资本主义的终局：最后一批将存在的企业

然后他们突然意识到。

And then they're like, oh, wait a second.

我们实际上需要产生一些收入，以便偿还为筹集这些资本而欠下的债务。

We actually need to make some revenue so that we can pay back the debts from this capital we need to raise.

而他们为了追赶那些已经拥有充足资本来部署和构建这些技术的巨头，一直在不断累积这种永久性债务。

And it's this perpetual debt that they are accruing just to catch up to these conglomerates that have the capital already to to deploy and to build these things.

因此，OpenAI 的智能问题确实存在，你也能看到他们在新产品上面临的挣扎，比如 Rumors 社交媒体信息流或 Sora 信息流。

So the intelligence problem with OpenAI is a real one, and you see the struggle as well with their new products like the the Rumors social media feed or the Sora feed.

这感觉有点迷失方向。

It just feels a little lost.

他们似乎更在意为了吸引用户和赚取收入而追求病毒式传播，而不是真正创造能带来全新价值的产品。

It feels like they're trying to go viral for the sake of earning users and earning revenue versus actually creating a really compelling product that generates net new value.

不只是 Web 2.0 那套，我们现在是在打造带广告的 AI 版 Facebook，但还加上了

Not just like web two point o, we're creating the AI Facebook now with ads on top, but with that

2.0，但把它变成 AI 版。

point o, but make it AI.

是的。

Yeah.

因为他们宣布正在开发一个经过人类验证的WorldCoin社交网络，由于其人类验证的特性，这可能会比我们目前的社交网络有重大升级。

Because they they announced that they are working on a, you know, a WorldCoin human verified social network, which is potentially a great upgrade from the social networks that we currently have because of its human verified.

但即便如此，老兄，这终究还是个社交网络。

But nonetheless, dude, it's a social network.

我的天，谁在乎啊？

Like, who the f cares, man?

我们已经有这些了。

We got those.

到2026年，这对人类有什么价值呢？

What value was that contributing to humanity in the year 2026?

是的。

Yeah.

这将是最大的悲剧，因为2000年代初到2010年代中期，大量的工程人才和资源都投入到了打造算法驱动的社交信息流上，目的是让人上瘾、多看广告、增加收入。

And it's it's it would be the ultimate tragedy, right, is because the so much of the engineering density and talent in the early two thousands, mid mid twenty tens went towards generating these algorithmic social feeds to get people hooked and addicted to see more ads, to generate more revenue.

而这种商业模式应用到人工智能上，显得如此悲哀和枯燥，因为人工智能有如此巨大的潜力。

And that business model applied to AI just seems so sad and so dry because there is so much upside with this AI.

它的上限是无限的。

It is uncapped.

你可以做许多令人惊叹的新事情，创造新产品，并在其基础上不断创新。

You can do so many amazing new things, create new products, innovate on top of it.

如果只是陷在旧模式里，继续在那里建设，那就太遗憾了。

It would be a shame to just get stuck back in that old place and continue to to build there.

好吧。

Okay.

所以，前两个拼图块是智能和资本。

So that those are the first two puzzle pieces, intelligence and capital.

我认为我们很好地阐述了这两者之间的关系。

And I think we did a good job illustrating the relationship between those two things.

我想把第三个宝石加入到资本主义的灭霸手套中——能源。

I wanted to add the third jewel to the Thanos gauntlet of capitalism, energy.

正如我们之前提到的，当你向大语言模型注入更多能量时，就能获得更强的智能。

And as we alluded to, you get more intelligence when you just pump energy through LLMs.

这其实是你教给我的。

That's actually something that you taught me.

也许你可以再多讲讲这一点。

Maybe you could talk a little bit more about that.

同时也谈谈能源在智能以及几乎所有方面的重要性。

Also talk about just like the importance of energy as it relates to both intelligence and also just like kinda everything.

是的。

Yeah.

在那副手套中，最重要、归根结底唯一真正关键的是能源这一支柱，因为没有它，就既没有资本，也没有智能。

In the gauntlet, the most important one, and at the end of the day, the only one that truly matters is this energy pillar because in the absence of it, there is no capital, there is no intelligence.

资本和智能都是能源的直接下游产物。

It is a direct downstream effect of energy.

我们在获取能源方面遇到了困难，这将是一个持续的挑战，很可能成为制约AI未来发展和进步的瓶颈。

And we have some trouble getting energy, and this is going to be a continued struggle and probably some sort of a bottleneck in terms of moving forward and how AI progresses.

但能源问题至关重要，因为能源输入与智能输出之间存在直接关联。

But the energy thing is super critical because there is that direct correlation between energy in and intelligence out.

最能利用这一点、实现吉瓦到太瓦级AI计算和训练的公司将会胜出，因为能源是最稀缺的资源。

And the companies that are most equipped to leverage this to get to multiple gigawatts to terawatts of AI compute and training, they're going to be the ones that that win because it is the scarcest resource.

获取能源实在太困难了。

It's so difficult to get energy.

在美国，我们每年的能源产量大约是1.3太瓦，而我们实际使用的也差不多是这个量。

In The United States, we produce what is I think we use about 1.3 terawatts of energy per year.

我们目前只能利用其中大约一半的能源，而到本十年末，AI数据中心的能耗预计将占据其中很大一部分，除非我们新建更多能源设施。

We're only able to use about half of that, and the projections by the end of this decade for AI data centers are going to consume a huge percentage of that if we don't build more energy.

因此，能源问题是非常现实的。

So the energy problem is very real.

从某种意义上说，在我们所暗示的这种丰裕世界中，能源也成为了资本的终极形态，最终的货币很可能是瓦特而非美元。

And in a way, energy, I believe, in a world of abundance like we are we're kind of hinting at, energy does become the end state of capital too, where, like, the end currency is likely wattage instead of dollars.

我认为这是一个重要的衡量标准，因为瓦特确实能带来实际价值。

And I think that's an important denomination because wattage can actually get you value.

瓦特数可以转化为任何东西。

Wattage can be converted into anything.

它可以转化为智能，因此也能转化为其下游的任何事物。

It could be converted into intelligence and therefore anything downstream of that.

而货币，说实话，非常不稳定。

Whereas currency, I mean, it's it's very ficky.

它真的很不稳定。

It's very, like, fickle.

它可以被贬值。

It can it can be debased.

它可以被改变。

It can be changed.

但能源就是能源，我们目前短缺，随着我们继续扩展这些AI系统，未来可能也一直会短缺。

But energy is energy, and we have a shortage of it, we probably always will as we continue to scale these AI systems.

你所暗示的是，当你集齐了所有四个拼图碎片时，你就拥有了能源。

Think what you alluded to is, like, when you have all four puzzle pieces, you have energy.

你拥有能源、劳动力、资本和智能。

You have energy, you have labor, you have capital, and you have intelligence.

你不再需要钱了，因为这四样东西是你在这个世界上做任何想做的事所需要的。

You don't need money anymore because those are the four things that you need to do anything that you want in the world.

你可能会觉得，资本其实就是钱，而钱本质上只是在协调资源。

You'd like capital has just been, money has been basically just trying to coordinate resources.

但如果你拥有了所有资源，并且能接触到所有资源，你就根本不需要自己去协调它们，因为你有足够的智能来完成这一切。

But if you own all the resources and you have access to all the resources, you don't actually need to coordinate them yourself because you have the intelligence to do so.

但我们有点跑题了。

But we are skipping ahead a little bit.

我总是忍不住想，再深入聊聊能源这个问题。

I do I keep I keep wanting to I wanna drill down a little bit more on the energy thing.

真的有数据中心里装着NVIDIA芯片却闲置着，因为没电吗？

Is it true that there are data centers out there that have, like, NVIDIA chips ready to go and they're just idle because they don't have power?

我不久前听说了这样的事。

I heard that, like, somewhere not terribly long ago.

这是真的吗？

Is that true?

确实有过这样的情况。

There there was a case where that's true.

我不确定现在是否还是这样，但我认为这将是一个迫在眉睫的问题。

I'm not sure that still is, but that will be an impending problem, I would imagine.

这意味着我们在生产更多的GPU吗？

Does that mean does that imply that we are producing more GPUs?

我们制造GPU的速度比生产足够电力来驱动它们还要快吗？

We're manufacturing GPUs faster than we can produce energy to power them?

目前是这样。

Currently.

看起来这确实是当前的趋势。

It it appears as if that is the the current trend.

