英伟达向CoreWeave投资20亿美元

欢迎收听本播客。

Welcome to the podcast.

我是主持人贾登·谢弗。

I'm your host, Jaden Shafer.

今天在播客中，我们将讨论英伟达刚刚达成的一项巨额投资交易。

Today on the podcast, we're talking about the massive investment deal that NVIDIA has just made.

20亿美元将投入CoreWeave，而这家公司背负着巨额债务。

$2,000,000,000 is gonna go towards CoreWeave, who has a ton of debt.

这笔资金将帮助他们增加5吉瓦的AI算力。

It's gonna help them add five gigawatts of AI compute.

这是一项极其重大的交易，考虑到CoreWeave目前所处的境况以及他们是如何走到这一步的，这个故事也相当有趣。

This is an absolutely massive deal, and this is also a it's kind of a an interesting story considering the situation that CoreWeave has found themselves in and how they got here.

所以今天我们将在播客中全面探讨这些内容。

So we'll be covering all of that on the podcast today.

但在开始之前，我想向大家介绍我们刚刚在AI Box中新增的最令人兴奋的功能。

But before we do, wanted to tell you about the newest feature that we have just added to AI Box that I am so excited about.

AI Box，正如你所知，是我自己的初创公司，专注于打造富有氛围感的AI工具。

AI Box, as you know, is my own startup and it's for vibe building AI tools.

我们刚刚新增了大家一直期待的附件功能。

And we just added the feature that is much requested of attachments.

现在你可以上传图片，在创建工具时也能附加文件，这意味着你可以创建大量新工具，比如上传PDF、电子表格或图片，这些工具能接收各种类型的文件作为输入，并输出结果，非常酷。

So, you're able to attach images, and when you build boxes, you're able to attach files to it, which means that there's a ton of new tools you can create where you're uploading, PDFs, or you're uploading spreadsheets, or uploading images, and you can actually make tools that can take inputs of all these types of files and spit out, outputs, which are really, really cool.

如果你想去aibox.ai尝试创建一个工具，而且你不是开发者，那就去试试吧。

So if you wanna go try building a tool on aibox.ai, if you are not a developer, go check it out.

每月20美元，你就可以访问所有工具以及40多个不同的AI模型，用来替代你所有的订阅服务。

It is $20 a month to get access to all of the tools and over 40 different AI models that you can use to replace all your subscriptions.

链接在简介中。

It is linked in the description.

我很想看看你创造了什么，也很乐意为你宣传。

I'd love to see what you make, and I'm happy to shout it out.

如果你做出了什么酷炫的东西，记得告诉我，我很乐意在播客里分享。

So let me know if you build anything cool, and I'm happy to share that on the podcast.

好的。

Alright.

让我们开始今天的节目吧。

Let's get into the episode today.

让我印象特别深刻的是，我和英伟达，当然，还有这个高达20亿美元的CoreWeave估值。

So what was what was really impressive to me is my and NVIDIA, of course, just like the price tag on this, this $2,000,000,000 for CoreWeave.

我认为他们之前就已经有相当紧密的合作关系了，因为CoreWeave是一家数据中心运营商，正试图在未来十年内迅速扩大其AI基础设施规模。

And I think they already had quite a close relationship, because CoreWeave is, you know, a data center operator, and they're trying to very quickly grow a lot of their AI infrastructure footprint, over the next ten years, really.

因此，这项投资将使CoreWeave的A类股票估值达到每股87.20美元，并旨在加速其在2030年前部署超过5吉瓦AI算力的计划。

So this investment is now going to value CoreWeave's class a shares at $87.20 apiece, and it is intended to basically speed up their plan to deploy more than five gigawatts of AI compute capacity by 2030.

这种规模的电力容量将使CoreWeave跻身全球最耗能的AI基础设施运营商之列。

So that level of power capacity is gonna basically place CoreWeave among the top, you know, the the most energy intensive AI infrastructure operators in the world.

