本周AI动态：英伟达20亿美元埃隆交易 | OpenAI用户规模庞大

欢迎回到《无限AI周报》。本周，英伟达正稳步迈向成为一家价值10万亿美元的公司。黄仁勋宣布向xAI投资20亿美元，这让埃隆·马斯克笑逐颜开。但萨姆·奥特曼也没闲着，他宣布ChatGPT周活跃用户总数达到8亿，相比上次更新猛增了3亿，这简直太疯狂了。

Welcome back to the Limitless AI Roundup. This week, NVIDIA is well on its way to becoming a $10,000,000,000,000 company. Jensen Huang announced a $2,000,000,000 investment in xAI, putting a massive smile on Elon Musk's face. But Sam Altman's been pretty busy as well. He announced 800,000,000 total weekly users on ChatGPT, which is a 300,000,000 jump from last update, which is absolutely insane.

乔什，苹果公司可能最终会放弃他们的Vision Pro。本周有很多内容要覆盖，也有很多要深入探讨，首先从AMD的这条突发新闻开始。AMD，对于没听说过的人来说，基本上是英伟达的主要竞争对手。他们制造一系列重要的芯片、GPU以及训练和推理新AI模型所需的基础设施。在这次公告中，他们与OpenAI达成了一项为期多年的总容量达六吉瓦的合作协议。

And Apple might be scrapping their Vision Pro after all this effort, Josh. Lots to cover this week and lots to get into, starting with this breaking news from AMD. AMD, for those of you who who haven't heard of it, is pretty much NVIDIA's, main competitor. They create a bunch of important chips, GPUs, and infrastructure needed to train and inference new AI models. And in this announcement, they secured a partnership with OpenAI of a total of six gigawatts over multiple years.

我认为具体是五年。但在我看来，最重要的是他们向萨姆·奥特曼提供的10%股权，乔什。你听说了吗？他们基本上是在向OpenAI提供自己公司的股份。

I think this specific is five years in total. But the most, important fact about this, in my opinion, is the 10% equity stake that they're offering Sam Altman, Josh. Did you did you hear about this? They're basically offering up, a stake for OpenAI in, their own company.

是的。这是一笔相当惊人的交易。在我们深入讨论这笔交易之前，你能调出图表让我们看一下吗？因为AMD的周线图太夸张了，我认为它很好地反映了这笔交易的规模和重要性。是的。

Yeah. This is a pretty outrageous deal. Before we get into the deal, can we just take a moment to look at the chart, if you don't mind pulling that up? Because the weekly chart of AMD is outrageous, and I think it's very reflective of the size and scale of this deal. Yeah.

所以，在消息爆出的当天，我认为股价上涨了近40%。收盘时可能涨了大约30%左右，或者某个惊人的数字，但市场很喜欢这个。我认为原因在于这是一项重大赌注和技术的垂直整合。OpenAI需要芯片，AMD需要有人买他们的芯片，因为英伟达现在几乎碾压了所有人。

So, on the day the news broke, I think the stock was up almost 40%. It might have closed, like, around plus 30% or something outrageous, but the market loves this. And I think the reason why is because this is this is a big bet and a vertical integration of technology. Like, OpenAI needs chips. AMD needs someone to sell their chips to because NVIDIA is, like, kinda crushing everyone.

这笔交易可以从两个不同的角度来看。首先，AMD向OpenAI出售大量芯片。OpenAI承诺购买巨量的AMD专用AI芯片。它们被称为Instinct GPU，与英伟达的Blackwell不同。他们将从2026年使用足够消耗一吉瓦电力的芯片开始。

And the deal, you can kinda view it in two separate ways. So AMD first, they sell OpenAI a mountain of chips. OpenAI promises to buy a gigantic amount of AMD's special AI chips. They they're called the Instinct GPUs, think, which is different from NVIDIA's Blackwell. They'll start with enough chips to use one gigawatt of power in 2026.

这就像点亮整个西雅图这样一座城市所需的电力。吉瓦是个巨大的单位，以前没人达到过，但他们计划在2026年实现。随着时间的推移，总能耗将增长到六吉瓦，这简直是疯狂到离谱的事情。想象一下，你为学校项目建造一支机器人军队，你需要10万块乐高积木。

So that's like the electricity needed to light up a whole build a whole city like Seattle. Like, gigawatt is huge. No one's reached it before, but they're gonna try to do that in 2026. Over time, it'll grow to six gigawatts of total energy, which is like this crazy outrageous thing. Imagine you build a robot army for a school project, you need a 100,000 LEGO bricks.