我不认为这种情况会持续下去，因为在未来几年内，如果我们没有更多晶圆厂投产，就会遇到芯片产能瓶颈。

I don't suspect that will continue to be true because we are we are very much going to hit a chip wall in the next couple of years should there not be more fabs that come online.

但就能源问题而言，这是真的，而且原因有两点。

But in terms of the the energy issue, it is true, and it's it's due to two things.

其中之一是电网没有足够的电力来供应这些数据中心，同时不破坏其他所有用电需求。

One of it is just that the grid doesn't have a sufficient amount of energy to supply these data centers without ruining everything else.

这些数据中心非常耗电。

It's like these data centers are very power hungry.

我认为，XAI 新建的 Colossus 数据中心之一，耗电量几乎相当于旧金山这座城市。

A single data center I believe one of one of XAI's new Colossus data centers consumes about as much electricity as San Francisco, the city.

所以，大量的能源集中在一处，给电网带来了沉重负担。

So it's a tremendous amount of energy density packed into one place, and it does have a burden on the grid.

这个问题的第二部分是建造数据中心，以及获得许可、审批和所有其他接入电力所需的条件。

The second part of this problem is actually building the data centers and getting the rights and the approvals and all of the other requirements needed to get the power plugged into this thing.

它们在建设地的城市遇到了很多阻力，因为这对所在城镇或城市造成了巨大影响。

They're meeting a lot of resistance through the cities that they build in because it's a big impact on whatever town or city they're being built in.

这对电网造成了巨大压力。

It's a huge strain on the grid.

它们消耗大量资源。

They use a lot of resources.

它们产生很大噪音。

They make a lot of noise.

这对它们所在的城市或州来说负担很重，那里也出现了问题。

It's it's very taxing on a city or state that it's placed in, and they're having problems there too.

所以这是双重问题。

So it it is twofold.

我们确实在面临能源问题。

We're running up against the energy problem for sure.

会出现大量无法供电的GPU，因为我们根本供不起电。

There will be dark GPUs because we just can't power them.

但也存在一种情况，我们能供电，却缺乏GPU。

But then there's also a world in which we can power them and we don't have GPUs.

因此，这里的资源之间存在着极其脆弱的平衡。

So it's this very fragile balance between the resources here.

而且，当大家都喜欢谈论泡沫会在哪里形成、会如何发展时。

And, like, when everyone likes to to talk about a bubble and where it's gonna form and how it's gonna play out.

重要的是要关注所有相关因素以及整个技术栈。

And it's important to just kinda have your eye on all of the the elements that are involved and and down the whole stack.

甚至连内存和RAM都正成为巨大的瓶颈。

Even things like RAM and memory are becoming a big bottleneck.

因此，能源很重要。

So energy is important.

这很有挑战性，尤其是面对所有法规，以及那些并非为数据中心设计的老旧电网，这引发了很多问题。

It is challenging, particularly with all the legislation, particularly with the an outdated grid that isn't built for data centers, and it's causing a lot of problems.

是的。

Yeah.

好的。

Okay.

所以，你向AI模型输入更多能源，就能获得更强的智能。

So you pump more energy through an AI model, you get more intelligence.

另外，如果你向最后一个拼图块——劳动力——投入更多能源，你也会获得更多的制造业。

Also, if you pump energy through the last puzzle piece, labor, you also get more manufacturing.

所以，这些也与其它拼图块息息相关。

So these are the this is how this relates to the other puzzle pieces as well.

如果我们想在过去二十年、也就是过去二十年里推动的不是比特，而是原子，那么人类的创新全都发生在比特的世界里。

If we want to move not bits, but atoms over the last twenty years, the last two decades, human innovation has all happened in the world of bits.

相比之下，我们在原子世界里的创新其实并不多。

Like, we haven't really innovated by contrast in the world of atoms very much.

所有的信息创新都发生在屏幕背后。

Like, all of the info innovation happened behind screens.

比如，如果你想在过去二十年里寻找未来，你得去看你的手机。

Like, if you wanted to go find the future over the last twenty years, you had to go into your phone.

你得去看你的电脑。

You had to go look at your computer.

也许这是因为劳动力本身并没有太多创新。

Maybe that's because labor hasn't really been all that innovative.

能量如何作用于劳动力，我们接下来谈谈这个。

And the way that energy feeds into labor let's talk about that next.

我们来谈谈劳动力，因为有一些公司正在以与AI实验室（如Anthropic、OpenAI、Google、xAI）商品化智能相同的方式商品化劳动力。

Let's talk about labor because there are a few companies out there who are in the same way that the AI labs, Anthropic, OpenAI, Google, x AI are commoditizing intelligence.

有一些公司正在以完全相同的方式商品化劳动力。

There's a few companies out there who are commoditizing labor in the exact same way.

你能谈谈这个吗？

Can you talk about that?

是的。

Yes.

关于世界变化缓慢，你提到的这一点，我非常喜欢一个例子：一个出生于1870年、活到1920年的人，在这短短五十年间，见证了电力、电话、汽车和飞机的发明。

And to your point about the world's changing slowly, there's this great example that I love where someone who was born in 1870 and lived to 1920, just in those fifty years, over that time, saw the invention of electricity, the telephone, the automobile, the airplane.

他出生时还是马车时代，而到他50岁时，天空中已经飞满了飞机。

Like, were born and there was horse and buggies, and by the time you were 50, there was airplanes flying through the sky.

这一切都是源于那个令人难以置信的制造、工业化和创造全新事物的时代。

And that was downstream of this, like, unbelievable era of manufacturing and industrializing and building net new things.

从那以后，我们停滞了。

Since then, we've stalled.

我们是什么时候去的月球？大概五十年前吧？

We went to the moon in, like, fifth how many years ago?

但之后我们再也没有回去过。

And we haven't been able to go back.

在原子世界中，我们明显停滞了，而与此同时，比特世界却取得了进展——互联网作为一种发明和创造，彻底改变了我们的世界，但现实世界本身却几乎没有发生太大变化。

There's been this, like, very clear stalling in the world of atoms, you mentioned, in exchange for progress in the world of bits, where the Internet, as as an invention, as a creationist, changed the world as we know it, but the world itself hasn't really changed a whole lot outside.

劳动力以及劳动力成本的下降将极大地改变这一局面，我认为，除了资本之外，所有这些挑战的共同主题是正在发生通缩效应。

And labor and the decreasing cost of labor is going to change that in a huge way, and I think that's the theme between all of these gauntlets except for capital is that there is deflationary effects happening.

随着这些通缩效应的出现——劳动力成本、每千瓦电力成本、每个智能单元或每个token的成本不断下降——它释放出了巨大的丰裕潜力。

And as those deflationary effects happen, as the cost of labor, the cost of per kilowatt, the cost per piece of intelligence per token goes down, it unlocks this tremendous amount of abundance.

特别是在劳动力领域，人形机器人明确地取代了这种价值捕获方式。

And in the world of labor in particular, humanoid robots is the very clear displacement of of that value capture.

随着这些具备我们所讨论的智能的人形机器人实现规模化，它们能够彻底改变经济，以远低于人类劳动力的成本完成大量工作，从而释放出巨大的增长空间。

And with the advent of these humanoid robots at scale, with the intelligence that we spoke about, that pillar built into them, they are able to basically capitalize, just revolutionize the economy in a way that that offsets a lot of the labor that humans have for a significantly less cost, which therefore uncaps the upside.

因为一个国家的GDP实际上就是每个人产出的总和乘以人口数量。

Because the really, like, the GDP of a a country is really just like the the amount of the output of a person times the amount of people.

这就是你能创造多少产出、多少价值的方式。

And that's like how much output you're able to create, how much value you're able to create.

如果你用一个机器人来替代人，它能24小时不间断工作，从不出错，没有耐心，不会顶嘴，只是单纯地执行任务、达成目标。

And if you replace the person element with a robotic element that works twenty four seven, that works without making any mistakes, that has no patience, that doesn't talk back, that is just doing a job executing an outcome.

这从根本上释放了劳动力市场的潜力，使我们能够改变周围的世界，因为劳动力变得极其廉价。

It essentially uncaps what labor markets are capable of, and it allows you to change the world around us because labor is so cheap.

人们变得懒惰了。

And people have gotten lazy.

我们过上了好日子。

We got a good life.

我们很容易就退居幕后，刷着那些我们之前提到的、专为优化愉悦感而设计的信息流。

It's it's been very easy to sit back and scroll on those timelines that we talked about that were generated to optimize pleasure.

