处于人工智能泡沫核心的公司

拥有智能家居是个很酷的想法，但也让人望而生畏。

Having a smart home is a cool idea, but kind of a daunting prospect.

你得考虑买哪些设备，如何把它们全部连接起来。

You have to figure out which devices to buy, how to connect them all together.

这一切都太复杂了。

It's all just a lot.

但在接下来两周的Vergecast节目中，我们将尝试简化整个

But for two weeks on the Vergecast, we're trying to simplify all

过程。

of it.

我们会花时间解答你们关于智能家居的所有问题，然后逐个房间探索一栋真实的房子——我自己的房子，尝试让它变智能，并让这些智能功能变得合理实用。

We're gonna spend some time answering all of your questions about the smart home, and then we're gonna go room by room through a real house, my real house, and try to figure out how to make it smart and how to make all of that smart make sense.

以上内容及更多精彩尽在Vergecast，各大播客平台均可收听。

All of that and much more on the vergecast wherever you get podcasts.

本特别系列由家得宝赞助播出。

This special series is presented by The Home Depot.

大家好，欢迎来到《解码器》。

Hello and welcome to Decoder.

我是《The Verge》主编伊莱·帕特尔，《解码器》是我主持的一档探讨重大理念与其他难题的节目。

I'm Eli Patel, editor in chief of The Verge, and Decoder is my show about big ideas and other problems.

很多人认为AI是个泡沫。

So a lot of people think AI is a bubble.

包括OpenAI CEO萨姆·奥特曼在内的许多人——他一边四处奔走告诉所有人他认为AI是泡沫，一边却在筹集和挥霍着在旁人看来正是泡沫标志的巨额资金。

A lot of people, including people like OpenAI CEO Sam Altman, who keeps running around telling everyone that he thinks AI is a bubble, all the while raising and spending enormous amounts of money that seem like bubble indicators to everyone else.

这一切真是令人困惑。

This is all pretty confusing.

于是我们派出了《The Verge》的高级记者Liz Lopatto，去调查这个所谓的AI泡沫——它是否真实存在，可能如何破裂，以及这一切意味着什么。

So we set Verge senior reporter Liz Lopatto out to report on the supposed AI bubble, whether it's real, how it might pop, and what all this might mean.

而她今天加入节目，要讨论的正是处于这场风暴中心的一家公司。

And she's joining the show today to talk about one particular company that sits right in the middle of all of it.

这家公司名叫Core Weave。

That company is called Core Weave.

Liz花费了大量时间深挖它的历史、财务状况，以及这个精彩故事所揭示的现代AI热潮真相。

And Liz has spent considerable time digging into its history, its financials, and the truly fascinating story that all of it tells us about the modern AI boom.

实际上Liz下周将在《The Verge》发布关于Core Weave的重磅报道。

Liz actually has a big story about Core Weave running next week on the verge.

所以听完本期节目后若想获取更多细节，敬请关注那篇报道。

So if you want even more detail after you listen to this episode, just stay tuned for that.

但对我来说，Core Weave故事的基本脉络本身就足够引人入胜。

But to me, the basics of the Core Weave story are themselves fascinating.

Core Weave由前大宗商品交易员创立，最初是新泽西州一家从事加密货币的以太坊挖矿公司。

Core Weave was founded by former commodities traders as an Ethereum mining firm out in New Jersey in crypto.

就像科技行业的许多公司一样，当加密货币遇冷时，CoreWeave转向AI领域，开始将囤积用于以太坊挖矿的GPU转而建设自己的数据中心业务。

Like so much of the tech industry, when crypto sputtered, CoreWeave pivoted to AI and started to use the same GPUs it was stockpiling to mine Ethereum to build its own data center operation.

CoreWeave的最大创新在于融资模式——

CoreWeave's big innovation is creative financing.

它筹集巨额资金建设装满英伟达GPU的数据中心，然后将这些GPU回租给大型AI公司，以满足该行业对算力看似永不满足的需求。

It raises huge amounts of money to build data centers full of NVIDIA GPUs and then leases those GPUs back to big AI companies to feed the industry seemingly insatiable demand for compute.

这使得CoreWeave成为过去一年里颠覆市场的天价AI基础设施建设的支柱。

That's turned CoreWeave into a pillar of the obscenely expensive AI infrastructure build that's turned markets inside out over the past year.

CoreWeave本身在今年三月上市，如今市值已接近500亿美元。

CoreWeave itself went public this past March and it's worth almost $50,000,000,000 today.

仅今年一年，CoreWeave就与Meta、OpenAI等公司签署了价值数百亿美元的多份合同，为其数据中心提供访问权限。

And just this year alone, CoreWeave has signed multiple contracts worth tens of billions of dollars with companies like Meta and OpenAI for access to its data centers.

但Liz发现，CoreWeave的故事远比几个华尔街人士押对赌注一夜暴富要复杂得多。

But what Liz has found is that the story of CoreWeave is a lot more complicated than just a few Wall Street types making some smart bets and getting fully rich.

事实上，关于CoreWeave只是为AI淘金热提供‘铁锹’的叙事背后，隐藏着极其复杂的金融操作，以及与英伟达本身的特殊关系。

In fact, the narrative around CoreWeave, it's just making the picks and shovels of the AI Gold Rush, is built around some very complex financial maneuvering, as well as a unique relationship to NVIDIA itself.

Liz表示，从很多方面来看，如果没有英伟达的巨额资金投入，CoreWeave根本不可能存在。

In many ways, Liz says, CoreWeave could not exist without extraordinary levels of financial investment from NVIDIA.

