独白：AI数据中心的过度建设

这是iHeartPodcast。

This is an iHeartPodcast.

保证是真人。

Guaranteed human.

丽莎，跟他们说说我们的新节目。

Lisa, tell them about our new show.

我们有个新播客。

We have a new podcast.

名字叫《别骗我》。

It's called don't lie to me.

我是莱克西·伊瓦努。

I'm Lexi Iwanu.

我们是最好的朋友，相差二十二岁，但心贴得很近。

We're two best friends, twenty two years apart, but so close

在内心深处。

at heart.

我是Lisa DiGiulio，年纪大的那个，这意味着我们的成长背景完全不同，偶尔意见也不一样。

And I'm Lisa DiGiulio, the old one, which means we have wildly different references, occasionally different opinions.

到了我这个五十多岁的年纪，什么都经历过了。

And at this point in my fifties, I've lived it all.

而我三十多岁，作为

Now in my thirties, being the

New Shoes的创始人，好品味就是我的事业。

founder of new shoes, good taste is my business.

请在iHeartRadio应用、Apple Podcasts或任何你收听播客的地方收听。

Listen on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

我是Lacey Moseley，作家、演员、单口喜剧演员、创作者，也是Scam Goddess播客的主持人。

I'm Lacey Moseley, author, actor, stand up comedian, creator, and host of scam goddess podcast.

这个真实犯罪喜剧播客讲的都是抢劫、诈骗，

The true crime comedy podcast all about robbery, fraud,

以及那些干这些事的人。

and those who practice it.

每周我都会请来一位很棒的嘉宾。

Each week, I bring on an amazing guest.

哦，我把他们骗了。

Oh, I scammed them.

什么？

What?

我能跟你聊上一个小时。

I could talk to you for an hour.

你跟我一样，天生就是个话痨。

You're a generational yapper like me.

所以我才把聊天变成了工作。

That's why I had to make yapping a job.

现在你可以在YouTube上看到我节目的完整版。

And now you can catch full episodes of my show on YouTube.

开始收听《骗子女神》吧，

Start listening to scam goddess on

在iHeartRadio应用、Apple Podcasts或者任何

the iHeartRadio app, Apple Podcasts, or wherever

你收听播客的地方。

you get your podcasts.

骗子女神。

Scam goddess.

嗨。

Hi.

我是凯莉·布雷克曼。

This is Kylie Brakeman.

安吉拉·杰里托娜。

Angela Gerritonna.

杰里米·科尔汉。

Jeremy Colhane.

帕特里克·麦克唐纳。

Patrick McDonald.

我们是这个播客的主持人。

And we're the hosts of the

《艺术家对艺术家对艺术家对艺术家》播客。

artists on artists on artists on artists podcast.

这是一个即兴角色喜剧播客，每周都有新的艺术家小组讨论他们的技艺。

The improvised character comedy podcast where each week a new panel of artists discuss their craft.

加入我们，一起和欧洲歌唱大赛的参赛者们聊聊。

Join us for a panel of Eurovision participants.

有时候我们也会搞点万圣节的东西。

And we do a little Halloween sometimes.

教皇会打扮吗？

Does the pope dress up?

他在万圣节那天会眨眨眼。

He he winks on Halloween.

然后凑到麦克风跟前。

Gets right up next to the microphone.

你能听到吗？几乎能听到睫毛碰到一起的声音。

You can I hear you can almost hear the eyelashes hit each other?

你从没见过意大利这么安静，就像万圣节教皇站起来眨眼的时候。

You never seen Italy so quiet as when the pope gets up to wink on Halloween.

每周四，在iHeartRadio应用、Apple Podcasts或任何你收听播客的地方，收听《艺术家对艺术家对艺术家对艺术家》。

Listen to artists on artists on artists on artists every Thursday on

在iHeartRadio应用、Apple Podcasts，

the iHeartRadio app, Apple Podcasts,

或者任何你收听播客的地方。

or wherever you get your podcasts.

所以，差不多到11岁左右，至少70%的孩子都有手机了。

So, like, by age 11 or so, at least 70% have a phone.

我是乔伊医生，这里是《黑人女孩的心理治疗》。

I'm doctor Joy, and this is therapy for black Girls.

