ARK投资的布雷特·温顿谈人工智能的下一阶段、亚马逊药房的野心与iPhone延期 | 2025年11月11日

欢迎各位收看Informations TI TV。

Welcome everyone to the Informations TI TV.

我叫阿卡什·帕什里查。

My name is Akash Pashricha.

今天是11月11日，星期二。

It is Tuesday, November 11.

今天我们为大家准备了一场精彩的节目。

We have got a great show lined up for you today.

首先，我们将讨论我们独家报道的内容：由于首款型号销售未达预期，苹果正在推迟其下一代iPhone Air的发布。

First up, we are discussing our exclusive reporting that Apple is delaying the release of its next iPhone Air after the first model isn't selling as expected.

接着，我们将采访ARK投资的首席未来学家布雷特·温顿，探讨人工智能的估值以及下一轮重大科技创新将是什么。

We're then talking to ARK Invest's chief futurist, Brett Winton, about AI valuations and what the next big wave of tech innovation is going to be.

接下来，我们将转向我们发布的另一条独家新闻。

Next up, we'll turn to another scoop we published.

OpenAI的Stargate项目获得了Blue Owl Capital高达30亿美元的投资。

OpenAI's Stargate effort is getting a $3,000,000,000 investment from Blue Owl Capital.

此外，我还将与亚马逊药房的一位高管讨论该公司整体的药房战略。

And separately, I'm also talking to a top Amazon pharmacy executive about the company's pharmacy strategy at large.

最后，我们将与Epidemic Sound的首席执行官进行对话。

Finally, we're going to end the show with a conversation with the CEO of Epidemic Sound.

该公司推出了一项新工具，帮助创作者为视频配乐。

The company unveiled a new tool to help creators soundtrack their videos.

这是一期内容丰富的节目，让我们马上开始。

It's a busy show, so let's get right on into it.

苹果的新款iPhone Air销量未达预期。

Apple's new iPhone Air isn't getting the uptake that it hoped for.

这是根据我们在《The Information》上发布的一篇新报道。

That is according to a new story that we published in The Information.

我的同事报道说，由于首款型号的销售低于预期，公司正在推迟下一款iPhone Air的发布。

My colleagues reported that the company is delaying the release of its next iPhone Air as the first model sales have been below expectations.

我们还报道说，苹果可能考虑重新设计该型号，为第二代产品增加第二个摄像头镜头，而第一代仅有一个摄像头。

We also reported that the company could be looking to redesign the model to include a second camera lens after the first model had only one.

现在加入我们讨论这一话题的是吉尔·卢里亚，DA Davidson的董事总经理兼科技研究主管。

Joining me now to discuss that is Gil Luria, Managing Director and Head of Technology Research at DA Davidson.

吉尔，欢迎来到TI TV。

Gil, welcome to TI TV.

很高兴你能来到这里。

It's great to have you here.

谢谢你的邀请。

Thanks for having me.

那么，我们来谈谈我们昨天报道的这条新闻。

So let's talk about this news that we reported yesterday.

你对此有什么看法？

What did you make of it?

当我们第一次看到这款iPhone时，它看起来非常酷。

Well, when the when we first saw the iPhone there, it looked very cool.

它甚至看起来很有魅力。

It almost looked sexy.

对吧？

Right?

我们都想要更薄、更轻的手机，所以它确实有吸引力。

We all want a thinner, lighter phone, and so there was an appeal there.

但一旦顾客拿到手，他们就发现了几个问题。

But once customers put their hands on it, they realized a couple of things.

一是它其实并没有那么薄，因为摄像头凸起让它在最厚的地方和别的手机一样厚。

One is it's not actually that thin because the camera bump means that it's at its thickest point.

它的厚度和其他iPhone一样。

It's as thick as any other iPhone.

而且，牺牲的是电池续航。

And then also that the sacrifice is in the battery life.

而电池续航是消费者升级手机时最重要的考虑因素之一，如果不是最重要的的话。

And that's something that is one of the top, if not the top consideration for consumers to upgrade their phones.

所以，虽然它通过‘性感’的设计为苹果增添了一些光环和声望，但主要促使消费者购买了Pro和Pro Max型号，而这其实是个好事，因为这些是苹果定价更高的产品。

So while it did add some halo and cache to Apple by being sexy, it mostly drove consumers to buy the Pro and Pro Max models, which, by the way, is a good thing because those are higher priced items for Apple.

对。

Right.

那么，对于苹果来说，推出这款折扣型号有多重要？推出这么多不同版本的iPhone又有多重要？

Now so how important was it for Apple to unveil this discounted model, and how important is it that it comes out with all these different variations of the iPhone?

因为正如你所说，任何iPhone的销量对苹果都有好处。

Because as you said, the iPhone segment any iPhone sales are good for that.

是的。

Yeah.

推出iPhone Air的部分目的，是为了推高Pro系列的价格定位。

So part of the introduction of iPhone Air was to push the price points of Pro up a little bit.

通过在iPhone和Pro之间增加一个型号，他们成功地提升了Pro机型的定价。

By adding something between the iPhone and the Pro, they were able to then wedge up the Pro model.

这也在帮助他们。

So that's helping them as well.

因为今年恰好是大规模的升级周期。

Because it turns out that this year is a big upgrade cycle.

我们原本以为去年会迎来一次升级周期，因为有新的AI功能。

We thought the upgrade cycle would be last year with new AI functionality.

结果发现，消费者还没有准备好接受苹果提供的AI功能。

Turns out that consumers aren't ready for the AI Apple has to offer.

苹果所提供的AI功能与ChatGPT应用中的功能相比，并没有太大的增量改进。

Apple isn't really offering very, incremental AI to what you can get in the ChatGPT app.