这是一种非常容易陷入的陷阱，我们确实陷入了其中，这很可能导致了我们在物理世界中的实际创新能力大幅下降。

And it's this really easy trap that you get caught into, which we have, and it's probably led to a significant decline in terms of our actual innovation in the physical world.

尤其是当你把智能和劳动结合起来，比如蓝领和白领工作，你是说所有人类的工作吗？

Especially when you combine intelligence and labor, like blue collar and white collar, do you mean like all of the human jobs out there?

我不想说得太悲观，我其实并不想断言所有人类都会变得毫无必要。

Not to get dystopian, I don't really think I want to care to, like, say that, like, all humans are just gonna become unnecessary.

那是另一个播客节目的话题了。

That's a different podcast episode.

这期节目讲的是，我们今年正在实际制造机器人。

What this episode is about is, like, we are literally building robots this year.

今年就会有机器人走在我们中间。

There are going to be robots that walk among us this year.

我们已经花了四年时间研究人工智能，而这场竞赛不会很快停止。

We are have already been working on intelligence for four years, and that race isn't stopping anytime soon.

如果 anything，它正在变得更快。

If it if only it's getting faster.

把这两块拼图结合起来——智能机器人，那就是完整的劳动力。

You put those two pieces, those puzzle pieces together, intelligent robots, that's the complete labor force.

那就是支撑全球GDP的支柱。

That's the atlas that holds up the world's GDP.

如果你拥有一支不需要喂食、不需要上厕所、不需要支付医疗保障的劳动力，会怎样呢？

And what if you had a labor force that was you didn't have to feed, you didn't have to have bathroom breaks, you didn't have to pay for health care.

再次忽略那些反乌托邦的方面，那是另一个话题了。

Ignoring the dystopian stuff, again, separate episode.

从根本上说，由于成本下降，企业利润所对应的GDP将飙升。

Fundamentally, the GDP that comes out for the the profits of corporations will just go through the roof because the cost will go down.

智能的成本在下降，劳动力的成本也在下降，而你拥有一支可以7x24小时、全年无休工作的机器人舰队。

Intelligence, the cost of intelligence is going down, the cost of labor is going down, and you have a fleet of robots that can work twenty four seven, three sixty five.

到那时，我们就能做到各种事情。

We can just do things at that point.

到那时，我们就能做任何事情。

We can do anything at that point.

这里有两个相互叠加的效应，而且都是指数级的。

There's two compounding effects here that are both exponential.

一个是每单位劳动力和每单位智能的成本呈指数级下降，但同时还有每单位劳动力的产出。

One is the, I mean, the exponential decrease in cost per labor per intelligence, but then also the output per unit of labor.

人形机器人相对于人类的产出要高得多，而且成本低得多，这就同时产生了双重指数增长，从而在你能够创造的价值方面带来了极其陡峭的增长。

The output per humanoid versus human is so far higher at such a lower cost that you're getting these double exponentials at once, and it creates this very vertical growth in terms of how much value you're able to create.

我们现在开始看到的，正是这一趋势在比特世界、在数字赛博空间中的初步体现，通过那些存在于你电脑屏幕背后的AI智能系统。

And what we're starting to see now is the first part of that in this world of bits, in this digital cyberspace, through these AI intelligence systems that exist on your computer behind your screen.

这些工作流程的自动化正变得越来越成熟，以至于已经开始有人被裁员了。

And the automating of these workflows is is becoming it's getting to a point where come people are getting laid off.

公司正在发生变化，因为需求在改变，这些AI在实现可验证成果方面实在太有效了。

Companies are changing because the needs are changing because these AIs are so effective at reaching verifiable outcomes.

只要你能在电脑上验证你的知识型工作的成果，你就可以先训练AI有效地完成它，最终达到完美的状态。

So long as you can verify the outcome of your knowledge work on a computer, you can train an AI to do it effectively first and then perfectly at the end state.

这种趋势会延伸到物理世界，在那里你将这两者结合起来。

That continues out to the physical world where you combine these two.

你把智能部分与物理部分融合在一起。

You merge the intelligent part with the physical part.

你给智能体赋予一个与人类能力相当的物理形态，当大规模部署这些智能体时，你就解锁了杰文斯悖论的新维度——当你拥有如此多的劳动力时，你该做什么？

You give the intelligence a physical manifestation that is as capable as a human, and when you deploy these at scale, you you create this new unlock in Jevan's paradox where you what do you do now that you have all this labor?

你可以用这些劳动力去解决各种全新的难题，因为由此带来的丰裕是巨大的：没有任何任务过于困难、繁重或智力复杂，以至于机器人无法以每小时、每分钟几美分的成本完成——而人类却做不到。

Well, you have all these cool new problems that you could set it to solve, And there's so much abundance that comes from it because there is no task that is too difficult or too laborious or too intellectually complex that a robot can't do for a few pennies per per hour, per minute, whatever the cost is gonna be, that a human otherwise wouldn't be able to.

因此，我们将从中看到的巨大突破将是前所未有的。

So the unlocks that we'll see from that are going to be huge.

网上有个梗说，美国在制造业和基础设施方面都太差了。

There's a the meme out there that like The United States just sucks at manufacturing and also like infrastructure too.

比如，不久前在推特上疯传的一个视频，展示中国工人在周末建起了一座桥。

Like you go there's that video that was rocketing around Twitter not terribly long ago of like this Chinese laborers built a bridge in, a weekend.

他们封闭了这条高速公路。

And they, like, they shut down this highway.

他们拆除了旧桥。

They demolished it.

然后建起了新桥。

They built it up.

然后他们建了桥，浇筑了混凝土，整个过程就在一个周末内完成了。

And then they they built a bridge, poured the concrete, and then it was done in a weekend.

他们可能还得让混凝土多晾一会儿，但反正也没事。

And they probably had to let the concrete dry a little bit longer, but like whatever.

然后汽车就在很短的时间内开始在桥上行驶了。

And then cars were dry driving over it in a very short amount of time.

这跟监管有关，但关键是，想象一下，如果你有一支机器人队伍，可以随便派去执行任务，会怎样？

That's that's about regulations, but nonetheless, the idea is that like, what if you had a fleet of robots who you could just set on a task to go do things?

要实现这一点，你只需要机器人、为它们供电的能源，以及让它们变聪明的智能。

And all you need in order to do that is you need the robots, you need the energy to power them, and you need the intelligence to make them smart.

目前已经有公司正在研究这三方面。

And there are companies working on all three of those things.

而且我也有资金能够资助这些项目。

And I also have the capital to be able to finance that.

所以，当我们谈论塑造世界未来时，一旦我们能真正随心所欲地做事，这些反馈循环就会变得极其强大，而有些公司已经比其他公司更接近这一结局了。

And so, like, when we talk about shaping the future of the world, these feedback loops become very, very powerful when we can start to just literally do whatever we want, and there are certain companies that are closer to this outcome than than than others.

是的

Yeah.

当然了

Big time.

我的意思是，像特斯拉这样的公司今年直接取消了Model S和Model X的生产线，转而全力生产更多人形机器人，目标是达到一百万台。

I mean, a company like Tesla is they just canceled their Model s and x production lines in exchange for creating more humanoid robots this year so they can get to a million of those.

他们把汽车换成了机器人？

They traded cars for robots?

他们确实用汽车换机器人了，而且他们会继续走这条自动化道路，因为这无疑是实现丰裕世界最盈利、最有效的方式。

They traded cars for robots, and they're going to continue to go down this automation trajectory because it is by far the most profitable and most effective towards this world of abundance.

从极限角度来看，所有劳动最终都将由某种人工智能系统和相应的机器人载体来完成，因为归根结底，这种方式的优越性实在太明显了。

And it it makes sense that at the limit, basically, all labor will be done through some sort of artificial intelligence system and some sort of robotic vessel for it because it again, it's just at the limit, it is so far superior.

一家拥有诸多高价值成果的公司，比如会计业务，如果有100名人类员工和100个AI，显然100个AI的版本会胜出。

A company with a series of very valuable outcomes, like we mentioned, say accounting for example, that has a 100 humans and a 100 AIs is obviously gonna be by a 100 AIs.

但即使是一家拥有99个AI和1个员工的公司，也会输给拥有100个AI的公司，因为那一个员工就是拖累效率的短板。

But even the company with 99 AIs and one human is going to be beat by the company that has a 100 AIs because that one human is the loss function.