我们这期节目专门探讨你孩子脑子里在想什么，他们的群聊、他们的友谊，以及你能做些什么。

We spent our surviving middle school run getting into what's happening inside your kid's brain, the group chat and their friendships and what you can do about it.

你的孩子不是故意这么做的。

Your child is not doing this on purpose.

这正是我们从那些专门研究这个的心理学专家那里得到的观点。

That's the kind of perspective we've been pulling from the psychologists and psychiatrists who study this for a living.

欢迎在iHeartRadio应用、Apple Podcasts或任何你收听播客的地方收听《黑人女孩的心理疗愈》。

Listen to therapy for black girls on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

自我关怀九月提醒你放慢脚步，好好关照自己，放下手机，也许问问自己：我为什么会这样？

Self care September is your reminder to slow down, check-in with yourself, get off your phone, and maybe ask why am I like this?

我是Gemma Speck，播客《二十几岁的心理》的主持人，这个月我们帮你理解正在成长中的自己。

I'm Gemma Speck, host of the Psychology of Your Twenties Podcast, and this month we're helping you make sense of the person that you are becoming.

从设定界限、处理友谊，到理解焦虑、自信、人际关系和依恋模式。

From setting boundaries and navigating friendships, to understanding anxiety, confidence, relationships, and attachment.

有时候，我们做出的重大决定，甚至自己都没意识到，并不是出于我们真正想要的，而是为了避免感到孤独或独自一人。

Sometimes our biggest decisions without us even realizing it aren't driven by what we truly want, but by what will prevent us from feeling lonely or being alone.

欢迎在iHeartRadio应用、Apple Podcasts或任何你收听播客的地方收听《二十几岁的心理》。

Listen to the psychology of your twenties on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

没赶上最新的Roland Martin Unfiltered播客吗？

Didn't catch the latest Roland Martin unfiltered podcast?

这里是您错过的内容。

Here's what you missed.

人们一觉醒来，发现，哦，天哪。

People wake up and go, oh, damn.

等等。

Wait.

等一下。

Hold up.

他们把那些都改了吗？

They changed all of that?

是的。

Yes.

是真的。

It's real.

这是一次全面的攻击。

This is a wholesale attack.

它针对的是黑人和所有联邦机构。

It is targeting black people and every federal agency.

这很赤裸裸。

It's raw.

白人从来没有让清算持续超过十年。

White folks have never allowed a reckoning to last more than a decade.

收听罗兰·马丁在黑人信息网络上的每日评论。

Catch Roland Martin's daily commentary on the Black Information Network.

并在iHeartRadio应用、Apple Podcasts上下载Roland Martin Unfiltered。

And download Roland Martin unfiltered on the iHeartRadio app, Apple Podcasts

或者在你获取播客的任何地方。

Or wherever you get your podcasts.

欢迎收听《解码女性健康》。

Welcome to Decoding Women's Health.

我是伊丽莎白·波因特医生，纽约市阿德里亚健康研究所女性健康与妇科主任。

I'm doctor Elizabeth Poynter, Chair of Women's Health and Gynecology at the Adria Health Institute in New York City.

我将与顶尖研究人员和临床医生对话，为您带来关于中年女性健康的重要信息。

I'll be talking to top researchers and clinicians and bringing vital information about midlife women's health directly to you.

百分之百的女性都会经历更年期。

A hundred percent of women go through menopause.

即使是自然过程，我们为什么要忍受它呢？

Even if it's natural, why should we suffer through it?

请收听伊丽莎白·波因特医生的《解码女性健康》，可在iHeartRadio应用、Apple Podcasts或您获取播客的任何平台收听。

Listen to Decoding Women's Health with doctor Elizabeth Poynter on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

呼叫Zone Media。

Call Zone Media.

你好，欢迎收听本周的《Better Offline》独白节目。

Hello, welcome to this week's Better Offline monologue.

我是主持人Ed，订阅新闻通讯、高级版，你知道的，那些好东西。

I'm your host Ed subscribe to the newsletter, premium, you know, all that good stuff.

我会把链接放在里面。

I'll have the links in there.

这周我想聊聊我发的一篇免费通讯，题目是《AI芯片都去哪儿了？》。

And this week, I wanna talk about actually a free newsletter I put out called where are all the AI chips?

