那真是太糟糕了；一场噩梦。上周不仅仅是一场惨败；而是一次彻底的毁灭。我们注意到许多连续纪录和里程碑 — — 半导体连续九天下跌。天哪。CoreWeave回到了上次暴跌时的水平。哦不。内存崩溃。这不可能。但事实是，市场的表现，尤其是周五，就像数据中心故事，那个所谓的1.2万亿美元或12万亿美元“超级周期” — — 本身就是一个吻别死亡的术语，就像煤炭超级周期或水力压裂超级周期一样 — — 已经结束。
Boy, that was awful; a nightmare. Last week wasn't just a rout; it was an obliteration. We had a lot of streaks and milestones noted — semiconductors, down nine straight days. Egads. CoreWeave back to where it was when it plummeted last time. Oh no. Memory implosion. It can't be. But the simple truth is that this market acted, especially Friday, as if the data center story, the $1.2 trillion or $12 trillion "super-cycle" — in itself a kiss-of-death term, like the coal super-cycle or the fracking super-cycle — is over.
有史以来最伟大的股票故事已经结束；你最好把每一分钱都拿出来，买入高露洁-棕榄、可口可乐、宝洁公司，以及希望中的Kimberly-Clark。至于我们需要的那些AI工厂？“荒凉之路。”鲍勃·迪伦，说这不是真的。这一切怎么会发生得这么快？买家们都去哪儿了？我有一些答案。反应错了？让我们先从弄清楚原因之前的反应开始，因为它可能是错误的反应。
That the greatest stock story ever told is over; and you better take every penny and buy Colgate-Palmolive , Coca-Cola , Procter & Gamble and, hopefully, Club name Kimberly-Clark . And all of those AI factories we need? "Desolation Row." Bob Dylan, say it isn't so. How did this happen so quickly? Where were the buyers? I have some answers. Wrong reaction?
当我看待事态的分解时，上周被抛售的集中度， culminated in Friday's devastation，它与数据中心/AI论点的所有组成部分几乎完美地吻合了 — — 这个论点代表了市场市值的约25%，尽管不是其绝对百分比，而其绝对百分比远小于这个数字。为什么是反应而不是原因？马车放在前面？不，我不这么认为。我相信，我们会发现上周发生的事情是某些公司头寸的平仓，加上一个我们从未听说过的对冲基金的崩溃，该基金从一家尚未披露自身亏损的公司借了数十亿美元，而我们不知道这笔借款。
Let's start by figuring out the reaction ahead of the cause because it might be the wrong reaction. When I look at the breakdown, the concentration of what was being sold last week, culminating in Friday's devastation, it lined up rather perfectly with all of the components of the data center/AI thesis that represents about 25% of the market's market cap, albeit not its absolute percentage, which is far smaller than that. Why the reaction and not the cause? Cart before the horse? No, I don't think so.
我可以这样争论，就像我那位伟大的朋友和TheStreem最初的雇员Herb Greenberg所喜欢说的那样：这是由债务驱动的超规模骗局的终点站。
I believe that what happened last week we will find is an unwind of certain firms' positions coupled with the collapse of some hedge fund we have never heard of that borrowed billions that we don't know about from a firm that hasn't disclosed its own losses yet for lending the manager the money.
万亿美元的骗局，源自英伟达CEO黄仁勋被诱导的胆识，让他决定给每个人发钱，以便他们能够购买英伟达最新、最强大的图形处理单元（GPU），这导致了CoreWeave的公司债券收益率飙升。如今，边际买家正被残酷的债券市场挤出。请保留判断，不要急于评判黄仁勋。这让我想到我们在投资生活中每天都要面对的两种叙事。
I could argue that, as my great friend and original TheStreem hire, Herb Greenberg, is fond of: This is the end of the line for the debt-fueled hyperscale-hoax. The trillion-dollar con, spawned by Nvidia CEO Jensen Huang's induced moxie to give everyone money so they could buy the latest and greatest graphic processing units (GPUs) from Nvidia, led to CoreWeave's spike in corporate bond yields. That the marginal buyer is now being margined out by the cruel bond market. Reserve judgment Don't be so quick to judge Jensen.
