美国芯片管制收紧，长鑫存储押注新DRAM架构：一场被迫的豪赌
字幕摘录
| 时间 | 英文 | 中文 |
|---|---|---|
| 0:00 | So, um, imagine you get a job to build a really huge modern building. | 所以,嗯,想象你得到一个工作 建造一个非常巨大的现代建筑。 |
| 0:05 | Okay, like a tall city tower. | 好吧,像一座高大的城楼。 |
| 0:07 | Yeah, exactly. | 没错 |
| 0:08 | But there is a massive problem. | 但有一个巨大的问题。 |
| 0:11 | You are legally banned from buying modern cranes. | 依法禁止你购买现代起重机. |
| 0:15 | Oh wow. | 哦,哇呜。 |
| 0:16 | Right? | 对吧? |
| 0:16 | That makes it hard. | 这才难 |
| 0:17 | Right. | 对 |
| 0:18 | So do you just stop building? | 你只是停止建设? |
展开字幕全文（543 条）
| 序号 | 英文 | 中文 |
|---|---|---|
| 1 | So, um, imagine you get a job to build a really huge modern building. | 所以,嗯,想象你得到一个工作 建造一个非常巨大的现代建筑。 |
| 2 | Okay, like a tall city tower. | 好吧,像一座高大的城楼。 |
| 3 | Yeah, exactly. | 没错 |
| 4 | But there is a massive problem. | 但有一个巨大的问题。 |
| 5 | You are legally banned from buying modern cranes. | 依法禁止你购买现代起重机. |
| 6 | Oh wow. | 哦,哇呜。 |
| 7 | Right? | 对吧? |
| 8 | That makes it hard. | 这才难 |
| 9 | Right. | 对 |
| 10 | So do you just stop building? | 你只是停止建设? |
| 11 | No, you completely redraw the map. | 不,你完全重新绘制地图。 |
| 12 | You have to find a new way. | 你必须找到新的方法。 |
| 13 | Exactly. | 没错 |
| 14 | And today we are looking at how a big roadblock in the tech world | 今天我们正在研究科技界的路障 |
| 15 | is forcing a Chinese chip maker to do exactly that. | 迫使一个中国芯片制造者这么做 |
| 16 | It is a really wild story. | 这是一个非常疯狂的故事。 |
| 17 | It is. | 没错 |
| 18 | We have an article here from the South China Morning Post. | 本文有"南华早报"的文章. |
| 19 | It is about a company named CXMT. | 是关于一家叫CXMT的公司 |
| 20 | And they are trying a very risky new way to build computer chips. | 他们正在尝试一种非常危险的新方法来制造计算机芯片。 |
| 21 | Yes, they are. | 是的,他们是。 |
| 22 | So our mission in this deep dive is to understand why. | 因此,我们这次深潜的任务是理解原因。 |
| 23 | Why are they changing their basic design? | 他们为什么要改变基本设计? |
| 24 | Right. | 对 |
| 25 | And how does this change the global tech race? | 这如何改变全球科技竞赛? |
| 26 | Yeah, and most importantly, we will look at how this changes the devices you use every day. | 最重要的是,我们将研究这如何改变 你每天使用的设备。 |
| 27 | Like the phone in your pocket. | 就像你口袋里的电话 |
| 28 | Because it really will affect everyone. | 因为它真的会影响每个人。 |
| 29 | It will. | 会的 |
| 30 | Just a quick note before we jump in though. | 只是我们跳进去前的快讯 |
| 31 | This story involves trade rules between the US and China. | 这个故事涉及美中贸易规则. |
| 32 | Right, the export bans. | 没错 出口禁令 |
| 33 | Yeah, but we want to be super clear. | 是的 但我们想说清楚 |
| 34 | We are not taking any political sides here at all. | 我们完全不在这里采取任何政治立场。 |
| 35 | No, not at all. | 不,一点也不。 |
| 36 | We are just looking at the facts and the source material. | 我们只是看看事实和源材料。 |
| 37 | We just want to understand the technology. | 我们只是想了解技术。 |
| 38 | And the technology choices here are like really fascinating. | 这里的科技选择非常迷人。 |
| 39 | So set the stage for us. | 所以,为我们准备舞台。 |
| 40 | Why change the design at all? | 为什么改变设计? |
| 41 | Well, this company hit a wall. | 这家公司撞到墙了 |
| 42 | The United States has strict export rules now. | 美国现在有严格的出口规则。 |
| 43 | Right. | 对 |
| 44 | And these rules stop Chinese chip makers from buying a very specific machine. | 这些规则阻止了中国芯片制造商购买非常具体的机器. |
| 45 | What kind of machine? | 什么样的机器? |
| 46 | The industry calls them EUV machines. | 业界称它们为EUV机器. |
| 47 | EUV? | 欧维? |
| 48 | Yeah, it stands for extreme ultraviolet. | 是的,它代表着极端紫外线。 |
| 49 | But you can just think of them as very advanced light printers. | 但你可以把它们看作是非常先进的轻型打印机。 |
| 50 | Okay, light printers? | 好吧,轻打印机? |
| 51 | Right. | 对 |
| 52 | They come from a Dutch company named ASML. | 他们来自一家名叫ASML的荷兰公司. |
| 53 | Just one company? | 就一个公司? |
| 54 | Mostly, yeah. | 大部分时间,是的。 |
| 55 | And these machines use invisible light to print tiny, tiny circuits onto computer chips. | 这些机器使用隐形光将微小的电路打印到计算机芯片上。 |
| 56 | And a circuit is just a tiny path for electricity, right? | 而电路只是一条电路,对吧? |
| 57 | Exactly. | 没错 |
| 58 | It is just a path. | 这只是一条道路。 |
| 59 | And you need these special Dutch printers to make the paths small enough for modern phones. | 你需要这些特殊的荷兰打印机 使这些路径足够小 用于现代手机。 |
| 60 | Because the smaller the path, the more paths you can fit on one chip. | 因为路径越小,你就越能搭配到一个芯片上. |
| 61 | Yes. | 对 |