也离不开AI行业对英伟达股票及其芯片的估值，以及AI未来可能带来的整个经济体系的全面转型。

And from the value the AI industry is putting on NVIDIA stock on its chip and the wholesale transformation of the entire economy that AI is supposed to deliver at some point in the future.

所以我想问Liz：CoreWeave如此依赖英伟达意味着什么？AI产业如此依赖CoreWeave又意味着什么？如果AI泡沫真的破裂，世界会发生什么？

So I wanted to ask Liz, what does it mean for CoreWeave to be so dependent on NVIDIA, for the AI industry to be so dependent on CoreWeave, and what might happen in a world where the AI bubble actually pops?

好的。

Okay.

接下来是我们的资深记者Liz Lopato关于CoreWeave和AI泡沫的报道。

We're our senior reporter Liz Lopato on CoreWeave and the AI bubble.

开始吧。

Here we go.

Liz Lapato，你是The Verge的资深记者。

Liz Lapato, you're a senior reporter here at the Verge.

你在The Verge这里还做过许多其他工作。

You also had many other jobs here at the Verge.

你报道了大量关于人工智能及其背后资金的内容。

You're reporting a lot on AI and the finances behind AI.

欢迎来到Decoder节目。

Welcome to Decoder.

你好。

Hi.

很高兴来到这里。

It's good to be here.

我非常期待和你讨论这个话题。

I'm very excited to talk to you about this story.

当你开始报道人工智能时，最终总会被问及是否存在泡沫。

When you start reporting on AI, you end up being asked if things are a bubble.

而当你开始报道泡沫时，很快就会直接关注到Core Weave。

And when you start reporting on a bubble, you land at Core Weave pretty directly.

这是你的经历，也是我的经历。

That is been your experience, it's been my experience.

向大家解释为什么讨论AI泡沫这个概念会让你如此迅速地联想到Core Weave。

Explain to people why even talking about the idea of an AI bubble leads you to Core Weave as quickly as it does.

你常听到的一个听起来非常聪明复杂的投资理论是：'哦，我们不是在直接投资AI公司'。

So one of the investment theses you hear that sounds really smart and sophisticated is, oh, we're not investing in, like, direct AI companies.

我们投资的是基础设施（镐和铲）。

We're investing in the picks and shovels.

就像淘金热时期卖镐和铲子，这确实比亲自去淘金更明智。

Like, selling the picks and shovels during a gold rush, that that does seem smarter than trying to go pan for gold.

我想，好吧。

I thought, okay.

那我们去看看数据中心吧。

Well, let's go look at the data centers.

这些基本上就是现代的镐和铲子。

Those are, like, essentially the picks and shovels.

对吧？

Right?

Coreweave的特点在于它是被称为'新云'的一类公司之一。

The thing about Coreweave is that it's one of a group of companies called a Neo Cloud.

他们基本上就是建造数据中心，然后出租给别人使用。

And they essentially just, like, make data centers that they then lease out to people.

由于最近上市，相比其他很多公司，我们能获得更多关于它的信息，这要归功于他们提交的那份漂亮的S-1文件。

Because it has gone public recently, there is more information available about it than there is about a lot of other things because of that beautiful s one they filed.

我当时在想，或许可以了解下数据中心的经济效益。

I was thinking, oh, I'll just get some thoughts about, like, data center economics.

人们能收回投资吗？

Can people get their investment back?

这个会有回报吗？

Like, is this gonna have a return?

然后我就掉进了一个坑里。

And then I fell into a pit.

这就是我们现在在这里的原因，因为我还没从这个困境中爬出来。

That was why we are now here because I still haven't climbed out of this pit.

让我先把这个放在更广阔的背景下，你可以向我解释CoreWeave与我们熟悉的云服务提供商有何异同。

So let me just put this in like broader context, and you can explain to me how CoreWeave is the same or different than the cloud providers we're familiar with.

亚马逊建立了AWS，它至今仍是整个游戏中的领先云服务提供商。

Amazon built AWS, it is still the leading cloud provider in the entire game.

谷歌建立了Google Cloud，微软则有Azure。

Google built Google Cloud, Microsoft has Azure.

在我看来，这些既关乎运行在数据中心上的软件和服务，也同样关乎芯片。

Those in my mind were as much about the software and services that ran on the data centers, as they were about the chips.

对吧？

Right?

如果你使用AWS，亚马逊会为你提供一系列基础服务，你可以非常快速地利用亚马逊的资源部署一项服务、一个应用或后台程序。

If you use AWS, Amazon builds you a bunch of primitives, can just stand up a a service, an application, something in the background very very quickly using Amazon's resources.

他们会管理需求，如果需要还会为你调配更多计算资源，亚马逊将为你接管数据中心的许多运营工作，这样你就能专注于自己的应用。

They will manage demand, they will manage rolling more compute at you if you need to, and that's Amazon is gonna take over a bunch of parts of running the data center for you, so you can just focus on your application.

谷歌和微软也是如此。

Same with Google, with Microsoft.

而他们的锁定优势就在于所有这些服务配置。

And their lock in was all of that service provisioning.

对吧？

Right?

他们会为你完成所有这些工作。

They're gonna do all this work for you.

CoreWeave只是卖芯片的，对吧？

CoreWeave is just selling chips, right?

它没有大型云服务商所拥有的中间层基础设施。

It doesn't have that middle layer of infrastructure that the big clouds have.

难道真的只是把芯片放在房间里租给你，然后你想怎么用就怎么用？

Is it really just we're gonna put the chips in a room and lease them out to you and you can do whatever you want?

