亚马逊2000亿美元资本支出领跑AI竞赛

欢迎收听这个播客。

Welcome to the podcast.

我是你的主持人，杰登·谢弗。

I'm your host, Jaden Schafer.

所以我不打算隐瞒。

So I'm not gonna lie.

我觉得最近，在人工智能领域以及整个行业，似乎变成了一场谁能在数据中心上花更多钱的竞赛。

I feel like recently it feels a lot like in AI and just the industry in general in general, it's turned into a giant contest of who can spend the most money on data centers.

从逻辑角度来看，这似乎是有道理的。

I think it makes sense from a logical perspective.

对吧？

Right?

基本上，你拥有的数据中心越多，计算能力就越强。

Basically, the more data centers you have, you have more compute.

而你的计算能力越强，你的AI系统就越优秀。

And the more compute you have, your AI systems get better.

如果你的AI系统更优秀，最终你就能获得最多的客户，成为AI时代、科技时代真正的王者。

And if you have a better AI system, you're ultimately going to, you know, get the most customers and you're gonna be like the the king of the AI era of you know, the tech era of AI.

所以我觉得这种逻辑很有道理。

So I think it makes a lot of sense.

但这种逻辑，我认为在很多不同行业中是有局限性的。

And that logic though, I do think it has some limits in a lot of different industries.

你知道，你的长期成功最终还是取决于你赚了多少钱，以及你是否赚得比花的多。

You know, your long term success is still gonna come down to how much money you make and if you're making more money than you spend.

所以我认为，现在我们正处在一个有趣的时刻，对于大型科技公司来说，算力——你可以称之为军备竞赛，但本质上就是每个人都在疯狂烧钱，而这正是决定谁将成为赢家的关键。

So I think right now we have this kind of interesting moment where with big tech, the compute, you know, you could call it an arms race, but it's basically everyone is spending insane amounts of money and that's sort of what is dictating who's going to be the winner.

但如果这就是游戏规则，那么目前亚马逊显然处于领先地位。

So if that is the game though, I think right now Amazon is definitely in the lead.

所以今天在播客中，我想分析一下亚马逊最近的财报，同时也想聊聊谷歌和亚马逊在资本支出上的投入，因为这简直令人难以置信。

So today on the podcast, I wanna break down recent earnings reports from Amazon and also want to talk a little bit about Google and Amazon and how they are spending on CapEx because it is absolutely insane.

在深入这些内容之前，我想宣布一个关于我自己的初创公司aibox.ai的重大消息。

Before we get into all of that, I wanted to make a huge announcement about my own startup which is aibox.ai.

我们刚刚首次新增了多个定价层级，这意味着你可以获得更多的积分，有时甚至花更少的钱。

We've just, for the first time ever, added a bunch of new pricing tiers, which is basically the ability for you to get more credits and, in some cases, for less money.

因此，我们首次推出了每月9美元的套餐。

So for the first time ever, we've just launched a $9 a month tier.

每月9美元，你就可以访问40多个顶级AI模型，涵盖从Anthropic、Cohere、DeepSeek到Google、Meta、Mistral、OpenAI、Perplexity、QuenX，还有众多出色的图像生成器，比如Blackforce Labs、Ideogram，当然还有Eleven Labs和OpenAI的文本转语音音频模型。

So for $9 a month, you get access to over 40 of the top AI models, everything from Anthropic to Cohere to DeepSeek to Google to Meta, Mistral, OpenAI, Perplexity, QuenX, tons of cool image generators like Blackforce Labs, Ideogram, of course, eleven Labs and OpenAI's text to speech audio models.

你只需每月9美元，就能获得这一切，而不用分别在Anthropic、OpenAI和其他十几个平台上支付每月20美元的订阅费。

You get all of that for $9 a month instead of having $20 a month subscriptions on Anthropic and OpenAI and a dozen other platforms.

你只需每月花9美元，就能访问所有这些服务。

You can spend $9 a month and get access to all of it.

我们还新增了几个更高价位的套餐。

We've also added a bunch of higher priced tiers.

以前我们只有每月20美元的选项。

So we used to only be $20 a month.

现在我们新增了40美元和80美元的套餐。

We've now added a 40 and $80 tier.

我们之前有一批用户，他们是真正的重度使用者，大量地使用音频和图像功能。

We had a bunch of people that were, you know, really power users doing a lot of cool stuff with audio and image.