因此，我们最终得到的结论是：手机已经用了四五年了，所以消费者在更换手机。

And so all we ended up with is, hey, phones are about four or five years old, so consumers are upgrading them.

在他们更换手机的同时，苹果成功地推出了Air型号。

And as they're doing that, Apple was able to wedge in the Air model.

再次，为它增添一点吸引力，但更重要的是，抬高了Pro和Pro Max的价格，让消费者在升级时实际上购买了更昂贵的手机。

Again, give it a little bit of a sexy halo, but more importantly, push up the prices of Pro and Pro Max so consumers are actually buying a more expensive phone as they upgrade.

所以我们现在报道说，该公司计划推迟iPhone Air新机型的发布，无论它最终何时推出。

So now we've reported that the company is planning to delay the release of its new model for the iPhone Air whenever that ends up being.

你认为该公司最终彻底取消这个型号的可能性有多大？

What do you think the likelihood is that the company actually just ends up scrapping the model altogether?

好吧，我们得等到明年才能知道。

Well, we we won't find out until next year.

而且明年更有可能发生的是推出一款可折叠的iPhone。

And what's more likely to happen next year is a foldable iPhone.

我们已经拥有可折叠的三星手机一段时间了。

We've had foldable Samsung phones for a little while.

它们相当受欢迎。

They're quite popular.

因此，苹果只是想确保做得完美。

And so, Apple just wanted to make sure to get it right.

他们想确保它看起来好看。

They wanna make sure that it looks right.

它能顺利折叠。

It folds right.

它看起来很漂亮。

It looks nice.

它不会损坏。

It doesn't break.

而且听起来他们明年就会准备好，因此升级iPhone Air将不再是必需的。

And it sounds like they'll be ready for that next year, which is upgrading the iPhone Air won't be a necessity.

也许他们会在那之后的一年才这么做。

Maybe they'll do it for the year following that.

而且，这可能会包含更多的摄像头，或者减少一些功能，以便整体更薄。

And again, it'll be either including more cameras or maybe less functionality so it's thin throughout.

但我们可能还要再等一两年才能知道它会是什么样子，因为明年重点似乎是折叠屏手机。

But we may not find out for another year or two what that looks like because it looks like the emphasis for next year will be the foldable phone.

我很好奇，折叠屏手机。

I am curious, foldable phone.

你能谈谈你对苹果决定转向折叠屏手机的看法吗？

Can you think about tell us about what you make of Apple's decision to move towards a foldable phone.

正如你所说，其他公司已经推出了这种产品。

As you said, other companies already have it.

我不认为苹果是一家会刻意追赶竞争对手的公司。

I don't see Apple as really someone who is going after their competition necessarily.

他们一直以创新为傲。

They've always prided themselves on innovation.

你认为为什么折叠屏手机现在成为他们的重点呢？

Why do you think the foldable phone is such a focus for them right now?

因为他们看到三星已经取得了成功。

Because they see that Samsung has been successful.

所以在这种情况下，这完全是竞争驱动的。

So it is fully competition in that case.

是的。

Yeah.

是的。

Yeah.

苹果无疑是非常创新的。

And Apple is very innovative for sure.

他们的产品非常吸引人。

And their products are very compelling.

我们都对他们非常忠诚。

We're all very loyal to them.

但现实地说，他们已经很久没有推出新产品了。

But realistically speaking, they haven't introduced a new product in a long time.

他们一直在观察市场感兴趣的方向，然后以自己的水准去实现——这种水准能令消费者感到惊喜，再推出相应的产品。

They have been mostly seeing what the market is interested in and then doing it at their level, which is a level that delights consumers and then coming up with that.

这让许多三星安卓用户感到沮丧，他们觉得自己才是创新的先行者，率先推出好点子，而苹果只是抄袭他们。

Much to the frustration of Samsung Android users who feel like they are innovative and get things right first and then Apple just copies them.

苹果并不在意，因为只要他们做对了，就能卖出更多手机，赚取更多利润。

Apple doesn't care because if they get it right, they'll sell a lot more phones and generate a lot more profits.

如果我们稍微放宽视野，看看苹果过去几年在整体设备生态系统中的创新努力，我想听听你的看法：这是否是一个创新与推出客户真正想要的产品之间的挑战？

If we just zoom out a little bit and look at Apple's efforts to innovate in the device ecosystem at large over the past few years, One thing I wanted to get your take on is whether this is a challenge of innovation and delivering a product that customers actually want.

我们已经看到了像Vision Pro这样的产品，说实话，要打造出客户真正想要的东西很难。

We've seen with stuff like the Vision Pro, mean, it's hard to build things that customers want.

还是说，另一个挑战在于苹果公司规模太大，想要在如此庞大的公司中掀起波澜真的很难？

Or is the other end of the challenge just that Apple is so big, it's really hard to make a dent in a company that is the size of Apple?

这两个挑战中，你觉得哪一个现在更严峻？

Which one of those two strikes you as the bigger challenge right now?

嗯，这两个都是挑战。

Well, they're both challenges.

但对他们来说，更重要的是把事情做对。

But, really, for them, it's much more important to get it right.

所以，我反而认为Vision Pro在某种程度上是一次巨大的成功。

So I would actually say that the VisionPRO in its own way was a big success.

如果你去过苹果商店并体验过Vision Pro，那种体验非常引人入胜。

If you've been to an Apple store and did a trial with the VisionPRO, the experience is incredibly compelling.

你周围的整个世界都栩栩如生。

It's the the world around you is vivid.

你可以对它进行大量控制。

You have a lot of control over it.

这是对未来的一次精彩一瞥。

It's a great glimpse into the future.