它会出错。

It's it makes mistakes.

它会感到疲惫。

It gets tired.

它不确定。

It's not sure.

如果在这些过程中哪怕有一个普通人参与，就会成为劣势，因为人工智能机器和人形机器人要优越得多，或者很快就会优越得多。

If there is even a single human in the loop through a lot of these things, it becomes a disadvantage because the AI machines, the humanoid robots are so far superior, or they will soon be so far superior.

这就引出了一个问题：那我们之后会怎样？

And it begs the question is, like, what what happens to us then?

我认为这确实是一个变革时期。

And I don't think this is it it is a period of change.

这无疑是一个巨大的转型期，但其后续并不必然是悲观的结局。

It is a big transition period for sure, but the backside of it doesn't have to be the doomer take.

就像我最喜欢的一个例子，回到社交媒体动态，它们确实是真实存在的。

It's like my my favorite example going back to social media feeds is that their social media feeds are a real thing.

比如，现在有一种叫做网红的职业，你可以发表言论、环游世界、制作视频，因为人们有大量充裕的时间和资本来消费你的内容，并购买你代表其他公司推广的产品。

Like, there are things called influencers as a profession where you can say things and travel the world and make videos and you can create content because people have so much abundant time and capital to spend consuming your content and then buying into the things that you are selling on behalf of other companies.

这是我们创造出来的一种奢侈品行业，三十年前根本不可能存在，因为我们当时根本没有这些条件。

And that is this luxury industry that we have created that couldn't have existed even thirty years ago because we just didn't have that.

我们没有空闲时间。

We didn't have the free time.

我们也没有充裕的资本。

We didn't have the abundant capital.

我们所看到的许多其他事物也是如此。

And the same thing is is true for a lot of other things that we see.

比如F1，这是我喜欢的另一个例子，那个赛车联赛。

Like f one is another example that I love, the racing league.

今年早些时候我看了苹果公司拍的那部电影，这是一个规模庞大的活动，一个价值数十亿美元的产业，雇佣了数万人，每年吸引数百万观众，但它对社会的实际贡献却很小。

I watched the the movie from Apple earlier this year, and it's this huge event that is a multibillion dollar industry that employs tens of thousands of people that attracts millions of people per year, but it doesn't actually contribute much to society.

它只是很有趣而已。

Like, it's it's entertaining.

这就像体育运动。

It's like sports.

纯粹来说，很多这类事物并没有直接推动人类的进步，但它们有趣，能给人带来目标感。随着对劳动力的需求持续减少，普通人将拥有更多的空闲时间和可支配资本，可以用来享受并投入各种细分领域。

Pure and a lot of these things are they're not directly contributing to the forward progression of, I guess, humankind, but they're fun and they give people purpose, and there's going to be a lot more opportunities for that as the need for this labor continues to decrease and we have more abundance of free time and free capital as just normal people to enjoy it and spend it in different subcategories.

是的。

Yeah.

我认为，我们之前提到过一种令人恐惧的反乌托邦未来：当人工智能接管了我们所有的工作，白领工作被取代，机器人再接管蓝领工作，那么人类的精神还剩下什么？

I think, like, we've alluded to just, this dystopian future that I think people are very afraid of when like AI takes all of our jobs, all the all the white collar jobs, and then robots take all the blue collar jobs, and then and then what is left of the human spirit.

我觉得这种担忧是有道理的。

I I think that's a fine fear to have.

但与此同时，我不会被洗脑，相信当劳动和智能变得免费时，我的生活反而会变糟。

At the same time, like, I will not be psyoped into believing that when labor and intelligence are free that my life gets worse somehow.

我们可以讨论长期来看谁拥有什么，资本最终流向何处。

Like, I don't like, we can talk about the the the long term who owns what, where all the capital ends up.

但只要劳动、智能和能源都是免费的，我们就能真正免费地建造一切。

But like, if labor and intelligence and energy are all free, we can literally build stuff for free.

而且，我不确定。

And, like, I don't know.

这很奇怪，人们竟然认为这会给社会带来负面效应。

It does it it's weird to that people are thinking that that's going to, like, produce, like, a negative output to society.

而且，很难想象在一个我们拥有如此富足的世界里，你的处境会比今天更差。

And, no, it's difficult to imagine a world where we have this abundance and you aren't in a better place than you are today.

就像今天出生的人，即使是世界上最贫穷的人，也比一百年前最富有的人处境好得多，而且随着我们不断进步，这种情况还在持续。

It's like people born today, even among the poorest people born today, are in a far better position than the wealthiest people were a hundred years ago, and that continues to happen as we progress forward.

嘿，Banklist 族群。

Hey, Banklist Nation.

我是大卫。

It's David.

如果你正在听这个，那是因为你在收听

If you're hearing this, that's because you are listening to the

免费的 Banklist 播客频道。

free Banklist podcast feed.

你知道有一个高级的Bankless RSS订阅源吗？

Did you know that there is a premium bankless RSS feed?

高级订阅源包含我为个人研究所做的额外访谈，以及关于加密货币行业我想解答的更深入问题，这些问题能让我作为Bankless Ventures和我的个人投资组合中的投资者更加见多识广。

The premium feed has extra interviews that I do for my own personal research and just deeper questions that I want answered about the crypto industry, questions that I want to answer so I can be more informed as an investor both at Bankless Ventures and also just in my own personal portfolio too.

此外，没有广告，这意味着如果你选择高级订阅源而非免费版，每年能节省大约二十小时的时间，因为你选择直接支持Banklist。

Also, are no ads, which means if you listen to the premium feed instead of the free feed, you'll get about twenty hours of your life back every year because you choose to support Banklist directly.

所以如果你有兴趣在跳过广告的同时获取额外内容

So if you're interested in getting extra content all while skipping the

广告，或者你只是欣赏我们所做的

ads or you just appreciate what

如果你支持我们的工作并希望我们继续下去，我们会非常感激你注册Bankless高级会员，节目说明中有链接可以开始。

we do here and want us to keep doing it, we'd appreciate it if you signed up for bankless premium, and there is a link in the show notes to get started.

为美好的2026年干杯。

Cheers to a good 2026.

如果你能用交易加密货币的工具和速度来交易黄金、外汇和全球市场，会怎么样呢？

What if you could trade gold, forex, and global markets with the same tools and speed that you use for crypto?

这正是 Bitget TradFi 所实现的功能。

That's exactly what Bitget TradFi unlocks.

在经历强劲的测试期需求后，包括单日黄金交易量超过一亿美元，Bitget TradFi 现已向所有用户开放。

After strong beta demand, including over a $100,000,000 in single day gold trading volume, Bitget TradFi is now live for all users.

在你现有的 Bitget 账户内，你可以交易涵盖外汇、贵金属、指数和商品的 79 种金融工具，所有交易均以 USDT 直接结算。

Inside of your existing Bitget account, you can trade 79 instruments across forex, precious metals, indices, and commodities, all settled directly in USDT.

无需切换平台，也无需法币兑换。

No platform switching and no fiat conversions.

这就是 Bitget 普适性交易所愿景的体现。

This is Bitget's universal exchange vision in action.

加密货币与传统金融并肩共存。

Crypto and traditional finance side by side.

你将获得深厚的流动性、低滑点，以及最高达 500 倍的杠杆，让你能够将加密策略应用于宏观市场。

You get deep liquidity, low slippage, and leverage up to 500 x, letting you apply crypto strategies to macro markets.

刚接触传统金融？

New to TradFi?

从黄金开始。

Start with gold.

黄金兑美元货币对流动性强，受宏观因素驱动，是加密货币与传统市场之间的天然桥梁。

The gold USD pair is liquid, macro driven, and a familiar natural bridge between crypto and traditional markets.

立即在 bitget.com 上交易黄金。

Try trading gold on Bitget now at bitget.com.

点击节目说明中的链接获取更多信息。

Click the link in the show notes for more information.

这不是财务建议。

This is not financial advice.

好的，乔什。

Okay, Josh.

这就是那四个关键要素。

So those are the four puzzle pieces.

是的。

Mhmm.

智能、能量、资本、劳动力。

Intelligence, energy, capital, labor.

这里的理念是我的个人投资理念：哪家公司能同时拥有这四者，就能赢得比赛。

And the idea here, this my my investment thesis for my own financial portfolio is whichever company gets all four just wins the game.

