饮食TBPN：2025年11月7日

时间线因'又来了？又是Sam Altman'而一片混乱。'安全网门'事件持续发酵。争论焦点在于Sarah Fryer使用了'安全网'一词的言论。她后来收回了关于安全网的评论，表示'我并不是在要求OpenAI股权提供安全网'。

The timeline is in turmoil over Again? Sam Altman again. Backstop gate continues unabated. The main point of debate is over Sarah Fryer's comments that she used the word backstop. She backtracked on her backstop comment and said, I wasn't asking for a backstop of OpenAI equity.

我当时在倡导这个美国制造计划。这是一周前提交的OpenAI文件，他们主张将数据中心支出纳入美国制造计划范畴。他们特别提倡联邦贷款担保，而Simfer Satoshi说Sam欺骗了所有人。

I was advocating for this American manufacturing plan. Here's an OpenAI document submitted one week ago where they advocate for including data center spend within the American manufacturing umbrella. They specifically advocate for federal loan guarantees, and Simfer Satoshi says Sam lied to everyone.

我们来具体看看细节。

Let's read the specifics.

是的。具体内容包括生产和AI数据中心。扩大AMIC（美国先进制造业投资税收抵免）的覆盖范围将降低资本的实际成本，降低早期投资风险，并释放私人资本以帮助缓解瓶颈并加速美国AI建设。通过降低美国制造业扩张风险来对抗中国。为制造商提供快速扩大生产所需的确定性和资本，联邦政府还应部署赠款、成本分摊协议、贷款或贷款担保，以扩大工业基础产能和韧性。

Yes. The specifics are production and AI data centers. Broadening coverage of the AMIC, which is the American Advanced Manufacturing Investment Credit, will lower the effective cost of capital, derisk early investment, and unlock private capital to help alleviate bottlenecks and accelerate the AI build in The United States. Counter the PRC by derisking US manufacturing expansion. To provide manufacturers with certainty and capital they need to scale production quickly, the federal government should also deploy grants, cost sharing agreements, loans, or loan guarantees to expand industrial base capacity and resilience.

所以这里的情况是，OpenAI在一周前——大家都可以去阅读这份公开信，它在官网上是公开的——建议政府应该将数据中心或AI（是的）或代币工厂纳入制造业范畴，使其有资格获得与传统制造业（是的）、国防科技等类似的激励政策。

So what's happening here is OpenAI, a week ago, and everyone can go and and read this Yes. Letter, which is publicly available on their website, was making the recommendation that the government should effectively treat data centers or AI Yes. Or token factories, put them in the manufacturing bucket, which would qualify them for similar incentives that traditional manufacturing Yeah. Defense tech, etcetera.

我不反对向政府寻求资助。我认为这实际上是最佳实践。这实际上符合股东利益。你有受托责任尽可能向政府寻求帮助。你有充分动机要求你在国会的代表、参议员和华盛顿的官员尽一切可能支持你的使命。

I don't have a problem with asking the government for a handout. I think that that's actually, like, best practice. It's actually in your shareholders' responsibility. Like, you have a fiduciary duty to ask the government for as much help as possible. You have every incentive to ask your person in congress and your senator, the people in Washington, to do everything they can to support your mission.

这在Elon和特斯拉身上曾取得成功，但在Solyndra案例中失败了。游戏规则是：如果有纳税人的钱在棋盘上流动，你希望让这些资金流向与你利益一致的领域。

This has worked out in the past with Elon and Tesla. It didn't work out in the case of Solyndra. The game on the field is if there are taxpayer dollars that are me moving around the board, you wanna get those into And so that are aligned with you.

人们对此提出质疑的关键点在于

And so the thing that people are taking issue with

是的。

Yes.

在他昨天发布的消息开头部分，他说：'我们不希望也不需要政府为OpenAI的数据中心提供担保。'

Is that in the opening of his message yesterday, he said, we do not want we do not have or want government guarantees for OpenAI data centers.

是的。

Yes.

这似乎与他们一周前撰写并仍挂在官网上的公开信内容相矛盾。

And that seems to conflict with the message, the letter that they wrote a week ago that is still up on their website.

你当然可以主张改变游戏规则，比如'我们不该这样做'，或者'我更倾向于自由市场经济'。但在非自由市场经济的现实世界里，你总希望自己的公司能赢，对吧？

You can always advocate for, like, we should change the game. Like, we shouldn't be doing this. Like, I would prefer a more of a free market economy. But in the world where we're not in a free market economy, you want to have your company win. Right?

这才是理性选择。不过在场边谈论比赛规则确实显得怪异。当你说'哦，这是种互惠互利的情况'，或者'有了后盾我们就能更激进'时，这种表态非常奇怪。

That's just rational. Now it is weird optics to talk about the game on the field, and that's very odd. When you say, oh, yeah. This is a one hand wash to the other situation, or, oh, yeah. This is a this is a situation where, you know, a backstop will allow us to be more aggressive.

这感觉就像银行家说，哦，是的。我知道政府在08年肯定会救我们，所以我故意批准那些贷款，比如某人的第五套房子，明知道他们还不起。我没问他们的工作，没问他们的收入，也没问他们的资产情况。

That feels like the banker saying, oh, yeah. I knew that the government was gonna bail us out in o eight, so I was intentionally underwriting loans where it was somebody's fifth house, and I knew that they couldn't pay it. I wasn't asking about their job. I wasn't asking about their income. I wasn't asking about their assets.

