贝索斯计划为人工智能筹集1000亿美元，而英伟达面临挑战

欢迎收听本播客。

Welcome to the podcast.

我是您的主持人杰登·谢弗。

I'm your host, Jaeden Schafer.

今天节目中，我们来讨论TechCrunch提出的一个有趣问题：AI代币是新的签约奖金，还只是做生意的正常成本？

Today on the show, there is a hilarious question that TechCrunch is asking, which is, are AI tokens the new signing bonus or just a cost of doing business?

随着公司向工程师发放代币配额，这方面出现了许多有趣的现象。

As companies are giving token allocations to engineers, there's a lot of funny stuff that's going on with that.

我们还将讨论为什么华尔街并未被英伟达最近举办的大型发布会真正打动。

We're also gonna be talking about why Wall Street wasn't really won over by Nvidia's big conference that they recently did.

我们将带您独家参观亚马逊的Tranium实验室。

There is an exclusive tour of Amazon's Tranium Lab.

这款芯片赢得了Anthropic、OpenAI甚至苹果公司的青睐。

This is the chip that won over Anthropic, OpenAI, and even Apple.

我们想聊聊这里面究竟发生了什么。

We wanna talk about what's going on inside of that.

而且，wordpress.com 现在允许 AI 代理撰写和发布文章等，老实说，我至今仍在使用 WordPress。

And wordpress.com is now letting AI agents write and publish posts and more, which honestly, I actually use WordPress to this day.

据报道，杰夫·贝佐斯正在寻求筹集一千亿美元，用于收购并用 AI 改造传统制造企业。

Jeff Bezos is also reportedly looking for a $100,000,000,000 to buy and transform old manufacturing firms with AI.

今天我们将在播客中深入探讨所有这些内容。

We're gonna get into all of that on the podcast today.

但在开始之前，我想提一下，AI Box 已经在我们的平台上的游乐场和构建器功能中推出了视频功能。

But before we do, I wanted to mention that AI Box has just launched video on our platform for both of our playground and our builder.

所以，如果你想打造包含视频自动化功能的酷炫工具，无论是使用 OpenAI 的 Sora、Google 的 VO3，还是其他众多出色的视频生成模型。

So if you wanna build really cool tools that have automations that include video, whether that's OpenAI Sora, Google VO three, or a ton of other cool video generation models.

快去体验 aibox.ai 平台吧，这是我自己创办的初创公司。

Go check out the aibox.ai platform, my very own startup.

每月仅需 8.99 美元即可开始使用。

It's only $8.99 a month to get started.

你可以生成视频、图像、音频、文本以及大量其他酷炫内容，包括音乐和音效。

You can generate video, image, audio, text, and tons of other cool stuff, including music and sound effects.

所以去 aibox.ai 花 8.99 美元试试吧。

So go check out aibox.ai for $8.99.

你就能访问几乎所有主流的 AI 模型，而这些模型通常每个都要花 20 美元。

You get access to basically all of the top AI models that you'd usually have to spend $20 each for.

希望这能帮你省下大量金钱，也省去在不同标签页间来回切换的时间。

So, hope that's something that saves you a ton of money, and also a ton of time juggling between different tabs.

你可以在 aibox.ai 上一站式获取所有这些功能。

You can get it all in one central place on aibox.ai.

好了。

Alright.

我们开始今天的节目吧。

Let's get into the episode today.

我想聊的第一个话题是 AI 代币补偿。

The first story I wanted to talk about is AI token compensation.

目前，用于生成 AI 内容的代币，比如云代币，正成为开发者和公司支付用户报酬的重要预算组成部分。

So right now, token is becoming a really tokens to generate AI things, like cloud tokens, are becoming a really big budget layer, to what a lot of developers and companies are paying their users for.

这简直就像周边礼品。

It's it's almost like swag.

对吧？

Right?

就像你加入一家公司时，他们会说：嘿，来看看。

Like when you go join a company and they're like, oh, check out.

你会得到一个免费的酷炫背包、免费的酷炫连帽衫，还有各种东西。

You get this free cool backpack and free cool hoodie and all this stuff.