这很可能在我们这一代人的时间里发生。

And this is gonna happen probably in our lifetime.

我们很可能解决这些问题。

Like, we're probably going to solve these things.

我们会解决能源问题。

We're gonna solve energy.

我们会解决智能问题。

We're gonna solve intelligence.

我们会解决劳动力问题，然后资本就会变得无关紧要。

We're solve labor, and then and then that makes capital irrelevant.

所以我把这称为资本主义的事件视界。

And so I'm calling this like the event horizon of capitalism.

这就像是当一家公司在这四个方面的发展达到一个临界点，所有增长都变得自我强化且结构上不可逆转。

It's just like when a company reaches a point where its growth on all four of these things just becomes like self reinforcing and just structurally irreversible.

它们赢得了垄断游戏。

They win the monopoly board.

我想和你聊聊一些公司，因为我觉得有些公司已经比其他公司更接近这个状态了。

And there are some companies I wanna talk to you about because I think there are some that are closer than others.

我们谈过谷歌，但也许我们再重新梳理一遍。

We talked about Google, but maybe let's just trace over it again.

在我看来，谷歌拥有智能和资本。

Google has, in my mind, intelligence and capital.

它拥有谷歌 Gemini，并且能够获取数据——这一点非常关键，而其他任何人工智能实验室都不一定具备。

It has Google Gemini, and it's being able to and it also has the data, very important, which is not necessarily what any other AI lab has.

它拥有的数据量和对互联网每个人的连接度，远超任何其他人工智能实验室。

It has far more data and connectivity to everyone on the Internet than any other AI labs.

因此，谷歌在智能方面拥有其他人工智能实验室所不具备的优势。

So it has intelligence in a way that other AI labs don't.

而且它还拥有资本。

And it also has the capital.

所以在我看来，谷歌在四个灭霸宝石中的两个方面占据了非常有利的位置。

So Google to me feels very well positioned to have a very strong grasp on two of the four Thanos jewels.

你怎么看？

What do you think about that?

是的。

Yes.

而且在某种程度上，它们在能源方面也占据优势，因为虽然它们不能生产能源，但它们有非常高效的方式从能源中提取智能。

And in a way, they're they're well positioned for energy too because while they cannot generate the energy, they have very efficient ways of extracting intelligence from it.

这主要归功于它们所研发的硬件芯片。

And that's mostly due to the hardware chips that they've built.

它们拥有这些芯片，被称为TPU，即张量处理单元。

The they have these, they're called TPUs or tensor processing units.

而谷歌的另一个优势，回到前面提到的优势，就是在那里工作的那些人——那些实际的员工和研究人员——所具备的卓越智慧。

And one of the advantages, going back to the advantages of Google, is the just sheer intelligence of the humans that work there, of the actual people and and researchers that have worked there.

而且，他们不仅发明了Transformer，还发明了TPU，意识到并理解到AI的训练将使用与传统GPU擅长的完全不同的一套数学方法。

And, again, they invented the transformer, but they also invented the this TPU, realizing and understanding that AIs were going to be trained using a very different set of math than traditional GPUs are good at.

所以当NVIDIA制造GPU时，它是一个通用处理器。

So when NVIDIA makes a GPU, it's a a general purpose processing unit.

同一个GPU既可以用来玩视频游戏，也可以用来训练AI。

It's it's the same GPU that can play video games that is training AI.

而谷歌的TPU则高效得多，因为它专长于训练这些模型所需的特定类型数学运算。

With a Google TPU, it is far more efficient because it is good at the specific type of math required to train these models.

因此，谷歌拥有这种TPU，能够比传统公司以更低的功耗获得更多的计算能力。

So Google has this TPU, and they're able to get a lot more compute per watt out of their AI models than their traditional companies.

所以，虽然他们并不实际产生能源，但他们能更高效地利用能源，通过降低每令牌的能耗来降低成本。

So while they aren't actually generating the energy, they are far more efficient at extracting it and decreasing that cost per token through energy.

所以我认为他们在各个方面都独具优势。

So I think they're they're uniquely equipped across the board.

我们谈到了他们所拥有的数据垄断。

We talked about the the data monopoly that they have.

我们谈到了他们在资本效率上的优势，以及能够推动这一进程的卓越人才。

We talked about the capital efficiency that they have, and just the the raw human intelligence that's capable of pushing this forward.

但这种情况直到2024年年中之前都不存在。

And this hasn't been the case up until maybe mid twenty twenty four.

谷歌实际上在将这项技术产品化方面遇到了巨大困难。

Google was actually having a really difficult time productizing this.

尽管他们拥有所有必要的组件，却无法真正推出一款优秀的消费者产品。

And while they had all of the pieces, they couldn't actually generate a a good consumer product.

拉里和谢尔盖重返公司，尤其是谢尔盖，他成功扭转了局面。

Larry and Sergei came back to the company, particularly Sergei, and he kind of turned the ship around.

现在，他们的产品发布速度令人难以置信地惊人。

And now they are shipping at a velocity that's that's unbelievably impressive.

他们所拥有的资本带来的一个衍生优势，是能够为研究人员提供时间和空间，以实现这些突破性的创新，谷歌DeepMind就是一个绝佳的例子——谷歌早在很久以前就收购了DeepMind，并允许它独立运作，与谷歌其他部门几乎没有任何关联。

And a downstream effect of the capital that they have is is their ability to give people time and space to do research, to unlock these novel breakthroughs, and Google DeepMind is a really great example of that because Google acquired DeepMind a very long time ago and let them just exist in this weird place that was not really connected with Google.

他们只是说，嘿。

They were just like, hey.

他们保护了他们。

They protected them.

他们保护他们免受扎克伯格试图收购并将其封存为资本化产品的影响，或其他任何人的企图；据我理解，DeepMind的人只是说：伙计，我就想研究人工智能。

They protected them from, like, Zuck trying to buy them out and just shelve them in a capitalistic package or anyone else trying they just like to as I understand it, the DeepMind people were just like, dude, I just want to, like, research AI.

就想做研究。

Just wanna research.

谷歌说：太好了。

And Google's like, great.

这是钱。

Here's money.

别让任何人收购你们。

Don't let anyone else buy you.

这是用来做研究的钱。

Here's money to go do research.

然后他们为自己争取了时间。

And then they just bought themselves time.

当他们需要将DeepMind的研究成果从货架上取下并整合到Google的技术栈中时，他们因为拥有资本而具备了这种选择权。

And when it came time for them to take the DeepMind research off the shelf and put it into the Google stack, they had that optionality available to them because they had capital.

是的。

Yeah.

他们买了一个选择权。

They they bought an option.

因为他们有足够强的资产负债表，能够承担这笔资本投入，等待数年，直到德米斯和DeepMind团队如今掌管Google的AI部门，并主导了Google全部的AI业务，逐渐取代了原有的内部Google Brain团队。

And because they had the balance sheet to do that, they could afford to put that much capital on out on loan for however many years it took until Demis and the DeepMind team now run Google's AI intelligence division, and it runs, basically, all of Google's AI, kind of eating the in house Google Brain team that was existing.

因此，他们拥有一个难得的全方位机会来实现这一目标，而且这是很长时间以来首次真正付诸行动——他们正在发布产品、研发硬件，并掌控着在规模化竞争中获胜所必需的全栈垂直整合体系。

So they have this this really unique opportunity across the board to do it, and for the first time in a long time, they're they're acting on it, and they're releasing the products, they're building the hardware, and they're controlling this full vertically integrated stack that is required to win at scale.

很难想象Google不会成为首批跨越这一事件视界门槛的公司之一。

And it it's very difficult to make a case that Google won't be, if not the first among the very first to cross this event horizon threshold.

根本没有人比他们更适合。

There's just no one better equipped.

是的。

Yeah.

那英伟达呢？

What about NVIDIA?

你会把英伟达归到哪一类？

Where would you classify NVIDIA?

因为如果我们把数据也算在内的话，比如谷歌在智能领域几乎垄断了数据，你或许也可以把英伟达归入这一类，因为它在芯片生产上也拥有某种程度的垄断地位，而要实现智能，你确实需要芯片。

Because I suppose if we're putting data, like if Google has like an almost monopoly on data in the intelligence category, I suppose you could also put NVIDIA there because it has like a somewhat monopoly on chip production, and you need you need chips to have intelligence.

要想获得智能，你需要能源、数据、芯片，还需要模型本身。

Like, in order to get intelligence, you need energy, you need data, you need chips, and you need the model itself.