| 62 | And more paths mean a much faster phone. | 更多的路径意味着一个更快的电话。 |
| 63 | Okay, but CXMT is blocked from buying these Dutch printers. | 好吧,但CXMT 被阻止购买这些荷兰打印机。 |
| 64 | Right. | 对 |
| 65 | They are cut off from the best standard tools. | 他们与最佳工具隔绝。 |
| 66 | So here is my question. | 这就是我的问题 |
| 67 | If everyone else in the world uses these highly advanced printers, what is CXMT doing? | 如果世界上其他人都使用这些高度先进的打印机,CXMT在做什么? |
| 68 | They have to get creative. | 他们得有创意 |
| 69 | But are they, you know, trying to build a modern sports car using old hand tools? | 但是,他们,你知道, 试图用旧的手工具建造现代的运动车吗? |
| 70 | Because that seems like a bad idea. | 因为这似乎是一个坏主意。 |
| 71 | Yeah, that would be a guaranteed way to fail, but they are not doing that. | 是啊,这将是一个保证 失败的方式,但他们没有这样做。 |
| 72 | Oh, they are not? | 是吗? |
| 73 | No. | 没有 |
| 74 | They are not trying to force old tools to do the new job. | 他们并不是要强迫老工具来做新的工作。 |
| 75 | They are actually changing the shape of the chip itself. | 他们实际上正在改变芯片本身的形状。 |
| 76 | Wait, changing the shape? | 等等,改变形状? |
| 77 | Yeah. | 对 |
| 78 | They are changing the basic layout. | 他们正在改变基本布局。 |
| 79 | Think back to your building example. | 回想一下你的建筑榜样。 |
| 80 | The one with no crane? | 没有起重机的那个? |
| 81 | Right. | 对 |
| 82 | If you do not have the crane, you cannot build a tall tower. | 如果你没有起重机, 你不能建造一个高塔。 |
| 83 | So what do you do? | 那你怎么办? |
| 84 | You buy a wider piece of land. | 你买一块更大的土地。 |
| 85 | And you build a flat, wide building instead. | 你建一座平坦的宽阔建筑 |
| 86 | Ah, I see. | 原来如此 |
| 87 | You change the design to fit the tools you actually have. | 你改变设计是为了适应你实际拥有的工具。 |
| 88 | Exactly. | 没错 |
| 89 | By changing the physical layout of the chip, | 通过改变芯片的物理布局, |
| 90 | CXMT can use older printing machines that they already have. | CXMT可以使用他们已经拥有的更古老的印刷机. |
| 91 | And they still get a fast chip in the end? | 他们最后还是得到一个快芯片? |
| 92 | That is the goal, yes. | 这就是目标,是的。 |
| 93 | Wow. | 珃 |
| 94 | So they are drawing a completely new map for the tiny city inside the chip. | 他们正在为芯片里面的小城市绘制全新的地图。 |
| 95 | A completely new map. | 一张全新的地图 |
| 96 | Okay. | 摆 |
| 97 | To understand this new map, we need to know what they are actually building. | 为了了解这个新地图,我们需要知道它们实际上正在建造什么。 |
| 98 | Right. | 对 |
| 99 | The type of chip. | 芯片的类型。 |
| 100 | Yeah. | 对 |
| 101 | The source says they make a specific kind of chip called DRAM. | 消息来源称他们制造了一种特定的芯片,称为DRAM. |
| 102 | DRAM. | 梦。 |
| 103 | Yes. | 对 |
| 104 | Dynamic Random Access Memory. | 动态随机访问内存. |
| 105 | But we can skip the long name. | 但我们可以跳过长的名字。 |
| 106 | Yeah. | 对 |
| 107 | Let us keep it simple. | 让我们保持简单。 |
| 108 | Just think of DRAM as the fast working memory for your computer. | 把DRAM当作你的电脑的快速存储器 |
| 109 | Fast working memory. | 快速工作记忆. |
| 110 | Right. | 对 |
| 111 | It is kind of like the physical desk in your office. | 这有点像你办公室的办公桌 |
| 112 | Oh, I like that analogy. | 哦,我喜欢这个类比。 |
| 113 | Yeah. | 对 |
| 114 | So your hard drive is like the big filing cabinet in the corner of the room. | 所以你的硬盘就像房间角落的大档案柜 |
| 115 | Okay. | 摆 |
| 116 | That is where I keep all my old files. | 那是我保存所有旧文件的地方 |
| 117 | Exactly. | 没错 |
| 118 | It holds everything long term. | 它能长期维持一切。 |
| 119 | But when you need to actually work on a file, what do you do? | 但当你需要实际工作 在文件,你做什么? |
| 120 | I pull it out and put it on my desk. | 我拔出来放桌上 |
| 121 | Right. | 对 |
| 122 | You put it on your desk so you can reach it quickly. | 你把它放在桌子上 这样你就可以快速到达它。 |
| 123 | So the DRAM is the desk. | 所以DRAM是办公桌。 |
| 124 | Yes. | 对 |
| 125 | When you open a map app on your phone, it goes from the cabinet to the DRAM desk. | 当你在手机上打开一个地图应用程序,它从柜子到DRAM办公桌. |
| 126 | Because the phone needs to read that map instantly while I am driving. | 因为手机需要立刻读完地图 在我开车的时候 |
| 127 | Exactly. | 没错 |
| 128 | And as artificial intelligence gets more popular, computers need much larger desks. | 随着人工智能越来越流行,计算机需要更大的办公桌. |
| 129 | Because AI uses so much data. | 因为AI使用这么多的数据. |
| 130 | Yes. | 对 |
| 131 | An AI system reads billions of pieces of data at the same time. | 一个AI系统同时读取数十亿块数据. |
| 132 | It needs a massive amount of fast working memory. | 它需要大量的快速工作记忆. |
| 133 | Okay. | 摆 |
| 134 | So everyone wants to build a bigger desk. | 所以大家都想建一个更大的办公桌 |
| 135 | Right. | 对 |