所以现在，如果你需要更多的令牌或积分，尤其是当你开始使用构建器、自动化你的工作流程时，你可以选择更高阶的套餐。

And so now you have the ability to get higher tiers if you want more tokens, more credits, especially once you start building things on the builder, automating a lot of your workflows.

所以这里有一些非常令人兴奋的功能。

So there's some really exciting stuff.

如果你想了解并获取所有这些功能，就去查看 aibox.ai 吧。

If you wanna go check it out and get access to all of that, go check out aibox.ai.

你现在可以选择每月9美元的套餐，也可以选择每月40美元和80美元的更高阶套餐以获得更多令牌。

You can now get a $9 month tier and also 40 and $80 month tiers for more tokens.

我会在描述中留下 aibox.ai 的链接。

I'll leave a link in the description to aibox.ai.

我认为现在很多人讨论的焦点是亚马逊在AI支出上的规模。

I think right now, the thing that a lot of people are talking about is Amazon when it comes to how much is getting spent on AI.

他们在周四发布了最新的财报，表示预计今年在资本支出上将投入约2000亿美元。

They did a recent earnings report on Thursday, and they said that they expect to spend about $200,000,000,000 on CapEx through this year.

所以，这些资金中有很大一部分花在了人工智能、定制芯片、机器人以及低地球轨道卫星上。

So that's you know, a lot of that's getting spent on AI, custom chips, robotics, and also low earth orbit satellites.

我认为，这比去年的1310亿美元有了大幅增长。

I think this is a really big increase from the $131,000,000,000 it spent last year.

所以，我的意思是，虽然还没达到翻倍，但从1310亿增加到2000亿，确实是一个巨大的跃升。

So, I mean, almost I mean, it's not quite double, but 200,000,000,000 from 131,000,000,000, like, is a massive jump.

现在，考虑到当前行业形势，他们和任何公司都可能倾向于把所有资金都投入人工智能，但亚马逊在众多科技公司中略显独特，因为除了强大的科技形象，他们还拥有庞大的实体业务基础。

Now while think I it would be tempting for them and any company to give all of that to AI with just given the industry and everything that's going on right now, I think Amazon is a little bit unique among a lot of other tech companies because they also in in addition to having, you know, a big tech face, they also operate a massive physical footprint.

对吧？

Right?

其中很大一部分正在被改造用于机器人和自动化，这使得人们很难忽视他们资本预算中非人工智能的部分。

A lot of that is being retrofitted for robotics and automation, which makes it a lot harder to kinda dismiss the non AI portion of their capital budget.

但如果你看看这个领域其他公司的情况，比如谷歌，也紧随其后。

But if you look at other companies too in the space, like Google's not far behind.

谷歌在周三表示，今年预计资本支出将达到1750亿至1850亿美元，高于前一年的910亿美元。

Google said on Wednesday that they're expecting to spend 175,000,000,000 to 185,000,000,000 in CapEx this year, which up is from 91,000,000,000 the year before.

我的意思是，从910亿美元增加到1750亿美元，或者从1300亿美元增加到2000亿美元，这些增幅都太大了。

I mean, these are massive jumps going from 91,000,000,000 to 175,000,000,000 or from 130,000,000,000 to 200,000,000,000.

这些数字简直在疯狂飙升。

Like, these these numbers are just going crazy.

Meta上周也发布了财报，预计今年的资本支出将在1150亿到1350亿美元之间。

Meta also had their earnings report last week, and they projected between 115 and 135,000,000,000 in CapEx for this year.

甲骨文曾经被认为是AI基础设施热潮的主要受益者之一。

Oracle, which was, you know, once seen as basically, you know, one of the major beneficiaries of the AI infrastructure boom.

我的意思是，他们正在与OpenAI和软银合作，投入500亿美元建设数据中心和基础设施。

I mean, they have the $500,000,000 OpenAI and SoftBank deal that they're doing to to build data centers and infrastructure.

但就连甲骨文今年的预期支出也不到500亿美元。

But even Oracle is expecting to spend, you know, not quite as much, $50,000,000,000 this year.

我认为微软尚未公布今年的支出计划，但最近几个季度的资本支出总额大约是375亿美元，如果保持这个节奏，年化后大约是1500亿美元。如果他们持续当前的投入水平，微软的支出应该会达到约1500亿美元。

So I think Microsoft hasn't yet said how much they're gonna be spending this year, but I think a bunch of, you know, recent quarterly CapEx expenditure totals were about 37,500,000,000.0, and that kinda annualizes to about a $150,000,000,000 if, you know, if they're spending that they have going on right now, if that remains steady, they Microsoft should be at about a 150.