对吧？

Right?

所以也许英伟达也应该属于智能领域？

And so maybe maybe NVIDIA belongs in the intelligence category?

英伟达当然属于智能领域，但要捍卫它的护城河会很困难，而且要达到像谷歌那样的规模增长也会很困难。

NVIDIA certainly belongs in the intelligence category, but it's going to be difficult to defend that moat, and it's gonna be difficult to grow at a scale that a company like Google will.

因为，再说一遍，关于TPU，谷歌有自己的AI训练芯片，拥有自己的芯片对掌控整个技术栈非常有利。

Because, I mean, again, to the point about the TPUs, Google has their own AI training chips, and it is very advantageous to have your own because you own the whole stack.

所以他们可以从中榨取额外的利润。

So they can squeeze out extra margins.

他们可以像针对自家芯片那样优化软件，而这是无法用英伟达芯片做到的。

They can optimize the software for it in a way that you can't with NVIDIA chips.

英伟达是一家规模庞大、增长潜力巨大的公司，但所有市场力量都在与之对抗。

Now NVIDIA is a huge company with a tremendous amount of upside, but all of the market forces are working against it.

像苹果这样拥有M系列芯片的公司已经证明，你可以制造出性能更好的芯片。

Companies like Apple with m series chips that they itch they prove that you could build a chip that works better.

谷歌正在研发自己的芯片。

Google is building their chips.

特斯拉即将建设一座TeraFab工厂。

Tesla is about to build a TeraFab.

许多公司都在尝试自主研发芯片。

A lot of companies are are trying to in house their chips.

就连OpenAI和Anthropic也在与其他公司合作，试图开发自己的芯片，因为拥有整个技术栈具有巨大优势。

Even OpenAI and Anthropic are working with other companies to try to develop their own because they're such an advantage of owning the entire stack.

而对于那些无力承担此成本、没有自有晶圆厂或无法自行制造这些芯片的少数公司，英伟达将会为它们提供支持。

And for the few companies that can't afford to do this, that don't have their own custom fabs or that that can't actually build these things, NVIDIA is going to be there for them.

英伟达将成为训练或推理计算的唯一供应商，这是一个巨大的市场。

And NVIDIA will be the sole supplier of, if not training, then inference compute, and that's a huge market.

但它不像谷歌那样拥有未被发掘且无上限的增长潜力。

But it's not this untapped, uncapped upside that a company like Google can actually reach.

是的。

Yeah.

所以我认为，无论哪家公司率先跨越这个资本主义的事件视界，它们都将自行研发芯片。

So I think if we're saying whichever company wins this race to get over this capitalistic event horizon, they are going to be building their own chips.

嗯。

Mhmm.

可能性最大。

It's most likely.

而且因为它们能够大规模地做到这一点。

Also because they could do that at scale.

而垂直整合和规模效应，最终会带来胜利。

And the vertical integration, the scale, it it all leads to a successful winning game.

对。

Right.

好的。

Okay.

我们来谈谈特斯拉。

Let's talk about Tesla.

你认为特斯拉在这场竞赛中处于什么位置？

Where do you see Tesla in this race?

好的。

Okay.

特斯拉。

Tesla.

有人说我有偏见，但在我看来，特斯拉显然是赢家

I call me biased, but Tesla is the clear winner to me in terms

对于听众来说，乔什是个狂热的特斯拉支持者。

For for the listener, Josh is a huge Tesla moon boy.

但这是有道理的。

But rightfully so.

我确实听过。

And I have listen.

这并不是这样，我可以解释为什么。

This is not this is this is like I have I can explain why.

特斯拉擅长其他公司都不具备的一件事，那就是以大规模生产全新的产品。

And Tesla is good at something that no other company is, and that's that's manufacturing at scale with net new products.

如果你看看谷歌或者苹果这样的公司，它们可能是全球生产消费类产品最多的公司之一，比如iPhone、Pixel和安卓手机。

If you look at a company like Google or you look at a company like Apple, who probably creates some of the most consumer products of any company in the world through these handsets, like the iPhones and the the Pixels and the Androids.

它们非常擅长制造小巧而具体的产品。

They're very good at building small and specific things.

比如，谷歌成功转型生产了Pixel Buds这样的产品，你得到的是耳机这类小型消费电子产品。

Like, Google was able to pivot to building these Pixel Buds where you get earphones, and you you get these small consumer product devices.

但要大规模制造复杂大型的产品是一个极其困难的问题。

But to manufacture large complicated things at scale is a tremendously difficult problem.

而且为了

And in order to

抵消这种人力成本，你必须这样做。

offset that labor pillar, you must do this.

你必须创造某种机器人，某种这种智能的物理体现。

You must create some sort of robot, some sort of physical embodiment of this intelligence.

我想你刚才描述的差异，也就是谷歌硬件和苹果硬件之间的区别，在于它们是比特导向的硬件。

I suppose the difference that you just described, the difference between Google's hardware and Apple's hardware, is it's bit oriented hardware.

它们是互联网导向的硬件，而特斯拉是原子导向的硬件。

It's Internet oriented hardware, where Tesla is atom oriented hardware.

我认为这就是你所强调的重点。

I think that's kind of the emphasis you're making.

没错。

That's right.

是的。

Yeah.

这么说很到位，特斯拉为现实世界的基础设施和生产力创造产品。

It's a great way to put it is, yeah, Tesla creates products for real world infrastructure and productivity.

自动驾驶出租车将消除人们每天因驾驶和交通拥堵而损失的大量生产力。

It's like the the cyber cab is going to displace all of the lost productivity that people experience every day through driving and traffic.

这是一次巨大的突破。

That's a huge unlock.

不仅如此

And not only that

特斯拉网络的自动驾驶出租车将取代优步、Lyft等所有网约车服务，因为特斯拉自己生产汽车，而交通成本几乎等于特斯拉的成本——几乎是免费的，因为他们拥有电池和能源等资源。

for The cyber the cyber cab Tesla network that's going to displace Uber, Lyft, all ride sharing maybe because Tesla just produces its own cars, and now transportation is essentially at the cost of Tesla, which is free because they have batteries and energy and stuff.

是这样吗

Is that

你就是这个意思吗？

kinda what you're talking about?

是的。

Yeah.

但，为什么就此止步呢？

But, like, why stop there?

为什么不考虑UPS、联邦快递和邮政服务呢？

Why not UPS and FedEx and the postal service?

很多东西都需要运输。

Like, many things get transferred.

有太多东西在移动了。

So many things move.

有无数的人、箱子、包裹和信息通过物理世界被运送。

Like, so many people and boxes and packages and information gets moved through the physical world.

如果你能把每公里的成本降低到远低于人类所能提供的价格，那么其他任何产品和服务在经济上就都变得不可行了。

And if you decrease the cost per kilometer to a price that is so far lower than any price that a human could do it for, it becomes economically unviable to to go with any other product and service.

没有哪家公司比特斯拉更适合做这件事了，因为他们已经证明了自己能够大规模部署这些系统，并且他们极度优化了每公里成本的降低，因为这能让你取代整个交通运输行业。

And there's no company better suited to do this because they have proven that they could roll these things out at scale than Tesla, and they're hyper optimized on getting that cost per kilometer down because it allows you to displace transportation.

未来，你根本不需要再开车，甚至今天出生的人也永远不会需要拥有汽车，因为使用这项服务的成本如此之低，拥有汽车在经济上变得毫无意义。

Like, there you will not need to ever drive a car in the future, but also someone born today will never need to own a car because it'll be economically inefficient because the cost to use this service is so low.

这还只是就交通而言。

And that's just in terms of transportation.

而今年，他们正转向人形机器人领域。

And then the humanoid element is what they're pivoting to this year.

他们将从一百万台开始，然后逐步提升到一千万、一亿，乃至十亿台。

They're gonna start with a million, and then they're gonna ramp up to 10,000,000, a 100,000,000, a billion.

他们的理念是让每个人拥有多个类人机器人，因为它们将如此便宜、如此普及、制造起来又如此简单。

And the idea is to have multiple humanoid robots per human because they will be so cheap, so abundant, and so easy to manufacture.

最终，世界将进入一个自我复制的机器时代——机器制造机器，而人类劳动则被大规模的人形机器人和自动驾驶出租车彻底取代。

And there eventually reaches a world where there's this, like, self propagating machine where the machine builds the machine, but the the labor thing gets offset so much by these humanoid robots, by these cybercabs at scale.