| 136 | And the standard way to design this memory is an old map. | 而设计这种内存的标准方法是老地图. |
| 137 | The industry calls it the 6F squared layout. | 业界称之为6F平方布局. |
| 138 | Yeah, 6F squared. | 对,6F平方 |
| 139 | We do not need to worry about the complex math here. | 我们不需要担心这里复杂的数学。 |
| 140 | The letter F just represents the size of the basic parts. | 字母F只是代表基本部分的大小. |
| 141 | Right. | 对 |
| 142 | It's just a unit of measurement. | 这只是一个计量单位。 |
| 143 | So the standard map uses six squares of space to hold one single piece of memory. | 因此标准地图使用6平方的空间来持有单片记忆. |
| 144 | Exactly. | 没错 |
| 145 | Six squares. | 六正显. |
| 146 | And it is a very reliable design. | 这是一个非常可靠的设计。 |
| 147 | Every major company in the world uses it. | 世界上每个大公司都使用它. |
| 148 | Because it works really well. | 因为它的工作真的很好。 |
| 149 | It works perfectly with those advanced Dutch EUV printers we talked about. | 它和我们讨论过的先进的荷兰EUV打印机是完全一样的。 |
| 150 | You just use that incredible light to print those six squares smaller and smaller and smaller. | 你只要用那不可思议的光来打印这六个方块大小越来越小. |
| 151 | Right. | 对 |
| 152 | But as we know, CXMT cannot do that. | 但正如我们所知,CXMT不能这样做。 |
| 153 | Because they cannot buy the machines. | 因为他们不能买机器。 |
| 154 | Right. | 对 |
| 155 | So they are moving to a new map. | 因此,他们正在向一个新的地图移动。 |
| 156 | This new layout is called 4F squared. | 这种新的布局被称为4F平方形. |
| 157 | 4F squared. | 4F方块。 |
| 158 | So they drop from six squares down to four squares. | 因此它们从六个方块下降到四个方块. |
| 159 | Yes. | 对 |
| 160 | They save about one third of the space. | 他们节省了大约三分之一的空间. |
| 161 | That is a massive difference. | 这是一个巨大的差异。 |
| 162 | It is. | 没错 |
| 163 | I mean, on a microscopic scale, saving one third of the space is huge. | 我的意思是,在微观尺度上, 节省三分之一的空间是巨大的。 |
| 164 | It is. | 没错 |
| 165 | They can pack way more memory into the exact same physical area. | 他们可以把更多的记忆 包装到完全相同的物理区域。 |
| 166 | And the best part for them is the factory side, right? | 他们最好的部分是工厂方面,对吧? |
| 167 | Exactly. | 没错 |
| 168 | Because this basic layout is so tight and dense, | 因为这种基础布局非常紧凑和密集, |
| 169 | they can hit their speed goals without the banned EUV printers. | 没有被禁用的EUV打印机,他们就能打出速度目标. |
| 170 | They just use older printers on a tighter map. | 他们只是用老打印机 在更紧的地图。 |
| 171 | Right. | 对 |
| 172 | They save 33% of the space right away. | 他们立刻节省了33%的空间. |
| 173 | It is like taking a normal neighborhood of spread out houses. | 这就像一个正常的邻里 分散的房子。 |
| 174 | And turning them into a dense, tall apartment building. | 然后把它们变成一个密集的,高大的公寓建筑. |
| 175 | Right. | 对 |
| 176 | You fit far more people into a single city block. | 你把更多的人安置在一个城市街区里 |
| 177 | Exactly. | 没错 |
| 178 | But hold on, though. | 但坚持,虽然。 |
| 179 | What? | 你说什么? |
| 180 | If you tear down a regular neighborhood to build a massive dense apartment block, | 如果你拆掉一个普通的街区来建造一个巨大的密集的公寓街区, |
| 181 | you eventually have problems. | 你最终会有问题。 |
| 182 | Like what? | 比如说? |
| 183 | Well, you overwhelm the system. | 好吧,你覆盖了系统。 |
| 184 | The water pipes break. | 水管断绝. |
| 185 | The traffic gets terrible because there are too many cars. | 车流量太高了 因为车太多了 |
| 186 | Right, right. | 对,对。 |
| 187 | Doesn't packing a computer chip this tightly cause massive physical problems in the factory? | 包装一个电脑芯片 会不会在工厂里造成巨大的物理问题? |
| 188 | Yes. | 对 |
| 189 | It creates huge manufacturing problems. | 这造成了巨大的制造问题. |
| 190 | I knew it. | 我早知道会这样 |
| 191 | Building that dense apartment block requires many, many more steps than building a simple house. | 建造这个密集的公寓区 比建造一个简单的房子需要很多,更多的步骤. |
| 192 | So the 4F square design forces the factory to do extra steps. | 因此4F方形设计迫使工厂采取额外的步骤. |
| 193 | A lot of extra steps. | 很多额外的步骤。 |
| 194 | Extra production cycles. | 额外的生产周期。 |
| 195 | Let us walk through the factory floor then. | 那让我们穿过工厂的地板 |
| 196 | Because making a chip is not just one step. | 因为制造芯片不只是一步 |
| 197 | No, it takes weeks. | 不,这需要几个星期。 |
| 198 | You have to wash the silicon. | 须洗硅. |
| 199 | You bake it. | 你来烤吧 |
| 200 | You print on it with light. | 你用光印着它。 |
| 201 | You wash it with chemicals. | 你用化学剂洗它。 |
| 202 | Right. | 对 |
| 203 | You do this hundreds of times over and over. | 你做这几百次 一遍又一遍。 |
| 204 | So if this new dense map adds, say, 50 extra steps. | 因此,如果这个新的密集地图增加了,比如说,50个额外的步骤. |
| 205 | That is 50 more chances for something to go terribly wrong. | 还有50次机会出错 |