所以我认为这种增长也引起了投资者的关注，但我觉得微软在总支出上仍然远远落后于亚马逊和谷歌，因为这两家公司的支出简直太疯狂了。

So I think that increase has also drawn investor scrutiny, but I think still Microsoft is far behind Amazon and Google just in overall spending because those companies are you know, it's it's absolutely insane.

所以我认为，如果你从科技行业内部来看，就能理解他们为什么这么做。

So I think if you look at it from inside the tech industry, the it it's like it makes sense why they're doing this.

如果人工智能真的要重塑经济，那么谁拥有最多的高端计算资源，谁就能掌握这种极其稀缺的资源，而且这在全球范围内都成立。

If AI is really gonna reshape the economy, then whoever has the most high end compute is gonna become, you know, the person that has this really scarce resource, and this is globally too.

所以我认为，现在能掌控自己供应链的公司才能生存下来。

So I think whatever company controls their own supply right now is gonna survive.

其他所有公司都会被淘汰。

Everyone else is gonna get bitted out.

这将会变得相当疯狂，真的会变得非常疯狂。

It's it's gonna get pretty, you know, it's gonna get pretty crazy.

甚至连埃隆·马斯克这样的人都在计划在太空中建造数据中心，因为他认为地球上的计算能力和能源都不够用，还有其他各种问题。

And you even have people like Elon Musk who's, you know, gonna be building data centers in space because he said, you know, the earth doesn't have enough compute and energy and, yeah, you know, all this other stuff.

所以这算是他的解决方案。

So that's kind of his solution.

因此，我认为在这一切背景下，目前一些最大的科技公司正在拼命为未来做准备，而那个未来将由计算资源的稀缺所定义。

So I think with all of that, some of the biggest tech companies right now are just trying to race to prepare for basically a future that is defined by compute scarcity.

我觉得这是我们不得不接受的事情。

This is something I think we're gonna have to live with.

投资者对此并不太热衷，因为几乎每家公司都在做出这些巨大的承诺。

Investors are, I mean, not very enthusiastic about this because basically every company is making all of these massive commitments.

而这些公司做出承诺后，它们的股价在财报发布后都出现了下跌。

And all of them after doing that, they've all seen their stocks drop following their earnings reports.

我认为跌幅最惨重的是那些声称要花最多钱的公司。

I think the steepest decline was coming from the people who said that they were gonna spend the most amounts of money.

感觉投资者们都在翻白眼，心想：天啊，我们真有必要花这么多钱吗？

So it feels like investors are, you know, kinda rolling their eyes like, oh, man, do we really need to spend all that?

这感觉像是一种风险。

It it feels like a risk.

但我要说，这种投资者的抵触情绪并不仅限于那些还在寻找明确AI商业模式的公司。

But I will say that kind of some of this investor pushback isn't just limited to companies that are still looking for a clear AI business model.

我认为它也延伸到了像亚马逊和微软这样已经运营着非常盈利的云业务的公司。

I think it also is extended to companies like Amazon and Microsoft that already operate, you know, really highly profitable cloud businesses.

他们对如何将人工智能变现有着非常明确的计划。

They have a really straightforward plans for monetizing AI.

我认为，至少从华尔街的角度来看，问题在于规模。

The problem, I think, at least from if you're if you're kinda looking at from, a Wall Street perspective is just the scale.

对吧？

Right?

涉及的数字简直巨大无比。

The numbers involved are just enormous.

就像，这些数字大得惊人。

Like, they're huge.

所以我认为，这种怀疑情绪在一定程度上拖慢了整个行业的发展。

So I think there's a lot of skepticism that has kind of, like, slowed down the industry in a bit.

如果你相信人工智能真的会改变几乎每一个行业，那么仅仅因为投资者在短期内感到不安就退缩，那是不合理的。

Like, if you believe AI is going to really transform almost every single sector, I think it'd be irrational to pull back just because investors are kind of uneasy in the short term.

所以我认为这是其中一面，但许多大型科技公司现在确实面临着巨大压力，需要为自己的支出提供合理解释，或者至少缓和一下支出带来的负面观感。

So I I think that's one side of it, but I think still a lot of these big tech companies are now kind of getting a lot of pressure to justify or maybe at least soften some of optics of their spending.