所以价格必须降低。

So the price point has to come down.

头戴设备必须变得更轻。

The the headset has to get lighter.

电池组可以是外置的。

The battery pack can be external.

这些都将是他们要解决的问题，因为正如你所说，他们需要另一个重磅品类。

These are all things that they will work out because to your point, they need another big category.

我们知道它会是什么样子。

And we know what that's gonna look like.

它将包含两个要素。

It's gonna be two elements to it.

将会有一种头戴式护目镜，取代台式机、电视等设备。

There's gonna be the, the overhead goggles that will will replace desktops and TVs and things like that.

当我们静止不动时，我们将佩戴头戴式护目镜，沉浸在虚拟世界中，而不是通过24英寸、27英寸甚至90英寸的电视屏幕来观看。

When we're stationary, we're gonna have overhead goggles that immerse us in a world instead of viewing it through a 24 inch screen or a 27 inch screen or even a 90 inch TV.

我们会完全沉浸其中。

We'll get immersed in that.

而另一类则是可穿戴设备，比如眼镜。

And then the other category is gonna be the wearables, the glasses.

在这方面我们已经知道它会是什么样子了，因为Meta已经在引领这一方向，其Ray-Ban眼镜正变得越来越好，帮助我们在移动中增强现实体验。

And there we already know what that looks like because meta's actually leading the way with the meta Ray Bans that are getting better and better at providing us the ability to enhance our our world and as we are in motion.

因此，当我们不坐在办公桌前或沙发上时，这两个新平台就显得尤为重要。

So when we're not sitting down at our desk or on our couch, those are the two place new platforms that are important.

Meta正在全力争夺这两个领域的主导权。

Meta's racing to win in both of those categories.

我们可以期待谷歌以自己的方式参与其中。

We can expect Google to participate in its way.

嗯。

Mhmm.

苹果做对了。

Apple gets it right.

苹果让消费者感到愉悦。

Apple delights consumers.

如果你试过Vision Pro，就会发现它是一种非常引人入胜的体验，等到它正式上市时，我们就能享受到这种体验。

And if you've tried the Vision Pro, it is a very engaging, experience that we are going to have when it's ready for prime time.

巴里，快点，在你走之前，你提到了谷歌。

Barry, quickly, before we let you go, you mentioned Google.

我们有报道称，苹果正在推进与谷歌的合作，以在其Siri中使用其AI技术。

We have the reporting that Apple is marching towards a deal with Google to use its AI in Siri.

你认为这反映了苹果整体AI战略的什么信息？

What do you think that tells you about Apple's AI strategy at large?

实际上，这是一件好事，因为他们拥有消费者，可以灵活选择向消费者提供哪种模型，而无需做出长期承诺。

Actually, it's a good thing that they have the consumer so they can choose what model to feed the consumer and they don't have to make long term commitments.

他们会现在就给消费者提供他们能提供的东西。

They'll give the consumer what they what they can give them right now.

他们会关注这些模型的发展。

They'll watch the models develop.

他们会弄清楚谁是赢家，然后将该模型提供给他们的消费者。

They'll figure out who the winner is and they'll supply that model to their consumers.

因此，他们无需在数据中心和模型交易上投入数千亿美元，而是让谷歌和OpenAI去做，让他们相互竞争，看谁能赢得苹果消费者的青睐。

So instead of spending hundreds of billions of dollars on data centers and trading models, they're letting Google do it and OpenAI do it and they're letting them race each other to see who gets access to the Apple consumer.

这就是我们目前所处的状况。

And that's where we're at.

我们根本不在乎这是哪个模型，只要Siri能理解我们在说什么并提供帮助就行。

We just we don't care whose model it is as long as Siri can understand what we're talking about and be helpful.

很好。

Great.

吉尔，很高兴你能来参加我们的节目。

Well, Gil, it's great to have you on the show.

谢谢你前来做客。

Thank you for coming on.

我们非常感谢。

We really do appreciate it.

谢谢。

Thank you.

好的。

Okay.

今天人工智能板块承压。

The AI trade is under pressure today.

有几则新闻成为焦点。

A couple stories in the spotlight.

CoreWeave在下调全年收入预测后股价下跌。

CoreWeave is falling after lowering its full year revenue forecast.

该公司表示，由于一家第三方数据中心开发商进度滞后，导致项目出现延误。

The company reported it is facing delays due to a third party data center developer who is behind schedule.

软银出售了其在英伟达的全部股份，套现超过50亿美元，英伟达股价也因此下跌。人工智能板块曾推动大型科技股大幅上涨，推高了创纪录的估值。

NVIDIA is also in the red after SoftBank sold its entire stake in the company for more than $5,000,000,000 The AI trade has powered a massive run up in mega cap stocks, fueling record valuations.

但经过数月的狂热之后，投资者仍在质疑，这种炒作是否超前于基本面。

But after months of euphoria, investors are still asking if the hype is ahead of the fundamentals.

为此问题提供解答的是ARK投资的首席未来学家布雷特·温顿。

Here to help answer that question is Brett Winton, Chief Futurist at ARK Invest.

布雷特，欢迎来到TI TV。

Brett, welcome to TI TV.

很高兴你能来到这里。

It's great to have you here.

谢谢邀请我参加。

Thanks for having me on.

我想先从整体的AI投资谈起。

So I want to start with the AI trade at large.

你如何看待这些巨头公司的股价表现以及更广泛的市场走势？

What do you make of the way that these mega cap companies are trading right now and the broader market movements at large?

我认为现在显然存在一种说法，认为AI是一个泡沫，不值得在数据中心资本支出上投入这么多钱，因为我们无法从这些技术中获得实际效益。

I think there's clearly this narrative out there that AI is a bubble, that you shouldn't spend so much money on data center CapEx, that we're not gonna get utility out of these things.