你现在通过云支出和代币分配获得的就是这些东西。

It is what you're getting now with cloud spend and token allocation.

那些被称为高影响力领域的工程师，会获得更多的代币分配。

Engineers that are in what are called, quote unquote, like high impact areas are getting larger token allocations.

这意味着他们能使用更好的工具、更快的输出，甚至可能拥有更好的职业发展路径。

And that means that they get better tools, they get faster outputs, they get, you know, basically potentially a better career path.

而那些处于不太重要岗位的人呢，

People that are in, you know, a role that they're like, well, it's not quite as important.

他们会削减这方面的投入，而你基本上就被挡在了最快的职业发展路径之外。

They're gonna they're gonna cut back on that, and you basically are getting held back from the fastest career growth path.

所以，这对许多公司来说是个有趣的难题。

So, it's kind of an interesting conundrum for a lot of these companies.

我想讲的第二个故事是，英伟达的GTC大会并没有给华尔街留下深刻印象。

The second story I wanted to cover is that Nvidia's GTC Conference really didn't impress Wall Street.

这挺有意思的。

This is kind of interesting.

在主题演讲期间，英伟达的股价实际上下跌了。

The Nvidia stock actually dropped during the keynote.

我觉得这是一个相当强烈的信号。

Think that's a pretty strong signal.

即使他们发布了许多积极的消息，但依然不足以达到预期，这恰恰说明了英伟达的崛起有多么迅猛——我不想说被过度炒作，但它的增长确实太惊人了。

Even, you know, they had a lot of positive announcements, and even that wasn't enough to really meet expectations, which just goes to show how, I don't wanna say hyped up, but like Nvidia has had such a meteoric rise.

它已经增长太多了。

It's grown so much.

我觉得人们对它期望太高了。

I think people expect so much from it.

英伟达现在是一家价值4万亿美元的公司，但由于规模如此庞大，增长预期变得极其苛刻。

And Nvidia right now is a $4,000,000,000,000 company, but it is under a lot of pressure because at that scale, growth expectations are really extreme.

我认为仅仅领先于人工智能是不够的。

I think it's not enough to just lead AI.

他们必须持续不断地主导这一领域。

They have to really dominate it continuously.

我的意思是，你听到了不少来自黄仁勋在大会上接受采访的内容，他们有着宏大的计划、远大的愿景和众多的想法。

I mean, you heard a lot of the interviews coming out of Jiang Zhen Huang at the conference, they have huge plans, huge vision, huge ideas.

他想主导许多行业，但人们觉得这还不够，或者认为他无法实现这些目标。

He wants to dominate a lot of industries, and people just don't think it's enough, or they don't think he can pull it off.

我觉得现在投资者感到担忧，因为有很多其他超大规模企业正在开发自己的工具。

I think right now investors are worried because there's a lot of other hyperscalers that are building tools right.

我们有亚马逊、谷歌，还有许多其他公司正在研发自己的芯片。

We have Amazon, we have Google, we have a lot of others that are building their own chips.

我认为这基本上是英伟达主导地位面临的最大长期威胁。

I think that's basically the biggest long term threat to Nvidia's dominance.

我认为目前的AI需求可能正在达到峰值。

I think AI demand right now might be, you know, potentially peaking.

人们担心当前的支出是提前透支的。

I think there's concern that current spending is front loaded.

一旦一些公司的初始基础设施建成，许多公司可能会放缓投资。

A lot of companies may slow investment once some of their initial infrastructure is built.

因此，我认为叙事正在从增长转向可持续性。

And so, I think the narrative is shifting a bit from kind of growth to sustainability.

华尔街在问，当竞争者大量进入市场时，英伟达还能否维持利润率？

And Wall Street is asking, you know, can Nvidia maintain the margins when competition is really entering the market?

所以，这是我们接下来需要关注的。

So that's gonna be what we have to look out for there.

接下来我想谈的是亚马逊的Tranium实验室。

The next thing I wanna cover is Amazon's Tranium Lab.