地球上没有任何其他公司能像特斯拉一样，证明自己有能力制造出如此深刻影响现实世界的复杂产品，或许只有中国的一些公司能与之相比，这似乎是一个真正的威胁。

And there's no other company on Earth that has proven they can build complex products like this that impact the real world other than Tesla and perhaps some companies out of China, which seems like a real threat.

但在美国公司中，就我们今天能够投资的企业而言，特斯拉无疑是这场实体制造竞赛中的绝对赢家。

But Tesla, in terms of just American companies that we can invest in today, they are by far the winner in this physical manufacturing race.

所以我们谈到了为什么谷歌相比Anthropic或OpenAI具有优势，是因为他们拥有资本。

So we talked about why Google has the advantage over Anthropic or OpenAI is because they have capital.

他们可以像谷歌一样跌倒后还能恢复。

They can scale like Google can stumble and recover.

但如果OpenAI跌倒了，比如他们投资了十亿美元去打造一个没人用的社交网络，这对他们来说几乎是灾难性的，因为他们已经在这上面投入了太多。

But if OpenAI stumbles, like if it invests a billion dollars into building a social network that no one uses, that's almost like catastrophic for them because they just invested so much into that.

我认为我们可以把同样的思路延伸到人形机器人上。

I think we can extend that same idea to humanoid robots.

市场上还有其他机器人初创公司，你知道，特斯拉是其中之一，但特斯拉并不是一家初创公司。

There are other robot startups out there and, you know, Tesla is one of them, but Tesla's not a startup.

特斯拉是一家市值数千亿美元的上市公司。

Tesla has a publicly traded multi trillion dollar company.

所以那些机器人初创公司，我不知道你能不能说出几个，可能会因为跌倒而失败，因为他们走在刀锋上，因为他们是初创公司，但特斯拉并不是这样的。

And so the robot startups, I don't know, maybe you could you could name a few, might stumble and fail because they are walking on a razor edge edge because they are a startup, but that's not what Tesla is.

特斯拉是一家盈利的公司。

Tesla is a profitable company.

对吧？

Right?

是的。

Yeah.

所以从某种意义上说，Figure就像是Rivian之于特斯拉。

So a figure in a way is kind of like the Rivian to Tesla.

当谈到特斯拉时，Figure就像是它的替代品。

It's like the Figure is the the alternative to Tesla when it comes to Oh, Tesla.

是的。

Yeah.

就像是，嗯。

It's like the yeah.

没错。

Exactly.

所以，就像Rivian之于汽车，Figure之于人形机器人，也就是Optimus项目。

So, like, what Rivian is to the cars, Figure is to the humanoids, the Optimus program.

Figure 的问题在于，首先，他们在实际的自主性方面面临困难，因为他们没有像特斯拉那样数千万英里的驾驶数据；其次，他们实际上不具备大规模制造的能力。

And the problem with Figure is that, well, one, they are having a difficult time with the actual autonomy because they don't have the data set of tens of millions of these miles driven with these cars, and then two, they don't actually have the manufacturing capabilities at scale to do this.

这个方程式中被低估的一部分其实是创始人——埃隆，以及他所带领的团队。

And an underrated part of this equation is actually the founder, is Elon, is the team that is building this.

世界上没有人能像他一样，在制造领域完成如此多看似不可能的任务并实现规模化成功。

There is no one in the world who's who's accomplished so many impossible missions and succeeded at scale when it comes to manufacturing.

这确实是一个极其艰巨的挑战。

And it's just it's remarkably difficult challenge to do.

因此，Figure 不仅在资本、制造能力以及真正的原始智能方面处于劣势，而且在经验上也处于劣势。

So not only is figure out a disadvantage when it comes to capital, when it comes to manufacturing capability, when it comes to actual, like, raw intelligence, but they're also at a disadvantage of experience.

他们以前从未做过类似的事情，而特斯拉在2017年和2018年生产Model 3生产线时，曾多次濒临破产，仅差几周就撑不下去了。

They've never done anything like this before, and there were many moments in time in which Tesla was weeks away from bankruptcy in 2017, 2018 when it came to building the model through production line.

要制造出足以将每台人形机器人成本降至所需水平的规模化人形机器人，这将是一项极其艰难的 uphill 任务——一旦你犯错，比如供应链中任何一个环节出现断货，比如来自台湾某第三方公司当天未能发货的一个电机，整条生产线就会停滞。

And for someone to build humanoid robots at the scale required to get the cost per humanoid down to a required level, it's going to be this unbelievably difficult uphill task where if you do make a mistake, if if you have a single glut in your supply chain that prevents you from getting a single motor that from this third party company in Taiwan that's not shipping that day, the entire production line stops.

特斯拉意识到了这一点，因此在美国建立了首个电池正极工厂，还建立了前沿的精炼厂，因为他们意识到供应链，尤其是国际供应链，所带来的生存威胁。

And Tesla's realized this, and they've created a a battery a cathode plant in The United States, the first one, and they've created a refinery in The United States, the cutting edge of this because they realized the existential threat that is supply chains, particularly international.

像Figure这样的公司根本没有足够的资源和资本来如此广泛地布局，并确保拥有大规模成功制造所需的冗余能力。

And a company like Figure just doesn't have the resources and capital to spread their wings out that far and actually make sure that they have the redundancies needed to build this at scale successfully.

但这并不意味着Figure不会成为一个极其成功的初创公司。

Now that's not to say that, like, Figure won't be an incredibly successful startup.

OpenAI也是如此。

Same thing with OpenAI.

对吧？

Right?

我的意思是，据说OpenAI有望以一万亿美元的估值上市，这可能有点超出了我的承受范围。

I mean, granted OpenAI is like rumored to go public at a trillion dollars and so that might be a little bit rich for my blood.

但以Figure为例，它在2025年9月的最后一次估值是390亿美元，而特斯拉的市值是多少呢？

But like Figure for example, its last valuation in September 2025, $39,000,000,000 compared to, like, what was what's the Tesla market cap?

今天是1.3万亿美元。

1,300,000,000,000.0 today.

1.3万亿美元。

1,300,000,000,000.0.

但我们讨论的这个想法，就像是在争夺灭霸的无限手套。

But so, like, the the idea that we're talking about is, like, this race to fill out the Thanos gauntlet.

嗯。

Mhmm.

但你们也必须明白，听众也得理解，市场会给事物定价，这些公司的估值正是基于它们在资本主义的灭霸手套竞赛中的进展。

But you also have to we also have to understand listeners also have to understand that the market prices things and these companies are being priced according to their race towards the Thanos gauntlet of capitalism.

所以并不是说Figure是个糟糕的投资，只是Figure离完成它在苏格兰的征程还差得太远，根本不在考虑范围内。

And so it's not to say that, like, figure's a bad investment, but it's just, like, Figure is is so far away from finishing its fit down in Scotland that's not even in in consideration.

我们谈论的是那些试图垄断资本主义方方面面的公司。

Like, we're talking about these companies that are, like, working on monopolizing basically everything about capital ism.

所以我想顺便提一下这个推论。

So I kinda wanna just throw that corollary in there.

是的。

Yeah.

没错。

That's correct.

这绝不是在贬低那些做出卓越工作的公司。

This is in no way degrading towards the companies who are doing great work.

Figure是一家了不起的公司，正在打造非常困难的技术并解决复杂的挑战。

Figure is an amazing company, building very difficult things and solving complex challenges.

只是存在不同层次，而特斯拉在这场通往资本主义终极目标的游戏中处于不同的层次。

It's just there are levels, and Tesla is just at a different level in this game of reaching that capitalistic end game goal.

资本主义的终极目标。

The end game of capitalism.

是的。

Yeah.

是的。

Yeah.

我们来谈谈SpaceX。

Let's talk about SpaceX.

好的。

Okay.

哦，是的。

Oh, yeah.

今天确实发生了一件有趣的事，就在我们开始录音的时候。

Interesting happened today, actually, right, as we started recording.

有一个传言流传出来了。

There's a a rumor that went out.

有传言称埃隆·马斯克的SpaceX正在与xAI洽谈合并事宜，而这一消息在他们上市前显得尤为相关，与我们今天的讨论息息相关。

To Elon Musk's SpaceX is in talks to merge with x AI before their IPO, which is inherent it did incredibly relevant to this conversation.

是的。

Yeah.