还是说这家公司有其他护城河优势？

Or is there something else about this company that makes it have more of a moat?

实际情况比这稍微复杂些，但也没复杂太多。

It's a little more sophisticated than that, but not a lot more than that.

因为事实证明，芯片的连接方式对获得良好性能和稳定运行至关重要。

Because it turns out, like, the way that you connect the chips really matters in terms of being able to get good performance and good uptime out of them.

而且获取能源确实非常困难，这目前是AI发展的主要供应限制。

And it also turns out that it's really hard to get energy, and that is a major supply constraint on AI right now.

CoreWeave在获取能源方面有良好记录，同时工程能力较强，因此能对芯片租赁收取溢价。

CoreWeave has a pretty good track record of getting energy, and they also have relatively good engineering and can charge a premium for renting out their chips, essentially.

但这算不上宽阔的护城河。

But it's not a huge moat.

这也不一定是持久的护城河——如果他们失去现有的工程优势，就会陷入价格战的困境，而这是他们无法取胜的局面。

It's not necessarily a durable moat either because if they lose the engineering advantage they've got, then they are basically in a situation where they're competing on price, and that's not a situation where they can win.

对我来说这才是问题的核心。

That's really the heart of it for me.

亚马逊、谷歌和微软作为超大规模云服务商，除了价格之外还有很多竞争手段。

Amazon and Google and Microsoft, they compute to be these huge hyperscaler clouds, but they have many many ways to compete beyond just price.

CoreWeave似乎在价格上展开竞争，或许还包括数据中心建设、能源合同或优惠能源定价。

CoreWeave is competing on what feels like price, maybe data center build out or energy contracts or preferential energy pricing.

它的客户是谁呢？

Who are its customers for that?

这真有意思。

So this is fun.

他们的客户包括微软、英伟达、Meta以及其他许多需要额外计算资源的地方。

Their customers are Microsoft, NVIDIA, Meta, and a number of other places that need essentially overflow compute.

比如你和微软签了合同，结果微软可能根本没有能力执行你的合同。

So if you do a contract with, let's say, Microsoft, it may turn out that Microsoft just doesn't have the capacity to execute your contract.

于是他们就会把某些业务——可能是价值较低的部分——转移出去。

And so then they will push something, maybe something less valuable.

谁知道呢？

Who knows?

我不清楚他们如何做出这些决策。

I don't know how they make these decisions.

他们会把业务转给CoreWeave来执行合同。

They'll push it over to CoreWeave in order to execute the contract.

真正让我感兴趣的是，既然有这么多差异化的云计算提供商，而它们都决定需要标准化的计算资源，CoreWeave就是专门为它们提供这种标准化服务的。

That's the thing that really interests me, right, is that if you have all of these differentiated cloud computing providers, and they've decided they need a commodity in compute, CoreWeave just exists to provide that commodity to them.

难道没有其他供应商能提供这种标准化服务吗？

Can't something else provide that commodity to them?

他们不应该自己建数据中心吗？

Shouldn't they build their own data centers?

问题是，他们确实如此。

Well, they are, is the thing.

而这正是CoreWeave有趣的部分原因。

And this is part of what makes CoreWeave interesting.

对吧？

Right?

因为如果数据中心建设按我预期发展，他们的许多客户可能会变成竞争对手。

Because a lot of their customers may potentially become their competitors if the data center build out goes the way that I think it's going to go.

这确实暗示了CoreWeave面临的一个长期问题：目前他们的大客户都是巨头企业。

It does suggest a long term problem for CoreWeave, which is that right now, their customers are the big guys.

虽然他们也有其他客户，但微软是他们最大的客户，其次是英伟达。

Like, they have other customers besides the big guys, but Microsoft is their biggest customer followed by NVIDIA.

明白吗？

You know?

所以那些需要AI云托管服务的客户不会直接找CoreWeave。

And so people who are wanting an AI cloud host aren't going directly to CoreWeave.

他们会选择微软，而微软通过提供溢出计算能力来留住这些客户。

They're going to Microsoft, And Microsoft is keeping those customers by doing overflow compute.

因此某种程度上，微软基本上能获得这些计算资源，却不用承担全部风险。

So there's this sort of, like, way in which Microsoft is basically getting to have this compute without having to take on some of the risks of it.

CoreWave是从哪里来的？

Where did Core Wave come from?

这是微软的发明吗？

Is this an invention of Microsoft?

这是一家有历史背景的公司吗？

Is this a company with a history in the past?

它是因为AI才全新出现的吗？

Is it wholly new because of AI?

是加密货币。

It is crypto.

对。

Yeah.

当然，它是

Of course, it's

加密货币。

crypto.

就是加密货币。

It's crypto.

所以这些人在2022年加密货币热潮后基本上就转向了，不再搞加密货币。

So these guys pivoted basically after, the crypto boom in 2022 away from crypto.

因为在2021年，CoreWeave的所有收入都来自以太坊挖矿。

Because in 2021, CoreWeave made all of its money by mining Ethereum.

所以他们基本上把握住了时机，真正搭上了AI热潮的快车。

So they basically timed things correctly to really ride the AI boom.

而且，你知道，在AI领域你看到的一个让我觉得是警示信号的现象是，有很多人从加密货币转向AI，因为他们认为AI会更赚钱。

And, you know, one of the things that you see in the AI space that strikes me as a flag is that there are a lot of people who pivoted from crypto into AI because they thought it was going to be more profitable.