这是一个非常重要的故事，原因有很多，但Tranium已经被许多顶级AI实验室使用，包括Anthropic、OpenAI，甚至苹果也被报道在使用它。

This is a huge story for a number of different reasons, but Tranium is already being used by a lot of the top AI labs, Anthropic, OpenAI, even Apple are all reportedly using it.

亚马逊正在对AI栈进行垂直整合，这意味着他们不仅托管模型，还在研发芯片、运行基础设施，并分发模型。

Amazon is vertically integrating the AI stack, which means that they are not just hosting the models, they're actually building the chips, they're running the infrastructure, and they're distributing the models.

目前，Tranium的优化重点是成本效率，而不是峰值性能。

Tranium right now is optimized for cost efficiency, not for peak power.

因此，Nvidia在性能上占据优势，而亚马逊则在每令牌成本上竞争。

And so Nvidia is really winning on the performance, and Amazon is competing on price per token.

目前，AWS正逐渐演变成一个多功能市场平台。

AWS right now is turning into kind of this multimodal marketplace.

你有Tranium，有Bedrock，还有多个模型提供商，亚马逊正成为默认的AI分发层。

You have Tranium, you have Bedrock, you have multiple model providers, and Amazon is becoming basically the default AI distribution layer.

我认为这才是真正的Nvidia威胁。

I think this is the real, you know, Nvidia threat.

这并不是一家初创公司。

It's not really a startup.

对英伟达来说，最大的竞争对手不仅仅是一家新的芯片初创公司。

The biggest competitor is not just another chip startup for Nvidia.

而是这些超大规模云服务商掌控了大量需求，如果人们通过AWS购买英伟达的产品，而亚马逊开始自主研发一些芯片和其他可能威胁英伟达的工具，他们就能轻松地向用户推广这些产品，直接从英伟达手中抢走大量市场份额。

It's that these hyperscalers control a lot of demand, and if people are, you know, let's say they're buying access to Nvidia through AWS, well, if Amazon starts to create some of their own chips and some of their other, you know, tools that threaten Nvidia, it's really easy for them to market that and promote that to their users and steal a lot of, market share straight from Nvidia.

接下来我想谈谈WordPress的AI发布功能。

The next thing I wanna cover is WordPress's AI publishing.

我认为，WordPress目前正试图实现完全自主的内容循环。

So I think WordPress right now is trying to unlock fully autonomous content loops.

基本上，我们已经知道，像Open Claw这样的工具可以让智能体自主完成研究、撰写和发布博客，无需任何人工干预。

Basically, we know with tools like Open Claw, agents can now research, write, publish with no human required blogs.

过去已经出现过很多帮助实现这类功能的AI工具。

And there's been a lot of like AI tools that have helped to do this type of stuff in the past.

我也曾多次尝试过这些工具。

I've even messed around with it a lot.

我的意思是，你最终还是会面临互联网上一直存在的同一个问题：如果你生产的是低质量内容，人们并不欣赏或使用，如果他们不会在你的网页上停留很久阅读内容，如果他们因为内容看起来质量低劣或像是AI生成的而迅速离开，那么你的网站在谷歌搜索排名中就会下滑。

If you I mean, basically, it's you're gonna have the same problem as you've always had on the Internet, which is if you create low quality things that people don't really appreciate, use, if they don't have a lot of you know, if they don't stay on your web page for a long time reading your content, if they bounce quickly because it looks low quality or AI generated, then you're just gonna fall in the Google rankings.

所以，我认为有必要指出的是，作为这个故事的一个前提，WordPress 确实正在更多地融入 AI，但如果你生产低质量的内容，它并不会真正对你有帮助。

So, I think it's important to to note that, like, as a caveat to this whole story, you know, WordPress sure is is kinda getting in more with AI, but if you make low quality content, it's not gonna really gonna benefit you.

对于那些过于担心 AI 粗制滥造内容的人，如果用户不喜欢网站，看到 AI 粗制内容，谷歌在搜索结果中会很好地降低这类内容的排名。

And for people that are too concerned about AI slop, if people don't like the website, if you see AI slop, it gets very you know, Google does a good job of demoting that in the search results.

