BV1BpXbByECN · 浑水摸鱼清源1 00:00:00,080 --> 00:00:00,560 你好 2 00:00:00,560 --> 00:00:04,360 大家好，欢迎回到另一期头脑风暴节目 3 00:00:04,360 --> 00:00:05,019 这周 4 00:00:05,019 --> 00:00:06,679 我们来聊一聊太空制造 5 00:00:06,679 --> 00:00:11,640 如果我说全球半导体产业增速不够快 6 00:00:11,640 --> 00:00:16,350 至少根据埃隆·马斯克今天所说，我们正在讨论的内容 7 00:00:16,350 --> 00:00:17,410 什么是可信的 8 00:00:17,410 --> 00:00:19,570 什么是纯粹的天方夜谭 9 00:00:19,570 --> 00:00:20,589 山姆·布雷特 10 00:00:20,589 --> 00:00:22,850 帮我们拆解一下吧 11 00:00:25,719 --> 00:00:26,199 好的 12 00:00:29,519 --> 00:00:34,459 文明未来的三大要素 13 00:00:36,219 --> 00:00:38,140 太阳能板 14 00:00:38,939 --> 00:00:44,200 发射能力与半导体 15 00:00:44,359 --> 00:00:47,960 前两项SpaceX正在推进 16 00:00:47,960 --> 00:00:54,119 星舰将把低地球轨道成本降至约每公斤一百美元 17 00:00:54,119 --> 00:00:56,988 甚至可能更低 18 00:00:56,988 --> 00:01:02,850 特斯拉和SpaceX都在扩大太阳能制造 19 00:01:02,850 --> 00:01:05,549 中国早已领先 20 00:01:05,549 --> 00:01:06,489 但在美国 21 00:01:06,489 --> 00:01:07,170 美国公司 22 00:01:07,170 --> 00:01:08,120 和科技企业 23 00:01:08,120 --> 00:01:11,459 而第三项才是缺失的关键 24 00:01:11,459 --> 00:01:15,620 这就是太空制造的意义 25 00:01:17,810 --> 00:01:18,170 是的 26 00:01:18,170 --> 00:01:19,430 这总结得很到位 27 00:01:19,430 --> 00:01:26,448 埃隆总能预见几年后的瓶颈 28 00:01:26,448 --> 00:01:27,129 嗯 29 00:01:27,129 --> 00:01:32,308 这些瓶颈会阻碍他的战略计划和芯片获取 30 00:01:32,308 --> 00:01:34,748 正是他预见到的瓶颈 31 00:01:34,748 --> 00:01:38,700 因为他认为只要投入合理但不过度的资本 32 00:01:38,700 --> 00:01:42,480 他就能将 33 00:01:42,480 --> 00:01:44,019 万亿瓦级能量送入太空 34 00:01:45,019 --> 00:01:45,560 嗯 35 00:01:45,560 --> 00:01:47,859 需要万亿瓦级芯片配合能量 36 00:01:47,859 --> 00:01:52,879 以便用于训练和推理 37 00:01:52,879 --> 00:01:57,000 这些 38 00:01:57,000 --> 00:01:57,560 强大的 39 00:01:57,560 --> 00:01:58,579 人工智能模型 40 00:01:58,579 --> 00:02:01,028 和嗯 41 00:02:01,028 --> 00:02:03,930 万亿瓦级芯片无法买到 42 00:02:03,930 --> 00:02:05,730 所以他说道 43 00:02:05,730 --> 00:02:08,110 如果无法满足我的需求 44 00:02:08,110 --> 00:02:09,639 我就自己扩大产能 45 00:02:09,639 --> 00:02:14,838 我需要补充的是 46 00:02:14,838 --> 00:02:16,699 他在电池领域也做过类似事情 47 00:02:16,699 --> 00:02:21,528 并垂直整合电池供应链 48 00:02:21,528 --> 00:02:25,599 当他投入资本进行垂直整合时 49 00:02:25,599 --> 00:02:26,159 整个供应链随之加速发展 50 00:02:26,159 --> 00:02:28,379 与他的努力形成竞争态势 51 00:02:28,379 --> 00:02:31,599 这也可作为谈判策略 52 00:02:31,599 --> 00:02:37,409 促使这些公司加大对所需资本投入 53 00:02:37,409 --> 00:02:39,210 为了满足 54 00:02:39,210 --> 00:02:39,710 我喜欢 55 00:02:39,710 --> 00:02:40,849 我喜欢这里的框架 56 00:02:40,849 --> 00:02:43,150 但或许退后一步 57 00:02:43,150 --> 00:02:43,889 布雷特 58 00:02:43,889 --> 00:02:46,169 解释半导体供应链 59 00:02:46,169 --> 00:02:52,959 对于可能不了解aria fab对供应链意义的人 60 00:02:54,800 --> 00:02:55,759 我的天啊 61 00:02:55,759 --> 00:02:58,360 我不完全确定不同组件的区别 62 00:02:58,360 --> 00:03:00,819 据我理解，关税fab 63 00:03:00,819 --> 00:03:07,479 这是一个占地三倍中央公园面积的巨大建筑 64 00:03:07,479 --> 00:03:09,159 并将容纳 65 00:03:09,159 --> 00:03:10,099 所有不同的 66 00:03:10,099 --> 00:03:10,740 你知道的 67 00:03:10,740 --> 00:03:11,300 呃 68 00:03:11,300 --> 00:03:15,919 半导体制造的供应链组件 69 00:03:17,000 --> 00:03:17,539 是的 70 00:03:17,539 --> 00:03:17,840 我的意思是 71 00:03:17,840 --> 00:03:20,280 比如光刻与晶圆厂 72 00:03:20,280 --> 00:03:23,349 与芯片设计在同一栋楼 73 00:03:23,349 --> 00:03:24,270 内存生产 74 00:03:24,270 --> 00:03:26,610 先进封装和测试全部在一栋楼内 75 00:03:27,810 --> 00:03:29,849 这就是想法 76 00:03:30,009 --> 00:03:30,669 是的 77 00:03:30,669 --> 00:03:31,129 是的 78 00:03:31,129 --> 00:03:31,650 是的 79 00:03:31,650 --> 00:03:32,449 这个理念 80 00:03:32,449 --> 00:03:32,810 我的意思是 81 00:03:32,810 --> 00:03:34,949 正如埃隆的所有雄心 82 00:03:34,949 --> 00:03:37,990 要垂直整合价值链所有环节 83 00:03:37,990 --> 00:03:44,150 让产品需求方能紧邻制造设备 84 00:03:44,150 --> 00:03:45,490 正在生产产品的设备 85 00:03:45,490 --> 00:03:48,490 然后获取制造设备的产出 86 00:03:48,490 --> 00:03:52,740 可立即测试并据此迭代 87 00:03:52,740 --> 00:03:54,080 在半导体领域 88 00:03:54,080 --> 00:03:59,300 我们谈论的是每台机器数亿美元 89 00:03:59,300 --> 00:04:04,909 高度精密复杂的资本设备 90 00:04:04,909 --> 00:04:06,169 维护和运行 91 00:04:06,169 --> 00:04:07,110 他的 92 00:04:07,110 --> 00:04:09,930 嗯，而在 93 00:04:09,930 --> 00:04:11,229 生长晶圆 94 00:04:11,229 --> 00:04:12,610 蚀刻晶圆 95 00:04:12,610 --> 00:04:14,490 后端生产芯片 96 00:04:14,490 --> 00:04:17,110 配备测试设备进行检测 97 00:04:17,110 --> 00:04:18,069 确实 98 00:04:18,069 --> 00:04:18,730 至少 99 00:04:18,730 --> 00:04:24,029 按此规划堪称人类历史上前所未有的计划 100 00:04:24,029 --> 00:04:25,329 在雄心方面 101 00:04:25,329 --> 00:04:28,750 我指的是标价报告是二十亿美元 