你可能还记得英伟达开始这段惊人涨势的原因之一，就是所有这些加密货币公司都在购买它的芯片。

And you may remember one of the reasons why NVIDIA was starting to have this incredible run was because all of these crypto companies were buying its chips.

英伟达当时对此并不太热衷，因为

And NVIDIA was, like, not the most enthused about that because

实际上一点也不热衷。

Not at all, actually.

没错。

No.

我觉得他们甚至削弱了某款芯片的性能。

I think they even nerfed one of their chips.

但他们并不看好加密货币。

But they didn't love crypto.

当AI热潮真正兴起时，特别是在ChatGPT发布后，对英伟达和许多人来说，这比‘我们要挖比特币’听起来靠谱多了。

And when AI started to really get frothy, especially after ChatGPT was released, I think that was something that felt better to Nvidia and to a lot of other people than, yeah, we're gonna mine Bitcoin.

所以Horoweave恰逢其时地抓住了这个机会。

And so, like, Horoweave was in the right place at the right time to really ride that.

抽象层面我能理解，对吧？

I understand that abstractly, right?

作为英伟达，你肯定不想卷入一场单纯为挖比特币或参与以太坊区块链的算力竞赛，因为长期来看专用芯片（ASIC）会打败你。

If you're Nvidia, you don't wanna be in a race to do literally a single calculation to mine Bitcoin or to participate in the Ethereum blockchain because you'll get beaten by ASICs or other specialized compute over time.

也许你根本不想参与这种竞争，更不愿因此削弱对长期价值更高的其他市场的供应能力。

And maybe you just don't want to participate that and you certainly don't want to reduce your ability to supply other markets that you think are more valuable long term.

但如果你是Core Weave，可能你真相信那是未来，并且擅长采购和连接GPU来搭建针对特定任务的数据中心。

If you're Core Weave, maybe you do believe that that is the future and you've gotten really good at buying and connecting GPUs to build data centers for specific tasks.

AI浪潮来袭时，英伟达说‘就是它了’

AI comes along, Nvidia says this is the thing.

这就是我们芯片设计所面向的多功能计算解决方案，如果你相信的话，它能以多种方式推动经济发展。

This is the multi purpose compute solution that our chips are made for, that can if you, you know, believe it, can accelerate the economy in various ways.

这里将是增长所在。

This is where the growth is going to be.

然后你看看CoreWeave就会觉得，既然你们已经非常擅长用我们的GPU搭建数据中心，现在就来为我们想要的目标继续努力吧。

And then you look at CoreWeave and you're like, well, you're already pretty good at building data centers full of our GPUs, now do it for the thing that we want you to do it for.

于是你就发现CoreWeave应运而生了。

And then you just arrive at CoreWeave exists.

没有英伟达，CoreWeave根本不可能存在。

It's just not possible for CoreWeave to exist without NVIDIA.

我们稍后再深入讨论这点。

And we can get more into that in a minute.

但想想这家公司的运作方式——查看S1文件会发现，他们宣称拥有超过25万颗英伟达芯片，这规模相当惊人。

But just, like, thinking about how this company works, if you look at the s one filing, they say that they own more than 250,000 NVIDIA chips, which is big.

要知道，这些基础设施正是运行AI模型所必需的。

And that's, you know, infrastructure that's necessary for running AI models.

事实上他们只用英伟达芯片。

And in fact, it only has NVIDIA chips.

其他品牌一颗都没有。

It doesn't have any others.

更重要的是，英伟达本身就是CoreWeave的投资方。

So on top of that, you have NVIDIA, which is an investor in CoreWeave.

今年三月当IPO受挫时，英伟达立即出手收购了股份。

When, you know, the IPO faltered in March, Nvidia swooped in and bought shares.

英伟达以多种方式为CoreWeave提供支持，包括承诺购买任何未售给客户的剩余算力。

Nvidia has backstopped CoreWeave in a number of ways, including saying that they will buy any excess compute that isn't sold to customers.

所以他们实际上是在提供收入保障，这非常了不起。

So they're essentially providing a guarantee of income, which is remarkable.

这就像是，你知道的，拥有一个骑士。

It's like having, essentially, you know, a knight.

对吧？

Right?

就像，英伟达是国王，CoreWeave就是骑士。

Like, NVIDIA is the king and CoreWeave is the knight.

而且，你知道，英伟达在很多方面保护着CoreWeave，而CoreWeave在很多方面似乎也在为英伟达服务。

And, you know, so NVIDIA is protecting CoreWeave in many ways, and CoreWeave in many ways seems to serve NVIDIA.

我很欣赏这个王室比喻，但更直接地说，英伟达是CoreWeave的主要供应商。

So I appreciate the royal metaphor, but more directly on the ground, NVIDIA is CoreWeave's main supplier.

整个商业模式就是：我们以X美元购买这些芯片，然后以Y美元出售这些芯片的使用权，而Y长期来看必须大于X。

And the entire business is we will buy these chips for x amount of dollars and sell the use of those chips for y, and y over time has to be more than x.

但如果你的供应商同时也在购买需求，这不是会让事情变得非常复杂吗？

And doesn't that get very complicated if your supplier is also buying the demand?

这里面有些东西简直让我脑子转不过弯来。

Like, there's something there that just breaks my brain.

这就是为什么你偶尔会收到我发的私信，说我感觉自己要疯了。

It's one of the reasons why you have been getting the occasional DM from me where I just say I feel crazy.

因为我看着这些东西，我...我感觉自己要疯了。

Because I look at this stuff and I I I feel crazy.