它的规模将与谷歌、OpenAI以及其他正在自建基础设施的大型企业相当，CoreWeave将与它们并驾齐驱。

It's gonna be on par with the same scale that you see, you know, some of these hyperscalers like Google and OpenAI and all of these other people that are, you know, building their own infrastructure, CoreWeave is gonna be right there with them.

作为这笔交易的一部分，英伟达和CoreWeave表示，他们将共同开发他们所称的‘AI工厂’。

So, as part of this deal, NVIDIA and CoreWeave say that they're going to kind of jointly develop what they're calling AI factories.

这些是大规模的数据中心，具有非常特定的用途，即用于训练和运行人工智能模型，而且所有这些都将使用英伟达的硬件。

These are large scale data centers, and they have a very specific purpose, which is for training and running AI models, and that's all of it is of course gonna be using NVIDIA's hardware.

因此，对英伟达来说，这是一个明智的举措。

So NVIDIA, this is a smart move for them.

他们正在寻找数据中心领域的主要玩家之一，并表示：我们会投资你们，但你们必须锁定使用我们的硬件。

They are finding one of the bigger players in data centers, and they're saying, look, we'll invest in you, but you're gonna lock in with our hardware.

当然，还有他们的软件栈。

And of course also their software stack.

因此，CoreWeave 也将在其平台上全面采用英伟达的技术。

So CoreWeave is also gonna standardize on NVIDIA technologies across its platform.

这包括即将推出的 Rubin GPU 架构，该架构预计将在 Blackwell 之后推出，以及英伟达的 Bluefield 网络和存储系统。

So that includes the upcoming Rubin GPU architecture, which is expected to come after the Blackwell, along with Bluefield networking and storage systems in NVIDIA's.

他们最近还推出了一条新的 CPU 系列，名为 Vera。

They kinda have this new CPU line that they just announced, which is called Vera.

因此，CoreWeave 将与所有这些技术进行整合。

So CoreWeave's going be integrating with all of that.

我认为这同样是一项非常有趣的投资，因为这本质上是对CoreWeave的一种信心投票，尽管该公司一直因其资本结构问题而受到大量质疑。

I think it's also a really interesting investment because it's basically a vote of confidence when CoreWeave has been under a lot of scrutiny for its, basically the structure of its capital structure.

对吧？

Right?

因此，他们大量依赖债务来资助扩张，坦白说，这家公司的增长速度确实令人印象深刻，但他们用自己拥有的GPU作为所有债务的抵押品。

So, they've leaned really heavily on debt to finance their expansion, which of like, let's be honest, this company is growing at an absolutely impressive rate, but they're using their own inventory of GPUs as collateral for all of the debt that they've taken on.

所以，如果你查看PitchBook并参考那里的部分数据，截至九月，CoreWeave的债务总额约为188.1亿美元，同时他们在第三季度的收入约为13.6亿美元。

So, according to if you look at PitchBook and you kinda look at some of the data over there, CoreWeave has about $18,810,000,000 in debt as of September, and so at the same time, they also have about $1,360,000,000 in revenue in the third quarter.

因此，你可以看到，尽管他们的服务需求旺盛，但债务规模接近190亿美元，假设他们每年收入50亿美元，他们也必须偿还所有这些债务。

So, I think you can see there's also, you know, a lot of demand for their services, but also, like almost $19,000,000,000 in debt, and let's say they're making like $5,000,000,000 a year, you know, they they have to go service all of that debt.

因此，在建设大量AI基础设施的同时，他们承受着巨大的压力。

So, there is a lot of pressure there while they're building a lot of this AI infrastructure.

他们的首席执行官是迈克尔·因特拉德，他多次为他们的商业模式辩护。

Their CEO is Michael Intrader, and he has repeatedly defended their model.

说真的，150亿美元的收入？

I mean, to be fair, $15,000,000,000 in revenue?

我的意思是，最近一次是2036年1月，过去可能低得多，但他很可能在预测未来的增长。

I mean, like, let's just say, the most recent was 01/1936, and so, probably a lot lower in the past, but he's probably projecting growth into the future.