这就是为什么我期待更多讨论转向补贴、激励和引入新能源的方向。

Which is why I kind of expect a lot more of the narrative and to shift towards subsidizing and incentivizing and bringing new energy You

你们昨天就在讨论这个。

were talking about that yesterday.

这直接惠及实验室和所有建设数据中心的人。但这也非常符合美国的广泛利益。理论上普通美国民众也能受益。

Which directly benefits the labs and and and anybody building a data center. But it also feels very much in America's, interests broadly. It theoretically would benefit the average American too.

OpenAI需要非常清楚地认识到他们所处的地位——他们是全球最炙手可热的公司。市场对他们的股票需求无限。他们可以上市，也可以继续募集私募资金。他们需要站在风险曲线的另一端，而不是那些没人愿意投资的、可能十年都亏损的美国晶圆厂项目。

OpenAI needs to be crystal clear about the position that they're in, which is that they are the hottest company in the world. There is unlimited demand for their shares. They could be a public company. They could go raise more private capital. They need to be on the opposite end of the risk curve from the stuff that's like, no one really wants to invest in an American fab that might lose money for a decade.

想想那些短期内能赚钱的领域，就是SaaS（软件即服务），对吧？AI就是为SaaS服务的。没错。

Like, when you think about the things that make money in the short term, it's SaaS. Right? AI is for SaaS. Yeah.

人们的担忧确实如此。对于政府支持的数据中心贷款项目

People's concern Yeah. For, government backed data center lending

是的。

Yeah.

是的。你是在用芯片作为抵押品贷款，而芯片的贬值速度非常快。没错。我们也不清楚。如果你现在看看Core Weaves的情况，他们的公司信用违约互换已经达到500个基点左右了。

Yeah. Is that you're lending against chips, which have a really fast depreciation schedule. Yep. We don't know. And if you look at, right now, Core Weaves, corporate default swaps are now sitting around 500 basis points.

大幅飙升。这是领先的新云服务商之一

Jumped up dramatically. This is one of the leading Neo Clouds

确实如此

They are

仍保持在白金级别。新云服务商仍处于白金级。但人们开始担心他们了，对吧？是的。

still in the platinum. Neo Cloud in the platinum platinum tier. And people are worried about them. Right? Yep.

可能面临某些破产风险。所以如果开始做政府担保的数据中心贷款，可能会出现大量新建数据中心却找不到明确投资回报途径的情况。这会让整个产业链变得过于狂热。回想周三莎拉·弗莱尔的访谈，她还觉得市场不够狂热呢。

Potentially, you know, having some bankruptcy risk. And so if you start doing government guaranteed data center lending, you could get in a situation where there's a bunch of new data centers that come online that really don't have a clear pathway to ROI. Yeah. And it just incentivizes the entire stack to just really exuberant. Going back to Sarah Fryer's interview on Wednesday, she felt the market was not exuberant enough.

今年已经有很多疯狂和愚蠢的事发生了。也许我们不需要更多了。联邦住房金融局局长威廉·波尔特表示，房利美和房地美正在考虑入股科技公司。这位局长说两家公司正在研究如何获取科技公司股权。我们有些最大的科技上市公司正提议用股权换取房利美和房地美与其业务合作。

There's been a lot of insanity this year, silliness. Maybe we don't need more. The director of the Federal Housing Finance Agency, William Polt, says Fannie and Freddie eyeing stakes in tech firms. Bill Polt, the director of the Federal Housing Finance Agency, said that Fannie Mae and Freddie Mac are looking at ways to take equity stakes in technology companies. We have some of the biggest technology and public companies offering equity to Fannie and Freddie in exchange for Fannie and Freddie partnering with them in our business.

我们正在考虑入股那些愿意接受我们投资的公司，因为房利美和房地美对整个生态系统具有巨大影响力。确实如此。

We're looking at taking stakes in companies that are willing to give it to us because of how much power Fannie and Freddie have over the whole ecosystem. So Yeah.

这次《华尔街日报》的活动催生了许多文章。你们换了一整套新的调音台吗？调音台那边是什么情况？

This, this Wall Street Journal event is just so many articles came out of this. Did you get a whole new soundboard? What's going on with the soundboard?

是的，我保留了所有经典设备。

I did. I still got all the classics.

保留了所有经典设备。

Got all the classics.

正在尝试一些新素材。让我们为全球商界人士喝彩。

Working with some new material. Let's give it up for global businessmen.

让我们鼓掌欢迎。你有新邻居了。显然汤姆·佩蒂曾住在你加州马里布的社区里。他当时以1120万美元的价格住在马里布一栋8744平方英尺、七间卧室带音乐工作室的豪宅里。买家是精神分析学家兼作家史蒂文·斯莱德·蒂恩。

Let's give it up. Let's give it You have new neighbor. So Tom Petty apparently lived in your neighborhood in Malibu, California. He was living in Malibu in a $11,200,000 house, 8,744 square feet, seven bedrooms, has a music studio. The buyer is Steven Slade Thien, a psychoanalyst and author.

蒂恩没有回应置评请求。我在想他是否愿意有空一起喝啤酒、聚聚。我们应该联系一下

Thien didn't respond for a request for comment. I wonder if he you'd wanna, you know, get beer sometime, hang out. We should reach out to

请他发表评论。

him for comment.

我参加了第一届Outside Lands音乐节。

I was at the first ever Outside Lands.