结果发现，答案要么是在仓库里，要么是在没通电的数据中心里。

And it turns out that the answer is either in warehouses or unpowered data centers.

在和《卫报》合作的一项调查中，艾莎·唐和我一起做了这个报道，她特别棒。我报道说微软只有220万块GPU在运行，远低于人们以为的数量。

In an investigation with The Guardian, Aisha Down over there worked on it with me, she's awesome, I reported that Microsoft only had 2,200,000 GPUs in service, far less than people believed.

虽然当时因为还要追查一些线索没法报道，但芯片的总功率估计约为19.93亿瓦，而这些GPU的安装总成本大约在500亿到600亿美元之间。

And while I couldn't report it at the time because I still had to run down some leads, the total power of the chips was estimated at about 1.993 gigawatts of capacity, with the overall cost of those GPUs installed somewhere in the region of 50 to $60,000,000,000.

几周前，彭博社报道说，尽管微软声称在过去三个季度每个季度都增加了一千兆瓦的电力——这里说的“声称”是指微软在财报电话会议上亲口说的——但他们的总数据中心容量是12千兆瓦，其中只有2千兆瓦是专门用于AI的。

A few weeks ago, Bloomberg reported that despite reporting that Microsoft had added a gigawatt in each of the last three quarters, and by reporting, I mean the things that Microsoft literally said on their earnings call, that only two gigawatts out of their total 12 gigawatts of data center capacity was specifically for AI.

换句话说，尽管自2022年初以来，微软在资本支出上已经花了超过2650亿美元，但实际投入使用的GPU只有大约500亿美元。我估计，价值高达1060亿美元的GPU和相关硬件现在要么躺在没通电的数据中心里，要么在未完工的数据中心里，要么在仓库里，而且整个行业几乎没有新的AI算力上线。

In other words, despite having spent over $265,000,000,000 on capital expenditures since the beginning of twenty twenty two, Microsoft has only put about $50,000,000,000 worth of GPUs into service, and I estimate that as much as 106,000,000,000 worth of GPUs and associated hardware are now sitting either unpowered in data centers, incomplete data centers, or in warehouses, and that very little AI compute is coming online throughout the entire industry.

微软CEO萨钦·阿德拉去年11月也差不多承认了这一点，当时他告诉一位采访者，他有一堆芯片放在库存里插不上电，这是他的原话。

Microsoft CEO Sachin Adela had sort of admitted this in November of last year when he told an interviewer that he had a bunch of chips sitting in inventory that he couldn't plug in, and that's a quote.

虽然他没提到仓库里囤积的规模，也没说实际启动的数量有多少。

Though he didn't mention the sheer scale of the warehousing or indeed how little he'd actually turned on.

我这个人好奇心重，就去微软之外找更多证据，结果在几家超大规模云厂商和新晋云服务商的资产负债表里找到了线索——就藏在‘在建工程’这一项下面，公司通常就把没装好的GPU和没建完的数据中心藏在这儿。

Well, I'm a curious little critter, so I went looking for more evidence of the problem outside of Microsoft and found it buried in the balance sheets of multiple hyperscalers and neo clouds under the construction in progress line on the balance sheet, which is specifically where companies bury their uninstalled GPUs and incomplete data centers.

而且这个数字在很多家那里都大幅增长了，增长幅度相当惊人。

And this number has grown across a ton of them, well, by a remarkable amount.

谷歌、Meta、甲骨文、亚马逊、SpaceX、特斯拉，还有CoreWeave、iHeart这类新晋云服务商，以及Core Scientific、Applied Digital这些机柜托管公司，加起来在建工程的金额大约3740亿美元。这个数可能还偏低，因为亚马逊那部分约710亿美元的贡献，数据只更新到2025年底。

Google, Meta, Oracle, Amazon, SpaceX, Tesla, Neo Clouds like CoreWeave and iHeart and colocation companies like Core Scientific and Applied Digital all have about $374,000,000,000 of construction in progress, a figure that's likely lower than the true number because Amazon's contribution, is about $71,000,000,000, is only current as of the end of twenty twenty five.

而且接下来又有两三个季度都是这样。

And there have been two now, very soon, three more quarters of that.

这家公司只公布年度数据，就像我之前说的。

And the company only reports annual, like I said.