我们就假设是50亿或60亿美元吧。

So, let's just say like 5 or $6,000,000,000.

在180亿美元的债务基础上，用六年时间还清其实并不需要太多年。

On six, you know, on $18,000,000,000, it doesn't take that many years for them to pay that off.

但你也要知道，他们正在快速增长，不断扩大规模，从英伟达那里获得了大约20亿美元，他们可能会把这笔钱全部，或大部分用于偿还债务和其他开支。

But you also have to know that they are growing rapidly, they're scaling, they're taking in, you know, like $2,000,000,000 from Nvidia, and they're gonna probably go spend all of that, or spend a big chunk of it on servicing the debt and all of that.

所以，他们的首席执行官表示，人工智能迫使企业重新思考传统的融资方式和合作伙伴结构。

So, their CEO has said that AI has forced companies to rethink traditional financing, and also partnership structures.

他基本上说，人工智能生态系统中并不存在循环投资的担忧。

He basically said, you know, there's no concerns about circular investment in the AI ecosystem.

他指出，这本质上描述了供应链需求的剧烈变化，他称这种变化需要芯片制造商、基础设施提供商和众多人工智能实验室之间的大量协作。

He's pointed to what basically describes as a quote, violent change in supply chain demand, one that he said requires a lot of collaboration between chip makers, infrastructure providers, and a lot of these AI labs.

因此，他基本上是在说，如果你想消除瓶颈，这是唯一可行的方式。

And so, he's basically saying, if you want to remove the bottlenecks, this is the only way that you're going be able to do it.

过去几年他们的发展轨迹表明，人工智能市场正在迅速重塑自身。

Their trajectory over the last few years, I think, just shows how quickly the AI market is reshaping itself.

CoreWeave，对于那些忘记的人来说，CoreWeave 最初其实是一家加密货币挖矿公司。

CoreWeave, and for those that forget, right, CoreWeave actually originally was like a crypto mining company.

当基于 GPU 的计算需求急剧上升时，他们开始 aggressively 转向基础设施，我认为很多从事加密货币挖矿的人，后来遭遇了加密货币熊市、寒冬——无论你怎么称呼它——这一切都始于 FTX 崩溃，随后许多加密货币价格暴跌。

So they started really aggressively pivoting to infrastructure when the demand for GPU based compute was just going through the roof, and I think there's a lot of people that were in crypto mining, and then, you know, the kind of the crypto, the crypto bear market happened, the winter, crypto winter, whatever you want to call it, it kind of all started with FTX collapsing, and then the price of a lot of cryptocurrencies was crashing.

许多矿工因此失去了盈利能力。

And a lot of these miners were not profitable.

但就在同一时间，人工智能开始迅速扩张。

But at the exact same time, we had AI, which was starting to scale rapidly.

很多人需要 GPU，而 CoreWeave 就是那些拥有大量基础设施的公司之一，他们说：‘好吧，兄弟们，我们转去做 AI 训练了。’

A lot of people needed this GPU, and so CoreWeave was one of those big companies that had all of this infrastructure, and they're like, okay guys, we're moving to training AI.

自三月上市以来，他们迅速扩展，不再局限于单纯的算力服务。

So since going public in March, they've moved really quickly to expand beyond just kind of raw compute.

他们正在构建更广泛的 AI 开发栈，并通过大量收购来实现这一目标。

They're kind of building a broader AI development stack, and they're doing this through a lot of acquisitions.

所以他们实际上正在收购大量公司。

So they're actually going and buying a lot of companies.

我认为这些交易包括许多知名公司。

I think those deals include a lot of high profile companies.

其中之一是Weights and Biases，这是一个被广泛使用的平台，用于跟踪和管理机器学习实验。

One of those is Weights and Biases, which is like a really widely used platform for tracking and managing machine learning experiments.

他们还收购了一家名为Open Pipe的强化学习初创公司。

They also had the acquisition of the reinforcement learning startup, which is called Open Pipe.