我认为这完全错了。

I think it's flat wrong.

从上至下和从下至上都有理解它的方法。

There's, like, top down ways and bottoms up ways to understand it.

但总体而言，今年我们将拥有十亿AI聊天机器人用户，到2020年代末、2030年之前，这一数字将达到40亿至50亿。

But at a very high level, we're gonna go from a billion AI chatbot users this year to roughly four to 5,000,000,000 by the late 2020s and before 2030.

我们认为，像ChatGPT这样的AI生产力工具如今能为知识工作者带来20%的生产力提升。

And we think that AI productivity tools like ChatGPT deliver knowledge workers a 20% productivity boost today.

我们预计这一数字在今年内将翻三倍，然后再次翻三倍。

We think that's going to triple over the course of this year and then triple again.

因此，在两年内，他们的定价能力应提升约9倍，而用户数量在稍长一段时间内将增长5倍。

And so, you can over two years, their pricing power should increase by around 9x, and the knowledge the number of users over a little longer should increase by 5x.

因此，你可以将两者叠加，得出整个领域的定价能力大约提升了40到50倍。

So you can compound the two to say the pricing power in the entire space is roughly, you know, 40 to 50 x higher.

所以，如果你看看OpenAI或Anthropic的收入，我认为人们所讨论的收入预测相对于它们实际拥有的定价能力而言是保守的，因为这些神奇的工具将带来巨大的生产力提升，足以覆盖它们所计划投入的数据中心支出。

So if you look at OpenAI's revenue or Anthropix revenue, I think actually the revenue projections that people are talking about are conservative relative to the pricing power that they have because these magical artifacts are gonna deliver amazing productivity advances more than enough to underwrite the data center spend that they're talking about engaging in.

但让我问你一个问题。

But let me ask you this.

你谈到了定价权。

You talked about pricing power.

麦肯锡最近发布了一份报告，称全球39%的企业员工表示人工智能对运营利润产生了一定影响。

McKinsey put out a report recently that said that 39% of employees at corporations around the world say that AI has had any impact on operating profit.

而他们最多只将5%的运营利润归因于人工智能的使用。

And at most, they attribute 5% of their operating profit to the use of AI.

我之所以问你这个问题，是因为盈利能力无疑是个关键问题。

And I ask you that because profitability has certainly been the question.

你是认为定价权将是这些公司实现盈利的方式吗？还是你如何看待这一情况的发展？

Is pricing power the way you see these companies ending up generating a profit, or how do you see that playing out?

是的。

Yeah.

首先，据我们所知，从毛利润角度来看，这些公司实际上已经实现了非常健康的利润率。

Well, one, as far as we know on a gross profit basis, some of these companies are actually printing very healthy margins.

要理解这一点，你和我都是知识工作者。

The way to understand it, it's you and I are knowledge workers.

我肯定你用过ChatGPT。

I'm sure you use ChatGPT.

我用它来完成复杂的任务，就像一个思维伙伴，用于代码生成、数据查询，真正为ARC构建预测模型。

I use it to do complex tasks as kind of like a thought partner and something to do kind of code generation, data querying, to really build out forecasting models on behalf of ARC.

ARC里的每个人都在使用它。

And everybody in ARC uses it.

所以我能指出，哦，这个查询带来了多少利润吗？

And so am I able to point to, oh, this query led to this profitability?

不能。

No.

但我可以说，我完成的工作量比没有AI辅助时多得多。

But I'm able to say, hey, I got a lot more done than I otherwise would able be able to get done in a work free AI.

企业为提高生产力的软件付费时，并不会全额支付。

And the way businesses pay for productivity enhancing software, they don't pay full freight for it.

他们大约支付10%。

They pay roughly 10%.

我们认为，到2030年，AI软件的支出将接近一万亿美元。

Well, we think almost $10,000,000,000,000 is gonna be spent on AI software by 2030.

因此，你可以计算一下，我们假设提供给知识工作者的生产力提升足以支撑这一预测。

So you can do the math on kind of the productivity that we assume is going to be delivered to knowledge workers to justify that forecast.

但这足以支持我们今天所看到的数据中心建设。

But it's more than enough to support the data center build out that we see today.

我想问问你ARC有多少种策略。

I want to ask you about the number of strategies that ARC has.

我的意思是，你无处不在。

I mean, you're everywhere.

你在软件、机器人、数字资产、医疗保健等领域都有涉足。

You're in software, robotics, digital assets, healthcare.

在研究这些领域时，你目前花最多时间的是哪个类别？

Which of these categories are you personally spending the most time on right now as you study these spaces?

在我们的风投基金中，我们对人工智能有大量布局。

Well, certainly in our venture fund, we have a lot of exposure to AI.

在我们的前几大持仓中，我们持有OpenAI、XAI和Anthropic。

In our top holdings, we have OpenAI, XAI, and Anthropic.

因此，我们认为，到本十年末，基础模型平台 collectively 的价值将达到15万亿到20万亿美元的规模。

So we think that the foundation model platforms are gonna be remarkably valuable on the order of 15 to $20,000,000,000,000 by the end of the decade collectively.

而在ARCW——也就是我们专注于人工智能的基金中——

And, you know, in ARCW, which is really our AI focused fund.

显然，我们对平台即服务（PaaS）有大量敞口，包括Palantir等公司，同时也涉及基础设施即服务（IaaS）。

There's obviously a lot of platform as a service exposure, including Palantir and others, but also infrastructure as a service.

而特斯拉实际上是全球最大的人工智能项目。

And then Tesla is is really the largest AI project in the world.