这笔巨额资本支出里有一大块是谷歌的，大概一千二百二十亿美元，这真的让人震惊。

A large chunk of that larger CIP number is Google's, which sits around a $122,000,000,000, which is truly shocking.

顺便说一句，在所有这些案例中，这个数字每次报告时都在增长，过去十二个季度一直如此，亚马逊的情况因为物流业务有些波动。

And in every case, by the way, the number has grown every single time it's been reported for the last twelve quarters with some fluctuation in the case of Amazon because of their logistics operations.

但说实话，目前它还在持续增长。

But really right now, it's only growing.

不过，并不是所有公司都会披露在建工程的数据。

Now not everybody reports construction in progress.

所以这个数字并没有涵盖微软、房利美、夏朗、Equinix、Nebius、Poolside，像Vantage这样的私人运营商，中东或欧洲的主权AI建设项目，为OpenAI和Anthropic打造的私人项目，以及甲骨文或Meta那些表外建设计划——我估计这些项目的金额高达数百亿美元。

So that number doesn't include Microsoft, firmness, Sharon, Equinix, Nebius, Poolside, private operators like Vantage, any of the sovereign AI build outs in The Middle East or Europe, the private projects built for OpenAI and Anthropic, or any of Oracle or Meta's off balance sheet construction projects, which were likely, I think, numbering in tens of billions of dollars.

因为这些特殊目的实体都放在资产负债表之外，所以它们不会体现在在建工程的统计里。

And because these are special purpose vehicles kept off of their balance sheet, that wouldn't be in construction in progress.

要是让我猜的话，在建工程的总金额大概在4000亿到5000亿美元之间。

Now if I had to guess, the total construction in progress number is somewhere between 400 and $500,000,000,000.

根据我所有的分析，我估计大约50%已经售出的AI硬件和芯片都被囤在仓库里，完全消化掉需要两年以上的时间。这意味着基本上英伟达卖出去的所有东西，都像是预售订单，要到2028年、2029年甚至更晚才能交付。

And based on everything I've analyzed, I estimate that approximately 50% of all AI hardware and chips be that have been sold is being warehoused and will take more than two years to fully ingest, which means that basically everything you've seen NVIDIA sell has been a preorder campaign that arrives sometime in 2028, 2029, or maybe beyond that.

换句话说，我认为英伟达从这些提前了24到36个月就卖出去的产品中，赚了大概2031万亿美元的收入。而目前AI算力的稀缺，主要是因为大部分产能都立刻卖给了Anthropic和OpenAI，留给其他公司的所剩无几。再加上新产能的建设周期长得让人抓狂，除非你是那些顶尖的AI实验室，否则根本排不上优先级。

In other words, I think NVIDIA has made somewhere between $203,100,000,000,000 dollars in revenue on stuff that's been sold probably twenty four to thirty six months in advance, and that any scarcity around AI compute is a result of most of the capacity being sold immediately to Anthropic and OpenAI, leaving very little for the rest of the world with more capacity taking agonizingly long to build and not being prioritized unless you're one of the AI labs.

所以，这一切到底意味着什么呢？

So what does all of this mean?

首先，基本上任何关于超大规模数据中心建了多少容量的公告都值得怀疑。

First of all, basically any announcement of how much capacity a hyperscaler has built is suspicious.

微软故意模糊它建了多少AI容量，这本身就说明整个AI行业都在这么做，而且实际可用的AI容量比我们想象的要小得多得多。

The fact that Microsoft is intentionally obfuscating how much capacity, AI capacity, I mean, it has built is a sign that the entire AI industry is doing so and that the available AI capacity is much, much, much smaller than we think.

那些说12或15吉瓦即将上线的估计，全是胡扯。

All of those estimates of 12 or 15 gigawatts coming online bollocks.

瞎扯。

Wank.

废话。

Tosh.

我不知道。

I don't know.

我想不出别的词了。

I can't remember other terms for it.

第二，如果微软——全球资本最雄厚、经验最丰富的数据中心开发商之一——都难以让大部分AI容量上线，那其他所有人也都一样。

Second, if Microsoft, one of the most well capitalized and experienced data center developers in the world, is having trouble bringing the majority of its AI capacity online, everyone is.

说实话，我觉得亚马逊可能是唯一一家，或者加上谷歌，拥有 comparable 经验的公司，更不用说经验更丰富的了。

Really, I think Amazon is the only company, maybe Google, that has comparable experience, let alone more.