我认为理解这一点的一种方式是，英伟达确实希望CoreWeave能够存在，因此他们会采取必要措施确保实现这一目标。

I think what's one way to understand this is that NVIDIA really wants CoreWeave to exist, and so they're going to do what they need to do in order to make that happen.

这其中是有原因的。

There's a reason for that.

部分原因在于数据中心的建设以及对过度建设的担忧。

Part of it is the data center build out and the fear that one might overbuild.

另一个原因是CoreWeave刚签下大单并扩大合作的重量级客户之一就是OpenAI。

And part of it is that one of CoreWeave's big customers who've just signed a huge contract with them and then expanded it is OpenAI.

据我所知，OpenAI目前尚未盈利，一直在亏损。

And OpenAI, as far as I know, is not profitable and hasn't been profitable and has been losing a lot of money.

因此如果你是某类人或银行，可能会对OpenAI这样没有盈利的大客户能否支付账单感到担忧。

And so if you are a certain kind of person or bank, you might be nervous that a major customer like OpenAI doesn't have profit and may not be able to pay its bills.

而英伟达的这项担保就是在说：嘿，

And so what this guarantee from NVIDIA does is it says, hey.

这些账单都会得到支付。

These bills are going to get paid.

如果你是股东、银行或其他相关方，其实谁在付账对你并不重要。

And if you are a shareholder or a bank or whoever, it doesn't really matter to you who's paying the bills.

这不是你需要操心的问题。

That's not, like, your problem.

关键是要确保这些账单得到支付。

It's just a matter of making sure those bills get paid.

我们需要短暂休息一下。

We need to take a quick break.

我们马上回来。

We'll be right back.

商界有句话说，你可以选择更好、更便宜或更快，但只能三选二。

In business, they say you can have better, cheaper, or faster, but you only get to pick two.

但如果能同时拥有这三者呢？

What if you could have all three at the same time?

这正是Coher、汤森路透和Specialized Bikes在升级到新一代云平台——甲骨文云基础设施(OCI)后的体验。

That's exactly what Coher, Thomson Reuters, and Specialized Bikes have since they upgraded to the next generation of the cloud, Oracle Cloud Infrastructure.

OCI是满足您基础设施、数据库、应用开发和AI需求的极速平台，您可以在高可用性、持续高性能环境中运行任何工作负载，且花费比其他云平台更少。

OCI is the blazing fast platform for your infrastructure, database, application development, and AI needs where you can run any workload in a high availability, consistently high performance environment and spend less than you would with other clouds.

为什么更快？

How is it faster?

OCI的块存储提供更高的每秒操作数。

OCI's block storage gives you more operations per second.

更便宜？

Cheaper?

OCI的计算成本降低达50%，存储降低70%，网络降低80%。

OCI costs up to 50% less for compute, 70% less for storage, and 80% less for networking.

更好？

Better?

在反复测试中，OCI客户报告称相比其他云平台具有更低延迟和更高带宽。

In test after test, OCI customers report lower latency and higher bandwidth versus other clouds.

这是为AI和您最重要工作负载而构建的云平台。

This is the cloud built for AI and all your biggest workloads.

此时此刻，无需任何承诺，免费试用OCI。

Right now, with zero commitment, try OCI for free.

前往oracle.com/vox。

Head to oracle.com/vox.

网址是oracle.com/vox。

That's oracle.com/vox.

我们继续连线《边缘》杂志资深记者莉兹·拉帕托。

We're back with Verge senior reporter Liz Lapato.

广告前我们讨论了CoreWeave的起源及其与英伟达的特殊关系。

Before the break we were discussing the origins of CoreWeave and its unique relationship to Nvidia.

现在我想更深入地探讨整个AI行业幕后正在进行的一些精妙会计操作——这些操作为昂贵的基础设施建设提供了资金支持。

Now I wanted to dive deeper into some of the clever accounting that's going on behind the scenes across the entire AI industry to pay for some of this costly infrastructure.

你在文章中指出，CoreWeave作为一个实体存在'具有某种糟糕的合理性'，它通过承担其他公司的风险来对冲风险并提高利润，将风险转嫁给这家公司，同时把数据中心建设失败的风险转移到其他地方，从而避免伤及大公司自身。

You write in your piece that CoreWeave as an entity makes quote a horrible kind of sense that it exists to hedge other companies risks and juice profits by offloading that risk into this company and offloading the risk of building data centers somewhere else that can fail without hurting the the big companies themselves.

请向大家解释这个概念。

Explain that to people.

它承担了建设数据中心的部分风险，而这本身就是一项艰巨任务。

It's taking on some of the risk of building data centers, and like that is an undertaking.

建设工程项目——我想大家都知道——经常无法按计划进行，原因与施工人员的技术水平毫无关系。

Construction projects, as I think everybody knows, often don't happen on schedule for reasons that have nothing to do with how good people are at building things.

比如遇到恶劣天气时，你就是没法施工。

Like, sometimes if the weather sucks, you just can't build.

这种情况时有发生。

That's just a thing that happens.

因此，制造东西本身就存在风险。

And so there is inherently risk in building stuff.

除此之外，芯片这类产品往往很快就会过时。

And then on top of that, if you think about chips, they frequently and very quickly become outdated.

所以如果你在这一代芯片上投入巨资，而英伟达明年又推出了新一代更好的芯片，你就得再次大笔投入。

And so if you are doing a big spend on this generation of chips and then NVIDIA comes out with a new, better generation of chips next year, you're doing another big spend.

这也算是一种风险。

That is also something of a risk.