| 206 | Wow. | 珃 |
| 207 | And a single speck of dust can ruin the whole chip, right? | 一块灰尘会毁了整个芯片,对吧? |
| 208 | Exactly. | 没错 |
| 209 | A speck of dust, a tiny change in temperature, anything can break it. | 灰尘的斑点,温度的微小变化,任何东西都能打破它. |
| 210 | So more steps mean more broken chips. | 所以更多的步骤意味着更多的碎芯片。 |
| 211 | Yes. | 对 |
| 212 | And this brings us to a critical word in the chip industry. | 这给我们带来了芯片行业的关键词。 |
| 213 | What is the word? | 什么词? |
| 214 | The word is yield. | 词是屈服。 |
| 215 | Yield. | 义德曰. |
| 216 | Yes. | 对 |
| 217 | Yield is simply the percentage of good working chips you get at the very end of the line. | ield只是你得到的 优秀工作芯片的百分比 在最端。 |
| 218 | Okay. | 摆 |
| 219 | So think about a car factory. | 想想汽车厂 |
| 220 | Right. | 对 |
| 221 | If you try a complicated new way to build engines and half the engines explode during testing. | 如果你尝试一种复杂的新方法来制造发动机,一半的发动机在测试中爆炸. |
| 222 | Then your yield is only 50 percent. | 那你的收益只有50% |
| 223 | And that is terrible. | 这是可怕的。 |
| 224 | Yes. | 对 |
| 225 | Because you still paid for the metal and the labor for those broken engines. | 因为你仍然支付金属 和那些破引擎的劳动力。 |
| 226 | Yes. | 对 |
| 227 | You just have to throw them all in the trash. | 你只要把它们都扔进垃圾桶 |
| 228 | So the price of the good engines has to double just to cover your losses. | 所以好引擎的价格要翻一番 才能弥补损失 |
| 229 | Exactly. | 没错 |
| 230 | The good chips have to pay for the bad chips. | 好芯片必须支付坏芯片. |
| 231 | And with this new tight 4F squared map, the extra steps push the yield way down. | 有了这张新的紧凑的4F平方图,额外的阶梯将产量推低. |
| 232 | The risk of broken chips goes up. | 芯片碎裂的风险上升。 |
| 233 | So the yield drops. | 所以产量下降。 |
| 234 | And the cost per chip goes up really fast. | 每块芯片的成本会迅速上涨。 |
| 235 | Right. | 对 |
| 236 | Now the source material shows CXMT plans to send some test chips to customers | 现在源材料显示CXMT计划向客户发送一些测试芯片 |
| 237 | by the end of this year. | 在今年年底之前。 |
| 238 | Okay. | 摆 |
| 239 | Test chips. | 测试芯片。 |
| 240 | Are these for phones? | 这是电话用的吗? |
| 241 | No. | 没有 |
| 242 | These are high-end chips for big computer servers. | 这些是大计算机服务器的高端芯片. |
| 243 | But wait. | 但等一会 |
| 244 | Making a few perfect test chips in a clean lab is entirely different from | 在干净的实验室里做几块完美的测试芯片 和 |
| 245 | running a profitable factory. | 经营一家盈利工厂。 |
| 246 | Very different. | 非常不一样。 |
| 247 | Making 10 chips is easy. | 制作10个芯片很简单 |
| 248 | Making 10 million is hard. | 赚一千万很难 |
| 249 | The article clearly notes that real mass production depends entirely on the market. | 文章明确指出,真正的大规模生产完全取决于市场。 |
| 250 | They have to see if customers will actually pay the higher price. | 他们得看看顾客是否真的会付出更高的代价. |
| 251 | This sounds like a financial nightmare for the company. | 这听起来像一个财务噩梦 对于公司。 |
| 252 | It does sound bad. | 这听起来确实很糟糕。 |
| 253 | I mean, the yields are low. | 产量很低 |
| 254 | The factory steps are long and complicated. | 工厂的台阶既长又复杂. |
| 255 | The risk of throwing chips in the trash is super high. | 将芯片扔进垃圾桶的风险极高. |
| 256 | Yeah. | 对 |
| 257 | How can CXMT even afford to try this? | CXMT怎么能买得起这个? |
| 258 | Are they just burning through all their cash to survive the trade rules? | 他们只是用他们所有的现金来维持贸易规则? |
| 259 | Actually, no. | 事实上,没有。 |
| 260 | The numbers in the report tell a very different story. | 报告中的数字是一个截然不同的故事。 |
| 261 | Oh, really? | 是吗? |
| 262 | Yeah. | 对 |
| 263 | They are not struggling for cash at all. | 他们根本没有为现金而挣扎。 |
| 264 | They're actually growing at an unbelievable speed. | 它们的生长速度令人难以置信。 |
| 265 | Wow. | 珃 |
| 266 | Okay. | 摆 |
| 267 | Give us the numbers. | 把号码给我们 |
| 268 | Let us look at their third quarter results. | 让我们看看它们第三季度的成果。 |
| 269 | Their total money coming in, their revenue, hit 118.71 billion yuan. | 他们的进货总额,收入达到1187.1亿元. |
| 270 | 118 billion. | 118亿. |
| 271 | Let me translate that. | 让我翻译一下 |
| 272 | That is about 17.8 billion US dollars in a single quarter. | 这大约是178亿美元在一个季度。 |
| 273 | Right. | 对 |
| 274 | It's a huge amount of money. | 其资甚多. |
| 275 | That is shocking. | 这是令人震惊的。 |
| 276 | And what about their growth rate? | 他们的增长率呢? |
| 277 | The growth is even wilder. | 增长更狂野. |
| 278 | The source says that number is up 613% from the previous year. | 消息来源说,这一数字比上一年上升了613%。 |
| 279 | Wait, 613%? | 等等,613%吗? |
| 280 | Yes. | 对 |