而且我认为微软也是第一家在2020年为OpenAI提供基于GPU的AI超级计算机的公司。

And I think Microsoft was also the first AI supercomputer powered by GPUs back in 2020 for OpenAI.

可能有人在此之前做过类似的事，只是为了故意跟我作对。

There's probably someone who did it before just to mess with me.

不过，如果他们都在搞砸，那我想所有人都是如此。

Nevertheless, if they're failing, I think everyone is.

最后，大家对AI算力需求无穷无尽的看法完全是扭曲的。因为实际上线的容量非常少，可能比大家以为的少了几个吉瓦，而且几乎全都流向了Anthropic和OpenAI。这制造了一种稀缺的假象，但实际上只是两家靠风投撑着的、不可持续的AI实验室在疯狂吞噬现有的所有资源。

And lastly, any beliefs that anyone has about the insatiable demand for AI compute are completely distorted because very little actual capacity is coming online, likely gigawatts less than people believe, and almost all of it is flowing directly to Anthropic and OpenAI, creating an illusion of scarcity when it's actually just two unsustainable venture backed AI labs sucking up whatever exists.

这也意味着我们现在处于一个严重的过度建设局面。

And this also means we're in a terrible overbuild situation.

目前，大家都在扩建AI数据中心，是因为微软、谷歌、亚马逊、甲骨文和CoreWeave这些公司有着巨大的收入积压订单。

Right now, everyone building more AI data center capacity is doing so because of the massive revenue backlogs across Microsoft, Google, Amazon, Oracle, and Core Wave.

但大家忽略了一个事实：这些积压订单中有一半到75%直接来自Anthropic和OpenAI，而这两家公司根本付不起这笔钱。

Ignoring the fact that anywhere from half to 75% of those backlogs are directly from Anthropic and OpenAI, neither of whom can actually pay for them.

任何超大规模云服务商如果签了合同要把这些算力给他们，那这些合同很可能可以取消。

And any hyperscalers signing contracts to give that capacity to them can probably cancel the agreements.

从外面看，整个行业看起来欣欣向荣，对AI GPU和AI算力的需求都很大，大家都觉得这些GPU马上就能变成钱。

To the outside world, everything looks like a thriving industry with tons of demand both for AI GPUs, which they think are turning into money immediately, and for AI compute.

微软还发布了一些明显有问题的声明，说每个季度都要增加一千兆瓦的算力，让人觉得它的收入增长是靠增加这些算力、花掉那些资本开支，再加上多样化的需求才实现的。

With Microsoft's clearly crooked statements around adding a gigawatt of capacity every single quarter, creating the illusion that its revenue growth is from adding all of that capacity and from spending all of those capital expenditures and from diverse demand on top of it.

但实际上，这些增长主要不是来自OpenAI的算力支出，这部分占了微软2026财年AI收入的70%，占同一财年总收入的7%。

Rather than mostly from OpenAI's compute spend, made up 70% of Microsoft's AI revenues in fiscal year twenty twenty six and seven percent of its overall revenues in that same fiscal year.

具体来说，大家以为每个季度都有大量计算机上线，而且这些计算机马上就能以高价租出去。

To be very specific about what I mean, people are assuming that lots of computers coming online every quarter and that all of that computers being immediately rented out for high rates.

所以他们认为建更多AI数据中心是显而易见的选择，就像印钱一样，因为他们觉得有大量被压抑的需求等着被满足，而且满足这些需求只需要把数据中心上线就行，微软从外面看就是这样，证明了一个季度能增加一千兆瓦。

As a result, they think that building more AI data centers is an obvious choice and akin to printing money because they believe there's unbelievable amounts of pent up demand waiting to be met and that meeting that demand will be as simple as bringing a data center online, which Microsoft is from the outside, proving can be done a gigawatt a quarter.

但实际上，上线的并不是AI数据中心的算力。

Except it's not AI data center capacity that's coming online.

真正的情况是，用于AI服务的算力增加得非常少。

What's actually happening is very little capacity is coming along for AI services.