你可以看看即将发生的Robotaxi变革，然后推断：四到五年后，我们将在人形机器人上看到同样的能力进步，从而理解为什么马斯克那近乎疯狂的愿景很可能改变世界。

And I think you you can look at what's about to happen with Robotaxi, and you can say, Hey, four to five years after, we'll see the same capability advance in humanoid robots and understand why Elon's maniacal vision is likely to change the world.

我想稍后再回到特斯拉的话题。

I want to come back to Tesla in a minute.

但除了人工智能之外，我问这个问题是因为Arc的战略非常多元化，目前你在哪些领域投入了最多的时间？

But outside of AI, and I ask this question because Arc has such a multifaceted strategy, outside of AI, where do you find yourself spending the most time right now?

是的。

Yeah.

我们相信，五大基础创新平台正在彻底改变经济世界。

We believe that five fundamental innovation platforms are basically changing the economic world.

包括加速一切的人工智能在内，我们还关注多组学，即治愈方法和生物数据将如何改变我们治疗健康的方式，以及能源存储和公共区块链。

So inclusive of AI, which is accelerating everything, we also focus on multi omics, which is really how cures and biological data are going to change the way that we treat health, energy storage, as well as public blockchain.

因此，我们拥有比特币ETF，以及大量数字资产和数字钱包的敞口。

So we have our Bitcoin ETF as well as a lot of digital assets exposures and digital wallets exposures.

但所有这些都正在被人工智能加速。

But it's all being accelerated by artificial intelligence.

这一轮周期中一个非常显著的因素是，一切进展都比我们预期的还要快，因为人工智能的发展速度实在太快了。

One of the really striking factors about this cycle is everything is moving forward faster than even we anticipated because AI is moving so quickly.

现在，如果我们聚焦到机器人领域，我们已经邀请过很多人上节目，许多研究这一领域的人士都在谈论机器人领域令人难以置信的估值。

Now, if we zoom in on robotics, we've had people come on the show, a lot of people studying this space, talking about the the bonkers valuations in robotics.

机器人显然与特斯拉有关，而特斯拉是一家我知道你密切关注并大力支持的公司。

And robotics is obviously tied to Tesla, which is a company that I know you you follow and and and back very closely.

你认为人们在看待机器人趋势及其未来发展时，普遍误解了什么？

What do you think people get wrong when they think about the robotics trend and how that is going to play out?

老实说，我担心人们低估了人形机器人将面临的困难和复杂性。

I mean, honestly, I worry that people are underestimating how difficult and complex a problem humanoid robots are going to turn out to be.

我们认为，如果你能大规模商业化机器人出租车——我们相信特斯拉今年就能做到——你将长期产生数百亿美元的现金流。

We think that if you commercialize robotaxia at scale, which we believe Tesla's going to do this year, you generate hundreds of billions of dollars in cash flow over time.

而你必须将其中很大一部分投入AI计算和制造，以实现人形机器人达到同等通用化规模。

And you'll have to spend a lot of that on AI compute and manufacturing to get humanoid robots to the same generalizable degree of scale.

我们认为，这些估算的误差范围很大，但总体上，商业化人形机器人比商业化机器人出租车难约二十万倍，意味着你需要投入那么多额外的AI算力。

We think it's, you know, call it a wide error bar on these estimates, but we think it's net a 200,000 times harder problem than commercializing Robotaxi, meaning you have to throw that much additional AI compute power at it.

虽然你现在在架构改进和芯片性能提升方面取得了巨大进展，但要让人形机器人达到人们所期望和承诺的水平，很可能需要投入数千亿美金的投资周期。

Now you're making great architectural improvements and kind of the quality of the chips is improving, but it really is a multi $100,000,000,000 probably investment cycle to get humanoid robots to the state that people are hoping for and promising.

我们已经报道过特斯拉在推动人形机器人研发、实现其预期进度时所面临的某些挑战。

We've reported on some of the challenges that Tesla is facing getting those humanoid robots to develop at the pace that they would have hoped.

你如何看待这些挑战？

What do you make of those challenges?

是的。

Yeah.

我的意思是，关节的灵活性、执行器、软件栈、为大规模制造而设计，这些都是巨大而棘手的工程难题。

I mean, the articulation of the joints, the actuators, the software stack, designing for manufacturing scale, these are all huge, thorny engineering challenges.

我们认为，最终投资人形机器人将在未来十年内带来惊人的资本回报。

And we think ultimately investing in humanoid robots will be an amazing return on invested capital measured over the course of the decade.

目前看来，人形机器人的商业化实现大规模应用可能要到2029至2030年左右。

We think commercialization is probably in a meaningful scale a twenty twenty nine to 2030 event at this point.

对我们而言，对特斯拉的估值主要基于Robotaxi。

And for us, the underwrite on Tesla is mostly robotaxi.

我的意思是，你能买到全球最畅销的汽车，只需多花几千美元，就能获得随时可用的机器人司机选项，这真的令人惊叹。

I mean, don't the fact that you can buy the the best selling car in the world and pay a few thousand extra dollars, and it comes with the option to have a robotic chauffeur that you can use at any time is really astounding and amazing.

这彻底改变了特斯拉的商业模式，从卖车转变为通过行驶里程收取平台费用。

And and that transforms the business model of Tesla from selling cars to collecting revenue platform fees off miles driven.

如果我们转向谈谈芯片行业和英伟达，你认为哪家公司最有可能挑战像英伟达这样的公司？

If we pivot to talking a little bit about the chip sector and about NVIDIA, who do you think has the best chance of challenging a company company like NVIDIA?