好的。

Okay.

所以现在发生的是这样一件事。

So here's what's happening.

WordPress 目前支撑着全球 40% 的网站。

WordPress, right now, they power 40% of the web.

我觉得我 personally 使用的所有网站几乎都是基于 WordPress 的。

I think basically all the websites I personally use are running off of WordPress.

我很少用其他平台，除了我在 Lovable 上自己开发的定制代码内容。

I don't use a lot of other, I mean other than just custom coded stuff I build on Lovable.

这算是主要的应用场景之一。

That's kind of one of the primary places.

而且这并不是一个细分领域。

And so it's not like a niche.

它基本上是网站分发的最大平台。

It's basically the biggest distribution for websites.

但这里最重要的故事是，wordpress.com 正在拥抱 AI 代理。

But the biggest story here is that basically wordpress.com is embracing AI agents.

当我提到 wordpress.com 时，并不是指每一个 WordPress 网站。

And when I say wordpress.com, that doesn't mean every WordPress website.

因为 wordpress.com 是一个你可以去托管和发布网站的地方，而 wordpress.org 则是免费运行在网站上的软件。

Because wordpress.com is a place that you can go and host and publish websites on, and wordpress.org is kind of the software that runs on websites for free.

所以我只是想区分一下这两者，但 wordpress.com 已经决定，作为一家网站和组织，将全面拥抱 AI 代理。

So I just wanna differentiate the two of them there, but wordpress.com has decided that they, as a website and organization, are gonna embrace AI agents.

他们为此发布了一整篇文章。

They put out a whole, kind of post about this.

但简单来说，这意味着你将拥有一个侧边栏可以与之对话，AI 代理将能够为你起草、编辑和发布网站上的所有内容。

But basically, what that means is that there's gonna be a side panel that you can talk to, and basically, AI agents are gonna be able to draft, edit, publish all of your content on your website.

它们将能够管理您的评论。

They're gonna be able to manage your comments.

它们将能够更新和修复元数据，并使用标签和分类来组织内容。

They're gonna be able to update and fix metadata and organize content with tags and categories.

许多过去需要手动完成的繁琐任务，现在都可以由您网站上的AI代理来完成。

A lot of this stuff that was quite menial to do in the task or, you know, in the past is now gonna be able to be accomplished by these AI agents on your website.

说实话，我非常支持。

Honestly, I'm a huge fan.

我觉得这太棒了。

I think this is amazing.

我要说的是，wordpress.com的托管版本，只占WordPress所驱动的43%网站中的一小部分。

What I'll say is that the hosted version of wordpress.com, it's only a really small fraction of the 43 of websites that WordPress, powers.

尽管如此，它的影响力仍然相当大。

Still, it is a pretty big footprint.

每月大约有200亿次页面浏览。

There's about 20,000,000,000 page views.

每月大约有4亿900万独立访客通过wordpress.com托管的网站访问。

There's about 409,000,000 unique visitors every month that go through websites that are hosted via wordpress.com.

因此，这些网站将能使用这些功能。

So those websites will get access to this.

但这并不是全球所有网站，也不是占整个网站40%的完整WordPress生态都获得了这些工具的访问权限。

But it's not every website in the world, and it's not the full kind of WordPress's 40% of the entire website that's kinda getting access to all of these tools.

好吧。

Alright.

今天最后一个消息是，杰夫·贝佐斯计划投资1000亿美元用于人工智能制造整合。

The last story of the day is that Jeff Bezos is looking to spend a $100,000,000,000 on AI manufacturing roll ups.

本质上，他将设立一个全新的基金，用于收购各大工业领域的众多公司。

Essentially, he would be starting a brand new fund, which is gonna be used to buy companies in basically a lot of the major industrial sectors.

我认为，他最终的目标是用人工智能实现这些行业的现代化和自动化。

I think ultimately he's trying to modernize and automate them with AI.