102 00:04:28,750 --> 00:04:29,750 这是一笔巨额资金 103 00:04:29,750 --> 00:04:30,170 可以说 104 00:04:30,170 --> 00:04:31,699 没有任何设施 105 00:04:31,699 --> 00:04:34,220 与这些数字相比 106 00:04:34,220 --> 00:04:38,579 我们有时显示二十到二十五亿美元实际上似乎偏低 107 00:04:38,579 --> 00:04:40,660 我知道这是一个疯狂的论断 108 00:04:43,069 --> 00:04:45,329 好吧这是为了让你感受规模 109 00:04:45,329 --> 00:04:48,870 特斯拉今年的资本支出计划是二十亿美元 110 00:04:48,870 --> 00:04:51,620 其中部分用于 111 00:04:51,620 --> 00:04:55,740 他所说的先进晶圆厂 112 00:04:55,740 --> 00:05:00,769 我认为是与三星合作建造他们的AI五芯片 113 00:05:00,769 --> 00:05:04,449 因此特斯拉已经在投资部分资本设备 114 00:05:04,449 --> 00:05:09,920 但这个设施需要十吉瓦电力 115 00:05:09,920 --> 00:05:11,759 嗯而且 116 00:05:11,759 --> 00:05:14,459 在这一背景下 117 00:05:14,459 --> 00:05:18,139 实际上这个资本支出估计偏低 118 00:05:18,139 --> 00:05:18,939 你需要考虑 119 00:05:18,939 --> 00:05:23,228 十吉瓦电力相当于建造十吉瓦核电站 120 00:05:23,228 --> 00:05:24,468 就像萨姆所说 121 00:05:24,468 --> 00:05:25,189 没错就在那里 122 00:05:25,189 --> 00:05:28,620 十亿美元或更多 123 00:05:28,620 --> 00:05:31,180 我意思是这取决于是否高效建设 124 00:05:31,180 --> 00:05:31,600 是的 125 00:05:31,600 --> 00:05:32,019 是的 126 00:05:32,019 --> 00:05:34,180 另一个思考方式可能是 127 00:05:34,180 --> 00:05:34,519 是的 128 00:05:34,519 --> 00:05:35,639 二十到二十 129 00:05:35,639 --> 00:05:37,100 五亿美元 130 00:05:37,100 --> 00:05:39,939 将颠覆多少万亿美元市场 131 00:05:39,939 --> 00:05:43,360 涵盖所有半导体公司的总市值 132 00:05:43,360 --> 00:05:43,720 嗯 133 00:05:43,720 --> 00:05:44,040 不不 134 00:05:44,040 --> 00:05:45,339 不是二十亿美元 135 00:05:45,339 --> 00:05:47,339 不是仅启动就需要二十亿 136 00:05:47,339 --> 00:05:50,879 二十亿用于破土动工开始建设 137 00:05:51,449 --> 00:05:52,149 是的 138 00:05:52,149 --> 00:05:53,389 你实际投入更多 139 00:05:53,389 --> 00:05:55,670 但这属于启动资本 140 00:05:55,670 --> 00:05:59,339 我还看到萨姆·多明戈的新文章 141 00:05:59,339 --> 00:06:01,699 Impulse Labs创始人 142 00:06:01,699 --> 00:06:06,459 工厂蕴含的知识与产出能力 143 00:06:06,459 --> 00:06:08,259 我认为这很相关 144 00:06:08,259 --> 00:06:11,278 将这些整合在同一地点 145 00:06:11,278 --> 00:06:14,738 才能真正掌握这项能力 146 00:06:14,738 --> 00:06:16,319 或许这就是 147 00:06:16,319 --> 00:06:17,059 你知道的 148 00:06:17,059 --> 00:06:22,079 其他尝试者未能同样成功的原因 149 00:06:22,079 --> 00:06:23,019 但我们正在前进 150 00:06:23,019 --> 00:06:23,600 我们正在偏离方向 151 00:06:23,600 --> 00:06:23,839 真的吗 152 00:06:23,839 --> 00:06:25,439 我想回到布雷特的观点 153 00:06:25,439 --> 00:06:30,069 这并非埃隆第一次这样做 154 00:06:30,069 --> 00:06:34,269 在这里成功有多种路径 155 00:06:35,810 --> 00:06:37,250 就是这三种要素 156 00:06:37,250 --> 00:06:38,790 如果埃隆能全部做到 157 00:06:38,790 --> 00:06:39,990 太令人惊叹了 158 00:06:39,990 --> 00:06:43,959 你知道这个计划不会幸存 159 00:06:43,959 --> 00:06:45,180 与敌人接触 160 00:06:45,180 --> 00:06:46,139 我可以说 161 00:06:46,139 --> 00:06:48,259 但就像电动汽车 162 00:06:48,259 --> 00:06:50,209 其他人也曾做过电动汽车 163 00:06:50,209 --> 00:06:52,189 但埃隆做到了 164 00:06:52,189 --> 00:06:56,689 现在特斯拉是否是全球生产最多电动汽车的公司 165 00:06:56,689 --> 00:06:57,129 不 166 00:06:57,129 --> 00:06:59,420 但存在电动汽车产业 167 00:06:59,420 --> 00:07:02,620 未来几乎肯定是电动化 168 00:07:02,620 --> 00:07:04,259 同样适用于电动卡车 169 00:07:04,259 --> 00:07:06,660 这曾是可能的事吗 170 00:07:06,660 --> 00:07:08,800 事实上所有人都说不可能 171 00:07:08,800 --> 00:07:09,980 然而 172 00:07:09,980 --> 00:07:10,480 看吧 173 00:07:10,480 --> 00:07:11,480 这是可能的 174 00:07:11,480 --> 00:07:14,379 而这并无不同 175 00:07:14,379 --> 00:07:14,839 没错 176 00:07:14,839 --> 00:07:16,560 除非我理解错了 177 00:07:16,560 --> 00:07:18,399 他提出的其他内容 178 00:07:18,399 --> 00:07:20,779 只是将一切纳入同一屋檐下 179 00:07:20,779 --> 00:07:22,519 技术并无差异 180 00:07:22,519 --> 00:07:22,779 没错 181 00:07:22,779 --> 00:07:25,100 他要推进到2纳米制程 182 00:07:25,100 --> 00:07:25,540 这正是 183 00:07:25,540 --> 00:07:26,379 这才是关键 184 00:07:26,379 --> 00:07:27,339 规模不同 185 00:07:27,339 --> 00:07:30,879 人们研究机器人已有二十年 186 00:07:30,879 --> 00:07:33,579 然后特斯拉突然说 187 00:07:33,579 --> 00:07:35,839 我们要垂直整合并投资于此 188 00:07:35,839 --> 00:07:41,509 如今已有六百家人形机器人公司获得大量融资 189 00:07:42,509 --> 00:07:43,730 仅仅因为你说规模 190 00:07:43,730 --> 00:07:44,790 你指的是什么 191 00:07:44,790 --> 00:07:45,680 因为并非 192 00:07:45,680 --> 00:07:50,100 整个半导体供应链也看到了趋势 193 00:07:50,100 --> 00:07:51,600 如果他们能满足需求 194 00:07:51,600 --> 00:07:52,160 他们会的 195 00:07:52,160 --> 00:07:56,399 只是建造这些存在困难 196 00:07:56,399 --> 00:07:57,059 嗯 197 00:07:57,059 --> 00:07:57,379 实际上 198 00:07:57,379 --> 00:07:58,098 他解决了这个问题 199 00:07:58,098 --> 00:07:59,738 他直截了当地面对这个问题 200 00:07:59,738 --> 00:08:00,620 