| 281 | So they multiplied their revenue by more than six in just 12 months. | 因此,他们在短短的12个月内,把收入增加了6倍以上。 |
| 282 | Exactly. | 没错 |
| 283 | And they are turning a massive profit too. | 并且他们也带来了巨大的利润。 |
| 284 | How much profit? | 赚多少钱? |
| 285 | They reported 54 billion yuan in pure profit. | 他们报道了540亿人民币的纯利润. |
| 286 | And here's the important part. | 这是重要的部分。 |
| 287 | They are taking a huge chunk of that money and pouring it directly into research. | 他们把一大笔钱 直接投入研究 |
| 288 | So they are betting on the future. | 因此,他们赌注未来。 |
| 289 | Yes. | 对 |
| 290 | Their research and development budget was over 11 billion yuan for the quarter. | 其本季度研发预算超过110亿元人民币. |
| 291 | 11 billion just for research. | 110亿只用于研究。 |
| 292 | Right. | 对 |
| 293 | That research spending is up 73%. | 研究开支增加了73%。 |
| 294 | So they are definitely not burning their last dollars. | 所以他们肯定不会烧掉最后的美元。 |
| 295 | They have a massive bank account. | 他们有一个庞大的银行账户。 |
| 296 | They have so much cash. | 他们有这么多现金。 |
| 297 | Which means they can actually afford to have a terrible factory yield right now? | 也就是说他们现在可以负担 糟糕的工厂产量? |
| 298 | Exactly. | 没错 |
| 299 | They are paying for the time to learn. | 他们花时间去学习。 |
| 300 | Ah. | 摆 |
| 301 | They can throw away broken chips every day | 他们每天都可以扔掉破碎的薯片 |
| 302 | while their engineers slowly figure out how to improve the new map. | 而他们的工程师们却慢慢地想出了如何改进新地图的方法. |
| 303 | Right. | 对 |
| 304 | They are buying time and they are buying experience. | 他们正在争取时间,他们正在争取经验。 |
| 305 | The high costs do not scare them. | 高昂的成本并不吓唬他们。 |
| 306 | No. | 没有 |
| 307 | Because this is their only path forward. | 因为这是他们唯一的前进道路。 |
| 308 | They have the motivation and they have the cash to survive the pain. | 他们有动机和现金 来忍受痛苦。 |
| 309 | OK. | 摆 |
| 310 | So CXMT is fully committed to this tight 4F squared layout. | 因此CXMT完全致力于这种紧凑的4F平方布局. |
| 311 | They are all in. | 他们都进来了。 |
| 312 | They are packing the memory as close together as possible. | 他们正在尽可能地把记忆放在一起。 |
| 313 | But what is the rest of the world doing? | 但世间众生作甚么. |
| 314 | That is where it gets really interesting. | 这才是真正有趣的地方。 |
| 315 | Because the US and China are not the only players in this game. | 因为美国和中国不是这个游戏的唯一玩家. |
| 316 | Are other giant tech companies copying this tight map? | 其他巨型科技公司 复制这张紧凑的地图吗? |
| 317 | No. | 没有 |
| 318 | The global industry is completely divided on this issue. | 全球工业在这个问题上完全分裂。 |
| 319 | Really? | 真的吗? |
| 320 | Yeah. | 对 |
| 321 | There is no agreement on the future of memory chips. | 对于内存芯片的未来没有达成一致. |
| 322 | Let us look at the United States first. | 让我们先看看美国。 |
| 323 | OK. | 摆 |
| 324 | The largest memory chip maker in the US is a company called Micron Technology. | 美国最大的记忆芯片制造商是一家名为Micron Technology的公司. |
| 325 | Micron? | 微纳? |
| 326 | Yes. | 对 |
| 327 | And according to the source, | 根据消息来源 |
| 328 | Micron strongly dislikes the 4F squared approach. | Micron强烈厌恶4F平方方法. |
| 329 | They dislike it? | 他们不喜欢吗? |
| 330 | Yes. | 对 |
| 331 | They are openly rejecting it? | 他们公开拒绝? |
| 332 | Yes. | 对 |
| 333 | The article quotes a guy named Scott DeBoer. | 文章引用了一个叫Scott DeBoer的人的话. |
| 334 | 他是Micron的主管 | |
| 335 | OK. What did he say? | 摆 他说什么? |
| 336 | He did not hold back at all. | 他丝毫没有退缩。 |
| 337 | He said the 4F squared design is, quote, | 他说4F平方设计是 |
| 338 | more complicated, it is more complex, it is more costly. | 更复杂 更复杂 更昂贵 |
| 339 | Wow. | 珃 |
| 340 | He really does not like it. | 他真的不喜欢 |
| 341 | No. | 没有 |
| 342 | And he speaks from direct experience. | 而他从直接的经验说话。 |
| 343 | Micron不仅仅是猜测 | |
| 344 | Oh, they tried it? | 哦,他们试过了? |
| 345 | Yeah. | 对 |
| 346 | They actually spent 20 years researching this exact dense layout. | 他们花了20年时间来研究这个精确的密集布局. |
| 347 | 20 years? | 20年吗? |
| 348 | 20 years. | 20岁 |
| 349 | They spent millions of dollars, | 他们花了数百万美元 |
| 350 | they built their own test chips in the lab, | 他们在实验室制造了自己的测试芯片 |
| 351 | and then they completely walked away from the idea. | 然后他们完全放弃了这个想法。 |
| 352 | Wait, why did they quit? | 等等,他们为什么辞职? |
| 353 | I mean, if a giant company like Micron spent 20 years on a tight map and gave up, | 我的意思是,如果像Micron这样的巨型公司 在一张紧凑的地图上花了20年时间 然后放弃了, |
| 354 | what did they hit? | 他们打了什么? |
| 355 | They hit the laws of physics. | 他们打击了物理定律. |
| 356 | Physics? | 物理吗? |