真的吗？你多大了？

Really? How old are you?

汤姆·佩蒂是周六的压轴演出。那是在2008年8月，那是我第一次闻到大麻的味道。我不停地问朋友父母，这是什么

Tom Petty headlined Cool. On Saturday. This was in August 2008, and that was my first time smelling cannabis. And I kept asking, like, his friend's parents, what is

那种难闻的

that stinky

味道？太臭了。我们怎么也躲不开。是啊，真搞笑。

smell? It's so stinky. Every we can't we can't get away from it. Yeah. That's hilarious.

埃隆的万亿美元薪酬方案已成定局。已签署。已获批。我肯定会在法庭上受到质疑。这种事总是会被起诉的。

Elon's trillion dollar pay package is done. It's signed. It's approved. I'm sure it will be contested in the courts. It's always contested in the courts.

如果埃隆达成所有不同阶段的业绩目标，他可能获得价值1万亿的特斯拉股票。他现在身价已达5000亿美元，但还直接持有4.14亿股特斯拉股票，2018年又获得了3亿股的奖励，而即将获得的奖励是分12个阶段发放的4.24亿股。基本上，他们是在给他已经拥有的同等数量股票。他需要完成一系列任务：必须将市值推得非常高，此外还有这些定性运营目标，或者说我认为是定量目标。

Elon could get 1,000,000,000,000 in Tesla stock if he hits all these different tranches. He's worth half a trillion now, but he also owns 414,000,000 Tesla shares outright, got another award in 2018 of 300,000,000 shares, and this next award is 424,000,000 across 12 tranche tranches. Basically, what he already had, they're giving him the same amount again. And there's a bunch of things that he has to do. He has to get the market cap really, really high, and then there's also these, like, qualitative operational goals, or I guess they're quantitative.

500亿的EBITDA（息税折旧摊销前利润），交付2000万辆汽车，售出100万台机器人，运营100万辆自动驾驶出租车，1000万份全自动驾驶订阅。显然其中有些目标比其他更容易操作。机器人的定义是什么？如果他推出非常廉价的机器人并大量销售，但那更像是玩具，这真的算达成目标吗？自动驾驶出租车的定义是

50,000,000,000 in EBITDA, 20,000,000 cars delivered, 1,000,000 robots sold, 1,000,000 robotaxis in operation, 10,000,000 full self driving subscription. Now some of those are obviously more gameable than others. What's the definition of a robot? If he comes out with a really cheap robot and he sells a bunch of those because it's more a toy, does that really fulfill the goal? What's the definition of a

什么？符合条件吗？就是一辆启用了

robotax? Qualifies. It's it's a Tesla that is enabled for

现在你开启全自动驾驶功能，我朋友就能乘坐它。

And now you turn on FSD and my friend rides in it.

实际载客量有要求吗？

Is there anything for actual rides?

但这些目标比其他更容易操作，而市值目标确实不行。

But these are more gameable than others, but the market cap really isn't.

现在有多少自动驾驶订阅用户？

How many self driving subscriptions are there today?

我查过了，现在大概在一到三百万之间。所以他至少得扩大三倍规模。他还没卖过任何机器人，所以一百万将全是新机器人。显然他交付了很多汽车。

I saw I I looked that up. It's somewhere between, like, one and three million right now. So he has to he definitely has to, like, triple the size at least. He hasn't sold any robots, so a million would be entirely new robots. He's obviously delivering a lot of cars.

在EBITDA方面，500亿的EBITDA。公司去年大概是13亿，所以这是EBITDA的巨大增长。我是说，500亿EBITDA是很大一笔钱。他只需要把市值做到8.5万亿美元，而特斯拉已经价值一万亿了。所以生活在万亿富翁的世界会变得很奇怪。

And on the EBITDA front, 50,000,000,000 in EBITDA. Company did, like, 13 last year, so that's that's a huge increase in EBITDA. I mean, 50,000,000,000 in EBITDA is a lot of money. He's he only has to take the market cap to 8,500,000,000,000.0, and Tesla's already worth a trillion. So it's gonna be weird to live in the world of the trillionaire.

不过我们正在接近这个目标。这不仅仅会在我们有生之年发生，肯定就在未来十年内。我好奇这会如何重塑我们的文化，比如美国社会。因为当亿万富翁变得如此普遍和显赫时，百万富翁就少了很多压力。

Like, well but we are getting close. Like, that's going to happen not just within our lifetime, like, definitely within the next decade. I wonder how that's gonna reshape our our culture, like the world in America. Because when billionaires became so prevalent and prominent, there was a lot of heat that was taken off the millionaire.

亿万富翁就是挡箭牌。

Billionaires are the heat shield.

没错。所以百万富翁变得更触手可及，而亿万富翁成了社会甩锅的对象

Yeah. So the millionaire became more accessible, and the billionaire became the thing that the society scapegoats

为所有问题。

for all the problems.

大约9.4%至9.5%的美国成年人是百万富翁。

Approximately 9.4 to 9.5% of American adults are millionaires.

是的。比如，当万亿富翁出现时，亿万富翁会怎样？就像，当你不得不说万亿富翁才是真正的政策失败，但亿万富翁也是政策失败时会发生什么。百万富翁还算可以接受，但也不理想。这就像事情变得复杂多了。

Yeah. Like, what happens to the billionaire when trillionaires come in? Like, what happens when you have to say, like, well, like, trillionaires are the real policy failure, but billionaires are also the policy's failure. And millionaires were, like, kind of okay with, but it's not great. It's like it becomes much more complicated.