也就是说 201 00:08:00,620 --> 00:08:02,740 如果你在半导体行业工作过 202 00:08:02,740 --> 00:08:04,480 你身上积累了大量伤疤 203 00:08:04,480 --> 00:08:08,009 因为这个行业长期处于繁荣与衰退的循环中 204 00:08:08,009 --> 00:08:09,250 他在说 205 00:08:09,250 --> 00:08:10,930 现在他正在签空白支票 206 00:08:10,930 --> 00:08:13,560 他和许多人正在下注 207 00:08:13,560 --> 00:08:14,399 他们在下注 208 00:08:14,399 --> 00:08:16,699 对智能技术的需求无限 209 00:08:16,699 --> 00:08:19,740 这就是我认为我们所处的境况 210 00:08:19,740 --> 00:08:23,300 而他对风险的承受程度远超这些人 211 00:08:23,300 --> 00:08:24,500 当然就像 212 00:08:24,500 --> 00:08:25,819 如果你在半导体行业 213 00:08:25,819 --> 00:08:29,019 可以说我们将以负责任的方式扩大规模 214 00:08:29,019 --> 00:08:32,730 因为我们认为这可能不会发生 215 00:08:32,730 --> 00:08:35,090 因此我们要保护业务 216 00:08:35,090 --> 00:08:36,850 而埃隆·马斯克则说 217 00:08:36,850 --> 00:08:39,169 这挺有趣的 218 00:08:39,169 --> 00:08:40,649 如果你试图回撤 219 00:08:40,649 --> 00:08:44,210 你知道每年向股东返还10% 220 00:08:44,210 --> 00:08:44,769 他说道 221 00:08:44,769 --> 00:08:49,529 我正在努力在月球建立质量驱动器并殖民其他星系 222 00:08:49,529 --> 00:08:50,990 这是完全不同的游戏 223 00:08:50,990 --> 00:08:52,649 不同的规模 224 00:08:52,870 --> 00:08:53,509 其实 225 00:08:53,509 --> 00:08:54,230 其实这是 226 00:08:54,230 --> 00:08:54,429 是的 227 00:08:54,429 --> 00:08:57,409 潜在增长的速度 228 00:08:57,409 --> 00:09:00,389 我认为更受 229 00:09:00,389 --> 00:09:04,210 人们意志的影响而非普遍认知 230 00:09:04,210 --> 00:09:08,169 所以就像山姆说的台积电 231 00:09:08,169 --> 00:09:08,649 你知道的 232 00:09:08,649 --> 00:09:09,690 甚至对他们的投资者 233 00:09:09,690 --> 00:09:10,889 如果你查看他们的投资者会议 234 00:09:10,889 --> 00:09:12,169 他们不得不拒绝 235 00:09:12,169 --> 00:09:16,950 我们对合作伙伴带来的需求充满信心 236 00:09:16,950 --> 00:09:22,309 并根据两年内高信用客户的订单进行调整 237 00:09:22,309 --> 00:09:24,289 比如亚马逊AWS之类的企业 238 00:09:24,289 --> 00:09:25,110 或者说 239 00:09:25,110 --> 00:09:30,788 因此他们不愿无意中过度扩张产能 240 00:09:30,788 --> 00:09:33,599 因为就像山姆说的周期性煤炭行业 241 00:09:33,599 --> 00:09:34,938 如果过度扩张产能 242 00:09:34,938 --> 00:09:35,759 你知道的 243 00:09:35,759 --> 00:09:37,789 这会摧毁你的利润率 244 00:09:37,789 --> 00:09:40,710 于是从健康的业务突然 245 00:09:40,710 --> 00:09:42,789 变为业务岌岌可危 246 00:09:42,789 --> 00:09:44,190 因为你投入了大量资本 247 00:09:44,190 --> 00:09:46,509 却未充分利用产能 248 00:09:46,509 --> 00:09:49,309 而这轮周期的风险用词危险 249 00:09:49,309 --> 00:09:50,450 就像这次不一样 250 00:09:50,450 --> 00:09:56,250 但这次的差异在于我们确实有需求 251 00:09:56,250 --> 00:09:56,870 你知道的 252 00:09:56,870 --> 00:09:58,269 即使在知识领域 253 00:09:58,269 --> 00:09:58,789 暂且不提 254 00:09:58,789 --> 00:10:02,200 具身机器人学以及生物学中的一切进展 255 00:10:02,200 --> 00:10:03,700 在知识领域 256 00:10:03,700 --> 00:10:08,190 我们的需求曲线看起来不会无限延伸 257 00:10:08,190 --> 00:10:12,240 但它会持续增长，暗示着 258 00:10:12,240 --> 00:10:13,059 与其说 259 00:10:13,059 --> 00:10:15,519 我们要每年提升20%产能 260 00:10:15,519 --> 00:10:18,740 你可以大幅削减增长速度 261 00:10:18,740 --> 00:10:22,600 然后投入巨额资本 262 00:10:22,600 --> 00:10:25,210 我们相信能获得良好回报 263 00:10:25,210 --> 00:10:26,549 所以马斯克说 264 00:10:26,549 --> 00:10:29,009 这些人对我来说不够快 265 00:10:29,009 --> 00:10:33,589 我要用自己的资产负债表融资来做这件事 266 00:10:33,589 --> 00:10:40,539 我认为这实际上会引发行业反制 267 00:10:40,539 --> 00:10:42,740 如果他开始取得进展 268 00:10:42,740 --> 00:10:43,899 他们会观察并说 269 00:10:43,899 --> 00:10:44,629 天啊 270 00:10:44,629 --> 00:10:46,950 我们现在要被这些公司甩开差距了 271 00:10:46,950 --> 00:10:52,440 他们将面临巨头科技公司的压力 272 00:10:52,440 --> 00:10:58,320 需要提供竞争武器对抗马斯克集团 273 00:10:58,320 --> 00:11:00,289 以特斯拉 274 00:11:00,289 --> 00:11:03,730 SpaceX和xAI为例 275 00:11:03,730 --> 00:11:04,730 尚未合并 276 00:11:04,730 --> 00:11:06,330 但可以视为 277 00:11:06,330 --> 00:11:06,870 嗯 278 00:11:06,870 --> 00:11:11,659 协同实体将获得计算优势 279 00:11:11,659 --> 00:11:16,320 相对于整个行业和其他领域 280 00:11:16,320 --> 00:11:20,090 行业将要求我们配备竞争武器 281 00:11:20,090 --> 00:11:20,809 面对这个 282 00:11:20,809 --> 00:11:22,129 太空探索 283 00:11:22,129 --> 00:11:22,870 庞然大物 284 00:11:22,870 --> 00:11:25,379 嗯 285 00:11:25,379 --> 00:11:25,820 嗯 286 00:11:25,820 --> 00:11:28,679 供应链会收到更强的需求信号 287 00:11:28,679 --> 00:11:30,679 供应链开始扩产 288 00:11:30,679 --> 00:11:35,259 然后如果他执行得当 289 00:11:35,259 --> 00:11:39,269 就能灵活获取中性供应 290 00:11:39,269 --> 00:11:40,169 如有必要 291 00:11:40,169 --> 00:11:41,129 我认为这就是 292 00:11:41,129 --> 00:11:44,470 如果这损害他在Grock的定位 293 00:11:44,470 --> 00:11:47,389 如果他们能生成所有这些 294 00:11:47,389 --> 00:11:49,629 你知道新的供应 295 00:11:49,629 --> 00:11:51,990 需求将转向Anthropic和OpenAI 296 00:11:51,990 --> 00:11:53,090 AI而非Croc 297 00:11:53,090 --> 