| 357 | Yes. | 对 |
| 358 | You simply cannot break the rules of nature. | 你根本不能违反自然规则 |
| 359 | Memory chips store data as tiny microscopic bits of electricity. | 记忆芯片将数据存储为微小的微缩电位. |
| 360 | OK. | 摆 |
| 361 | If you pack the memory cells too close together on the flat map, | 如果你把记忆细胞 放在平面图上 |
| 362 | the electricity starts to jump across the empty spaces. | 电力开始跳过空地 |
| 363 | It leaks from one cell to the next cell. | 它从一个细胞泄露到下一个细胞. |
| 364 | Exactly. | 没错 |
| 365 | Oh, I see. | 原来如此 |
| 366 | Think of it like two neighbors in a cheap apartment building. | 把它当成一个廉价公寓楼的两个邻居 |
| 367 | Right, with very thin walls. | 对,墙很薄 |
| 368 | Yeah. | 对 |
| 369 | If one neighbor plays loud music, | 如果有一个邻居演奏响亮的音乐 |
| 370 | the other neighbor hears it perfectly through the thin wall. | 另一个邻居从薄壁上完全听到 |
| 371 | Exactly. | 没错 |
| 372 | But in a computer chip, that loud music is electricity. | 但是在电脑芯片里 那响亮的音乐就是电 |
| 373 | And if the electricity leaks through the thin walls, your data gets mixed up. | 如果电流从薄壁漏出 你的数据就会被混淆 |
| 374 | Yes. | 对 |
| 375 | Your data is corrupted and your phone crashes. | 你的数据被破坏 你的电话崩溃。 |
| 376 | And the industry calls this problem what? | 行业把这个问题称为什么? |
| 377 | They call it crosstalk. | 他们称之为交叉说话。 |
| 378 | Crosstalk. | 交叉说话 |
| 379 | OK. | 摆 |
| 380 | So Micron decided they could not build the walls any thinner. | 所以Micron决定他们不能建造更薄的墙。 |
| 381 | The crosstalk was just too bad. | 交叉说话太糟糕了 |
| 382 | They believed the factory yields would never get high enough to make a real profit. | 他们相信工厂的产量 永远无法达到真正的利润 |
| 383 | So they threw the tight map in the trash? | 他们把紧凑的地图扔进垃圾桶? |
| 384 | Entirely. | 彻底. |
| 385 | But wait, if they refuse to pack the memory tighter on the flat ground, | 但是等等 如果他们拒绝在平地上收紧记忆 |
| 386 | where do they go? | 他们去哪儿? |
| 387 | They still have to make better, faster chips for AI. | 他们还得为AI做更好,更快的芯片. |
| 388 | Yes, they do. | 是的,他们这样做。 |
| 389 | So Micron is moving in a completely different direction. | 因此Micron正在向一个完全不同的方向移动. |
| 390 | What are they doing? | 他们在干什么? |
| 391 | They want to use something called 3D DRAM. | 他们想用一个叫3D DRAM的东西 |
| 392 | 3D吗? | |
| 393 | This goes right back to our city building analogy. | 这可以追溯到我们的城市建筑类比。 |
| 394 | It does. | 没错 |
| 395 | If the old map is a regular neighborhood and CXMT is building dense, flat apartment blocks. | 如果旧地图是普通街区,CXMT正在建造密集,平坦的公寓街区. |
| 396 | Then Micron is building tall skyscrapers. | 然后Micron正在建造高大的摩天大楼. |
| 397 | Ah. | 摆 |
| 398 | Instead of packing the memory flat and side by side, | 而不是把记忆固定在一起 |
| 399 | they want to stack them vertically. | 他们想垂直堆放 |
| 400 | Exactly. | 没错 |
| 401 | They want to put one layer of memory directly on top of another layer. | 他们想把一层内存直接放在另一层的上方. |
| 402 | You go up into the sky. | 你上天来 |
| 403 | Right. | 对 |
| 404 | And by going up, you avoid the thin wall problem entirely. | 通过上楼,你完全避免了薄墙问题。 |
| 405 | Because the apartments are not squeezed next to each other anymore. | 因为公寓不再互相挤压 |
| 406 | You just build a taller building on the same piece of land. | 你只是在同一块土地上 建造了一座更高的建筑 |
| 407 | Yes. | 对 |
| 408 | It avoids the physics trap of crosstalk. | 它避免了交叉说话的物理陷阱. |
| 409 | OK. | 摆 |
| 410 | So the US says build tall skyscrapers. | 美国说建造高大的摩天大楼 |
| 411 | China says pack the apartments tight on the ground. | 中国说把公寓收拾好 |
| 412 | Right. | 对 |
| 413 | What about South Korea? | 韩国呢? |
| 414 | Because they produce a massive portion of the world's memory chips. | 因为他们生产了大量的世界记忆芯片. |
| 415 | They really do. | 他们真的有。 |
| 416 | South Korea has giants like Samsung and SK Hynex. | 韩国拥有三星和SK Hynex等巨型. |
| 417 | And what are they doing? | 他们在做什么? |
| 418 | They are playing a very safe, cautious game. | 他们正在玩一个非常安全,谨慎的游戏。 |
| 419 | They are watching the global experiment play out before they spend their money. | 他们在花钱之前 正在观看全球实验的进行 |
| 420 | They are waiting to see who wins. | 他们等着看谁赢了 |
| 421 | Pretty much. | 相当多。 |
| 422 | The source notes that Samsung is keeping its options open. | 消息人士指出,三星保持了自己的选择权. |
| 423 | Oh, really? | 是吗? |
| 424 | Yeah. | 对 |
| 425 | They actually built a working 4F2 test chip back in February. | 他们实际上在2月份建造了一台有效的4F2测试芯片. |
| 426 | So they prove their engineers can handle the complex, tight map. | 所以他们证明他们的工程师可以处理复杂的,紧凑的地图. |
| 427 | Yes. | 对 |