亚马逊的云业务正解释了为什么这家公司愿意如此积极地推进。

Amazon's cloud business is, you know, kind of explaining why their company is willing to push so aggressively.

AWS在2025年第四季度实现了三年多来的最强季度增长率。

AWS ended 2025 with the strongest quarterly growth rate in more than three years.

在第四季度，AWS的收入达到了356亿美元。

So in the fourth quarter, AWS generated 35,600,000,000 in revenue.

这相当于同比增长24%，也是13个季度以来最快的增速。

That is a 24% year over year increase, and it's the fastest growth in 13 quarters.

这基本上意味着AWS的年化收入达到了1420亿美元。

So basically that puts AWS on an annualized revenue run rate of a $142,000,000,000.

我认为，这一业务板块的运营收入也有所上升。

I think operating income for, you know, the segment that particular segment also rose.

该季度的收入从一年前的106亿美元增长至约125亿美元。

It went up to about 12,500,000,000.0 in the quarter from 10,600,000,000.0 a year earlier.

我的意思是，这些数据增长得太快了。

So, I mean, these things are growing so fast.

在整个财报电话会议中，首席执行官安迪·贾西一直在强调这一增长的重要性及其规模化效应。

During the whole earnings call, the CEO, Andy Jassy, he was kind of emphasizing just the significance that, you know, of that growth and how it's scaling.

我认为，在1420亿美元的年化收入基础上实现24%的增长，这和在小基数上实现更高百分比增长完全是两回事。

I think growing 24% on a $142,000,000,000 run rate is just you know, I mean, this is way different than just posting higher percentage growth of a much smaller base.

对吧？

Right?

比如你看到一家小公司，可能会说：哇塞。

Like, you see a small company and it's like, wow.

这家公司今年的规模翻了一倍。

This company, you know, doubled in size this year.

这听起来非常惊人。

That sounds phenomenal.

但如果你能在1420亿美元的收入基础上实现24%的增长，那才是真正了不起。

But, like, if you can grow 24% on a $142,000,000,000, that is huge.

因此，他在电话会议中指出，AWS将继续比任何其他竞争对手增加更多的新增收入和产能。

So he was arguing in his call that AWS is gonna continue to add more incremental revenue and capacity than any other competitor.

而且我认为，他们无疑是这个领域中的领导者之一。

And I I mean, I think they are just, you know, by and far, they're one of the the leaders in the space.

我认为他们在Anthropic身上做了很多非常明智的投资，比如在早期就给了他们40亿美元。

I think they've done a lot of really smart investments into Anthropic, for example, giving them, you know, in the early days, $4,000,000,000.

他们还在持续向他们投资。

They keep investing into them.

当然，这些投资很大程度上依赖于Anthropic后续回流到AWS的支出。

And, of course, a lot of that investment is contingent on them coming back and spending with AWS.

所以我认为他们的增长前景非常好。

So I think their growth is looking really good.

我认为他们看到的大部分增长，那24%的增幅，来自于与众多大公司和政府客户的全新合作协议。

And I think a lot of the growth that they saw, that 24%, was part of a whole bunch of new agreements with big companies and government customers.

我认为Salesforce、BlackRock、Perplexity和美国空军都是新近与他们达成合作的客户。

I think Salesforce, BlackRock, Perplexity, and the US Air Force are all new customers making new deals with them.

他们的首席执行官还指出，美国前500强初创企业中，使用AWS作为主要云服务商的数量，超过了接下来两家提供商的总和。

Something else that their CEO was pointing out is that more of the top 500 US startups rely on AWS as their primary cloud provider than on the next two providers combined.

仅在第四季度，AWS 就为其数据中心网络增加了超过一吉瓦的电力容量。

So in the fourth quarter alone, AWS actually added more than a gigawatt of power capacity to their data center network.

一吉瓦真的非常惊人。

And a gigawatt is like really, really impressive.

根据他们的说法，AWS 的大量需求仍然来自企业客户，这些企业正在将本地系统迁移到云端，我认为这种趋势还会持续下去。

So according to them, a huge chunk of AWS demand is still coming from their enterprises, which are migrating from on premise systems to the cloud, which I mean, I think this is a trend that you'll continue to see.

过去，很多人会自己运行所有的计算和数据中心等设施。

A lot of people used to run their own run all of their own compute and data centers and stuff like that.