它们值得被解读，也值得被批判性地分析。让我们以上周的行动作为范例。当黄仁勋首次接触到加速计算的概念时，他意识到这可能是实现某些本来并不存在事物所需的催化剂：即巨大的、即时的模式识别，这使得机器能够获得思考的能力。对一切事物进行即时模式识别。他的朋友、OpenAI的山姆·阿尔特曼看到了从那些需要人工智能的人身上获利的途径。
Which brings me to the two narratives that we live with every day of our investing life. They are worth parsing. They deserve critical analysis. Let's use last week's action as a paradigm. When Jensen first stumbled on the concept of accelerated computing, he realized that it could be the accelerant needed for something that really didn't exist: gigantic, instant recognition of patterns that unlocked the ability for machines to adopt thinking.
你会认为每个人都会意识到人们会为这项技术付费，但实际上只有少数人这样做了——而且至今仍是如此。大多数大型科技公司只是考虑后端、云端、管道以及为想要通过结果收费的人提供服务。如今，只有OpenAI、Anthropic以及在较小程度上通过谷歌Gemini持有投资组合的Alphabet实现了对这一趋势的规模化资本化。
Instant pattern recognition of everything. His pal, Sam Altman of OpenAI, saw ways to profit from those who need intelligence artificially. You would think that everyone would have recognized that people would pay for that, but only a few did — and do.
但 boy，哦 boy，他们确实在这方面大获成功。不过目前，鉴于OpenAI和Anthropic尚未上市，存在一个主要由英伟达资助的活动宇宙——英伟达是一个久负盛名的俱乐部成员——这使得人工智能、生成式人工智能，以及如今的代理型人工智能得以蓬勃发展。如果没有英伟达提供的种子资金，也许这终究会发生，只是时间问题。
Most big tech companies just thought of the back end, the cloud, the pipes, the service to those who wanted to charge for outcomes. As it stands now, only OpenAI, Anthropic and, to a lesser extent, portfolio holding Alphabet via Google's Gemini have capitalized on the movement at scale. But boy, oh boy, have they capitalized on it.
黄仁勋想要获胜。他希望英伟达的股东获胜。他希望美国获胜。因此，他打开了自己的金库，并以风险投资基金帮助孵化新企业的方式资助了这些公司。
Right now, though, given that neither OpenAI nor Anthropic has gone public, there is a universe of activity mostly funded by Nvidia, a longtime Club name, that has allowed the artificial and generative artificial, and now agentic artificial intelligence, to flourish. If Nvidia hadn't provided the seed money, maybe it would have happened eventually. Jensen wants to win. He wants Nvidia shareholders to win.
怀疑者、批评者、那些依赖事物崩溃、中心不稳的人，只是将詹森的活动视为朗讯的一个巨大遗产——朗讯是世纪之交的电信设备公司，它通过为其客户融资，掩盖了事物实际上有多脆弱，直到明显看出，没有这种融资，几乎没有什么业务可做。或者说，正如格林伯格及其追随者所热情相信的那样，詹森正在玩转盘游戏。
He wants the United States to win. So, he opened his vault and helped fund companies the way venture capital funds help start new enterprises.
他把钱放在桌上，转动它，你拿走它，然后你从他那里购买产品，大家都很开心。我曾站在转盘桌的另一边——无论是在我母亲端上火鸡和肉汁的时候，还是当我经营TheStreem（我的第一份工作）的时候，我的第一反应就是吹哨揭露这种活动，于是我这么做了。这显然是肮脏的。我不需要那些货物。我不想要它们。我也不想要那虚假的投资。
The skeptics, the critics, the people who depend on things falling apart and the center not holding, just view Jensen's activities as one big legacy of Lucent, the telco equipment company from the turn of the century that did the most to mask how weak things really were by financing their one world of clients until it was obvious that there was very little business to be had without it. Or, to put it in a way Greenberg and his acolytes believe, passionately, Jensen is playing a game of Lazy Susan. He puts the money on the table, he spins it, you take it, and then you buy product from him, and everyone's happy. Having been on the other side of the Lazy Susan table, both when my mother served the turkey and the gravy and when I ran TheStreem, my first instinct was to blow the whistle on the activity, and so I did. It was so obviously dirty. I didn't need the goods.