这对其他公司意味着什么？对OpenAI的Sam Altman意味着什么？他能采用类似的策略吗？他能去OpenAI的营利性董事会说，嘿，如果OpenAI的IPO估值达到2万亿美元，我要20%的股份吗？

What does this mean for other companies? What does it mean for Sam Altman at OpenAI? Can he run a similar playbook? Could he go to the OpenAI for profit board and say, hey. If if OpenAI IPO is at 2,000,000,000,000, I want 20%.

我们知道Sam有点效仿Elon的策略。他们曾一起做生意，共同创立了OpenAI。我也好奇那些普通独角兽公司会怎样。如果你只是一家50亿美元公司的CEO，在那儿混日子，然后说，嗯...

We know that Sam runs a bit of an Elon playbook. They were in business together. They cofounded OpenAI together. And then I also wonder what what will happen at the garden variety unicorn. If you're just the CEO of a $5,000,000,000 company and you're just kinda hanging out there and you say, like, yeah.

我刚开始时拥有公司30%的股份，现在被稀释到5%或10%。但我喜欢这家公司，想把它做起来。我能去董事会说，我们现在估值50亿，如果我能做到500亿，你们能把我股权翻倍吗？

I I had 30% of the company when I started. I've been diluted down to five or 10, But I like this company, and I wanna get it up. Could I go to the board and say, okay. We're at 5,000,000,000 now. If I get us to 50, will you double my equity position?

那些成长阶段的创投公司会... 怎么说... 对此有何感想？

How would the growth stage venture companies Yeah. Feel about that?

我认为创始人应该这样想：除非你决定出售，否则没人能拿走你的股份。是的。你的工作就是让股价上涨。是的。而且未来还会发行更多股票。

I think the right way for founders to think about that is, like, no one's taking your shares Yeah. Unless you decide to sell them. Yes. Your job is just to make the share price go up Yeah. And there's gonna be more shares issued over time.

是的。但如果你能让股价永远上涨，那其实无所谓。你也可以回购。对，还可以回购。

Yeah. But if you just make the share price go up Yeah. Forever, it doesn't really matter. And you can also buy back. Also, can also buy back.

是的。几乎没有其他CEO会接受这样的交易，因为它太雄心勃勃了。是的。所以我认为这是健康的。

Yeah. Almost no other CEOs would take a deal, like, because it's so ambitious. Yep. And so I think it's I think it's healthy.

我们的好朋友泰勒·科斯格罗夫为我们准备了一份幻灯片，试图帮助绘制'七大巨头'。实际上，我称之为TBPN前十强。TBPN前十强，大致是AI领域最重要的十家公司。对，就是七大巨头加上几个额外成员。

Our good friend Tyler Cosgrove has put together a slide deck for us that tries to help map the mag seven. Really, I call it the TBPN top 10. TBPN top 10, the 10 most important companies in AI loosely. Yeah. The mag seven plus a few bonus ones.

我们要尝试

We're gonna try

并注意你的头部动作时间。

and Watch your take head time.

完成。你看。你要击中要害。我们将尝试梳理这些公司，根据它们对AGI的接受程度和需求程度进行排名。

Through. And look. You're you're to hit the horse. We're gonna try and go through the various companies and rank them based on how AGI pilled they are and how much they need AGI. Is

是这样吗？所以基本上在横轴上，我们看它们对AGI的接受程度

that right? So, basically, on the horizontal, we have how AGI pilled they

是的。

are. Sure.

所以我觉得这相当不言自明。某种程度上你相信AI会变得像...它能产生相当于

So I feel like that's fairly self explanatory. You kind of believe that AI will become something like it it can produce the the median economic output of

一个人的经济产出中位数。

a person.

那么在纵轴上，我们标注他们需要AGI的程度。好吧。我觉得这个可能稍微难一点。我想先做个说明。是的。

So then on the vertical axis, we have how much they need AGI. Okay. So I think this is a maybe a little harder. I so I I wanna qualify this. Yeah.

我的意思是，这并不一定代表你认为AI会拥有类似人类的意识。更多是在这个语境下，AI的经济价值会持续增长。我们先从Sam Altman说起，他相信AGI。

So, I mean, this doesn't necessarily believe that you have this kind of sentient, you know, AI that's as good as a person. Yeah. But I I think it more so in this context just means that AI will continue to become more and more economically valuable. Let's start, with Sam Altman. He believes in AGI.

对吧？他经营着最大的AI公司。

Right? He he runs kind of the biggest AI company.

我认为至少在短期内，OpenAI越来越像一个超大规模服务商。他们算是初级版的超大规模服务商。很多人想说他们对当前水平的OpenAI持悲观态度。但最终，当你观察他们业务的发展...嗯...在我看来，即使模型性能停滞不前，他们也能过得去。

I think that more and more, at least in the short term, OpenAI looks like a hyperscaler. They're kind of a junior hyperscaler. And I think their actions are more think that OpenAI, a lot of people want to say that they're bearish on OpenAI at current levels. But ultimately, when you look at how their business is evolving Mhmm. They seem to me like they'd be fine if the models plateaued.

接下来，我们谈谈Dario。

Next, have Dario.