当我们谈论这些公司在这场竞赛中所玩的高风险游戏时，你觉得SpaceX处于什么位置？

So when we when we talk about the, the game of risk that is being played between some of these companies very far along on this race, where where do you see SpaceX?

讽刺的是，就在昨天特斯拉的财报中，他们对xAI进行了大笔投资。

And ironically, just in Tesla's earnings report yesterday, they made a large investment in x AI.

所以这两家公司之间即将出现某种融合

So there is going to be a convergence happening between the two of

这些公司。

these companies.

等等。

Wait.

等等。

Wait.

好的。

Okay.

所以等等。

So wait.

稍等。

Hold on.

今天出了个传闻，我刚提到过，就是SpaceX正在与XAI就IPO前的合并进行谈判，而你却说特斯拉买入了大量XAI的股票？

There's this rumor that came out today that I just said, SpaceX in talks to merge with XAI before IPO, and you're just saying that Tesla has bought a bunch of XAI shares?

这是本周最新财报的情况。

It's just no last earnings report this week.

是的。

Yes.

好的。

Okay.

所以这里已经出现了三方融合的趋势。

So there's already a three way convergence happening here.

这种趋势正在缓慢发生，而且非常合理。

It's slowly starting to happen and it makes a lot of sense.

如果你看看每家公司擅长什么，XAI 在智能方面极为出色，特斯拉在制造方面极为出色，而 SpaceX 在近地轨道及更远领域拥有垄断地位。

So if you look at what each company is good at, XAI is exceptional at intelligence, Tesla is exceptional at manufacturing, and SpaceX owns a monopoly on low Earth orbit and beyond.

SpaceX 大概处于第二阶段，可能还没完全到位，也许吧。

SpaceX is, like, about two, maybe hasn't quite got actually, maybe.

它的垄断进展到什么程度了？

How far along is it on its monopoly?

所以它确实已经拥有了。

So it it has.

猎鹰重型和猎鹰9号项目取得了成功。

The Falcon Heavy and the Falcon nine program is successful.

它们正在为私人客户发射卫星。

They're launching satellites on behalf of private customers.

它们正在将星链卫星发射到外太空。

They're launching the Starlink satellites into outer space.

已经形成了一个成功的星座。

There is a successful constellation.

现在几乎每天都有发射。

There is launches almost every day now.

我觉得它们发射了超过500颗。

I think they launch, like, over 500.

有大量载荷被送入地球轨道。

It's just like an outrageous amount of payload going into outer orbit.

但问题是，仅靠这些还远远不够，无法达到我们把大量智能设备送入外太空所需的成本效益规模。

But the problem is it's just it's not enough to make it cost effective at the scale we need to move a lot of this intelligence into outer space.

而这一切都得益于尚未完全投入运行的星舰计划。

And that's enabled through the Starship program, which isn't fully up and running.

它仍然在开发中。

It's still a work in progress.

星舰仍在进行试射。

The Starships are still going through test launch.

它们仍然会爆炸，但已经越来越接近成功了。

They're still exploding, but they're getting close.

预计到今年年底，星舰计划将投入运行，并能够运送大量有效载荷。

And the the expectation is that by the end of this year, the Starship program will be working, and it will be able to send a lot of this payload.

其中一部分将是人工智能数据训练中心，部署到轨道上，尤其是低地球轨道——目前有传闻称苹果正与SpaceX合作，让全球所有手机都能接入星链卫星。

Some of it will be AI data training center into orbit and particularly into low Earth orbit where now we have Starlink that Apple is rumored to be partnering with SpaceX where all the cell phones on Earth get access to Starlink space satellites.

因此，我一直以来梦想的完美手机就是：电池无限、连接无限，无论你身处世界哪个角落，都不会有信号盲区，而SpaceX能让全世界每个人都能大规模实现这一点。

So the perfect handset is one that I always dreamed of, which has unlimited battery and unlimited connection where no matter where you go in the world, you have no dead zones, And SpaceX enables that to happen at scale for everyone in the world.

它通过低地球轨道和外层空间中这个近乎国家规模的网络，将所有人连接起来。

It brings everyone online through this basically nation state resistance network that exists out in lower earth orbit and outer space.

好的。

Okay.

那么太空中的能源问题呢？

What about the energy equation a part of space?

因为你提到了AI数据中心，这与能源息息相关。

Because I you you mentioned AI data centers because this is relevant to energy here.

是的。

Yes.

这三者的融合，我刚才没说完：AI或XAI负责智能运行，特斯拉制造芯片，SpaceX将它们发射到外太空。

And the convergence between the three, which I didn't finish my thought, which was AI or XAI runs the intelligence, Tesla makes the chips, SpaceX launches them into outer space.

因此，用于训练AGI的芯片将是特斯拉的芯片，而XAI将成为整个系统的协调者。

So the chips that are going to be trained to build AGI are going to be Tesla chips, and the x a XAI is going to be the orchestrator of that.

嗯。

Mhmm.

将AI数据中心部署在太空，这是一个新的创意。

The AI data center in space thing is a new novelty.

这是一件最近几个月才出现的新事物，因为只有在星舰计划有望成功后，它才成为可能。

It's this is a a new thing that has kind of come online in the last few months because it's only recently been made possible through the Starship program hopefully working.

其理念是，你需要在太空中拥有足够的能源，才能使在太空中训练数据中心的成本低于在地面运行的成本，从而更具吸引力。

And the idea is that you need a certain amount of energy in space in order to offset the cost enough so that it's more lucrative to to train data centers in space than it is on the ground.

是的。

Mhmm.

当星链实现规模化后，如果每年发射约一万枚这样的火箭，每公斤的运输成本将低于将AI和AI数据中心送入轨道的成本，甚至低于在地面运行它们的成本。

At with Starlink at scale, if they launch about 10,000 of these rockets per year, the cost per kilogram gets lower than it does to send AI space and AI data centers into orbit than it does to run them on the ground.

这解决了许多问题。

And it solves a lot of problems.

其中一个就是散热问题。

One of them is the cooling issue.

因为在地面上，你不仅要应对审批问题，还要应对能源问题，同时还必须处理散热问题。

Because when you're on the ground, not only do have to deal with the permitting and you have to deal with the energy, but you also have to deal with the cooling.

需要大量的液冷系统。

There's a tremendous amount of liquid cooling that goes on.

有冷却器和风扇，占用了这些数据中心约20%到25%的空间。

There's coolers, there's fans that takes up about 20 to 25% of the footprint of these data centers.

事实是，太空是真空的。

Well, it turns out space is a vacuum.

在没有阳光的情况下，只要在这些芯片上方撑一把小伞，散热就相当容易。

And in the absence of sun, if you just put a little umbrella over these chips, it's fairly easy to cool.

你只需要把热量辐射到太空中去。

You just radiate the heat off into space.

维护其实根本不是问题，因为没有活动部件。

And the maintenance doesn't really it's not really a factor because there's not there's no moving parts.

没有大量水流经其中。

There's not a lot of water running through

温度波动是芯片退化的首要因素。

the Temperature fluctuation is the number one degrading force on chips.

是的。

Yes.

当然。

Absolutely.

如果温度太高或太低，它们的性能就会下降。

If it's too hot or if it's too cold, they don't work as well.

在太空中，你只需遮挡阳光，就能很容易地控制温度。

Space, you can control that very easily just by blocking the sun.

能源问题可以通过太阳同步轨道来解决，这基本上意味着当你将卫星送入太空时，使其轨道与太阳保持同步。

The energy problem gets solved through things like sun synchronous orbit, which basically means when you launch a satellite into space, you match the orbit of the sun.

因此，它始终持续处于阳光照射之下。

So it's always perpetually right in the sun's rays.

没错。

Right.

你有一个太阳能电池板，全天候、全年无休地朝向太阳。

You have a solar panel that is all with twenty four seven, three sixty five pointed at the sun.

正是如此。

Exactly.

而在外太空，它的效率大约是地球上的七到八倍，因为去除了大气层的影响。

And when it is outer space, it is I think it's seven to eight times more efficient than it is on Earth because you remove the atmospheric part of the equation.

嗯。

Mhmm.

嗯。

Mhmm.

所以能源方面

So energy

这样更高效。

That's more efficient.

这意味着太阳在太空中比在地球上强七到八倍吗？

Does that mean that the sun is seven to eight times more intense in the in space?

我想是的。

I think that's what Yes.

那

That