 --> 00:12:25,240 听完前面那三种估值推演 337 00:12:25,240 --> 00:12:28,120 很多人现在情绪可能已经开始有点上头了 338 00:12:28,120 --> 00:12:30,940 但这里呢我还是想先把大家拉回现实 339 00:12:30,940 --> 00:12:32,500 因为老观众们应该都知道 340 00:12:32,500 --> 00:12:34,980 我从来不会把任何一家公司吹成是 341 00:12:34,980 --> 00:12:36,200 稳赚不赔的神话 342 00:12:36,200 --> 00:12:37,940 资本市场里面没有圣杯 343 00:12:37,940 --> 00:12:40,720 所有高回报的背后一定对应着高风险 344 00:12:40,720 --> 00:12:43,200 所以在真正研究WYFI之前 345 00:12:43,200 --> 00:12:45,690 有两个核心风险你必须要了解 346 00:12:45,690 --> 00:12:47,970 第一就是大客户和扩产的背后 347 00:12:47,970 --> 00:12:49,730 极度依赖英伟达的供应链 348 00:12:49,730 --> 00:12:52,450 这是WYFI最大的底层风险之一 349 00:12:52,450 --> 00:12:53,909 作为一家新运营厂商 350 00:12:53,909 --> 00:12:56,809 WYFI本身是无法自己制造芯片的 351 00:12:56,809 --> 00:12:57,980 这我们大家都知道 352 00:12:57,980 --> 00:12:59,940 所以他未来能不能继续扩张 353 00:12:59,940 --> 00:13:01,960 能不能按时把算力交付出去 354 00:13:01,960 --> 00:13:05,220 很大程度上都取决于他还能不能持续拿到 355 00:13:05,220 --> 00:13:06,740 英伟达的GPU配额 356 00:13:06,740 --> 00:13:08,200 这一点呢非常关键 357 00:13:08,200 --> 00:13:10,139 如果未来英伟达新一代的芯片 358 00:13:10,139 --> 00:13:11,720 在产能上出现跳水 359 00:13:11,720 --> 00:13:14,120 或者在分配名额上优先照顾了微软 360 00:13:14,120 --> 00:13:15,159 亚马逊这些大厂 361 00:13:15,159 --> 00:13:16,840 或者是他自己的亲儿子 362 00:13:16,840 --> 00:13:19,820 导致了WYFI出现了供应链的断档 363 00:13:19,820 --> 00:13:22,889 那么它的扩展速度就会立刻受到严重的拖累 364 00:13:22,889 --> 00:13:25,049 这是我们必须要提前意识到的 365 00:13:25,049 --> 00:13:28,259 第二呢就是AI数据中心是一个极度烧钱的行业 366 00:13:28,259 --> 00:13:30,940 很多人看到WYFI手里已经锁定了 367 00:13:30,940 --> 00:13:32,179 10.8亿美元的订单 368 00:13:32,179 --> 00:13:33,399 会下意识的觉得 369 00:13:33,399 --> 00:13:35,100 那他未来不是稳了吗 370 00:13:35,100 --> 00:13:36,210 并不是这样 371 00:13:36,210 --> 00:13:39,730 这些订单真正的问题在于它不会自动变成收入 372 00:13:39,730 --> 00:13:40,549 也就是说 373 00:13:40,549 --> 00:13:44,610 WYFI必须先花真金白银把对应的GPU服务器 374 00:13:44,610 --> 00:13:47,068 电力和数据中心全部建出来 375 00:13:47,068 --> 00:13:49,928 而这背后呢是需要极其庞大的资本开支 376 00:13:49,928 --> 00:13:52,149 在他的自我造血完成之前 377 00:13:52,149 --> 00:13:54,078 他的净利润依然为负 378 00:13:54,078 --> 00:13:56,519 尤其是在未来疯狂扩张的过程中 379 00:13:56,519 --> 00:13:58,158 如果他为了继续购买硬件 380 