好的。他是极度支持AGI药丸的。这就是为什么他如此反华的原因。对吧？

Okay. He's extremely AGI pill. This is kind of the reasoning why he's he's so anti China. Yeah. Right?

因为他把这视为一场真正的竞赛。这关系到超级智能、核武器。这是国家安全问题，绝对的。

Because he sees it as an actual race. This is going to super intelligence. Nuclear weapons. It is a national, you know, security. Totally.

如果中国先达到目标就会是个大问题。无人机在线。Anthropic需要持续增长才能在经济上保持合理性。

A problem if China gets there first. UAV online. You you need a lot of continued growth for for Anthropic to keep kind of making sense economically, I think.

好的。明白了。

Okay. Yeah.

对。下一位是Larry。Larry。Larry。Larry。

Yeah. Next is Larry. Larry. Larry. Larry.

Larry现在处境很有意思。他处在一个不相信AGI但又需要...的奇怪位置

Larry's in kind of an interesting spot here. Yeah. So this is kind of a weird place to be where you don't believe in AGI, but you need

它。好吧。他是怎么走到那一步的？

it. Okay. How did he wind up there?

他看起来不像会相信某种超级智能的类型

He doesn't seem the type that is believes in some kind of super intelligent

是的

Yes.

那个即将降临的上帝，你知道的，会催生这个新事物。是的。人类将会崛起。好的。

God that that is gonna come, that's gonna, you know, birth this new thing Yeah. And humanity will will rise. Okay.

萨莎。好的。

Sasha. Okay.

我们开始吧

Here we go.

我认为这是个相当合理的地方。是的。显然，某种程度上他有点沉迷AGI（人工通用智能）。是的。或者可能不止是有点。

I think this is a fairly reasonable spot. Yeah. Obviously, there's, you know, there's some sense where he is slightly AGI pilled Yeah. Or maybe more than slightly.

相信技术的力量。是的。

Believes in the power of the technology. Yeah.

我是说，他在OpenAI的早期阶段。是的。他认为AI总体上会变得非常有用，但也许不会变得超级智能。也许它不会取代每个人。它只是一个有用的工具。

I mean, he's very early on OpenAI. Yeah. He he thinks that AI in general will become very useful, but maybe it won't become, you know, super intelligent. Maybe it's not gonna replace every person. It's just a useful tool.

这是下一个

It's the next

我经常引用的一句话是他说：我对AGI的定义就是更大的经济增长。所以给我看经济数据，那就是——这是个非常实用的

quote I always go back to is him saying, like, my definition of AGI is just greater economic growth. So show me the economic numbers, and that will be it's like, it's a it's a very practical

是啊。

Yeah.

定义。我认为人们觉得他很理性。他不会过度冒险。

Definition. I think people see him as very reasonable. He's not gonna over skis.

是的。我喜欢他处于中间位置，在网格的某个中心点。那看起来才是真实的

Yeah. I like him in the center I like him in the center of the grid somewhere. That that seems like a real

而且如果OpenAI成功了，他也会做得很好。当然。如果他们不成功，我想他也

He's also if OpenAI works out, he'll do very well. Sure. If they don't work out, I think he's also

他有所保留。

He's hedged.

他做得相当不错。如果Jensen非常沉迷于IGI 是的。我是说，他就是这样的人 他是这一切建立的基石

He's doing quite well. If Jensen was very IGI pilled Yes. I mean, he is the the kind of he he's the rock on which this all is built

是的。

on. Yes.

他拥有筹码。

He has the chips.

是的。是的。

Yes. Yes.

如果他沉迷于AGI，他就不会分发那些筹码。他会全部留给自己，并训练自己的模型。

If he was AGI pilled, he would not be doing out those chips. He would keep all to himself, and he would be training his own model.

好的。

Okay.

这就是为什么我认为他更倾向于肯定。他不相信通用人工智能。是的，昨天有新博客文章发布，基本上就是AI 20.7。

So that's why I think he's he's more on the Yes. Doesn't believe in AGI's. Yes. There was a new blog post yesterday. It was basically AI twenty point seven.

有新内容发布，是AI 2032。所以基本上是同一个团队？不同团队，但AI 2027团队在推广它。桑达尔。

There was new one. It was AI twenty thirty two. So it's basically team? Different team, but the the team of AI twenty twenty seven was promoting it. Sundar.

你认为他比萨蒂亚更相信通用人工智能。

He believes in AGI more than Satya, you think.

是的。我觉得你可以从某种程度上看出这点，他们在某些方面甚至比萨蒂亚更早，对吧，比如DeepMind。

Yeah. Well, I I think you can see this in kind of they were even earlier in some sense than Satya, right, with with DeepMind.

没错。有点像是，如果你真的相信通用人工智能，我们看到的行动就像是在扭动身体说'我必须参与'。不管我是落后1%、10%还是80%，我都必须参与。

Yeah. There is a little bit of, like, if you really believe in AGI, the actions that we see are you, like, squirming and being like, I gotta get in. I I it doesn't matter if I'm 1% behind or 10% behind or 80% behind. I gotta get in.

我认为桑达尔肯定也在这条线以下，因为谷歌一直做得很好。嗯。AI最初出现时，人们认为它会摧毁谷歌，对谷歌不利。所以如果AI不成功，谷歌就保持原状，依然发展得很好。

I I think Sundar is also definitely below this line because, you know, Google has been doing very well. Mhmm. AI was at at first, I mean, you people thought of AI as, like, oh, this is gonna this is gonna destroy Google. This is bad for Google. So if AI doesn't work, then Google is just in the spot they were before, which is doing very well.