现在不像那时了。那时和现在的区别是什么？那时，大约在2000年，很明显，所谓的“供应商融资”仅仅是在惠及供应商，因为我们根本不需要供应商所提供的最终产品。这就是起诉书的真正核心，再说一遍。英伟达只不过是朗讯的再来，而詹森只不过是里奇·麦金恩——那个总是通过日益荒谬的新客户来完成季度目标的骗子，而你从未梦想过这些客户竟有钱进行这些采购。
I didn't want them. I didn't want the bogus investment, either. Now is not like then The difference between back then and now? Back then, circa 2000, it was obvious that what was known as "vendor finance" was simply benefitting the vendor, as we didn't need the end product of what the vendor was providing. That's the actual crux of the indictment, again.
巴西电信公司就是例子。但当然，他们并没有钱，这只不过是朗讯配上了一笔虚幻的银行贷款。最后，你会疑惑：到底有什么被购买了，还是钱根本就没有离开转盘？它只是被原地旋转回来。如果你认为当下的这一切建筑毫无价值或价值远低于表面，那么你就必须相信朗讯的范式。它如此完美、如此优雅，以至于它落入了做空的逻辑之中：好得不像真的，不值得去查证。
Nvidia is no more than the second-coming of Lucent, and Jensen is simply Rich McGinn, the charlatan who always made the quarter with increasingly preposterous new clients that you never dreamed had the money for the purchases. Brazilian Telco. But of course, they didn't, and it was just Lucent with a chimerical bankroll. In the end, you wondered, was anything even purchased, or did the money never even come off the Lazy Susan at all; it was just spun right back. You have to believe the Lucent paradigm if you think that all of this current-day edifice is worthless or worth a lot less. It fits so perfectly, so elegantly that it comes under the short-selling rubric: too good to check out.
这一切的美妙之处在于：整个建设进程真正启动，源自迈克尔·因特拉托尔（Michael Intrator），一名前比特币矿工，现任CoreWeave首席执行官，他从英伟达那里拿到了数十亿美元，作为交换，他购买了大量英伟达芯片来建设数据中心，起初主要服务于微软，但如今则为大多数超大规模云服务商提供支持。根据空头的说法——他们只是想让格林伯格（Greenberg）闭嘴，因为他只是最爱发声的负面叙事者——这一整套循环操作只有在CoreWeave由一个该死的比特币矿工领导的情况下才能奏效，即它能够说服足够多的公共债务客户注资，以维持其在懒人转盘（Lazy Susan）上的份额。
The beauty of it all: the whole build-out really got going with Michael Intrator, a former bitcoin miner turned CEO of CoreWeave, taking billions of dollars from Nvidia in return for buying a ton of Nvidia chips to put up data centers, first largely for Microsoft, but now for most of the hyperscalers. This whole circular song, according to the bears — to let up on Greenberg, who is simply the most vocal of the negative narrators — only worked if CoreWeave, run by a darned bitcoin miner, was able to con enough public debt customers into giving it money to hold up its side of the Lazy Susan.
这几乎就像诗在运动。但本周，这首诗戛然而止，因为CoreWeave的债务收益率飙升至15%以上，此时CoreWeave的支出回报率太低，无法证明其建设的合理性。如果没有新建设，那就不会有新的英伟达订单；如果没有新的英伟达订单，那就就是游戏结束。这就是故事情节。尽管周四出现了不错的反弹，但这一情节在整个星期都得以维持。
It's almost poetry in motion. The poetry stopped this week when CoreWeave's debt yields blew up to over 15%, where the rate of return on CoreWeave's spend is too low to justify the build. If there is no new build, then there are no new Nvidia orders. If there are no new Nvidia orders, then it is game over. That was the storyline.