对吧？

Right?

从事这个行业，就会面临这些不断贬值的资产。

Being in this business, there are these assets that depreciate.

所以CoreWeave承担了这些风险，而微软只是租用他们的算力。

And so CoreWeave is taking on those risks, and Microsoft is just renting their compute.

它不必担心，比如这个数据中心是否需要永久运营下去？

It doesn't have to worry about, you know, is this data center going to be something that I need to continue to operate forever?

这些芯片明年会不会过时？

Are these chips gonna be outdated next year?

这些都不重要。

None of that matters.

它只是想着，哦，这里可以安置过剩的业务？

It's just, oh, this is a place where I can I can put excess business?

我们在其他地方也见过这种模式。

We've seen this idea pop up in other places.

Meta刚刚成立了一个所谓的特殊目的实体，用于自建数据中心。

Meta just did what's called a special purpose vehicle to build some data centers of its own.

那些与CoreWeave处境相似的公司，实际上NVIDIA也在投资它们。

Companies that are situated much like CoreWeave that actually NVIDIA is investing in as well.

这是否仅仅是为了应对自建数据中心的风险——成本可能飙升，而我们却无法预知回报？

Is this just the answer to the risk of we're gonna build a data center, the cost might balloon, and we don't know what the return is?

是否只是问题出现了，这就是最显而易见的解决方案？

Is it just there's a problem and this is the obvious solution?

可能是这样。

It might be.

我一直在深入思考这些特殊目的实体的用途。

One of the things that I've been thinking about here a lot are these special purpose vehicles.

企业选择使用这类实体有许多原因。

There are lot of reasons why you might choose to use one as a company.

其中之一可能是为了限制风险。

One of them might be to limit risk.

比如，某个部分破产的风险，或者可能受到公司其他部分影响的风险。

Like, you know, the risk that something goes bankrupt or the the risk that something, you know, might be affected by another part of the company.

因为这些就像独立的小王国，能让你对正在做的事情感到一定程度的安心，我认为许多公司都在采用这种方式。

Because these are, like, little contained fiefdoms, that lets you feel a certain amount of comfort with what you're doing in a way that I think that a lot of these companies are embracing.

作为Matt Levine的长期读者和仰慕者，他曾写过这些公司如何保持其完美信用评级的文章。

And I I as a long time reader and admirer of Matt Levine, he had written a thing about how these companies are maintaining their pristine credit ratings.

比如微软的信用评级，我想比美国还要高。

Like, Microsoft's is better than The US, I think.

他们通过特殊目的载体（SPV）进行操作。

They're doing it via the SPVs.

他们通过某些交易来实现这一点，在这些交易中他们承担债务却不会影响信用评级。

They're doing it via some of these deals where they're taking on debt without touching their credit ratings.

他还暗示，这是因为他们认为保持良好的信用评级非常有益——他确实没说错。

And he has suggested that this is because they think it's just really nice to have a good credit rating, and he's not wrong.

拥有良好的信用评级确实很棒，因为这意味着你能以更低利率借款。

It is really nice to have a good credit rating because it means you can borrow at a lower rate.

但我也在思考，这其中有多少是因为这些特殊目的载体不必像母公司那样受同类监管，且具有某些可能更有利的结构。

But I also wonder how much of this is because these special purpose vehicles aren't subject to the same kinds of regulations as parent companies and have certain kinds of structures that may be advantageous.

我理解为什么作为中间商，在你 pristine 的信用评级和经营企业的泥潭之间运作能获利。

I understand why it's profitable to be the sort of middleman operator between your pristine credit rating and like the muck of growing a business.

传统上，在每个行业和你能想到的每个领域，这个中间地带都蕴藏着巨大财富。

There's a lot of money in that middle zone traditionally across every industry and every category you can think of.

但Corweave是一家上市公司。

But Corweave is a public company.

问题是：你能靠'我们愿意承担风险，但愿AI能成功'来维持一家上市公司吗？

Like, can you sustain a public company on we will accept your risk and hopefully AI pays off?

我的直觉反应是不能。

My impulse is no.

不过，我们终究会知道答案。

But I mean, we're gonna find out.

毕竟，我也可能是错的。

Like, you know, I could be wrong.

这不会是我第一次犯错。

It wouldn't be the first time I've been wrong.

但我看到CoreWeave背负了那么多债务，这真的让我很紧张。

But I look at CoreWeave and I see how much debt they've taken on, and it really makes me nervous.

你知道，过去几个月里市场一直有个持续的主题，就是看到大量债务被用于AI基础设施建设。

You know, there's been sort of this ongoing theme, I would say, in the last couple of months where we've seen a lot of debt being taken out in the market for this AI build out.

而且，使用债务而非利润是有充分理由的。

And, like, there's good reason why you would use debt rather than profits.

主要是你能快速获得大量资金，这样就能快速完成建设。

Mainly, that you can get a bunch of money really quick, and that lets you do the build out really quick.

这可能会影响到芯片的使用寿命。

And then potentially, you know, that matters for the lifetime of the chips.

因为如果你建设得够快，就能最大限度延长这些芯片的使用周期，或许还能更快实现盈利。

Because if you do the build out really quick, then you get the maximum lifetime out of these chips, and maybe you you make your profits faster.

这就是期望所在。

Like, that's the hope.

对吧？

Right?

芯片这部分每次和你讨论或阅读相关资料时，都让我感到难以置信。

So the chips piece of this is the other thing that just constantly breaks my brain whenever I talk to you about it or I read anything about it.