如果AI成功了，那么Gemini嗯...是最好的模型之一。他们会做得很好。所以扎克伯格也处于一个有趣的位置。是的。

If AI does work, then, I mean, Gemini is Mhmm. One of the best models. They'll do very well. So Zuck is also kind of in an interesting spot. Yes.

我认为扎克实际上是一个立场右倾的人。

I think I think Zuck is actually someone who has shifted rightward.

这很有趣。感觉扎克当时的态度就像是‘哦，对，人工智能。这很酷，我要勾选这个选项。’

It's fascinating. It felt like Zuck was sort of like, oh, yeah. AI. Like, it's this cool thing. I'm gonna check the box.

我有自己的团队。我们做了个有趣的小副业项目，是个开源模型。我们算是找到了自己的小赛道，但并没有参与那场宏大的宇宙级竞争。

I got my team. We did this fun little side project. It's this, open source model. We kinda found our own, like, little lane, but we're not, like, competing in the big cosmic battle.

你觉得这只是为了在AI军备竞赛中另辟蹊径吗？比如宣称‘我们要把这个市场商品化，采取中国式策略’？

Do you think that was just a counter position to way to win like, try to win the AI warp? Go say, hey. We're just gonna try to commodit commodify this market and, like, take a Chinese approach?

说到埃隆，我认为他很久以前就深信AGI（通用人工智能）了。

So Elon, I mean, he's been AGI pilled, I think, a very long time.

超级AGI信徒。我同意这个说法。

Super AGI pill. I agree with that.

作为OpenAI联合创始人，甚至在那之前，我认为他在安全领域就已经相当有影响力了。

OpenAI cofounder. Even before that, he I think he was fairly big in the in the safety space.

完全正确。

Totally.

你甚至在乔·罗根的节目上看到他谈论AI安全问题。他依然坚信这一点。

You see him even on on Joe Rogan, he was talking about AI safety. He still believes it.

是啊。他并没有退缩。

Yeah. He hasn't backed off.

AI安全之所以重要，是因为它将变得超级智能。它会统治世界。完全正确。完全正确。

And AI safety is important because it's gonna become super intelligent. It's gonna over the world. Totally. Totally.

所以他昨天谈到人形机器人就像是无限金钱漏洞。我觉得你需要通用人工智能才能真正解锁这个无限金钱漏洞。

So he talked yesterday about about humanoids being sort of an infinite money glitch. And I feel like you kind of need AGI in order to kind of unlock the the infinite money glitch.

没错。但同时，核心业务非常强大。汽车不需要通用AI。火箭不需要通用AI。星链也不需要通用AI。

Yes. But at the same time, very strong core business. The cars don't need AGI. The rockets don't need AGI. Starlink doesn't need AGI.

所以，是的，所以他并不像基金会实验室那样完全押注在这上面。

So, yeah, so he's so he's not entirely indexed on it in the way the Foundation Labs are.

查德问，泰勒在图表上的哪里？感觉泰勒已经不是AGI药丸了。

Chad says, where's Tyler on the chart? Feels like Tyler isn't AGI pill anymore.

我们选

Let's pick

他吧。

him up.

其实我就在这图表上。

I actually am on the chart here.

我们去找泰勒。

Let's go to Tyler.

我就是啊。我觉得我充满了AGI特质，对吧？我已经准备好迎接戴森球了。

I am I yeah. I I think I'm very AGI filled. Right? I know I'm ready for the Dyson sphere.

是的。

Yes.

我觉得，你知道，这只是几年时间的问题，屈指可数。

I think it's, you know, only a matter of years, handful.

只是几年时间的问题。

Only a matter of years.

如果人工智能不奏效，宏观经济

If AI does not work, the macro economy

是啊。

Yeah.

看起来不太妙。所以我对自己的工作前景感到相当悲观。好吧，没有通用人工智能的情况下。

Is looking not good. So I feel pretty bad about my job outlook Okay. Without AGI.

当然。当然。当然。尽管你已经就业了。

Sure. Sure. Sure. Even though you're already employed.

突发新闻。

Breaking news.

梳理一下时间线。关键转折点是什么

Walk through the timeline. What's the breaking

新闻？

news?

这就是你的福特猛禽的问题所在。不是那个R的问题

This was the problem with your Ford Raptor. It wasn't the r.

1993年3月，零到六十英里加速

Zero to 60 in 03/1993

达到720马力。5.2升机械增压V8发动机

to 720 horsepower. 5.2 liter supercharged V eight.

关于奥尔特曼。我们在考虑出售算力，但需要尽可能多的资源。扎克，上周我们还能出售算力。我们现在是算力短缺还是过剩？因为双方都说在尽可能多地购买算力，同时也在考虑出售

Altman today. We're looking at selling compute, but we need as much as possible. Zach, last week, we could sell compute. Are we in a compute shortage or not? Because both are saying they're buying as much of it as they can and thinking about selling it.

我喜欢那个关于黄仁勋的故事，他理解这里的动态，但如果需求端出现抛售，你还是会被压垮

I like that that story about Jensen where, like, he understands the dynamic here, but there's still you get kinda crushed if there's a sell off in terms of demand.