而那一点点新增的算力，直接落入了两家公司手中，它们拥有近乎无限的资源，这些资源来自风险投资，有时甚至来自超大规模云服务商本身，这让它们能够签下合同，占用大量算力，并填满那些收入积压。

And what little does is going straight into the hands of two companies with near infinite resources provided by venture capital and in some cases, the hyperscalers themselves, which allows them to sign contracts to take up massive amounts of capacity and indeed fill those revenue backlogs.

微软说它正在以每季度一千兆瓦的速度增加数据中心容量，但并不是AI专用的数据中心容量，而实际的AI算力，最多也就是每季度增加几十到几百兆瓦，而且这真的是最好的情况了。

Microsoft is saying that it's bringing data center capacity, not AI specific data center capacity online a gigawatt at a time, but it appears the actual AI capacity is at best coming on at a rate of maybe tens or hundreds of megawatts a quarter, and that really is, like, the best case.

也许有其他人做得更快，但我确实找不到证据。

Perhaps there are others that are doing it faster, but I actually can't find evidence of them.

说实话，这个故事在我脑子里盘旋了好几个星期，我一直犹豫要不要说出来，因为它的影响太大了。

I'll be honest that I have had this story in my head for weeks and was hesitant to jump the gun because of its ramifications.

因为看起来所有人都把这个故事搞错了。

As it appears that everyone's got this story wrong.

这比互联网泡沫还要严重得多。

This is magnitudes worse than the dotcom bubble.

打个最明显的比方，这就好比光纤电缆根本没埋进地里，或者只接了一半，只不过AI基础设施的实际需求被OpenAI和Anthropic严重扭曲了，导致规划中的算力容量是实际需求的五到五十倍。

To make the obvious comparison, this is akin to the fiber optic cable not even being laid in the ground, or only being halfway connected, except the actual demand for AI infrastructure is so thoroughly distorted by OpenAI and Anthropic that there's anywhere from five to 50 times the amount of capacity in planning than there is demand for it.

而且即便如此，那个需求本身也是虚高的。

And even then, that demand is inflated.

更糟的是，这之后这些设施也不会被用上。

To make matters worse, it's not going anywhere after this.

这些设施建成后，根本不可能按设计规模投入使用。

There is no situation where this is going to be used at the scale it's built.

而且，两三年或五年后，完成一个AI数据中心的成本依然高昂，运行那些GPU的费用也不会降低。

It will also be just as expensive to finish an AI data center in two or three or five years, and just as expensive to run those GPUs.

谁知道那里面会有多少代产品呢？

And who knows how many generations will be in that?

但我不确定。

But I don't know.

依我之见，以我专业的看法，我觉得我们撑不过Vera Rubin或Feynman这一代，也就是Vera Rubin的下一代。

In my my opinion, my esteemed opinion, I don't think we make it past Vera Rubin or Feynman, which is the generation after Vera Rubin.

即便如此，我对Feynman是否真的能出货也存有疑问。

And even then, I have questions about whether Feynman actually ships.

不过谁知道呢？

Who knows though?

现在的情况真是疯狂。

Things are crazy.

我还要明确一点，就像这周我和保罗·库德罗夫斯基讨论的那样，GPU市场在泡沫破裂后并没有出路，至少按现在这种建设规模来说是这样，因为提供推理服务的成本就像航空公司一样，固定成本很高。

I'll also be clear that as I discussed with Paul Kudrowski this week, that there is no post bubble economy for GPUs, at least at the scale they're being built, because the cost of serving inference is similar to an airline with high fixed costs.

作为运营商，你必须在一段时间内购买容量，无论是几百张还是几千张GPU，不管你是否真的有那么多需求。

You as an operator must buy capacity, hundreds or thousands of GPUs over a certain period of time, whether or not you actually have the demand.

虽然有可能在客户数量恰到好处的情况下勉强赚点利润，但一旦需求超过你之前的预测，服务就会变得不稳定、不可靠，甚至会把客户拒之门外；而如果需求低于预测，你仅有的那点利润空间也会被消耗殆尽。

While there might be a situation we could eke out a profit of some sort with the perfect amount of customers, anything above the demand you foresaw means an unstable and unreliable service and turning customers away at the door, and anything below consumes every ounce of margin you might have.