我的意思是，我认为不会有一家公司能直接挑战他们的主导地位。

I mean, I don't think it's gonna be a single company that challenges their primacy.

相反，会有很多垂直整合的玩家逐步蚕食他们的市场份额。

Instead, there are gonna be a lot of vertically integrated players kind of nibbling at the edges of their market share.

AMD 提供了相当全面的产品组合，由于配备了更多的显存，显然正在获得越来越多的市场份额。

And AMD has a a a reasonably comprehensive offering that is clearly gaining traction because of the larger amount of attached memory.

我们评估和看待英伟达的方式是：到2030年，AI数据中心的支出将达到3到4万亿美元，我们确实认为英伟达的市场份额会随着时间推移而下降。

The way we underwrite and and consider NVIDIA is is there's gonna be 3 to $4,000,000,000,000 in AI data center spend by 2030, and we do think NVIDIA's share is going to fall over time.

但这个数字如此庞大，即使在当前价格下，它依然处于一个很有吸引力的位置。

And so but the number is so large that it's still in an interesting position even at these prices.

但芯片行业无疑是一个周期性行业。

But the chip space is definitely a cyclical space.

不可避免的是，我们会到达一个阶段，届时人们过度采购了芯片，市场将出现调整。

Inevitably, we'll get to a point where people have overbought chips and there will be a correction.

这可能仅仅基于电力供应的可用性。

It could be just on the basis of power availability.

嗯。

Mhmm.

谁会受到伤害？

And Who gets hurt?

在那次调整中谁会受到伤害？

Who gets hurt in that correction?

我的意思是，芯片供应商会受到冲击，英伟达也会受损。

Well, I mean, you know, the chip providers, NVIDIA would get hurt.

AMD也会受损。

AMD would get hurt.

如果你买到了芯片却无法安装，突然间就会出现库存周期。

You you it if you if you can't install the chips that you buy, then suddenly there's going to be an inventory cycle.

看看半导体行业的历史。

Look at the history of semis.

库存周期会发生，突然间毛利率下降，而不是持续向上增长。

Inventory cycles happen, and suddenly gross margins fall as opposed to going up into the right perpetually.

然后，人们会说这是周期的结束，但根据我们对中长期资本支出的预期，这实际上可能是投资这些领域的一个良好机会。

And and then there's a you know, what people will say is the end of the cycle, but it's really probably an interesting buying opportunity into some of those exposures given our expectations for CapEx spend over the medium and long term.

但我必须提醒大家，事实并非如此。

But I I I would caution people, it's not.

半导体行业并不是一直向上增长的。

Semis is not up into the right.

事情并不是这样运作的。

That's not the way it works.

这一轮周期也会是一样的。

And it's it's gonna be the same this cycle.

在这个过程中会出现一些波折。

There are going to be hiccups along the way.

因此，芯片业务的周期性特点，再加上特斯拉明显转向垂直整合芯片，以及Tranium与亚马逊、谷歌商业化TPU等因素。

And so combination of that cyclical nature of the the chip business and then plus, the likes of Tesla clearly pushing into vertically integrated chips and others, Tranium with Amazon and Google commercializing TPUs.

这些都是一些竞争因素，但并不意味着它们会立即让英伟达陷入困境。

These are all competitive vectors that it's not like they immediately pull the rug out from under NVIDIA.

相反，它们只是在某个时候对其增长施加了压力。

Instead, they just put pressure on their growth at some point.

在我们结束之前，我问你个简单的问题。

Let me ask you a quick question before we go.

我们最近邀请了黑石的金东贤做客节目。

We had Tony Kim from BlackRock on the show recently.

他提出了一个观点，即那些在资本支出上大手笔投入的大型科技公司，从长远来看，一种负担这些资本支出的方式可能是增加债务，而它们此前一直尽量避免这样做。

He floated this idea that the big tech companies that are spending heavily on CapEx, one way for them to afford that CapEx in the long run could be loading up on debt, which is something that they've largely been able to avoid.

你觉得这是个好主意吗？

Do you think that's a good idea?

你怎么看这个观点？

What do you think of that?

我认为，这些确实是可以通过举债来融资的资本性资产。

I mean, I do think these are capital assets that you could layer debt against.

历史上，创新领域的重大资本支出周期都是通过债务融资的，而债务市场的规模远大于股权市场。

Historically, big CapEx cycles in innovation have been debt funded, and the debt markets are simply much larger than the equity markets.

风险在于，这些资本支出投资的折旧周期很长，且具有周期性。

The danger is you have long depreciation schedules and the cyclical nature of these CapEx investments.

因此，肯定有人会陷入困境。

And so certainly people could get offsides.

我认为，数据中心的投资回报率（ROIC）将会非常惊人。

I think ROIC on data centers, the return that you get in investing in data centers is going to be amazing.

但这并不意味着对每个人来说都会如此出色。

But that doesn't mean it's going to be amazing for everyone.

我认为，不可避免地，一些玩家会过度扩张，最终将出现行业整合。

I think inevitably, some players are going to get over their skis, and there's going to be a consolidation event.

但当我们回顾这个商业周期的终点时，会庆幸资本市场的资金支持了这些卓越的、提升生产力的计算芯片的建设，使我们拥有了所有这些神奇的能力：自动驾驶汽车、上市成本降低75%且上市速度加快40%的药物，以及每天都能让生活更轻松的模型——所有这些都将是我们当前正在建设的数据中心的副产品。

But we'll look back at the end of this business cycle and say, Wow, thank goodness the capital markets funded this build out of these amazing, productivity delivering compute chips so that we have all of the magical capabilities, cars that can drive themselves, drugs that cost 75% less to bring to market and are at market 40% faster, models that literally make your life easier day in and day out, all of those will be byproducts of this data center build that we're seeing right now.