尽可能多买些以获得优先使用权，用不完的部分再转卖出去，这主意其实不算太糟。

It it's not the worst idea to buy as much as you can so that you have preferred access to it and then resell some if you have more than you need.

我是说，我们刚和Roblox的大卫·巴祖基聊过，他说他们原本有自己的本地服务器，但后来遇到了需求高峰。所以转向超大规模云服务商，因为他们能实现负载均衡——这边有人玩Roblox时，那边可能有人在看网飞，同时还要存储各种数据。不同时段会有不同工作负载。直接跳到卖算力这一步，我觉得时间线安排有点诡异。

I mean, we just talked to David Bazooki from, from Roblox, and he was saying that, look. We had our we had our own on prem, but then we had we had spikes of demand. And so we went to the hyperscalers for that because they can load balance across while people are playing Roblox here, and then maybe they watch some Netflix over there, and then they are storing, you know, all sorts of different data. And there's different workloads that happen at different times. Jumping straight to selling compute, I think that the timelines are a little bit funky on this one.

这看起来很奇怪。感觉很仓促。非常仓促。埃隆似乎已经确认特斯拉要建半导体晶圆厂了。

It seems odd. It seems rushed. It seems rushed. Elon apparently confirmed that Tesla is going to build a semiconductor fab.

但这还不够。所以我们可能得建个特斯拉'太拉工厂'。哇，这名字真棒。哦。

It's still not enough. So I think we may have to do a Tesla TeraFab. Woah. Great name. Oh.

这消息太劲爆了。

That is a bombshell.

就像'千兆工厂'的升级版——'太字节工厂'。他状态很好对吧？我想不出其他办法能达到我们需要的芯片产量。所以估计我们得建个巨型晶圆厂。

It's like Giga, but way bigger. Terabyte. Terabyte. He's feeling good I right I can't see any other way to get to the volume of chips that that we're looking for. So I think we're probably gonna have to build a gigantic trip out.

台积电表示：不必。其实我们产能充足，可以供货。三星也说：没问题，交给我们。

Morris, giant TSMC is like, no. I actually, I'm fine. I will I will supply you. Samsung is like, I'm good. I will do it.

把另一个视频调出来，把擎天柱的这个视频调出来。

Pull up this other pull up this other video of Optimus.

有点像那种

There's something like

这东西，这东西会动。

This thing this thing has motion.

是啊，还不错。

Yeah. It's not bad.

想象一下，你某天晚上在办公室加班，这东西突然走出来盯着你，还开始做这些动作。

Imagine you're working late at the office one night and this thing just walks out onto the floor and starts looking at you and doing these moves.

对。大多数买特斯拉的人本来就有车。对吧？所以这就像是一对一的替换。你要用它取代谁？

Yeah. Most people that bought Teslas already had cars. Right? And so it's just like a one for one slot swap. Who are you replacing with this?

嗯，用这个的话，你可以取代，比如说，一个异国舞者。

Well, with this, you can replace, like, a an exotic dancer.

是的，这是真的。埃隆·马斯克现在银行账户里有1万亿美元。这是1000个10亿美元。他给地球上每个人分10亿美元后还能剩下9920亿美元。

Yes. That's true. Elon Musk now has $1,000,000,000,000 in his bank account. That's a thousand times $1,000,000,000. He could give every single human on Earth 1,000,000,000 and still be left with 992,000,000,000.

好好想想吧。每次这种趣味数学出现时人们都爱转发。中国也有类似的人形机器人项目，不过要恐怖得多，因为它完全进入了终结者模式。进入了诡异恐怖的领域。这相当疯狂。

Let that sink in. People love this funny math whenever it drops. China has a similar humanoid robotic project, although it's way, way scarier because it went full Terminator mode on this. In spooky, spooky territory. This is pretty crazy.

过去十个月里，三位非常有才华的朋友分别加入了不同的早期热门初创公司担任高管，却在发现公司实际收入远低于创始人在面试过程中和网上分享的数字后离职了。

In the last ten months, three very talented friends have joined separate hot early stage startups in senior roles and quit after realizing that the company's actual revenue was significantly less than what the founder had told them during the interview process and and shared online.

是啊。这又要回溯到春季了，对吧？当时感觉每个创始人都承受着疯狂压力，必须展示出那种从100万到1000万的疯狂增长曲线。

Yeah. Again, this is going back to spring. Right? That just felt like there was every founder was feeling this insane pressure to show, like, one to 10 million of like, a one to ten million ramp that was just insane.

而奇怪的是，随着这种压力不断升级，你只会越来越有动力用社区调整后的年度经常性收入和实际上不靠谱的合同来造假。

And the weird dynamic is that, like, as that pressure ramps up, you just get more and more incentive to fake it with community adjusted ARR and contracts that don't actually stick. And

哈萨克斯坦已签署谅解备忘录，准备从英伟达购买最多20亿颗先进芯片。让我们为哈萨克斯坦敲锣吧。预热起来。预热起来。

Kazakhstan has signed an MOU Let's go. To buy up to 2,000,000,000 of advanced chips from NVIDIA. Let's hit the gong for Kazakhstan. Warm it up. Warm it up.

预热起来。砰。漂亮的一击。我们哈萨克斯坦的朋友们干得漂亮。很高兴看到他们加入这场游戏。

Warm it up. Boom. Great hit. Great hit for our friends in Kazakhstan. Good to see them getting into the game.