他们通常会在总部或某个数据中心内进行这些操作。

And they would do it at their headquarters or in some sort of data center.

现在，他们正将这些业务迁移到 AWS，让 AWS 来负责管理。

And a lot of them are just moving it over to AWS to have them manage it.

同时，人工智能工作负载也在不断增强企业迁移到云端的必要性，因为他们自身根本没有足够的资源来本地运行这些任务。

I think at the same time, AI workloads are also increasingly reinforcing why these companies need to move to the cloud because they just don't have the resources to do it on-site.

我认为许多客户正在 AWS 上运行规模庞大的人工智能模型，并且他们也预期自己的整体云使用规模需要进一步扩大。

I think a lot of customers are running really large AI models on AWS and they they kind of expect that their broader cloud footprint needs to expand.

在第四季度，AWS占亚马逊2134亿美元总收入的16.6%，我认为这表明AWS是公司未来的重要组成部分。

AWS accounted for 16.6% of Amazon's $213,400,000,000 in total revenue in the fourth quarter, which I think shows it has a like it's a huge chunk of the company's future.

但16%毕竟不是全部。

But I mean 16% isn't everything.

显然，亚马逊其他业务板块仍然贡献了大量收入。

Obviously, is still doing a lot of revenue from other segments.

但即便有如此庞大的收入和增长，我认为这仍不足以安抚部分投资者，因为在财报电话会后，亚马逊股价在盘后交易中仍下跌了约10%，市场对这一激进的资本支出计划做出了反应。

But even with all of that revenue and all of that growth, I think that still wasn't enough to calm down some of their investors because after the whole earnings call, Amazon shares still fell about 10% after hours as the market was kind of reacting to this really aggressive capital expenditure plan.

此外，他们的每股收益也未能达到预期。

And also they had a miss on earnings per share expectations.

所以，尽管AWS可能表现优异，但其他领域却不尽如人意。

So even though, you know, AWS might be outperforming, other areas were not.

因此，我认为从现在起，华尔街基本上是在告诉亚马逊：即使公司基本面看起来非常良好，投资者对花数百亿美元投资于尚未完全实现的未来，仍然存在一定的顾虑。

And so I think from now, like basically Wall Street is telling Amazon, you know, even when the business fundamentals look really good, I think there's still a limit to how comfortable investors are with spending hundreds of billions of dollars on a future that hasn't yet fully arrived for the company.

我认为，人工智能以及这些数据中心等事物，无疑是亚马逊未来的重要组成部分。

I think, you know, AI definitely and a lot of these, you know, data centers and all these things are a huge part of the future for Amazon.

但今天他们从其他领域也赚了很多钱。

But today they're making a lot of money from other areas as well.

所以我认为华尔街仍然希望优先发展这些领域。

So I think Wall Street wants to still prioritize a lot of those segments.

这会很有趣，我们看到亚马逊、谷歌以及众多其他公司都在投入大量资金。

It'll be interesting to see, but we see Amazon, Google, and so many others spending a lot of money.

到2026年，看看这些数据中心的建设进展以及它们对这些公司的影响，将会很有意思。

Gonna It's be interesting in 2026 to see some of those build outs and how that fares for a lot of these companies.

非常感谢你收听这个播客。

Thanks so much for tuning into the podcast.

如果你喜欢这一集，一定要去查看 aibox.ai。

If you enjoyed this episode, make sure you go check out aibox.ai.

我们推出了全新的每月9美元的计划。

We have our brand new $9 a month plan.

如果你之前还没注册，这绝对是不容错过的绝佳选择。

If you haven't signed up before, this is an absolute no brainer.

只需9美元，你就可以在一个地方访问40多个顶级AI模型。

For $9, you get access to over 40 of the top AI models in one place.

每当有新模型推出想试用时，只需前往即可对比两个不同的响应，比如ChatGPT与Anthropic、Google Gemini、Grok或其他任何模型的输出。

Anytime a new model comes out that you wanna test, head on over and you get access to comparing side by side two different, you know, responses from ChatGPT versus Anthropic or versus Google Gemini or versus Grok or any other model.

这是一个将你所有AI工具集中管理的绝佳平台。

It's an awesome place to keep all of your AI stuff in one spot.

所以简介中有aibox.ai的链接，我们下一期再见。

So links in this description to aibox.ai, and I will catch you in the next episode.