00:11:53,710 所以 298 00:11:53,710 --> 00:11:57,389 现在他们不再受计算力限制 299 00:11:57,789 --> 00:11:59,389 这在某种程度上像 300 00:11:59,389 --> 00:12:01,629 如果如果如果这是你在描述的brett 301 00:12:01,629 --> 00:12:02,429 比如他 302 00:12:02,429 --> 00:12:07,049 你知道这点燃了传统半导体市场供应链下的火苗 303 00:12:07,049 --> 00:12:10,580 他们都在他扩大taria fab规模前整理好各自事务 304 00:12:10,580 --> 00:12:12,899 然后计算能力变得广泛可用 305 00:12:12,899 --> 00:12:15,460 所有已存在的需求 306 00:12:15,460 --> 00:12:19,609 等待claude和chachi bt的立即被释放 307 00:12:19,609 --> 00:12:21,969 锁定价值来自企业人士 308 00:12:21,969 --> 00:12:26,139 我认为grok正拿着巨额资产负债表风险 309 00:12:26,139 --> 00:12:30,779 因为他们承担了十到四十亿美元的债务 310 00:12:32,320 --> 00:12:34,440 这真是个大问题 311 00:12:34,440 --> 00:12:36,720 我不认为埃隆会害怕 312 00:12:36,720 --> 00:12:39,070 但我们是否害怕这种竞争 313 00:12:39,070 --> 00:12:39,389 是的 314 00:12:39,389 --> 00:12:40,570 因为我们不是埃隆 315 00:12:40,570 --> 00:12:42,409 我们是埃隆公司的投资者 316 00:12:42,409 --> 00:12:43,169 像这类公司 317 00:12:43,169 --> 00:12:47,190 这是我们需要关注的重大风险 318 00:12:49,309 --> 00:12:50,889 我认为我认为那里 319 00:12:50,889 --> 00:12:52,289 我我我意思是 320 00:12:52,289 --> 00:12:54,980 我认为那里存在 321 00:12:55,620 --> 00:12:58,879 存在一种假设可以获取计算资源 322 00:12:58,879 --> 00:13:01,840 并将其转化为变现潜力 323 00:13:01,840 --> 00:13:03,210 嗯 324 00:13:03,210 --> 00:13:06,070 Anthropic和OpenAI处于谷歌层面 325 00:13:06,070 --> 00:13:10,210 尽管我们无法清晰看到gemini的营收转化 326 00:13:10,210 --> 00:13:14,690 但anthropic正在执行这些行动 327 00:13:14,690 --> 00:13:15,710 最为成功 328 00:13:15,710 --> 00:13:16,769 现在 329 00:13:16,769 --> 00:13:21,110 当前AI模型公司的粗略数学 330 00:13:21,110 --> 00:13:24,889 现在首先是数据中心 331 00:13:24,889 --> 00:13:27,690 占用一座吉瓦级数据中心 332 00:13:27,690 --> 00:13:31,359 可租给他人每年十亿美元 333 00:13:31,558 --> 00:13:32,038 好的 334 00:13:32,038 --> 00:13:33,278 所以像 335 00:13:33,278 --> 00:13:33,519 实际上 336 00:13:33,519 --> 00:13:38,438 你可以粗略计算该领域计算数学 337 00:13:38,438 --> 00:13:41,169 你可以像 338 00:13:41,169 --> 00:13:42,629 将卫星送入太空 339 00:13:42,629 --> 00:13:47,710 然后向anthropic或谷歌出售容量每年十亿 340 00:13:47,710 --> 00:13:49,909 并逐年上涨 341 00:13:49,909 --> 00:13:53,129 因为地球建数据中心越来越难 342 00:13:53,129 --> 00:13:55,250 作为AI模型公司 343 00:13:55,250 --> 00:13:58,190 租用价值十亿美元的数据中心 344 00:13:58,190 --> 00:13:59,730 通过你的模型运行 345 00:13:59,730 --> 00:14:05,600 然后可按三十亿美元每吉瓦收取访问代币费用 346 00:14:05,600 --> 00:14:06,620 这是你产生的收益 347 00:14:06,620 --> 00:14:08,409 这就是当前的粗略计算 348 00:14:08,409 --> 00:14:09,490 数据中心 349 00:14:09,490 --> 00:14:14,149 该数据中心租赁成本按前向计算每年增长约6% 350 00:14:14,149 --> 00:14:19,340 我们认为该计算能力的变现率 351 00:14:19,340 --> 00:14:23,120 压缩到模型中每年增长约20% 352 00:14:23,120 --> 00:14:24,279 这是一个很好的特性 353 00:14:24,279 --> 00:14:27,000 就像现在如果能让这数学成立 354 00:14:27,000 --> 00:14:28,120 这数学关系非常 355 00:14:28,120 --> 00:14:33,250 高度依赖于Anthropic和OpenAI当前的表现 356 00:14:33,250 --> 00:14:37,710 并非针对XAI当前的发展进行调校 357 00:14:37,710 --> 00:14:38,649 而我们并不清楚 358 00:14:38,649 --> 00:14:40,690 他的赌注是 359 00:14:40,690 --> 00:14:45,220 我可以引导XAI团队朝正确目标前进 360 00:14:45,220 --> 00:14:46,960 提供足够的计算资源 361 00:14:46,960 --> 00:14:49,289 他们也能让这数学成立 362 00:14:49,289 --> 00:14:52,450 从长远来看在企业领域 363 00:14:52,450 --> 00:14:55,490 至少会有几家参与者 364 00:14:55,490 --> 00:14:56,909 大概三到四家 365 00:14:56,909 --> 00:14:59,110 正在以某种方式让这数学成立 366 00:14:59,110 --> 00:15:00,799 并将彼此竞争 367 00:15:00,799 --> 00:15:01,840 但会保持稳定 368 00:15:01,840 --> 00:15:03,100 利润率结构良好 369 00:15:03,100 --> 00:15:04,279 就像基础设施 370 00:15:04,279 --> 00:15:06,710 作为消费者端的服务 371 00:15:06,710 --> 00:15:09,289 我认为我们仍处于 372 00:15:09,289 --> 00:15:10,070 嗯 373 00:15:10,070 --> 00:15:12,929 某种程度的未探索领域 374 00:15:12,929 --> 00:15:14,710 关于具体形态尚未明确 375 00:15:14,710 --> 00:15:16,928 但这大致的数学逻辑 376 00:15:16,928 --> 00:15:18,528 所以埃隆说如果 377 00:15:18,528 --> 00:15:20,889 如果我能让这数学成立 378 00:15:20,889 --> 00:15:22,769 我需要尽可能多的计算资源 379 00:15:22,769 --> 00:15:24,889 因为我认为这数学将持续有效 380 00:15:24,889 --> 00:15:26,789 无论投入多少吉瓦电力 381 00:15:26,789 --> 00:15:32,889 获取每年一太瓦的算力将为我 382 00:15:34,009 --> 00:15:36,549 你知道这相当于三十亿美元 383 00:15:36,549 --> 00:15:41,210 乘以一千就是三十万亿美元年营收潜力 384 00:15:41,210 --> 00:15:43,090 目前大致如此 385 00:15:43,090 --> 00:15:47,009 所以我总是关注大数字 386 00:15:47,009 --> 00:15:48,809 二十亿似乎 387 00:15:48,809 --> 00:15:50,090 但不管怎样 388 00:15:50,370 --> 00:15:50,769 好的 389 00:15:50,769 --> 00:15:51,110 布莱特 390 00:15:51,110 --> 00:15:55,289 这里的一个反对意见是 391 00:15:55,289 --> 00:15:57,149 地球没有时间表 