也许他们应该拍《波拉特3》关于数据中心的故事。英伟达的亏损加速至日内负5%，自周一高点以来已下跌16%。这意味着自周一起市值蒸发了8.8万亿美元。哇，这简直是疯狂抛售。

Maybe they should do Borat three about data centers. Breaking NVIDIA's losses accelerated to negative 5% of the day, now down 16% since Monday's high. That marks a drop of $8,800,000,000,000 since Monday. Wow. That is a wild sell off.

泰勒不是引用过这个然后问，这算是利好消息还是什么来着？

Didn't Tyler quote this and say, is this bullish or something like that?

不，那说的是DoorDash。他们跌了20%。

No. That was on the the DoorDash. They went down 20%.

所有股票都跌了10%到20%。

Everyone's down 10 or 20%.

无论业绩达标还是未达预期，股价都会下跌。海森堡正在分享...对，分享一点红色（跌势）。微软过去八天跌了10%。他到底在搞什么？

We need to Whether you meet whether you beat or miss, you're going down. Heisenberg is sharing Yeah. Sharing a little bit of red here. Microsoft down 10% in the last eight days. What the hell is he doing?

他在购物。

He's shopping.

微软跌了三个点。

Microsoft went down three points.

这确实很糟糕。Meta下跌18%，亏损加速扩大。我的收件箱里全是人们急切想解读这情况的消息，答案是肯定的。看看标普一年的走势

That's real good. Meta down 18%. Losses, have accelerated. My inbox is full of people breathlessly trying to interpret this, and it is Yes. One year look at spy

结束了，彻底结束了。泡沫破裂了。现在可以开始重建了。

It's over. It's just completely over. The bubble popped. Now it's ready to start rebuilding.

这是一次

It was a

不错的历程。我们可以从这里重新出发。

good run. We can go out from here.

哦，股票已经停止交易了。所以我们现在安全了。

Oh, the stocks have already stopped trading. So we're safe now.

我们安全了。他们再也伤害不了我们了。

We're safe. They can't hurt us anymore.

周末经济伤害不了你。美国银行确认，截至昨日已从全球金融危机中完全恢复19个百分点。哇，整整花了十九年

The economy can't hurt you on the weekend. Bank of America Yes. As of yesterday, fully recovered 19 from global financial crisis. Wow. Took nineteen

多年后，某一天，突然又陷入另一场危机。你能想象吗？众所周知，这次是由沃伦·巴菲特亲自兜底。

years. The day, and then it's, like, back in another crisis. Can you imagine? Famously backstopped by none other than Warren Buffett.

关于Snap还有更多信息

There's some more information on the Snap

Snapchat的交易。

Snapchat deal.

Snap获得40亿美元，这比Perplexity的总收入还多。嗯。Snap除了获得访问权外什么都不用付出

Snap gets 400,000,000, which is greater than Perplexity's total revenue. Mhmm. Snap gives nothing except access to

等等。等等。哇。更多？什么？

Wait. Wait. Woah. It's more? What?

他们支付的金额超过了总收入？这笔交易看起来对Snap极其有利。Snap获得的40亿美元比Perplexity总收入还多。现在Signal表示这很可能是Perplexity给Snap的3.99亿美元股权，而非现金。所以关于其中现金比例存在疑问，但有可能他们只是给了价值4亿美元的股权，就像股票赠予。

They're they're they're paying more than the revenue? The deal looks incredibly in Snap's favor. Snap gets $400,000,000 greater than Perplexity total revenue. Now Signal is saying this is most likely 399,000,000 of Perplexity equity to Snap, not cash. So there is a question about how much of it is cash, but it is possible that they're just giving $400,000,000 of equity like a stock grant.

我很惊讶Snapchat还没想出办法，将这些消费级AI公司凭借其庞大用户群所获得的巨额资本全部变现。

I was surprised that Snapchat has not figured out a way to just monetize all the capital that a lot of these, consumer AI companies have given their massive, massive user base.

Snap将把Perplexity直接整合到Snapchat的聊天界面中。作为交易的一部分，Perplexity将在一年内向Snap支付4亿美元现金加股票，并获得Snap 9亿月活跃用户资源。Snapchat可能会在无意间开发出AGI，就为了

Snap will be integrating Perplexity directly into Snapchat's chat interface. Perplexity will play will pay Snap $400,000,000 over a year in a mix of cash and stock as part of the deal and gets access to Snap's 900,000,000 monthly active users. Snapchat's gonna accidentally build AGI just trying to

让狗头滤镜能真实地眨眼。

make the dog filter blink realistically.

我喜欢这个。

I love it.

德意志银行正在探索对冲其对AI数据中心风险敞口的方法。他们考虑的方案包括做空一篮子AI相关股票，以及通过合成风险转移购买违约保护。所以

Deutsche Bank is exploring ways to hedge its exposure to AI to data centers. It's looking at options including shorting a basket of AI related stocks and buying default protection via synthetic risk transfers. So

就像，要参与这场游戏。就像几年前，就像那个，如果你有任何与AI相关的东西，就赶紧融资、扩张、包装、转手、重组、转型。

Like, is getting in on the action. Like, few years ago, like the like, it was, like, if you have anything that's related to AI, like, go, go, go, raise money, grow it, Spin it. Flip it. Turn it around. Pivot it.

想怎么玩都行。

Whatever you wanna do.

祝大家周末愉快。我们爱你们。再见。周一见。

Have a great weekend, everyone. We love you. Goodbye. See you Monday.