The Motley Fool 对 $META 的判断变化
我的第一个选择可能看起来很奇怪，因为它不是英伟达所告知内容的直接受益者。两个中的一个是，但我要谈的第一个是Meta。
Yeah. So my first one is might seem very strange because it's not a direct beneficiary of what Nvidia has told us. One of the two is but the first one I'll talk about is Meta.
Meta当然在英伟达芯片上花费了巨额资金，但它也在自研定制芯片上花费了巨额资金。但这里的联系是关于成本，而不是收入。
Now Meta of course spends an insane amount of money on Nvidia chips but it also spends an insane amount of money on making its own custom chips. But the connection here is about cost and not revenue.
当英伟达表示其新的Vera Rubin芯片每兆瓦吞吐量提升30倍，每token成本比上一代降低35%时，如果这是真的，且Meta能以更低的成本获得相同的人工智能工作负载，那么Meta的核心业务当然会看起来好得多。
So when Nvidia said that its new Vera Rubin chip delivers 30 times more throughput per megawatt and 35 times lower cost per token than its prior generation. Well, if that is true and Meta can get the same AI workloads for less money, of course, Meta's business, Meta's core business is going to look much much better.
目前，Meta的核心业务，基本上是广告业务，实际上表现非常出色。上一季度创造了594亿美元收入，同比增长27%，管理层表示这直接由人工智能排名驱动，这是增长的主要部分。无论你是在Instagram上还是在Facebook上，当然，如果广告定向变得越来越好，整个生态系统都会变得更好，广告商的广告支出回报率也会更好，这对广告商有利，也可能对消费广告的人有利，因为你现在会看到你可能感兴趣并且可能购买的广告。
Now, right now, Meta's core business, which is basically the ads business, is actually doing extremely well. has generated $59.4 $4 billion last quarter grew by 27% year-over-year and the management said that it was directly driven by AI rankings is that's basically the big part of the growth here whenever you go on on Instagram the res on Facebook you name it of course if ad targeting becomes better and better the whole ecosystem becomes better the advertisers the return on ad spend is better as well it's good for the advertisers it's probably good for for the people that consume the ads cuz now you're actually getting ads from things you might you might be interested in and you might maybe buy it as well.
所以这是一个胜利。你基本上可以免费使用服务。是的，你会被定向广告，但在我看来，这是双赢的。这当然可能会进一步推动Meta的电子商务业务。
So that's a win there. You get to use a service for free basically. Yes, you get targeted with advertising, but it's in my opinion a win-win everywhere. This could of course push the e-commerce play more and more for Meta.
但我知道这是一种间接的方式来解释Meta的业务如何从中受益。但我确实认为，因为Meta的核心业务今天已经如此盈利，如果突然他们购买的GPU也变得更高效、更具成本效益，那只会让他们的核心业务变得更好。
But I know this is a bit of an indirect way to explain how Meta's business benefits from all of this. But I do think that because Meta's core business today is already so profitable. If suddenly the GPUs that they're buying are becoming also more efficient, more cost effective for them, it will just make their core business even better.
然后下一步当然是他们自己的模型，对吧？MSpark几个月前发布了。我确实认为，是的，Meta花了一段时间才真正推出一个伟大的模型。我相信他们会在几周内推出下一个模型。其代号是watermelon。
And then the next step of course is their own models, right? MSpark came out a couple of months ago. I do think yes, it took it took time until Meta really came out with a great model. I believe they're going to come out in a couple of weeks with their next model. The code name there is watermelon.
但我确实认为，一旦他们意识到Muark对我们来说是一个很好的基础，当然它运行在英伟达的GPU上，他们就可以在此基础上进行构建，无论是用于智能眼镜、VR还是他们想做的其他一切。他们可能想向广告商、公司销售更多的人工智能服务，让它们在Meta自己的堆栈上运行。
But I do think that once they figured out that okay, Muark, that's a great baseline for us, which of course is running on Nvidia's GPUs as well. they can start building on top of it whether it's for their smart glasses VR or everything else that they want to do. They want to sell maybe more AI services to advertisers to companies to run their business on Meta's own stack.
如果他们能在硬件层变得越来越好的情况下做到这一点，那只会让整个公司受益。今天，我认为这是一个被极度低估的业务。它因为目前巨大的资本支出而被忽视。所以它现在给自由现金流带来了压力。
And if they can do this when the hardware layer becomes better and better, it just benefits the whole company. Today I view this as an extremely undervalued business. It is being overlooked because of the huge capex right now. So it is putting free cash flow under pressure today.
有一场审判正在进行。他们解决了这个问题，在我看来，他们也作为赢家走了出来，因为他们必须支付的部分款项只有在他们的主要竞争对手YouTube和TikTok参与这一变化时才会支付。
There was a trial that was going on. They settled that in my opinion. They they came out of it as as winners as well because there is part of the money that they have to pay that they're only going to pay if their big competitors YouTube and Tik Tok participate in the change.