392 00:15:57,149 --> 00:15:58,549 我是说 393 00:15:58,549 --> 00:15:59,090 确实存在 394 00:15:59,090 --> 00:15:59,570 你知道的 395 00:15:59,570 --> 00:16:03,149 显然存在阶段性时间线 396 00:16:03,149 --> 00:16:04,509 但没有明确时间表 397 00:16:04,509 --> 00:16:05,070 特斯拉 398 00:16:05,070 --> 00:16:10,190 汽车销售全球表现疲软 399 00:16:10,190 --> 00:16:11,269 嗯 400 00:16:11,269 --> 00:16:16,129 这对特斯拉筹集资金来说是一个很好的叙事 401 00:16:16,129 --> 00:16:16,690 是的 402 00:16:16,690 --> 00:16:22,230 特斯拉将与SpaceX的IPO同步筹集资金 403 00:16:22,230 --> 00:16:23,889 这些都是合理的理由 404 00:16:23,889 --> 00:16:27,409 嗯，我还没仔细考虑过 405 00:16:27,409 --> 00:16:28,940 我我我还不确定 406 00:16:28,940 --> 00:16:35,450 我认为这显然体现了特斯拉与SpaceX AI更紧密的协同 407 00:16:35,450 --> 00:16:39,320 这比之前至少讨论过的要更深入 408 00:16:39,320 --> 00:16:41,340 这是我首先想提出的观点 409 00:16:41,340 --> 00:16:43,259 至少从呈现情况来看 410 00:16:43,259 --> 00:16:48,110 双方正在联合进行这项巨额投资，即这个嗯 411 00:16:48,110 --> 00:16:51,750 工厂由特斯拉明确提供 412 00:16:51,750 --> 00:16:52,429 呃 413 00:16:52,429 --> 00:16:53,190 你知道这些芯片 414 00:16:53,190 --> 00:16:54,110 用于其机器人 415 00:16:54,110 --> 00:16:55,179 用于其汽车 416 00:16:55,179 --> 00:16:56,799 而这些汽车也是机器人 417 00:16:56,799 --> 00:17:00,820 以及SpaceX的卫星 418 00:17:00,820 --> 00:17:06,329 目前还不清楚谁将负责SpaceX X的太阳能制造 419 00:17:07,329 --> 00:17:10,490 而特斯拉或许可以 420 00:17:10,650 --> 00:17:15,589 我认为特斯拉在自动驾驶出租车实现现金流后 421 00:17:15,589 --> 00:17:17,269 现金流开始回笼 422 00:17:17,269 --> 00:17:24,759 并获得预期自动驾驶出租车的合理市场定价 423 00:17:24,759 --> 00:17:26,200 你知道现金流收入 424 00:17:26,200 --> 00:17:30,299 你知道特斯拉可以独立从SpaceX IPO融资 425 00:17:30,299 --> 00:17:35,929 或者出现一种组合方案作为融资工具 426 00:17:35,929 --> 00:17:36,969 很明显的是 427 00:17:36,969 --> 00:17:41,449 这些实体需要大量资金投入未来 428 00:17:41,449 --> 00:17:43,209 在他的演讲中 429 00:17:43,209 --> 00:17:50,450 另外值得注意的是，直到Optimus真正规模化之前 430 00:17:50,450 --> 00:17:53,789 计算能力和芯片将被送往太空 431 00:17:53,789 --> 00:17:56,589 大致是80%与20%的分配 432 00:17:56,589 --> 00:17:58,869 这非常依赖轨道运输 433 00:17:58,869 --> 00:18:01,369 但即使如此 434 00:18:01,369 --> 00:18:05,059 当Optimus规模化后所需的功率 435 00:18:05,059 --> 00:18:07,880 Optimus芯片的功率需求将远低于 436 00:18:07,880 --> 00:18:09,039 发射到太空的功率 437 00:18:09,039 --> 00:18:10,039 所以他 438 00:18:10,039 --> 00:18:10,759 你知道的 439 00:18:10,759 --> 00:18:11,599 他的观点是 440 00:18:11,599 --> 00:18:12,440 大概相当于 441 00:18:12,440 --> 00:18:15,240 每年售出的Optimus数量 442 00:18:15,240 --> 00:18:19,960 等于人类数量的五分之一或十分之一 443 00:18:19,960 --> 00:18:22,000 基本上每人配备一个Optimus 444 00:18:22,000 --> 00:18:22,720 然后你 445 00:18:22,720 --> 00:18:24,099 每五到十年更换一次 446 00:18:24,099 --> 00:18:26,029 我认为这就是他的大致估算 447 00:18:26,029 --> 00:18:28,029 所以算作一亿台Optimus 448 00:18:28,029 --> 00:18:28,390 嗯 449 00:18:28,390 --> 00:18:28,769 但它们是 450 00:18:28,769 --> 00:18:29,589 它们不在 451 00:18:29,589 --> 00:18:33,259 你知道那些即将进入太空舱的大芯片 452 00:18:33,259 --> 00:18:38,069 因此仍需大部分用于太空的机器人出租车 453 00:18:38,069 --> 00:18:40,269 嗯作为融资机制 454 00:18:40,269 --> 00:18:41,009 尼克 455 00:18:41,009 --> 00:18:44,289 你分享了一些有趣的统计数据 456 00:18:44,289 --> 00:18:45,369 效果很好 457 00:18:45,369 --> 00:18:48,190 对我个人使用非常有效 458 00:18:48,190 --> 00:18:49,029 我呃 459 00:18:49,029 --> 00:18:51,309 我认为这些对话的结果是 460 00:18:51,309 --> 00:18:52,509 我们每周的对话 461 00:18:52,509 --> 00:18:57,559 我最终用旧款特斯拉Model Y换购了新款 462 00:18:57,559 --> 00:19:03,259 嗯因此我获得了第十四代或最新版本 463 00:19:03,259 --> 00:19:09,278 FSD系统确实与之前车辆有明显差异 464 00:19:09,278 --> 00:19:14,119 那么你最近给它打几分满分十分 465 00:19:14,119 --> 00:19:16,480 但这是经典十分制 466 00:19:16,480 --> 00:19:17,880 当你与埃隆打交道时 467 00:19:17,880 --> 00:19:18,960 试图理解 468 00:19:18,960 --> 00:19:23,259 有时现实与传闻存在差距 469 00:19:23,259 --> 00:19:24,920 而我正在使用它 470 00:19:24,920 --> 00:19:25,680 它确实有效 471 00:19:25,680 --> 00:19:28,460 而机器人出租车规模较小 472 00:19:28,460 --> 00:19:32,609 他们仍落后于Waymo吗 473 00:19:32,890 --> 00:19:33,609 哦你是说 474 00:19:33,609 --> 00:19:35,690 机器人出租车产品对吧 475 00:19:35,690 --> 00:19:37,089 当你观察旧金山的情况 476 00:19:37,089 --> 00:19:40,700 是的而且这就像 477 00:19:40,700 --> 00:19:41,539 给他们一些时间 478 00:19:41,539 --> 00:19:45,500 但这些都是当前的潜在能量 479 00:19:45,500 --> 00:19:47,019 正如你所知 480 00:19:47,019 --> 00:19:48,420 德州可以生产 481 00:19:48,420 --> 00:19:49,380 你知道吗 482 00:19:49,380 --> 00:19:50,200 它怎么说 483 00:19:50,200 --> 00:19:51,599 我拥有五十万辆 484 00:19:51,599 --> 00:19:54,289 每年五十万辆左右 485 00:19:54,289 --> 00:19:55,589 成千上万辆 486 00:19:55,589 --> 00:19:57,579 嗯然后 487 00:19:57,579 --> 00:20:00,420 直到跨越商业化门槛 488 00:20:00,420 --> 00:20:02,160 才会显现规模 489 00:20:02,160 --> 00:20:07,240 然后会远超Waymo的规模 490 00:20:07,240 --> 00:20:08,220 嗯 491 00:20:08,220 --> 00:20:09,420 这就是我的观点 492 00:20:09,420 --> 00:20:11,119 我认为很有道理 493 00:20:11,119 --> 00:20:13,240 嗯然后 494 00:20:13,240 --> 00:20:16,130 这就像从 495 00:20:16,130 --> 00:20:16,430 太棒了 496 00:20:16,430 --> 00:20:18,170 我使用时确实有效 497 00:20:18,170 --> 00:20:19,369 这个也确实有效 498 00:20:19,369 --> 00:20:19,809 太惊人了 499 00:20:19,809 --> 00:20:22,430 这有点像是以一种可靠的方式搭建起来的 500 00:20:22,430 --> 00:20:22,990 路边停车区 501 00:20:22,990 --> 00:20:24,309 投递和接送服务 502 00:20:24,309 --> 00:20:26,369 一切似乎都被紧密串联在一起 503 00:20:26,369 --> 00:20:29,250 仍然是相当艰巨的努力 504 00:20:29,250 --> 00:20:30,960 我认为他们正在处理中 505 00:20:30,960 --> 00:20:37,099 嗯，我觉得FSD与每个产品相关 506 00:20:37,099 --> 00:20:42,009 我和成年朋友之间最大的生活方式差异 507 00:20:42,009 --> 00:20:42,710 现在 508 00:20:42,710 --> 00:20:43,730 他们不明白 509 00:20:43,730 --> 00:20:45,549 他们大多数人都没有特斯拉 510 00:20:45,549 --> 00:20:47,029 在那些拥有特斯拉的人中 511 00:20:47,029 --> 00:20:49,039 甚至大多数人都没用过FSD 512 00:20:49,039 --> 00:20:52,799 我完全无法理解他们 513 00:20:52,799 --> 00:20:56,430 因为他们还没真正尝试过 514 00:20:56,430 --> 00:20:57,890 在一周的时间里 515 00:20:57,890 --> 00:20:58,750 真正体验过 516 00:20:58,750 --> 00:21:03,109 因为他们开车，而我的车在载我 517 00:21:03,109 --> 00:21:05,910 所以这简直 518 00:21:05,910 --> 00:21:09,250 我的生活在这个维度上完全不同 519 00:21:09,250 --> 00:21:12,210 现在它确实是一款真正的高端产品 520 00:21:12,210 --> 00:21:13,710 即使在当前功能水平下 521 00:21:13,710 --> 00:21:14,769 它还在不断改进 522 00:21:14,769 --> 00:21:16,799 这有点诡异 523 00:21:16,960 --> 00:21:19,680 你说得对，确实如此 524 00:21:19,680 --> 00:21:21,019 这花了尼克这么久 525 00:21:21,019 --> 00:21:22,630 哦，谁啊 526 00:21:22,630 --> 00:21:25,190 因为我之前用的是 527 00:21:25,190 --> 00:21:25,990 我想说 528 00:21:25,990 --> 00:21:27,150 让我陷入危险 529 00:21:27,150 --> 00:21:28,089 这就是我说的 530 00:21:28,089 --> 00:21:28,369 就像 531 00:21:28,369 --> 00:21:29,369 产品本身没错 532 00:21:29,369 --> 00:21:30,170 比如我 533 00:21:30,170 --> 00:21:33,329 当产品持续开发时 534 00:21:33,329 --> 00:21:34,049 这简直最糟糕 535 00:21:34,049 --> 00:21:36,490 直到它变得完美 536 00:21:36,490 --> 00:21:38,569 所以人们说他们试过 537 00:21:38,569 --> 00:21:38,930 他们就说 538 00:21:38,930 --> 00:21:39,089 哦 539 00:21:39,089 --> 00:21:39,609 还挺不错的 540 00:21:39,609 --> 00:21:40,130 但其实 541 00:21:40,130 --> 00:21:43,089 你最近一个月试过吗 542 00:21:43,089 --> 00:21:43,609 是的 543 00:21:43,609 --> 00:21:44,509 最近两个月 544 00:21:44,509 --> 00:21:45,089 真的吗 545 00:21:45,089 --> 00:21:47,170 要么三个月前 546 00:21:47,170 --> 00:21:48,230 现在有新版本 547 00:21:48,230 --> 00:21:49,730 呃 548 00:21:49,769 --> 00:21:50,910 但14.5版本 549 00:21:50,910 --> 00:21:53,710 即将发布一个新版本更新 550 00:21:53,710 --> 00:21:56,509 我认为这就是当前的问题所在 551 00:21:56,509 --> 00:21:57,839 当前的核心问题 552 00:21:57,839 --> 00:22:00,200 关税的空间组成部分 553 00:22:00,200 --> 00:22:01,400 重磅公告 554 00:22:01,400 --> 00:22:03,680 月球上的质量驱动器 555 00:22:04,319 --> 00:22:06,469 太空数据中心与太空 556 00:22:06,469 --> 00:22:08,749 你能帮忙梳理这个框架吗 557 00:22:08,749 --> 00:22:10,028 什么是质量驱动器 558 00:22:10,028 --> 00:22:14,828 长期来看这意味着什么 559 00:22:18,440 --> 00:22:24,000 首先我认为要理解接下来会发生什么 560 00:22:24,339 --> 00:22:26,359 时间线围绕星舰展开 561 00:22:26,359 --> 00:22:27,240 看看星舰 562 00:22:27,240 --> 00:22:27,619 首先 563 00:22:27,619 --> 00:22:30,900 星舰的发展将是所有后续工作的先决条件 564 00:22:30,900 --> 00:22:33,329 一旦实现可重复使用 565 00:22:33,329 --> 00:22:39,420 那时发射成本将呈数量级下降 566 00:22:39,420 --> 00:22:41,869 这才是真正打开局面的关键 567 00:22:41,869 --> 00:22:43,049 所以是的 568 00:22:43,049 --> 00:22:46,900 我觉得现在人们最需要关注的就是这个 569 00:22:46,900 --> 00:22:51,180 接下来将部署大量项目 570 00:22:51,180 --> 00:22:52,420 包括星链计划 571 00:22:52,420 --> 00:22:55,869 直接销售和轨道数据中心 572 00:22:55,869 --> 00:23:02,929 然后你将看到星舰与擎天柱技术的结合 573 00:23:02,929 --> 00:23:06,740 因为要在月球上建造这些设施 574 00:23:06,740 --> 00:23:08,460 我们确实可以运送大量人员 575 00:23:08,460 --> 00:23:11,519 但在马斯克的演示中 576 00:23:11,519 --> 00:23:16,289 很可能由擎天柱和人类协同作业 577 00:23:16,289 --> 00:23:20,450 因此需要关注擎天柱的发展进程 578 00:23:20,450 --> 00:23:26,329 你将能清晰看到我们的时间轨迹 579 00:23:26,329 --> 00:23:28,359 然后是月球上的质量驱动器 580 00:23:28,359 --> 00:23:34,259 你可以建立类似电磁轨道发射卫星 581 00:23:34,259 --> 00:23:39,069 而非依赖地球上的推进方式 582 00:23:39,069 --> 00:23:42,349 因为月球重力仅为地球的几分之一 583 00:23:42,349 --> 00:23:47,960 之后可服务地球轨道及更远空间 584 00:23:49,440 --> 00:23:49,839 所以山姆 585 00:23:49,839 --> 