周五的崩盘抹去了周四的反弹，一些人认为这预示了周一早上将发生的事情。周五的交易陷入了一种“立刻撤离”的局面，除了菲利普莫里斯（Phillip Morris）、默克（Merck）以及强生公司（Johnson & Johnson）的股东之外，几乎所有人都在这种情绪中。但冒着可能被证明是正确的风险，让我来提供一个替代性的叙述。
It held up throughout the week even as there was a nice bounce from Thursday's bottom. The collapse on Friday that wiped out the Thursday reversal foretold, some would say, what happens Monday morning. Friday's session devolved into a case of "get out now" for all but the owners of Phillip Morris, Merck , and Club name Johnson & Johnson .
周四的反弹源自 Jensen 在被问及如何应对CoreWeave时所说的话。他的回答很简单：不需要做任何事情；事实上，情况比我们想象的要好。他的回答很妙：任何其他措辞，如果暗示 Jensen 准备充当后盾——就像他是“英伟达的联邦储备”——那么我们就会知道懒人转盘的桌子仍在运转，而 Jensen 对如今无所不能的批评者们浑然不知。
But, at the risk of being right, let me offer you the alternative narrative. The bounce Thursday came about from something Jensen said when asked what to do about CoreWeave. The answer was simple: Don't need to do anything; in fact, things were better than we thought.
当我在洛杉矶担任谋杀案记者时，我曾报道过瓜镇大屠杀。我曾与那些饮用原始Kool-Aid而遇难者的亲属交谈。
The answer was brilliant: any other wording that said Jensen was ready to backstop — like he was the "Federal Reserve of Nvidia" — and we would know the Lazy Susan table was still in play and Jensen was blissfully ignorant of the now all-powerful critics. I covered the Jonestown mass murder when I was a homicide reporter in Los Angeles.
所以叫我一个了解Kool-Aid的人，知道它是什么样子，它对你有什么作用。英伟达并不销售Kool-Aid。黄仁勋拥有的东西，既不是吉姆·琼斯所拥有的，也不是空头目前所拥有的：可信度。事实是，虽然公共债务市场——也就是CoreWeave需要不断扩大规模的那个市场——目前可能被暂停，但黄仁勋知道一件被仇恨者反复否认的真相：英伟达的芯片经得起考验。
I talked to the relatives of the deceased who drank the original Kool-Aid. So call me someone who knows the Kool-Aid, what it looks like, what it does to you. Nvidia does not market Kool-Aid. Jensen has something that neither Jim Jones nor the bears now had and have: credibility.
它们可能能够保持价值长达10年。这对英特拉来说是天赐之福，因为黄仁勋正以大约五年的期限将这些芯片作为抵押进行贷款。当这段时间结束时，他会收回这些芯片。如果它们的价格能够保持稳定，而且确实如此——他可以卖掉它们；他可以再次租赁它们。毫无疑问，它们经得起考验。有真相的基准。不久的将来，这些东西将会出现双向市场，就像抵押贷款支持债券和汽车贷款支持债券一样。
The fact is, while the public debt market, the one CoreWeave needs to get bigger and bigger, may be halted for now, Jensen knows something to be true that is denied repeatedly by the haters: Nvidia's chips hold up. They may hold their value for 10 years. That's a godsend for Intrator because Jensen is lending against them for about five years. When that period is up, he takes them back. If their prices hold up and they have — he can sell them; he can lease them again. There is no doubt that they hold up. There are benchmarks of truth.
贷款市场现在已经知道这一点。英特拉已经以远低于公共市场提供的利率对这些芯片进行了抵押贷款。说公共市场错了或被操纵了，或者两者都是，这种说法很少见。但我认为CoreWeave正是如此，因为它是论点的关键——即建设尚未结束。要么你相信它，要么你不相信。如果你相信它，你可能会在周一早上被诱惑去买入，因为该集团目前被严重超卖。
There will be a two-way market in these things soon, like mortgage-backed and auto loan-backed bonds. The loan market knows that now. Intrator has borrowed against them already at much lower rates than the public markets offer. It is rare to say the public markets are wrong or manipulated or both. But I think that's the case with CoreWeave, as it is the linchpin to the thesis that the buildout isn't at its end. You either believe it, or you don't.