因此，虽然Meta可能不是这里的直接受益者，但我确实认为他们间接地从即将到来的英伟达产品中获益。我认为这是对的。我的意思是，我们也容易忘记，没有实际的硬件，软件是无用的。我们看到扎克伯格基本上意识到，从其他公司租用服务器空间长期来看会是一个失败的策略。
所以Meta正在花费巨额资金来建设自己的数据中心。我认为这再次是为了让他们永远不必以那种方式依赖任何人。
And so while Meta might not be the direct beneficiary here, indirectly I do think they win from the upcoming uh Nvidia products. >> I think that's right. I mean also I I think it's easy to forget that the software is useless without the actual hardware, you know, to run it. And we saw that that Zuckerberg essentially realized that renting server space from other companies would be a losing strategy long term. And so Meta is spending an absolute fortune to build their own data centers. Um, and I think this is again so that they never really have to rely on anyone else in that way.
他们提高了今年的资本支出预期。根据他们公布的最新数据，我认为高达1450亿美元。其中很大一部分直接投入了Hyperion，这是他们在路易斯安那州的新500亿美元数据中心项目。但他们也在德克萨斯州埃尔帕索、印第安纳州、俄克拉荷马州积极建设人工智能优化的数据中心。所以在这方面有很多值得关注的。
You know, they've raised their capex expectations for the year. I think up to $145 billion as of the last numbers that they put out. And a significant chunk of that is going straight into Hyperion, which is their new $50 billion data center project in Louisiana. Um but they're also actively constructing AI optimized data centers in El Paso, Texas, in Indiana, uh in Oklahoma. So there is a lot to watch uh where that's concerned.
Meta当然也与私募股权和金融巨头如Blue Capital和Black Rock合作，为这些数十亿美元的项目提供资金。
And Meta has of course partnered with, you know, private equity and financial uh giants like uh Blue Capital and and Black Rockck to fund these multi-billion dollar projects as well.
所以当你看到这一点时，是的，他们在建设和芯片上花费了很多钱。我们看到他们的季度自由现金流暴跌至约7.84亿美元，而去年超过80亿美元。
So when you look at this, you know, yes, they're spending a lot of money on construction and chips. We saw their their quarterly free cash flow plummeted to about 784 million down from over 8 billion last year.
我认为底线是Meta需要证明这些投资确实能让他们的核心广告业务更有利可图。这不是他们能在一两个季度内证明的，对吧？这需要时间。有建设时间表，当然还有优化广告业务。但如果他们无法证明这一点，他们就会大规模烧钱。
I think the bottom line is Meta needs to prove these investments can actually make their core ad businesses more profitable. Now that's not something they can prove in a quarter or two, right? That takes time. There's construction timelines and then there's of course optimizing the ad business. But if they are not able to prove that, they'll just be burning cash at a huge scale.
但我确实倾向于认为Meta非常有可能在这一切之后变得更强大、更好。我认为这是它目前正在经历的成长烦恼。我认为它是在为增长而投资。但我们当然可以理解为什么一些投资者对他们最近财务报告中公布的一些数字并不满意。
But I do tend to think uh that Meta is very much going to come out on the other side of this a better and stronger company. I think these are the growing pains that it is experiencing right now. I think it's investing to grow. Um but we certainly can understand why some investors have not been thrilled by some of the numbers we've seen come out in their recent financial reports.
是的，你知道最坏的情况是他们将以高溢价出售计算能力，他们告诉我们他们对此有大量需求，他们可以收取溢价，这意味着如果他们决定比如说将大型GPU集群用于外部人员，那么他们可以告诉市场，看，这是我们能提供给你们的数字，你们对这个数字满意吗？
因为亚马逊、谷歌、微软的一大优势是云业务。是的，我们花费了数千亿美元，但看，我们的云业务正在加速，利润率正在扩大。对于Meta来说，现在他们只能说我们的核心业务正在变得更好，但他们不能真正告诉我们，哦，我们纯粹的广告业务、广告人工智能业务正在变得越来越好，因为他们没有云业务。
也许将来他们会细分人工智能驱动的广告服务之类的。也许订阅服务可能在接下来的12到18个月左右，但就目前而言，还没有。
>> Yeah. And you know what worst case scenario they are going to sell compute for a huge premium and and they've told us that they've had a lot of demand for it and they can charge a premium which means they can then if they do decide to let's say use a a big GPU cluster for outsiders then they can tell the market look this is the number that we can provide you with regards to that are you happy with that number because the big advantage that an Amazon a Google a Microsoft have is is the cloud business. Yes, we're spending hundreds of billions of dollars, but look, our cloud business is accelerating, margins are expanding. With a meta right now, they can just say our core business is getting better, but they cannot really tell us, oh, our pure ads business uh ads AI business is getting even better and better because there is no cloud business for them. Maybe in the future they'll break down at AIdriven ad services or something like that. maybe subscription services probably over the next maybe 12 to 18 months or so, but as of right now, not