达里奥·阿莫迪本人今年二月在接受杜阿基什·帕特尔采访时也说过类似的话，他说在一个理论场景下，他可以在2027年底开始购买价值一万亿美元的算力，但如果他的收入不是一万亿美元，哪怕只有八千亿美元，用他的原话说，没有任何对冲手段能阻止他破产。

Dario Amodei himself said as much in an interview back in February with Duakish Patel, saying that in a theoretical scenario, he could buy $1,000,000,000,000 of compute that starts at the end of twenty twenty seven, but if his revenue wasn't a trillion dollars, if it was even $800,000,000,000, there was, and I quote, no hedge on earth that could stop him from going bankrupt.

这就是每一家运营AI服务的公司背后的基本经济逻辑，既脆弱又动荡。

These are the underlying economics of every single company running an AI service, and they're equal parts brittle and volatile.

感谢收听。

Thanks for listening.

我们下周见。

I'll see you next week.

丽莎，跟他们说说我们的新节目。

Lisa, tell them about our new show.

我们有个新播客。

We have a new podcast.

名字叫《别骗我》。

It's called Don't Lie To Me.

我是莱克西·伊瓦努。

I'm Lexi Iwanu.

我们是最好的朋友，相差二十二岁，

We're two best friends, twenty two years apart,

但心贴得很近。

but so close at heart.

我是丽莎·迪朱利奥，年纪大的那个，这意味着我们的共同话题差别很大，有时意见也不同，而且到了我这个五十多岁的年纪，什么都经历过了。

And I'm Lisa DiGiulio, the old one, which means we have wildly different references, occasionally different opinions, and at this point in my fifties, I've lived it all.

我现在三十多岁，

Now in my thirties, being

作为新鞋品牌的创始人，好品味就是我的事业。

the founder of new shoes, good taste is my business.

请在

Listen on the

iHeartRadio应用、Apple Podcasts或任何你收听播客的地方收听。

iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

我是莱西·莫斯利，作家、演员、单口喜剧演员、创作者，也是《骗子女神》播客的主持人。

I'm Lacey Moseley, author, actor, stand up comedian, creator, and host of Scam Goddess Podcasts.

这是一档真实犯罪喜剧播客，内容全是关于抢劫、诈骗，

The true crime comedy podcast, all about robbery, fraud,

以及那些干这些事的人。

and those who practice it.

每周我都会请来一位很棒的嘉宾。

Each week, I bring on an amazing guest.

哦，我骗了他们。

Oh, I scammed them.

什么？

What?

我可以跟你聊一个小时。

I could talk to you for an hour.

你跟我一样，天生就是个话痨。

You're a generational yapper like me.

所以我才把聊天当成工作。

That's why I had to make yapping a job.

现在你可以在YouTube上看我节目的完整版了。

And now you can catch full episodes of my show on YouTube.

开始收听《Scam Goddess》吧，

Start listening to scam goddess on

在iHeartRadio应用、Apple Podcasts，或者任何

the iHeartRadio app, Apple Podcasts, or wherever

你收听播客的地方。

you get your podcasts.

骗子女神。

Scam goddess.

嗨。

Hi.

我是凯莉·布雷克曼。

This is Kylie Brakeman.

安吉拉·杰里托娜。

Angela Gerritonna.

杰里米·科尔汉。

Jeremy Colhane.

帕特里克·麦克唐纳。

Patrick McDonald.

我们是

And we're the hosts of

《艺术家对艺术家对艺术家对艺术家》播客的主持人。

the artists on artists on artists on artists podcast.

这是一个即兴角色喜剧播客，每周都会有一组新的艺术家来讨论他们的手艺。

The improvised character comedy podcast where each week a new panel of artists discuss their craft.

加入我们，一起和欧洲歌唱大赛的参赛者们聊聊。

Join us for a panel of Eurovision participants.

有时候我们也会搞点万圣节的小活动。

And we do a little Halloween sometimes.

教皇也会打扮吗？

Does the pope dress up?

他在万圣节那天会眨眨眼。

He he winks on Halloween.

他会凑到麦克风跟前。

He gets right up next to the microphone.

你几乎能听到睫毛碰在一起的声音，对吧？

You can I hear you can almost hear the eyelashes hit each other?

你从来没见过意大利这么安静，就是教皇在万圣节站起来眨眼的时候。

You never seen Italy so quiet as when the Pope gets up to wink on Halloween.