你不是一次买几个，而是与乐高店达成协议，在五年内获得数百万个。你可以这样理解与他们的交易。然后在交易的后端，OpenAI有机会拥有AMD的一部分股份。作为交易的一部分，AMD给了OpenAI一份认股权证，这就像一张特殊门票，可以以超低价购买最多10%的AMD公司股份。想象一下每股只要一美分。

Instead of buying a few at a time, you make a deal with the LEGO store to get millions over five years. That's kind of how you could think of this deal with them. Then on the back end of the deal, OpenAI gets a chance to own part of AMD. As part of the deal, AMD gave OpenAI a warrant, which is kind of like a special ticket to buy up to 10 of AMD's company super cheap. Think like a penny per share.

但OpenAI只有在达成目标时才能获得这个，比如实际购买并使用所有这些芯片且没有任何问题。所以我给我朋友解释的一个例子是，如果你在交换宝可梦卡，你现在给你朋友一堆稀有卡片，作为回报，他们承诺如果你的收藏变得足够大，你就可以免费拥有他们卡盒的一半。所以这是一种双赢。你现在得到你需要的东西，希望未来能赚取大量收入。你可以这样理解这笔交易，这是对他们双方未来的赌注，但目标是一致的。

But OpenEye only gets this if they hit the goals, like actually buying and using all of those chips without any problems. So like, one example I was explaining to my friend is that if you're trading Pokemon cards, you give them you give your friend like a bunch of rare cards now, and in return they promise if your collection grows big enough, you get to own half of their card box for free. So it's kind of a win win. You get what you need now in hopes of earning a lot of revenue down the line. And that's kind of how you could think of this deal, is it's it's a bet on both of their futures, but aligned to the same outcomes.

所以所有这些只有在OpenAI能够建立这些数据中心，并且AMD实际上能够提供制造这些数据中心所需的GPU时才会实现，这是一个巨大的双赢，市场也很喜欢。我的意思是，这张图表很好地反映了人们认为这笔交易有多么令人印象深刻。是的。有很多关于这种收入循环的讨论，比如OpenAI付钱给AMD，AMD借东西给OpenAI。

So all of this only plays out in the case that OpenAI is able to make these data centers, and then AMD is actually able to give the GPUs required to make these data centers, which is a big win win, and the market loved it. I mean, this chart is very reflective of how impressive Yeah. People think this deal is. There's a lot of talk about this, like, kind of, like, circle of revenue happening where, like, OpenAI is paying AMD. AMD is lending stuff to OpenAI.

但总体而言，市场似乎很喜欢。Ejes，你对此有什么看法吗？

But overall, the market seems to like it. Ejes, did you have any takeaways from this?

嗯，嗯，这正是我想探讨的。你可以像你说的那样把它描述为一种双赢，但它也可以被描述为一种可能随时崩溃的纸牌屋。正如你所说，OpenAI承诺支付这些东西，但他们实际上还没有数万亿美元来支付这些GPU。而从AMD的立场来看，他们也没有准备好芯片或基础设施，对吧？

Well well, that's kind of what I wanna get into. You can frame it as one way as a as a win win, as you said, but it could also be framed as kind of like a house of cards that might come crumbling down at at any point. As you said, like, OpenAI is promising to pay for these things, but they don't actually have the trillions of dollars needed to, to pay for these GPUs just yet. And in AMD's position, they don't have the chips or the infrastructure ready to go either. Right?

所以这有点像，嘿，我欠你的，我们承诺会把这个建起来。实际上，提供给Sam Altman的10%股权是为了让交易更诱人，无论如何就是说，听着，我知道NVIDIA是老大，但我们可以做到。我们真诚地承诺，只要你信任我们并且我们做到了，你就可以以给定的价格购买我们公司10%的股份。并且会有里程碑来逐步实现这一点，对吧？

So it's kind of like a, Hey, I owe you and we'll promise to build this out. And actually the, the 10%, equity stake that was offered to Sam Altman was to sweeten the deal, in any case saying like, listen, I know that NVIDIA is the big dog, but we can deliver on this. And we're promising you in good faith that you can buy 10% of our company at a given price if you just trust us and we deliver on it. And there'll be milestones that rank this up. Right?