很好。

Great.

而且

And

布雷特，在你走之前，我快速问一下：你现在最痴迷的一个反向投资思路是什么？告诉我们吧。

Brett, very quickly before I let you go, one contrarian trade idea that you're obsessed with right now, give it to us.

我不确定是否要说出具体的名字，但我认为华尔街低估了治愈性疗法的价值。

I don't know if I want to give you an individual name, but I think that Wall Street underestimates how valuable cures are.

人们认为，卖一种患者只需服用一次、无需重复使用的药物是个糟糕的商业模式，因为人们更喜欢持续性的收入。

As in people think that, Hey, it's a bad business to sell a drug that somebody just takes once and they don't have to take it again, because people like recurring revenue.

但如果你计算一下治愈性疗法的定价，你会获得患者整个生命周期的支出总额，再加上你为患者带来的效用。

But if you do the math on how cures are going to be priced, you get the entire lifetime value of spend of the patient upfront, plus the utility you deliver to the patient.

因此，治愈性疗法的价值是典型慢性病治疗的两倍。

So cures are two times more valuable than the typical chronic treatment.

那么，你现在最专注的治愈性疗法是哪一种？

And so what is the cure that you are laser focused on right now?

好的。

Okay.

这是一种你从未听说过的疾病，称为遗传性血管性水肿，美国有七千人患有此病。

Well, it's a disease that you've never heard of, but hereditary angioedema, seven thousand people have it in The US.

他们每年必须花费50万美元用于预防性治疗，以防止自己住院，但即便如此，他们有时还是会住院。

They have to spend $500,000 per year on prophylactic treatments just to prevent themselves from going to the hospital, and they sometimes go to the hospital anyway.

因此，遗传性血管性水肿有一种潜在的治愈方法。

So there is a prospective cure for hereditary angioedema.

如果合理定价，这种疗法的价格应在数百万美元的高位，大约300万到400万美元，因此这是一个200亿到300亿美元的市场，而开发这一疗法的公司Intellia，由于另一项完全不同的治疗方案的数据结果，股价大幅下跌。

It should price, if it were fairly priced in the high millions of dollars, but maybe $3 to $4,000,000 And so it's a $20 to $30,000,000,000 market, and the company that's delivering it, Intellia, is down a lot based on a data readout from an entirely different treatment.

我一直在等你是否真的会说出这家公司的名字。

I was waiting to see if you would actually say the company.

那么这家公司是？

So the company is?

Intellia。

Intellia.

Intellia Therapeutics。

Intellia Therapeutics.

是的。

Yes.

好的。

Okay.

好吧。

Alright.

布雷特，谢谢你来参加节目。

Well, Brett, I wanna thank you for coming on the show.

很高兴有你。

It's great to have you.

和ARC的人聊天总是很愉快，我们很快还会再邀请你来节目。

Always a pleasure chatting with folks from ARC, and we'll have you back on the show again soon.

非常感谢。

Thank you so much.

OpenAI的Stargate数据中心项目获得了Blue Owl Capital 30亿美元的投资。

OpenAI's Stargate Data Center project is getting a $3,000,000,000 investment from Blue Owl Capital.

蓝猫资本最广为人知的身份是华尔街最大的私人贷款机构之一，但其在支持数据中心项目时采取了略有不同的方式。

Blue Owl is best known as one of Wall Street's biggest private lenders, but it is taking a slightly different approach to backing data center projects.

现在邀请我们的人工智能与金融记者迈尔斯·克鲁帕加入讨论，他今天率先报道了这一消息。

Joining me now to discuss that is Miles Krupa, our AI and finance reporter who broke that story today.

迈尔斯，欢迎来到节目。

Miles, welcome to the show.

很高兴你来到这里。

It's great to have you here.

谢谢，阿卡什。

Thanks, Akash.

那么，我们来聊聊你今天早上发布的这条独家新闻。

So let's talk about this scoop that you published this morning.

关于蓝猫资本和Stargate的这笔交易，我们需要了解哪些关键信息？

What do we need to know about this deal with BlueOwl and Stargate?

是的。

Yeah.

所以，你知道，Stargate 最初被宣传为一家为 OpenAI 创建计算基础设施的大型公司。

So, you know, Stargate was kind of pitched as this really big company to create computing infrastructure for OpenAI.

但实际上，它已经演变成多个独立的数据中心项目，每个项目都分别获得融资。

And it's actually kind of turned into a lot of different individual data center projects that are being financed individually.

因此，目前的情况是，BlueOwl 正在向新墨西哥州正在建设的这个数据中心投入约30亿美元的股权，而甲骨文公司——作为 OpenAI 的大型合同方——将成为该数据中心的租户，随后 OpenAI 将从甲骨文租用计算能力。

And so what's happening here is BlueOwl is investing about $3,000,000,000 of equity in this data center in New Mexico that's being built where Oracle, which has a huge contract from OpenAI, Oracle will be the tenant at that data center, and then OpenAI will be renting computing power from Oracle.

所以，Blue Owl 的这笔投资是股权部分，他们还正从一组银行筹集180亿美元的贷款。

So this Blue Owl investment is the equity, and then they're also going out and raising $18,000,000,000 in loans from a group of banks.

所以，这些交易的结构对我来说很有趣，因为人们可能会以为，哦，BlueOwl 是在投资 OpenAI，但事实完全不是这样。

So the the system of these deals is kind of interesting to me because people might think, oh, BlueOwl investing in OpenAI, which is not the case at all.

这些数据中心项目采取的结构非常精细。

This is a very nuanced structure that these data center projects take on.