然后在十月，他们同意收购Marimo——这是一个Jupyter笔记本的开源替代品，以及另一家专注于AI的公司Monolith。

And then in October, they agreed to acquired Marimo, which is an open source alternative to Jupyter notebooks, as well as Monolith, another AI focused company.

所以我认为，他们收购的这些公司表明，这是一家非常有雄心的公司，他们正试图为开发者提供一个高度垂直且集成的环境，而不仅仅是以出租GPU作为唯一的收入来源。

So I think all of these companies that they're buying shows that they have they're like it's a very ambitious company, and they're trying to give developers a really vertical, kind of integrated environment rather than just renting GPUs as kind of their only revenue source.

他们还在持续与一些大型AI公司建立合作关系。

They also are kind of continuing to make a lot of relationships with some of the big AI players.

他们拥有大量超大规模云服务商和AI实验室作为客户。

They have a whole bunch of hyperscalers that are using them, and a bunch of AI labs.

这包括OpenAI、Meta和微软。

So that includes OpenAI, Meta, Microsoft.

他们最近还特别扩大了与OpenAI的云合作伙伴关系。

They also recently expanded their cloud partnership with OpenAI in particular.

所以我认为，所有这些合作关系都帮助CoreWeave成为了一个非常好的计算服务替代选择，毕竟市面上显然还有AWS、谷歌和微软这些最大的云平台，但有时候这些公司并不想使用竞争对手的云平台。

So I think all of those relationships have helped to really set CoreWeave as a it's like, it's a really good alternative if you're looking for a compute provider, and there's obviously all of the loud like all of the largest cloud platforms out there, there's AWS and Google and Microsoft, but sometimes those companies don't want to use their competitors' cloud platforms.

因此，CoreWeave似乎是一个不错的替代方案，它拥有巨大的规模，而且人们确实能够使用它。

And so Coreave seems to be kind of this good alternative where they have, you know, tons of scale, but there, you know, people are able to use them.

所以，我认为，除了股权投资之外，英伟达还表示将与CoreWeave紧密合作，共同争取土地资源，这挺有意思的，对吧？

So, I think beyond just the equity investment, Nvidia also said that they're gonna work really closely with CoreWeave on securing land, which is kind of interesting, right?

以及未来数据中心的电力和电网接入资源。

And then power and grid access for future data center sites.

英伟达显然是家巨头公司，看起来它在政府或其他领域拥有广泛的人脉。

NVIDIA, obviously, a massive company, and it seems like they have a lot of connections, whether that's in the government or other places.

所以，说实话，对于CoreWeave来说，与英伟达合作可能是个不错的选择，因为英伟达可以帮助他们获得那些流程繁琐、极难拿到的土地、电力和电网接入合同。

So, honestly, this might be a good deal for CoreWeave to work with NVIDIA, who can go and get them land and power and grid access contracts that are like very bureaucratic and hard to get.

所以，这确实是许多AI公司面临的挑战，即这种区域性的能源基础设施问题。

So, this is definitely a challenge that a lot of AI companies are having, like this kind of regional energy infrastructure issue.

我认为英伟达还计划整合一些CoreWeave的AI软件和系统，对其进行设计，并将所有这些整合到他们自己的参考架构中，这些架构将作为其他公司和云服务商大规模部署AI的蓝图。

And I think NVIDIA is also planning to incorporate some of Core Web's AI software and systems, designing them, and putting all of those into some of their own reference architectures, which are then gonna be used by other companies and cloud providers as blueprints for deploying AI at scale.

因此，这对英伟达和CoreWeave来说都是双赢的局面。

So this is really a win win for NVIDIA and for CoreWeave.

在这一20亿美元的投资公布后，市场反应非常积极。

The markets, after all of this, after this $2,000,000,000 investment, reacted very positively to the news.

CoreWeave的股价在消息公布后上涨了15%以上。

CoreWeave shares jumped more than 15% after it was all announced.

对英伟达而言，这符合他们更广泛的战略。

For NVIDIA, this kind of fits into their broader strategy.

对吧？

Right?