Corby背负了大量债务，正在申请巨额贷款。

Corby was taking on tons of debt, is taking tons of loans.

它用从英伟达购买的芯片作为这些贷款的抵押品。

It's collateralizing those loans with the chips it's bought from Nvidia.

这些贷款的回报在于芯片将承担极其有价值的工作负载，但芯片会损耗，而AI行业尚未证明其能完成大量真正高价值的工作任务。

The payoff of those loans is that the chips will do extraordinarily valuable workloads, but the chips wear out and AI as an industry has not yet proven that it can do a bunch of really really valuable workloads.

这里存在一个时间顺序问题，让我觉得AI在英伟达芯片全部物理损耗前都无法实现盈利。

There's a chronology problem here that suggests to me that AI won't pay off before all of the NVIDIA chips literally wear out.

OpenAI只会以如此速度进行亏损的视频生成，以至于在有人发明出能让花旗银行利润翻倍的AI工作负载之前，芯片就已经损耗殆尽。

That OpenAI will just do unprofitable sort of video generation at such rate that the chips will wear out before someone invents an AI workload that doubles the profits of Citibank or something.

我看不出这些时间线如何能真正对齐，使得这门生意能够成功。

Like I don't see how those timelines actually align such that this business can succeed.

难道就没有比这更好的论点了吗？

Is is there a a better thesis out there than that?

这正是我非常好奇的问题，如果你知道答案请随时联系我，因为芯片折旧周期在这里至关重要，而人们对实际周期似乎存在不少分歧。

Well, so this is the thing that I'm really curious about and obviously drop me a line if you know, because the chip depreciation cycle really matters here, and there seems to be a fair amount of disagreement about what it actually is.

关注迈克尔·伯里的朋友们可能记得，这位在《大空头》中精准把握金融危机时机的人，他正在论证我们当前看到的折旧时间表实际上过于乐观。

For those of you who follow Michael Burry, who you may recall from The Big Short as the guy who, like, nailed the timing on the financial crisis, he's arguing that the depreciation schedules that we're looking at right now are in fact too rosy.

这可能是事实。

That may be true.

说实话，我对此也只是略知皮毛。

You know, I've I've barely scratched the surface here.

但我发现很多因素都会影响结果。

But the thing that I have found is that a lot of things vary.

对吧？

Right?

比如芯片的存放位置、周边基础设施状况、维护方式等所有其他因素，以及使用强度如何。

So where the location of the chips are, what kind of infrastructure is around them, how they're being serviced, all of these other things, how heavy the use is.

看起来影响这些芯片损耗速度的因素有很多。

There seems like there are a lot of variables involved in how quickly these chips wear out.

所以对我来说，这是核心问题之一：这能否成功运作，关键在于这些芯片是否算得上优质抵押品？

And so that, to me, is one of the central questions about whether this works out or not is, is this good collateral?

如果从某个角度看，这些芯片实际上和卡车没什么区别。

If there's one way of looking at this where these chips are effectively the same as a truck.

我们经常为车辆提供贷款，而卡车会贬值。

And we have vehicles that get loans all the time, and trucks depreciate.

我们对你业务中卡车的贬值周期有相当准确的预估。

We have a pretty good schedule for the depreciation of the truck for your business.

所以你可以用卡车作为抵押获得贷款，这根本不是问题。

And so you can get a loan with the truck as collateral, and that's not really an issue.

但由于这类金融产品实在太新，很难判断这些芯片的真实状况。

But because this particular kind of financial product is so new, it's sort of difficult to figure out what's going on with the chips.

因此在我看来，这项业务能否成功的关键问题之一就是：这些芯片的质量究竟如何？

And so to me, that is one of the central questions here about will this succeed is, like, how how good are the chips?

它们的使用寿命有多长？

How long do they last?

我要澄清一点。

I wanna be clear.

不只是CoreWeave在做这种事。

It's not just CoreWeave that's doing this.

他们是这类贷款的开创者。

They pioneered this kind of loan.

他们是第一个这么做的。

They were the ones who did it first.

这是因为他们的CEO和其他几位高管并非技术出身。

And that's because their CEO and a couple of other folks in their senior management, they don't come from tech.

他们是金融背景的人。

They're finance guys.

他们设计出这款有趣的金融产品，我个人对此非常着迷。

They came up with this interesting financial product, which, like, I I personally am fascinated by.

随后一批类似的新型云公司立即效仿了他们。

Then a bunch of other similar neo cloud companies immediately followed in their footsteps.

比如Lambda、Crusoe这些公司。

So that's like Lambda, Crusoe.

还有其他公司现在正通过私募信贷获取债务来购买英伟达芯片。

There are other companies that are now sort of making these deals with private credit in order to get debt and and buy NVIDIA chips.

CoreWeave在这整个过程中有盈利的途径吗？

Does CoreWeave have a path to being profitable inside of all this?

即使AI最终没有回报？

Like, even if AI doesn't pay off.

对吧？

Right?

就算OpenAI一分钱都赚不到，他们仍会向CoreWeave支付算力费用。

And OpenAI never makes a dollar, they're still gonna pay CoreWeave for capacity.

CoreWeave能否走这样一条路：看，我们只提供基础设施。

Is there a path for CoreWeave to say, look, we just have the infrastructure.

在某个时间线上，我们会解决这个问题的。

On some timeline, we're gonna figure this out.

我们可以顺势而为，并在此过程中赚钱。

We can just ride the wave and make money along the way.