00:23:56,880 当你提到确实需要关注所有正在进行的布局策略 586 00:23:56,880 --> 00:24:00,730 因为我认为这是人们质疑马斯克的关键所在 587 00:24:00,730 --> 00:24:03,289 以及他的公司试图 588 00:24:03,289 --> 00:24:03,910 你知道的 589 00:24:03,910 --> 00:24:04,410 想象一下 590 00:24:04,410 --> 00:24:06,299 他描述的场景确实 591 00:24:06,299 --> 00:24:10,380 必须确保星舰能够规模化运作 592 00:24:10,380 --> 00:24:13,660 同时还要确保擎天柱能规模化应用 593 00:24:13,660 --> 00:24:17,099 这两者都是极具挑战性的难题 594 00:24:17,099 --> 00:24:20,039 但两者需同时成功才能 595 00:24:20,039 --> 00:24:24,789 进而实现月球质量驱动器和太空数据中心的规模化 596 00:24:24,789 --> 00:24:25,990 所有这些 597 00:24:25,990 --> 00:24:27,930 规模巨大 598 00:24:27,930 --> 00:24:29,029 建立在庞大基础上 599 00:24:30,190 --> 00:24:31,150 你在展示它 600 00:24:31,150 --> 00:24:34,450 你以一种非常确定的方式在展示 601 00:24:34,450 --> 00:24:36,190 正如你刚才描述的那样 602 00:24:36,190 --> 00:24:38,390 除非你在听这个听证会 603 00:24:39,128 --> 00:24:39,788 不不 604 00:24:39,788 --> 00:24:41,088 当然这是简化 605 00:24:41,088 --> 00:24:44,169 但这就是你构建模型的方式 606 00:24:44,169 --> 00:24:47,440 Three 并没有创造完美之物 607 00:24:47,440 --> 00:24:49,519 当然还有嵌套的专家 608 00:24:49,519 --> 00:24:52,259 他们拥有世界上最聪明最优秀的 609 00:24:52,259 --> 00:24:55,979 最努力的团队通过迭代进化 610 00:24:55,979 --> 00:24:56,818 就是这样 611 00:24:56,818 --> 00:25:01,369 如果我们要在火星生活并探索其他星球 612 00:25:01,369 --> 00:25:04,619 需要许多事情顺利进行 613 00:25:04,619 --> 00:25:06,539 但你似乎一无所获 614 00:25:06,539 --> 00:25:08,779 没有人第一次就能做对 615 00:25:08,779 --> 00:25:09,960 但你需要大量 616 00:25:09,960 --> 00:25:10,839 我认为还有 617 00:25:10,839 --> 00:25:11,380 这确实 618 00:25:11,380 --> 00:25:12,160 这里有个 619 00:25:12,160 --> 00:25:15,619 星舰完全可重复使用 620 00:25:15,619 --> 00:25:21,799 是他们直接瞄准并押注的重大工程转折点 621 00:25:21,799 --> 00:25:24,079 这打开了大量期权价值 622 00:25:24,079 --> 00:25:29,430 而太空AI算力似乎是最佳利用方式 623 00:25:29,430 --> 00:25:30,670 一个星链 624 00:25:30,670 --> 00:25:32,710 你知道这是很好的现金流 625 00:25:32,710 --> 00:25:33,730 生成式业务 626 00:25:33,730 --> 00:25:37,929 太空AI算力实际上边际上 627 00:25:37,929 --> 00:25:42,009 从我们观察来看更具吸引力 628 00:25:42,009 --> 00:25:43,729 即使规模扩大 629 00:25:43,729 --> 00:25:46,549 但区别在于它可以持续扩展 630 00:25:46,549 --> 00:25:48,750 而星链更像是 631 00:25:48,750 --> 00:25:51,269 需求曲线尾部开始下降 632 00:25:51,269 --> 00:25:52,450 所以你可以 633 00:25:52,450 --> 00:25:53,029 你拥有 634 00:25:53,029 --> 00:25:59,150 你可以获得大量资本投入的卫星发射选项 635 00:25:59,150 --> 00:26:05,250 因此使运载工具的价值远超 636 00:26:05,250 --> 00:26:09,640 原本预期因为能建造千台而非百台 637 00:26:09,640 --> 00:26:13,319 并开始降低发射周期 638 00:26:13,319 --> 00:26:17,359 可将大量物资送入太空 639 00:26:17,359 --> 00:26:20,859 只要持续产生收益 640 00:26:20,859 --> 00:26:24,900 我们认为会继续如此从而获得进一步期权价值 641 00:26:24,900 --> 00:26:26,599 现在我们有了资金开发月球 642 00:26:26,599 --> 00:26:30,920 然后我们将降低地球轨道星座的运营成本 643 00:26:30,920 --> 00:26:31,519 从 644 00:26:31,519 --> 00:26:37,619 你知道从每公斤七十到六十美元降至五到六美元 645 00:26:37,619 --> 00:26:42,500 两年前马斯克会说 646 00:26:42,500 --> 00:26:42,720 哦 647 00:26:42,720 --> 00:26:43,759 月球是分心 648 00:26:43,759 --> 00:26:45,559 我们去火星 649 00:26:45,559 --> 00:26:46,500 所以这并不 650 00:26:46,500 --> 00:26:47,240 他并不 651 00:26:47,240 --> 00:26:50,589 你知道他并不介意 652 00:26:50,589 --> 00:26:51,190 有点像 653 00:26:51,190 --> 00:26:56,109 根据对战略格局的评估调整方向 654 00:26:56,109 --> 00:27:01,230 我认为月球开发背后有很多逻辑支撑 655 00:27:01,230 --> 00:27:02,349 并且可能发生 656 00:27:02,349 --> 00:27:03,950 而且 657 00:27:04,829 --> 00:27:07,069 就像山姆说的那样他们确实 658 00:27:07,069 --> 00:27:09,630 他们必须先解决这些问题 659 00:27:09,630 --> 00:27:12,619 而存在一个非常大胆的隐含赌注 660 00:27:12,619 --> 00:27:14,059 对计算力的需求 661 00:27:14,059 --> 00:27:15,960 需求真的非常极端 662 00:27:15,960 --> 00:27:17,960 对智能计算的需求 663 00:27:17,960 --> 00:27:21,019 嗯不断扩展又扩展 664 00:27:21,019 --> 00:27:22,940 是的 665 00:27:23,980 --> 00:27:25,029 好的 666 00:27:25,029 --> 00:27:27,250 上周评论中有没有问题 667 00:27:27,250 --> 00:27:27,650 尼克 668 00:27:27,650 --> 00:27:29,329 有一个问题 669 00:27:29,329 --> 00:27:30,670 我觉得很有趣 670 00:27:30,670 --> 00:27:33,829 为什么尼克看起来总是惊讶 671 00:27:33,829 --> 00:27:36,990 答案就是这个节目 672 00:27:36,990 --> 00:27:39,190 这个节目没有剧本 673 00:27:39,190 --> 00:27:41,809 布雷特口中的一些内容 674 00:27:41,809 --> 00:27:43,609 总是让我惊讶 675 00:27:44,690 --> 00:27:47,490 我们讨论的一些数据和内容 676 00:27:47,490 --> 00:27:51,240 我不断从你们和节目嘉宾身上学习 677 00:27:51,240 --> 00:27:53,609 所以我保持一种状态 678 00:27:53,609 --> 00:27:56,289 接下来发生的事大概会让我惊讶 679 00:27:56,289 --> 00:27:56,869 我们的观点 680 00:27:56,869 --> 00:27:57,730 我们的分析 681 00:27:57,730 --> 00:28:00,509 以及我们对世界的看法 