| 428 | Samsung proved the science works in a lab. | 三星证明了科学在实验室的工作。 |
| 429 | But SK Hynex give a very interesting quote about the business side of things. | 不过SK Hynex给出了一个非常有趣的引文,提到事物的商业方面. |
| 430 | What did they say? | 他们说什么? |
| 431 | SK Hynex said their main goal right now is to keep pushing the old standard 6F2 map for as long as they possibly can. | SK Hynex说,他们现在的主要目标是尽可能长时间地推动旧的标准6F2地图. |
| 432 | Oh, they want to stick with the regular neighborhood. | 哦,他们想坚持 与普通的邻居。 |
| 433 | Yes. | 对 |
| 434 | They want to keep using the advanced Dutch EUV printers to just refine the old design. | 他们希望继续使用先进的荷兰EUV打印机来完善旧的设计. |
| 435 | They are perfectly happy where they are. | 他们很快乐 |
| 436 | Exactly. | 没错 |
| 437 | The source has a great quote from SK Hynex executives talking to their investors. | 消息来源引用了SK Hynex公司高管与投资者交谈时的一段好话。 |
| 438 | OK. | 摆 |
| 439 | They said, quote, the timing of technological development and the timing of commercialization are distinct from each other. | 他们说,引文指出,技术发展的时间和商业化的时间是不同的。 |
| 440 | Let us translate that into plain English. | 让我们将它翻译成简单的英语。 |
| 441 | So technological development means building one amazing, perfect chip in a clean laboratory. | 因此,技术发展意味着在干净的实验室里 建立一个惊人的,完美的芯片。 |
| 442 | You prove that the science is possible. | 你证明科学是可能的 |
| 443 | Exactly. | 没错 |
| 444 | But commercialization means building 10 million of those chips in a normal factory and actually making a profit. | 但是商业化意味着在正常的工厂里建造1000万个这样的芯片,并实际盈利. |
| 445 | Right. | 对 |
| 446 | Just because you can build it in a lab does not mean you can sell it in the real world and make money. | 仅仅因为你可以在实验室里建造它 并不意味着你可以在现实世界里出售它和赚钱。 |
| 447 | Exactly. | 没错 |
| 448 | SK Hynex说他们会等的 | |
| 449 | They will let CXMT and Micron fight it out and spend all the money. | 他们会让CXMT和Micron去对抗 花掉所有的钱 |
| 450 | Right. | 对 |
| 451 | They will only switch to the tight map or the tall skyscraper map when the business math is totally safe. | 只有商业数学完全安全时,他们才会切换到紧凑的地图或高大的摩天大楼地图. |
| 452 | Wow. | 珃 |
| 453 | It is a massive three-way split in the industry. | 这是行业的大规模三重分裂。 |
| 454 | It really is. | 没错 |
| 455 | You have the U.S. pushing for vertical 3D memory. | 美国正在推动垂直3D记忆 |
| 456 | You have China pushing for super tight flat memory. | 你让中国追求超紧的记忆 |
| 457 | And you have South Korea sitting in the middle squeezing the last drops of profit out of the old method. | 你让韩国坐在中间 从旧方法中挤出最后的利润 |
| 458 | Yes, that is the perfect summary. | 是的,这是完美的总结。 |
| 459 | I want to look at CXMT's chances here, though. | 我想看看CXMT在这里的机会 |
| 460 | OK. | 摆 |
| 461 | If a giant like Micron fought the laws of physics for 20 years and lost, is CXMT just doomed? | 如果一个像Micron这样的巨人 与物理定律战斗了20年却失败了 CXMT是不是注定要失败了? |
| 462 | That is the big question. | 这是一个大问题。 |
| 463 | I mean, are they throwing 11 billion yuan at an impossible science problem just because the trade rules forced them to? | 我的意思是,他们扔110亿人民币 对于一个不可能的科学问题 只是因为贸易规则强迫他们? |
| 464 | Well, CXMT actually talked about this skepticism directly in the article. | 实际上,CXMT在文章中直接谈到这种怀疑论. |
| 465 | They did. | 没错 |
| 466 | What was their defense? | 他们的辩护是什么? |
| 467 | They argued that every company is different. | 他们认为,每个公司都是不同的。 |
| 468 | Just because Micron failed does not mean the physics are impossible to beat. | 仅仅因为Micron失败并不意味着物理学是不可能战胜的. |
| 469 | Right. | 对 |
| 470 | CXMT pointed out that they have different engineers. | CXMT指出他们有不同的工程师. |
| 471 | They have different patents. | 他们有不同的专利。 |
| 472 | They have entirely different manufacturing processes. | 它们的制造工艺完全不同。 |
| 473 | And their motivation is completely different, too. | 他们的动机也完全不同 |
| 474 | Exactly. | 没错 |
| 475 | That is a huge point. | 这是一个巨大的问题。 |
| 476 | Micron奢侈地放弃了 | |
| 477 | Yeah, Micron can just buy the newest Dutch printers and stick to the old map if they want to. | 是的,Micron可以买到最新的荷兰打印机 并坚持老地图,如果他们愿意。 |
| 478 | Right. | 对 |
| 479 | But CXMT does not have a backup plan. | 但CXMT没有备份计划. |
| 480 | The dense map is their only way forward. | 密集的地图是他们唯一的前进道路。 |
| 481 | Yes. | 对 |
| 482 | And when survival is on the line, engineers tend to find very creative solutions to physics problems. | 而当生存在线上,工程师往往会找到非常有创意的解决物理问题的办法. |
| 483 | They are forced to innovate or die. | 他们被迫创新或死亡。 |
| 484 | Exactly. | 没错 |