这一消息源自《华尔街日报》发布的多篇报道，均与杰夫·贝佐斯的人工智能初创公司‘普罗米修斯计划’相关。

This is according to a bunch of stuff that was published on the Wall Street Journal, and this is all kind of in relation to Jeff Bezos' AI startup, which is called Project Prometheus.

最初报道称，杰夫·贝索斯在去年11月就与他们合作了。

It was originally reported that Jeff Bezos was working in with them in November.

他是联合创始人兼联席首席执行官。

He's kind the co founder and a co CEO.

他与前谷歌高管维克·巴贾吉并肩合作。

He's alongside the former Google executive, Vic Badjaj.

Prometheus项目在启动时获得了约62亿美元的资金。

And Prometheus, at launch, had about a $6,200,000,000 in funding.

他们专注于打造这些高端人工智能模型。

They were focusing on creating these kind of high level AI models.

具体来说，他们正努力改进制造和工程领域，以及航空航天、汽车及其他类似行业中的各项技术。

And specifically, were just trying to improve things that were happening inside of manufacturing and engineering, also inside of aerospace, automotive, a bunch of other sectors like that.

我认为，他目前正在筹集的这个新制造基金将支持这一使命，他们将开始收购一些公司，这些公司最终会使用Prometheus产出的某些模型。

I think the new manufacturing fund that he's raising right now is gonna support that mission, and they're gonna start buying up companies that are gonna ultimately use some of the models coming out of Prometheus.

如果你仔细想想，这算是一种有趣的策略。

It's kind of an interesting strategy if you think about it.

据《华尔街日报》报道，贝索斯最近前往新加坡和中东地区，为这项计划筹集资金。

According to the Wall Street Journal, Bezos was recently traveling to Singapore in The Middle East as he was trying to raise funds for this effort.

我认为，他现在的计划是收购航空航天、芯片制造和国防领域的公司。

And I think right now, his plan is to acquire companies in areas like aerospace, chip making, and defense.

这确实是一个有趣的策略。

And it's a it's an interesting strategy.

对吧？

Right?

他实际上是在收购那些本会成为客户的公司。

He's literally acquiring companies that would be customers.

因此，他不需要再去寻找使用他产品的客户，而是直接收购公司，然后将他的AI技术整合进去。

And so instead of having to go find customers to use his product, he goes, acquires the company, and then puts his AI inside of those.

这是一个相当引人入胜的策略，我想他打算筹集1000亿美元来实现这一目标，这很有趣，因为Project Prometheus最初启动时的融资额为62亿美元。

It's it's kind of a fascinating strategy, and I guess he wants to raise a $100,000,000,000 to go and do this, which is interesting because Project Prometheus initially launched with 6,200,000,000.0 in funding.

所以这家软件公司目前只筹集了62亿美元，但他正试图为新基金筹集1000亿美元，专门用于收购那些将使用该技术的公司。

So the software company is only, you know, has only raised 6,200,000,000.0 in funding, but he's trying to raise a new fund of 100,000,000,000 just for the acquisition of the companies that will be powered by this.

这真是一个非常有趣的故事。

So, really fascinating story.

各位，非常感谢你们收听这个播客。

Guys, thank you so much for tuning into the podcast.

如果今天的内容让你对人工智能行业有了新的了解，麻烦你们在播客上留下一个评分或评论，这对我来说意义重大。

If this was interesting and if you learned anything new about the AI industry today, it'd mean the world to me if you left a rating or review on the podcast.

一如既往，如果你想尝试我们平台最新推出的视频模型以及其他数十个模型，请务必访问 aibox.ai。

And as always, make sure to go check out aibox.ai if you wanna try our latest, video models that we've added to the platform as well as dozens of other models.

我们刚刚从大约40个模型扩展到超过70个顶尖AI模型，全部集中在一个平台上，每月仅需8.99美元，如果选择年费计划还能享受20%的折扣。

We just went from having about 40 models to over 70 of the top AI models all in one place for $8.99 a month, and you get 20% off if you get an annual plan.

好了，各位。

Alright, guys.

感谢收听，我们下一期再见。

Thanks for tuning in, and I will catch you in the next episode.