682 00:28:00,509 --> 00:28:01,880 嗯不 683 00:28:01,880 --> 00:28:02,200 我没有 684 00:28:02,200 --> 00:28:02,960 我没看到 685 00:28:02,960 --> 00:28:05,200 这里还有一个问题 686 00:28:05,200 --> 00:28:06,079 或许布雷特 687 00:28:06,079 --> 00:28:06,759 你可以回答这个 688 00:28:06,759 --> 00:28:11,079 我对数字奥特士的理解是它能驱动 689 00:28:11,079 --> 00:28:16,259 你的电脑执行任务和应用程序 690 00:28:16,259 --> 00:28:17,299 基于你的提示 691 00:28:17,299 --> 00:28:19,259 我这样理解对吗 692 00:28:21,079 --> 00:28:22,619 不我认为大致正确 693 00:28:22,619 --> 00:28:29,559 而当前正在发生的转变是从需要 694 00:28:29,559 --> 00:28:35,539 拥有非常受控的系统访问权限 695 00:28:35,539 --> 00:28:39,000 转向能够运行的代理集合 696 00:28:39,000 --> 00:28:43,700 基本作为用户与网页数字界面交互 697 00:28:43,700 --> 00:28:46,299 并很快适用于任何系统运行的应用 698 00:28:46,299 --> 00:28:51,779 这对现有软件即服务公司极具冲击 699 00:28:51,779 --> 00:28:54,920 所以想象一下我甚至不需要列举一个 700 00:28:54,920 --> 00:28:59,200 但想象一家绩效管理公司持续跟踪你的KPI 701 00:28:59,200 --> 00:29:02,019 你需要在那里输入季度评估和其他内容 702 00:29:02,019 --> 00:29:02,859 他们收集这些数据 703 00:29:02,859 --> 00:29:04,579 并将其整合到一个界面中 704 00:29:04,579 --> 00:29:05,839 所以任何用户都可以使用 705 00:29:05,839 --> 00:29:08,299 但无法通过程序访问 706 00:29:08,299 --> 00:29:10,279 因为这是他们的核心竞争力 707 00:29:10,279 --> 00:29:11,799 他们部分提供 708 00:29:11,799 --> 00:29:14,890 相关的洞察和可视化结果 709 00:29:14,890 --> 00:29:16,730 假设你想迁移到该平台之外 710 00:29:16,730 --> 00:29:18,920 或者想做自定义操作 711 00:29:18,920 --> 00:29:20,599 你可以请求他们的团队开发功能 712 00:29:20,599 --> 00:29:21,640 也许他们会处理 713 00:29:21,640 --> 00:29:22,160 也许不会 714 00:29:22,160 --> 00:29:24,289 或者作为用户 715 00:29:24,289 --> 00:29:29,220 可以系统性地从SaaS服务商处提取所有数据 716 00:29:29,220 --> 00:29:29,619 嗯 717 00:29:29,619 --> 00:29:32,539 Digital Optimus 将实现这种行为 718 00:29:32,539 --> 00:29:34,500 因为系统无法识别 719 00:29:34,500 --> 00:29:35,599 是否是登录用户 720 00:29:35,599 --> 00:29:37,859 还是Digital Optimus在操作 721 00:29:37,859 --> 00:29:41,049 我认为这是下一步动作 722 00:29:41,049 --> 00:29:42,179 嗯 723 00:29:42,179 --> 00:29:44,538 以及企业中的智能代理 724 00:29:44,538 --> 00:29:44,959 这 725 00:29:44,959 --> 00:29:47,878 还能与现有复杂嵌套系统兼容 726 00:29:47,878 --> 00:29:51,539 兼容人们已运行的所有棘手嵌套系统 727 00:29:51,539 --> 00:29:56,670 我认为这将对 728 00:29:56,670 --> 00:29:57,349 你知道的 729 00:29:57,349 --> 00:29:58,269 现有SaaS 730 00:29:58,269 --> 00:29:59,190 某些 731 00:29:59,190 --> 00:30:00,609 我不 732 00:30:00,609 --> 00:30:01,349 是的 733 00:30:01,349 --> 00:30:02,109 好的 734 00:30:02,109 --> 00:30:02,430 是的 735 00:30:02,430 --> 00:30:03,329 看下一页 736 00:30:03,329 --> 00:30:06,029 请继续发送问题 737 00:30:06,029 --> 00:30:07,210 我们会回答 738 00:30:07,210 --> 00:30:07,789 全部 739 00:30:07,789 --> 00:30:08,890 Arc投资管理 740 00:30:08,890 --> 00:30:10,230 有限责任公司是SEC 741 00:30:10,230 --> 00:30:11,789 注册投资顾问 742 00:30:11,789 --> 00:30:15,079 Arc金融与公共公司是独立实体 743 00:30:15,079 --> 00:30:17,720 彼此无关联 744 00:30:17,720 --> 00:30:21,079 不涉及任何公司的营销或销售 745 00:30:21,079 --> 00:30:25,650 Arc金融与公共公司不对 746 00:30:25,650 --> 00:30:30,430 明确不承担任何责任或义务 747 00:30:30,430 --> 00:30:32,769 或损失或损害负责 748 00:30:32,769 --> 00:30:33,839 或其关联方 749 00:30:33,839 --> 00:30:34,500 执法人员 750 00:30:34,500 --> 00:30:35,920 员工或代理人 751 00:30:35,920 --> 00:30:38,859 任何公司均不对任何间接事件承担责任 752 00:30:38,859 --> 00:30:39,599 附带的 753 00:30:39,599 --> 00:30:44,869 或由此播客及其内容衍生的特殊损害 754 00:30:48,630 --> 00:30:54,440 不应作为投资决策的依据，且可能随时变更 755 00:30:54,440 --> 00:30:57,920 不构成显性或隐性的 756 00:30:57,920 --> 00:31:00,730 艺术提供的任何服务或产品条款 757 00:31:00,730 --> 00:31:07,690 投资者应自行判断特定投资管理服务是否符合自身需求 758 00:31:07,690 --> 00:31:13,349 节目中提及的公司或证券的所有陈述均为arc的个人观点 759 00:31:13,349 --> 00:31:14,450 及节目嘉宾 760 00:31:14,450 --> 00:31:18,089 并不代表arc对任何公司或证券的认可 761 00:31:18,089 --> 00:31:20,089 或arc建议买入 762 00:31:20,089 --> 00:31:20,450 卖出 763 00:31:20,450 --> 00:31:22,429 或持有任何证券 764 00:31:22,429 --> 00:31:25,849 历史结果不代表未来表现 765 00:31:25,849 --> 00:31:30,230 本节目中部分陈述可能为未来预期声明 766 00:31:30,230 --> 00:31:33,109 及其他基于arc当前观点和假设的前瞻性声明 767 00:31:33,109 --> 00:31:34,609 并涉及已知或未知的风险与不确定性 768 00:31:34,609 --> 00:31:39,430 可能导致实际结果、业绩或事件与声明存在重大差异 769 00:31:39,430 --> 00:31:44,730 arc不承担更新任何前瞻性信息的义务 770 00:31:44,730 --> 00:31:49,329 arc及其客户及相关人员可能 771 00:31:49,329 --> 00:31:53,069 但未必对讨论的证券或发行人拥有财务利益 772 00:31:53,069 --> 00:31:58,930 所获信息源自可信来源 773 00:31:58,930 --> 00:32:02,839 然而 774 00:32:02,839 --> 00:32:03,259 arc不对第三方信息的准确性或完整性作担保 775 00:32:03,259 --> 00:32:08,880 arc及其关联方可能
评论采集失败
采集完成，该视频无弹幕