他们正试图强化整个AI供应链。

They're trying to reinforce the entire AI supply chain.

我认为当前人工智能热潮的主要受益者和推动者是英伟达。

And I think the primary beneficiary and enabler of the current AI boom is NVIDIA.

因此，过去一年里，他们进行了数十项极具战略意义的投资，包括云服务商、人工智能初创公司，比如OpenAI。

And so because of this, they have made dozens of really strategic investments over the past year, and that's, you know, cloud providers, that's AI startups, mean, OpenAI.

基本上，所有大型人工智能初创公司都在获得英伟达的资金，我认为许多基础设施合作伙伴也获得了大量英伟达的投资。如今，英伟达已成为全球市值最高的公司，虽然这些数值会有些波动，但英伟达确实从这些投资中获得了巨大收益。

Basically, all of the big AI startups are getting NVIDIA money, and I think a lot of the infrastructure partners are also getting a lot of this NVIDIA money, which NVIDIA is now one of the is the most valuable company in the world, and I mean, these things fluctuate a bit, but Nvidia is definitely getting a lot of benefit from all of these investments.

但整个生态系统也在从中受益。

But the whole ecosystem is also benefiting.

我认为英伟达明白，他们必须确保一切稳定运行。

I think Nvidia knows they need to make sure everything is stable.

这些公司不能破产，也不能出现重大问题，否则会影响整个市场。

These companies aren't, you know, gonna go bankrupt or have any major issues or that affects the entire market.

所以，我认为如果他们支持像CoreWeave这样的公司，不仅是为了确保对他们的芯片保持持续需求——因为CoreWeave会购买更多他们的芯片，同时也是在帮助消除电力、土地、开发等方面的实际限制，这些限制原本可能拖慢人工智能的普及速度。他们可以说：‘我们有资源，我们了解整个体系，能帮你拿到所有这些合同，助你完成数据中心的建设。’

So, I think if they're backing a lot of these companies like CoreWeave, they're, you know, not only just making sure that there is sustained demand for their chips, right, because CoreWeave's gonna buy more of their chips, but they're also helping remove any sort of practical constraints around power, land, development that could otherwise slow the growth of AI adoption, they're like, look, we have connections, we understand the system, we can get you all of these contracts and help you, like, with everything you need to build out your data center.

因此，英伟达是个非常优秀的合作伙伴，但同时，英伟达也从这些合作关系中获得了巨大收益。

So, NVIDIA's a really good partner, but also NVIDIA is benefiting a lot from these relationships.

这看起来是双赢的局面，而且市场似乎也认同这对两家公司都有利。

Seems like a win win, and it seems like the seems like the markets have kind of agreed that this is something that's gonna be beneficial for both of these companies.

非常感谢您收听本播客。

Thank you so much for tuning into the podcast.

如果您学到了新东西，或者喜欢这一集，我很希望您能在您收听播客的平台留下评分和评论。

If you learned anything new, or if you appreciated the episode, I would love it if you could leave a rating and review wherever you get your podcasts.

如果您在苹果平台，可以给几颗星，再写一条评论。

If they're on Apple, you can leave some stars and then a comment.

如果您在Spotify上，可以去‘关于’标签页。

If you're over on Spotify, it's the about tab.

不过，非常感谢大家所有的评论、评价和支持。

But yeah, really appreciate all of the all of the comments, all the reviews, all of the support on the podcast.

我希望这对您了解AI生态系统正在发生的一切有所帮助，而这些评价对我的节目成长帮助巨大。

I hope this is helpful to you as you are learning what is going on in the AI ecosystem, and the reviews help me out tremendously with the show and growing it.

再次非常感谢。

So thank you so much.

请务必前往 aibox.ai 体验我们新增的文件上传功能，你可以用它构建允许上传文件的工具，创造出一些非常精彩的东西。

Make sure to go check out aibox.ai and the new file upload feature that we added so you can build tools that allow you to upload files and create some really incredible stuff.

非常感谢你的收听，我们下一期再见。

Thank you so much for tuning in, and I'll catch you in the next episode.