这就是债务之所以重要的原因，因为芯片贷款只是CoreWeave所举债务中的一种。

This is why the debt is important because the chip loans are only one kind of debt that CoreWeave has taken out.

摩根大通还提供了一项循环信用额度。

There's a revolving credit line with JPMorgan.

其中还涉及一些供应商融资。

There's some vendor financing involved.

还有与投资者的担保贷款。

There are secured loans with investors.

所有这些类型的贷款都有不同的结构。

All of these types of lending have different structures.

根据贷款到期时间和结构安排，这将影响CoreWeave需要资金的时间表。

And depending on when the loans come due and how this is structured, that's gonna affect CoreWeave's timeline for when they have to have the money.

只要他们能进入信贷市场，或许还能稍微延长这个期限。

Maybe as long as they have access to credit markets, they can extend that some.

对吧？

Right?

比如他们可以借新债来偿还旧债的利息，为自己争取更多缓冲时间。

Like they can take out a loan to pay off the interest on the other loan and give themselves a little more runway.

但最终当你审视这家公司时，考虑到这些贷款的规模，你会认为AI技术必须实现非常快速的起飞。

But eventually, you start to look at the company and you think, given the size of these loans, we really need to have a very fast takeoff for AI.

这已经包含在当前的假设中了。

That is built into the assumptions here.

所以如果你把AI想象成一项普通技术，一项重要但普通的技术，那么爆发式发展不会以那种方式发生。

And so if you imagine AI is a normal technology, like a big and important technology, but a normal one, that takeoff isn't going to happen that way.

我希望你思考一下，比如网购。

I would like you to think about, like, online shopping.

对吧？

Right?

它花了大概25年才占到零售额的25%？

Like, it took, what, twenty five years to be 25% of retail?

而且你知道，当人们在1999年说未来会在网上购物时，他们并没有错。

And, like, you know, when people were saying in 1999 that you're gonna be shopping online, they weren't wrong.

比如pets.com其实并不是个坏主意。

Like, pets.com was actually not a bad idea.

只是太早了。

It was just early.

时机不对。

The timing was wrong.

对吧？

Right?

因为你看Chewy，它是什么？

Because if you look at Chewy, what is it?

本质上就是个晚出现很多年的pets.com。

It is basically a much later pets.com.

而且取得了巨大的成功。

And it was enormously successful.

所以，你知道，当人们思考如何围绕新技术开展工作，企业如何围绕新技术调整业务模式时，可能会认为技术采用会比较缓慢。

So, you know, you think about like how people work around new technology and the ways that businesses work around new technology, and you think, well, maybe there's gonna be a slower adoption.

比如说，我们以一家销售实体产品的普通企业为例。

So, like, let's take, for instance, a normal business, like, a that that sells physical objects.

我们就叫它小部件公司吧，为什么不呢？

We'll say widgets because why not?

这家小部件公司可能没有1%的收入可以立即投资于人工智能。

And this widget business may not have 1% of its income that it can invest in AI right away.

因此它会继续使用传统产品，直到无法继续使用或遇到特别好的年份。

So it's going to continue using legacy products until either it can't or it has an especially good year.

这将延缓技术的采用速度。

And that is going to slow the adoption.

如果我们观察当前人工智能被采用的模式，最先接纳这项技术的是那些业务相对成熟的企业。

If we look at sort of the ways that AI is being adopted right now, the people who are taking it on first are the people who have these relatively sophisticated businesses.

我们指的是银行、软件公司这类企业。

So we're talking about banks, We're talking about software companies.

如果你考虑实体行业，采用速度就没有这么快了。

If you think about the physical world, that's not happening at quite the same rate.

我们需要短暂休息一下。

We need to take another quick break.

马上回来。

We'll be right back.

作为创始人，你正快速朝着产品市场匹配、下一轮融资或首个大企业交易迈进。

As a founder, you're moving fast towards product market fit, your next round, or your first big enterprise deal.

但随着AI加速初创企业的开发和交付速度，安全合规要求也来得更快，且这些标准比以往任何时候都更高。

But with AI accelerating how quickly startups build and ship, security expectations are also coming in faster, and those expectations are higher than ever.

及时做好安全和合规工作能推动增长，若拖延太久则可能阻碍发展。

Getting security and compliance right can unlock growth or stall it if you wait too long.

Vanta是一个信任管理平台，帮助企业自动化完成SOC2、ISO27001、HIPAA等35种以上框架的安全合规认证。

Vanta is a trust management platform that helps businesses automate security and compliance across more than 35 frameworks like SOC two, ISO 27,001, HIPAA, and more.

通过专为快速发展的团队打造的深度集成和自动化工作流，Vanta能让你快速做好审计准备，并在你的模型、基础设施和客户发展过程中通过持续监控确保安全。

With deep integrations and automated workflows built for fast moving teams, Vanta gets you audit ready fast and keeps you secure with continuous monitoring as your models, infrastructure, and customers evolve.

这就是为什么像Langcheng、Ryder和Cursor这样的快速成长型初创企业都选择Vanta从一开始就建立可扩展的合规基础。

That's why fast growing startups like Langcheng, Ryder, and Cursor have all trusted Vanta to build a scalable compliance foundation from the start.

访问vanta.com/vox通过Vanta初创企业计划立即节省1000美元，加入已在使用Vanta的10000多家雄心勃勃的企业行列。

Go to vanta.com/vox to save $1,000 today through the Vanta for startups program and join over 10,000 ambitious companies already scaling with Vanta.

限时优惠：访问vanta.com/vox可节省1000美元。

That's vanta.com/vox to save $1,000 for a limited time.