每周四，收听艺术家们聊艺术家们聊艺术家们聊艺术家们，尽在

Listen to artists on artists on artists on artists every Thursday on

iHeartRadio应用、Apple Podcasts，

the iHeartRadio app, Apple Podcasts,

或者你收听播客的任何平台。

or wherever you get your podcasts.

所以，差不多到11岁左右，至少70%的孩子都有手机了。

So, like, by age 11 or so, at least 70% have a phone.

我是乔伊医生，这里是《黑人女孩的心理治疗》。

I'm doctor Joy, and this is Therapy for Black Girls.

我们这期《度过初中》系列，聊的是孩子脑子里在想什么、群聊里的那些事、他们的友谊，以及你能做些什么。

We spent our surviving middle school run getting into what's happening inside your kid's brain, the group chat, and their friendships, and what you can do about it.

你的孩子不是故意这么做的。

Your child is not doing this on purpose.

这就是我们从那些专门研究这个的心理学家和精神病学家那里得到的观点。

That's the kind of perspective we've been pulling from the psychologists and psychiatrists who study this for a living.

在iHeartRadio应用、Apple Podcasts或任何你收听播客的地方，收听《Therapy for Black Girls》。

Listen to Therapy for Black Girls on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

自我关怀九月提醒你放慢脚步，与自己对话，放下手机，也许问问自己：我为什么会这样？

Self care September is your reminder to slow down, check-in with yourself, get off your phone, and maybe ask, why am I like this?

我是Gemma Spegg，是《Psychology of Your Twenties》播客的主持人。

I'm Gemma Spegg, host of the Psychology of Your Twenties podcast.

这个月，我们帮你理解正在成长的自己，从设定界限、处理友谊，到了解焦虑、自信、人际关系和依恋模式。

And this month, we're helping you make sense of the person that you are becoming from setting boundaries and navigating friendships to understanding anxiety, confidence, relationships, and attachment.

有时候，我们最大的决定，甚至自己都没意识到，并不是由我们真正想要的驱动的，而是由什么能防止我们感到孤独或独处驱动的。

Sometimes our biggest decisions without us even realizing it aren't driven by what we truly want, but by what will prevent us from feeling lonely or being alone.

在iHeartRadio应用、Apple Podcasts或任何你收听播客的地方，收听《Psychology of Your Twenties》。

Listen to the psychology of your twenties on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

错过了最新的《Roland Martin Unfiltered》播客吗？

Didn't catch the latest Roland Martin unfiltered podcast?

这是你错过的内容。

Here's what you missed.

不应该有任何执法机构不配备随身摄像头。

There should not be a single law enforcement agency that does not have body cameras.

这是真的。

It's real.

黑人农民一直受到攻击。

Black farmers have been under attack.

这只是他们随便给什么都贴上DEI标签的最新例子。

This is just the latest example of them just slapping DEI on anything.

这很直接。

It's raw.

而且我

And I'm

坐在这里想，他们是在把你们当傻子耍。

sitting here going, they are playing y'all for fools.

收听罗兰·马丁在黑人信息网络上的每日评论。

Catch Roland Martin's daily commentary on the Black Information Network.

还有，在iHeartRadio应用、Apple Podcasts上下载Roland Martin unfiltered。

And download Roland Martin unfiltered on the iHeartRadio app, Apple Podcasts

或者在你收听播客的任何地方。

Or wherever you get your podcasts.

欢迎收听《解码女性健康》。

Welcome to Decoding Women's Health.

我是伊丽莎白·波因特医生，纽约市阿特里亚健康研究所女性健康与妇科主任。

I'm doctor Elizabeth Poynter, Chair of Women's Health and Gynecology at the Atria Health Institute in New York City.

我将与顶尖研究人员和临床医生对话，为你带来关于中年女性健康的重要信息。

I'll be talking to top researchers and clinicians and bringing vital information about midlife women's health directly to you.

100%的女性都会经历

A 100% of women go through

请收听《解码女性健康》节目，由伊丽莎白·波因特医生主持。

Listen to Decoding Women's Health with Doctor.

在iHeartRadio应用、Apple Podcasts或你收听播客的任何地方都可以找到。

Elizabeth Poynter on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

这是iHeart播客，保证真人。

This is an iHeart podcast, Guaranteed Human.