你能给我们解释一下 BlueOwl 投资 OpenAI 和 BlueOwl 投资这个特定项目之间的区别吗？

Can you explain to us the difference here between BlueOwl investing in OpenAI versus BlueOwl investing in this particular project?

是的，没错。

Yeah, that's right.

所以，这个项目是一个独立的实体，一个与OpenAI或Oracle分开的公司。

So basically, the project is its own entity, its own kind of company that's separate from OpenAI or Oracle.

你知道，BlueOwl将在这里作为股权投资者。

You know, BlueOwl is going to be the equity investor here.

因此，如果这个数据中心项目在长期内保持高度价值，他们就能获得潜在的回报。

So they get potential upside, if this data center project remains really valuable over the long term.

然后，在贷款方获得偿付后，他们还能从Oracle支付的租金中获得收益。

And then they also get paid, you know, from the lease payments that Oracle will be making after the lenders get paid.

所以，这与投资初创公司有点不同，但BlueOwl在这里仍然有获得巨大回报的潜力。

So it's, you know, it's a bit different from investing in a startup, but there is still, you know, potential for big upside here for BlueOwl.

这些数据中心项目之所以这样结构化，是因为它们本质上很大程度上是一种房地产投资。

And the reason that these data center projects are structured like this is because what they're largely sort of real estate play.

这有什么好处呢？

Why is this beneficial?

没错。

That's right.

是的。

Yeah.

它们介于房地产和石油天然气管道、风能和太阳能等大型基础设施项目之间。

They're sort of in between real estate and big infrastructure projects like oil and gas pipelines and, you know, wind and solar projects.

因此，它们的融资方式与这些其他项目非常相似。

So they're financed in a very similar manner as those other projects are.

你知道，目前投入的资金基本上是用来覆盖建设成本的。

You know, the the funding that's going in right now is going to cover basically the construction costs.

所以这相当于为接下来三到五年提供资金。

And so that's sort of funding for the next, you know, three to five years.

然后，一旦投入运营，它可能会再次进行再融资。

And then, you know, once it's up and running, it'll probably get refinanced again.

所以，你可以把这理解为资助建造一个超大规模数据中心的前期成本。

So, you know, this is kind of think about this as kind of funding the upfront costs of building a really massive data center.

因此，BlueOwl 进行了股权投资。

So BlueOwl making an equity investment.

BlueOwl，正如你所说，这家公司主要以在华尔街从事私人信贷业务而闻名。

BlueOwl, of course, as you said, it's a company that is mostly known on Wall Street for its private credit plays.

这是否是更大趋势的一部分？

Is this part of a larger trend?

在人工智能时代，私人信贷公司是否在增加对AI领域的股权投资？

Private credit companies making more equity investments in the AI era?

是的。

Yeah.

我认为，总体而言，华尔街一些最大的公司，如黑石、KKR、阿波罗，正越来越深入地涉足数据中心领域，随着人工智能热潮持续升温，Blue Owl 也不例外。

I would say overall, some of the biggest firms on Wall Street, Blackstone, KKR, Apollo, you know, they're going deeper and deeper into data centers as this AI boom sort of, continues on, and Blue Owl is no exception.

它们甚至可能是数据中心领域增长最快的企业之一。

And they're maybe even one of the sort of fastest growing players in the data center space.

它们以大约10亿美元收购了这家数据中心投资公司IPI Partners。

They bought this data center investor called IPI Partners for about $1,000,000,000.

这笔交易于一月完成，此后它们一直非常忙碌。

That transaction closed in January, and they've been really busy ever since.

除了这笔交易外，他们还与Meta达成了一项重大交易，投资了Meta在路易斯安那州建设的数据中心。

In addition to this deal, they've done a huge one with Meta for this data center Meta is building in Louisiana.

他们为该项目投入了25亿美元的股权资金。

They invested $2,500,000,000 in equity for that.

他们还投资了Crusoe为Oracle和OpenAI开发的另一个Stargate项目。

And they've also invested in another Stargate project that's being developed by Crusoe for Oracle and OpenAI.

因此，他们正迅速成为这一领域的主要参与者之一。

So they're very quickly becoming kind of one of the main players here.

在我们结束之前，最后一个问题。

And last question before we let you go.

每当有记者来向我们通报Stargate的新进展时，我们都必须问一个问题：这个项目整体进展如何？

Anytime a reporter comes on to tell us about a new development with Stargate, we have to ask the question, how is the project going overall?

它是否按计划进行？

Is it on track?

他们现在进展到哪一步了？

How far are they?

你对现场情况进展有何感受？

What's your sense on on on how things are going on the ground?

是的。

Yeah.

嗯，我认为项目的实际形态与我们最初设想的有些不同，这算是一次调整。

Well, I would say it's taken a bit of a different shape than maybe what we first thought it would look like, and that was a bit of a pivot.

但看起来，OpenAI 正在顺利推进，争取获得其所说的约 10 吉瓦的算力。

But, you know, it does seem like OpenAI is well on its way to securing what it says could be about 10 gigawatts of computing power.

他们正在新墨西哥州、德克萨斯州和威斯康星州建立站点。

You know, they're standing up sites here in New Mexico, in Texas, Wisconsin.

他们还在与软银合作开发其他站点，而软银是 Stargate 项目的重要组成部分。

They're working on other sites with SoftBank, which is a big part of Stargate.

所以，尽管可能没有完全按照最初的规划进行，但他们确实成功为 Stargate 争取到了相当重要的数据中心站点，正如我们刚刚报道的，还获得了相应的资金支持。

So while maybe it isn't exactly how it was first laid out, you know, they are managing to secure, you know, pretty significant data center sites for Stargate and, as we just reported, securing the funding for them.

对。

Right.