如果你认为这一切都是一个巨大的骗局，那么就做空或者卖出。结论是，我不认为这是一个骗局。让我再说一遍：我不认为这是一个骗局。CoreWeave不是朗讯。英伟达不是当时的霸主供应商诺特尔。我们可以后退一步说，这只是在数据中心的支持者与反对者之间巨大战争中的一场小冲突。如果数据中心在中期选举中真的落后到连得克萨斯州和宾夕法尼亚州都不想要它们的程度，那么大战将在这次中期选举中爆发。
If you believe it, you are tempted to buy Monday morning as the group is ridiculously oversold. If you think it is all one big sham, then short away or sell. Bottom line I don't think it's a sham. Let me repeat that: I don't think it's a sham. CoreWeave is not Lucent. Nvidia is not Nortel, the almighty vendor of that era. We can step back and say this is just some skirmish in the giant war for or against the data center.
上周的情况看起来像是市场已经耗尽的一记重击，而选举正在对数据中心不利。我认为两者都不是事实。
The big battle will come with the midterm election if it turns out that data centers really fall behind in the election to the point that even Texas and Pennsylvania don't want them. Last week made it look like a one-two punch of the market's tapped out and the election is going against the data center.
我认为市场正在进行一次应得的喘息，并且总会有地方欢迎数据中心。我认为最终的建设规模将达到10万亿美元到12万亿美元，就像摩根士丹利首席执行官泰德·皮克在CNBC记者莱斯莉·皮克的出色采访中所说的那样。此外，我认为Anthropic和OpenAI都将上市，任何需要获得报酬的人都会得到报酬。而且，我认为禁止数据中心的州将会后悔他们当时拒绝的所有资金——这些资金如今将流向尚未投票或已投赞成票的其他州。
I think that neither is the case. I think the market is taking a well-deserved breather, and there will always be localities that welcome data centers. I think that the ultimate build-out will come to $10 trillion to $12 trillion, like Morgan Stanley CEO Ted Pick threw out in CNBC Leslie Picker's excellent interview . Moreover, I think that both Anthropic and OpenAI will come public, and anyone who needs to get paid will get paid.
超大规模云服务提供商将提供的交易令人震惊。事情已经形成闭环：我的观点有许多信徒。太多了。因为显然，其中许多人借钱来进行这项投资——不，是为了下这个赌注。他们甚至可能不理解把我们带到如今这一步的复杂链条。唉，他们就是问题所在。是对冲基金出了问题，而不是范式出了问题。如果我是对的，那么这团乱子将很快被理清；那些头寸在他们之下被平仓的大型对冲基金将暴露出来。
And, I think that the states that ban the data centers will rue the day they turned down all the money that is now going to flow to the remaining states that haven't voted or voted in favor. The deals the hyperscalers will offer are insane. Full circle: My view has many believers. Too many. Because obviously many of them borrowed money to make, not that investment, but that bet. They may not even understand the sophisticated chain that gets us to where we are now. They are, alas, the problem. Broken hedge funds, not broken paradigms.
CoreWeave的债务价格将下降，因为它一直通过购买信用违约互换被人为推高；甚至甲骨文的股价也会上涨。不要得意忘形。熊市也需要赚取他们的一年。但要意识到英伟达的芯片能用多久。要明白，不久将出现一个公开交易算力的市场，因为贝莱德——作为全球最大的资产托管机构——已经促成了这一点。要认识到，生成式人工智能无论对人类构成何种威胁，都将取得胜利。
If I am right, then the morass clears up pretty quickly; the bigger hedge funds that had their positions sold out underneath them will be revealed. The price of CoreWeave's debt goes down, as it was always juiced by credit default swap buying ; and even Oracle's stock goes up. Don't get complacent. The bears need to make their year, too. But be aware of how long Nvidia's chips last. Understand that there will soon be a public market trading compute because BlackRock , the largest repository of assets, has willed it to be.
并且要接受历史的教训：那些对第四次工业革命过于激进、资本不足或缺乏竞争力的人——就像人们常用的1870年代铁路类比——最终未能到达应许之地，将成为历史的注脚。()
Recognize that generative artificial intelligence, no matter what the threat to mankind, will triumph. And accept that history will be littered with people who thought about the fourth industrial revolution too aggressively or were thinly capitalized or not competitive — like the overused rail analogy from the 1870s — and didn't make it to the promised land. ()