| 485 | That desperation changes the math on the factory floor. | 绝望改变了工厂地板上的数学 |
| 486 | They might just accept a terrible profit margin for years just to establish their independence. | 他们可能只是接受一个可怕的利润率 几年来仅仅建立他们的独立性。 |
| 487 | Yes. | 对 |
| 488 | They probably will. | 他们可能会的。 |
| 489 | So let us bring this all back to you, the listener. | 所以让我们把这些都带回给你,听者。 |
| 490 | Why does all this matter? | 为什么这一切都很重要? |
| 491 | Exactly. | 没错 |
| 492 | Why does this complex fight over chip layouts matter to your daily life? | 为什么这种复杂的关于芯片布局的争吵 对你日常生活很重要? |
| 493 | Why should you care if the memory inside your phone is packed flat or stacked high? | 你为什么在乎手机里的记忆是平板的还是高堆的? |
| 494 | You should care because this memory is the foundation of the modern internet. | 你应该关心,因为这个记忆 是现代互联网的基础。 |
| 495 | Right. | 对 |
| 496 | The DRAM chips are the fast desks where all the world's processing happens. | DRAM芯片是世界上所有处理的快台. |
| 497 | Every single time you ask an AI on your phone to write an email or make an image or translate a language. | 每次你要求AI在你的手机上写邮件或制作图像或翻译语言. |
| 498 | It requires huge amounts of fast memory. | 它需要大量的快速记忆。 |
| 499 | The software is getting incredibly heavy. | 软件越来越重了 |
| 500 | And if the hardware cannot keep up, the whole AI revolution hits a brick wall. | 如果硬件无法跟上,整个AI革命就撞上了砖墙. |
| 501 | Exactly. | 没错 |
| 502 | The company that wins this layout war will control the future supply chain. | 赢得这场布局战的公司将控制未来的供应链. |
| 503 | If CXMT makes the dense map cheap and reliable, Chinese chips will flood the global market. | 如果CXMT让稠密的地图便宜可靠,中国芯片就会淹没全球市场. |
| 504 | And if Micron masters the 3D skyscraper, the US will hold the advantage in fast AI servers. | 而如果Micron掌握3D摩天大楼,美国将在快速AI服务器中占据优势. |
| 505 | The winner dictates how fast your next laptop runs. | 赢家决定你下台电脑的运行速度 |
| 506 | Right. | 对 |
| 507 | They decide how powerful your phone battery needs to be. | 他们决定你的手机电池需要多大的威力 |
| 508 | And they ultimately decide how much you pay for your devices. | 他们最终决定你为设备付多少钱。 |
| 509 | This invisible war over tiny city maps changes the global economy. | 这场针对小城市地图的无形战争改变了全球经济。 |
| 510 | It really does. | 确实 |
| 511 | And I want to leave you with a final thought to chew on. | 我想给你留一个最后的想法 咀嚼。 |
| 512 | Okay. | 摆 |
| 513 | What is it? | 这是什么? |
| 514 | This entire situation started because of strict trade limits. | 整个局面之所以开始,是因为严格的贸易限制。 |
| 515 | Right. | 对 |
| 516 | The bans. | 禁令。 |
| 517 | The rules were designed to slow a company down. | 规则是为了让公司减速 |
| 518 | But we are seeing a really strange side effect. | 但是我们看到一个非常奇怪的副作用。 |
| 519 | What do you mean? | 你说什么? |
| 520 | Well, by taking away the standard easy tools, you force a company to redraw the map entirely. | 通过拿走标准的简单工具 迫使公司重新绘制地图 |
| 521 | You force them to learn the hard way. | 你强迫他们学习艰难的方法。 |
| 522 | Exactly. | 没错 |
| 523 | It reminds me of learning to do math without a calculator. | 它提醒我学会数学,没有计算器。 |
| 524 | Right. | 对 |
| 525 | It is super painful at first. | 一开始很痛苦 |
| 526 | You make a lot of mistakes. | 你犯了很多错误。 |
| 527 | Your yield is terrible. | 你的回报是可怕的。 |
| 528 | Yes. | 对 |
| 529 | But eventually, you understand the numbers on a much deeper level than the person who | 但最终,你理解的数字 更深层次的人 |
| 530 | just pushes the easy buttons. | 只是按一下简单的按钮。 |
| 531 | Wow. | 珃 |
| 532 | That is a great point. | 这是一个伟大的观点。 |
| 533 | So the big question for the next decade is this. | 所以下一个十年的大问题是这个。 |
| 534 | Do strict limits actually cripple a company? | 严格的限制真的使公司瘫痪吗? |
| 535 | Or do they forge a much stronger, more dangerous innovator for the future? | 还是他们造了一个更强大、更危险的革新者? |
| 536 | If CXMT solves the physics problem that defeated Micron, they won't just catch up. | 如果CXMT解决了击败Micron的物理问题,他们不会只是追赶. |
| 537 | They might actually leap ahead. | 他们可能会跳跃向前 |
| 538 | And it all started because they were legally banned from buying a crane. | 这一切的开始是因为他们被合法禁止购买起重机. |
| 539 | That is fascinating. | 这是令人着迷的。 |
| 540 | Thank you for joining us on this deep dive into the source material. | 谢谢你加入我们 深入到源材料中 |
| 541 | We hope this changes how you look at the tiny, invisible cities | 我们希望这能改变你对隐形小城市的看法 |
| 542 | ricking hard inside your phone. | 用力地在你的电话里。 |
| 543 | Until next time, keep exploring. | 直到下次,继续探索 |
该视频共有字幕 544 条。解锁更多字幕为会员功能，请移动到 价格