00:13:58,158 --> 00:14:00,078 而不得不进行大规模的负债 381 00:14:00,078 --> 00:14:02,720 或者是在二级市场里面增发融资 382 00:14:02,720 --> 00:14:04,220 在当前市场整体利率 383 00:14:04,220 --> 00:14:06,039 依然维持一个高位的情况下 384 00:14:06,039 --> 00:14:08,159 这会在短期内对它的现金流 385 00:14:08,159 --> 00:14:09,929 造成明显的阶段性压制 386 00:14:09,929 --> 00:14:12,610 那换句话说就是他虽然拥有高成长性 387 00:14:12,610 --> 00:14:15,629 但他也必须持续烧钱才能维持这种成长 388 00:14:15,629 --> 00:14:18,589 这是所有AI技术设施公司共同的问题 389 00:14:18,589 --> 00:14:20,428 所以WYFI这种公司 390 00:14:20,428 --> 00:14:23,220 它不适合用传统价值股的方式去理解 391 00:14:23,220 --> 00:14:23,580 它 392 00:14:23,580 --> 00:14:24,860 更像是一种高波动 393 00:14:24,860 --> 00:14:26,480 高赔率的成长性资产 394 00:14:26,480 --> 00:14:28,600 就像是25年的AREN一样 395 00:14:28,600 --> 00:14:31,370 谁也没有想到他竟然一年之内翻了五倍 396 00:14:31,370 --> 00:14:33,330 那WYFI也是一样 397 00:14:33,330 --> 00:14:35,049 如果未来行业继续爆发 398 00:14:35,049 --> 00:14:36,919 它的弹性会非常夸张 399 00:14:36,919 --> 00:14:38,860 但如果供应链融资环境 400 00:14:38,860 --> 00:14:41,200 或者是AI需求出现阶段性的变化 401 00:14:41,200 --> 00:14:43,809 它的股价波动也一定会非常剧烈 402 00:14:43,809 --> 00:14:46,529 这也是为什么真正做成长型投资的人 403 00:14:46,529 --> 00:14:48,149 永远不会只看上涨空间 404 00:14:48,149 --> 00:14:49,129 更重要的是 405 00:14:49,129 --> 00:14:51,970 你有没有提前看清楚它背后的风险结构 406 00:14:51,970 --> 00:14:55,210 那像WYFI这种高成长高波动的标题 407 00:14:55,210 --> 00:14:58,039 如果你对于他们的操作策略没有把握的话 408 00:14:58,039 --> 00:15:00,399 可以点击评论区的置顶链接加入我们 409 00:15:00,399 --> 00:15:02,360 获取为你量身定制的方案 410 00:15:02,360 --> 00:15:04,559 那本期的视频到这里就结束了 411 00:15:04,559 --> 00:15:06,480 咱们已经把风险看得清清楚楚 412 00:15:06,480 --> 00:15:08,340 也把价值算的明明白白 413 00:15:08,340 --> 00:15:11,279 那你认为WYFI在接下来的日子里 414 00:15:11,279 --> 00:15:14,000 能够完美复制NBS的爆发神话吗 415 00:15:14,000 --> 00:15:16,639 或者说在今天拆解的这三个场景里 416 00:15:16,639 --> 00:15:18,860 你个人更倾向于哪一种结局呢 417 00:15:18,860 --> 00:15:20,820 欢迎在评论区留下你的思考 418 00:15:20,820 --> 00:15:22,139 咱们评论区见 419 00:15:22,139 --> 00:15:24,860 那同时也欢迎您点赞订阅ruby投资笔记 420 00:15:24,860 --> 00:15:25,899 并打开小铃铛 421 00:15:25,899 --> 00:15:27,500 不错过下一期精彩视频 422 00:15:27,500 --> 00:15:28,259 我是ruby 423 00:15:28,259 --> 00:15:30,220 咱们一起也会下期再见拜
评论采集失败
采集完成，该视频无弹幕