我想快速深入了解一下数学，因为利润率，正如你可以想象的，非常疯狂。所以承诺的六吉瓦总单位数是500万个AMD处理器。每个GPU的价格大约是2万美元。所以，AMD在这方面的总收入将约为1000亿美元。这是一个，这是一个很大的数字。

I wanna dig into the math very quickly because the margins are, as you can imagine, pretty nuts. So the total number of units that are promised for six gigawatts is 5,000,000 units for AMD processors. And the price per GPU is around $20. So, total revenue for AMD in this would be around a $100,000,000,000. That's a, that's a big number.

而且我知道OpenAI目前还没有那笔钱。没错。AMD从中获得的毛利率是50%，也就是500亿美元。但有人会说，好吧，但他们出让了公司10%的股份，而根据周五收盘价，AMD的10%股权价值——我认为是在这个公告正式发布之前，或者可能是发布当天——是210亿美元。所以你在假设，如果你能以210亿美元的估值买入，等到这些千兆瓦级别的算力交付给OpenAI时，它们的价值可能会略有增加。因此实际上，AMD花了210亿美元，在短期内赚了500亿美元，这是一笔相当不错的交易。

And I know OpenAI doesn't have that money just yet. Right. And the gross profit that AMD is taking from this is 50%. That's $50,000,000,000 But then someone is saying, okay, but they're giving 10% of the company, but then, 10% of AMD's equity at Friday's close, which is I think, before this announcement was officially made, or maybe on the day of this announcement was $21,000,000,000 So you're assuming that if you can buy in at a $21,000,000,000 valuation by the time that, these gigawatts are served to OpenAI, they'll probably be worth a marginal amount more. So really AMD spent $21,000,000,000 to make $50,000,000,000 in the immediate terms, which is a pretty good deal.

我想到的另一件事是，这对OpenAI尤其是Sam Altman来说是一个意外的举动，因为他一直与Jensen Huang和NVIDIA关系非常好。而Jensen一直以统治地位著称，你不想惹恼他。这直接违反了这一点。事实上，Josh，我不知道你是否看到了，他有一些所谓的评论，在这条推文中有详细描述，他说：我看到了这笔交易。考虑到他们对下一代产品如此兴奋，这很独特且令人惊讶。

One other thing that I thought of is this is an unexpected move from OpenAI and Sam Altman specifically, because he's been in really good cahoots with Jensen Huang and NVIDIA. And Jensen's been kind of known to rule the roost, and you don't wanna ruffle feathers with him. And this is a direct violation of that. In fact, Josh, I don't know if you saw this, but he he had some, some supposed commentary, which is detailed in this tweet where he goes, I saw the deal. It's unique and surprising considering they were so excited about their next generation product.

我很惊讶他们会在公司甚至还没建立起来之前就放弃10%的股份。不管怎样，我想这很聪明。所以我不直接知道那个引述是从哪里来的，但你可以看出Jensen有点像坐在他的王座上，观察着发生的事情，并确保孩子们行为规矩。最后一点是，我看到一条推文在流传。我找不到它，据称是从OpenAI的一份声明中摘录的，基本上说：我们正在与AMD建立这种合作伙伴关系，但这绝不违反我们与NVIDIA的合作伙伴关系，我们非常信任和热爱NVIDIA。

I'm surprised they would give away 10% of the company before they even built it. Anyway, it's clever, I guess. So I don't know directly where that quote is taken from, but you can just tell that Jensen is kind of like sitting on top of his throne and he's watching what's happening and he's making sure that the, that the kids behave. And one final point is, I saw a tweet going around. I I I couldn't find it, which, was apparently taken from a statement from OpenAI basically saying, we're entering this partnership with, AMD, but this by no means, violates our partnership with NVIDIA, who we trust and we love so much.

我们期待在未来进一步推进与NVIDIA的合作关系。他实际上提到NVIDIA的次数比提到AMD还多，而这次公告本是关于AMD的。所以这是一种谨慎的试探，好吧，我们要用谁的GPU，这是Sam一直在采用的策略，不仅仅是在GPU领域，而是在整个计算领域。记得不到一年前，他过度依赖微软，现在他已经将其分散到了Oracle和其他一些提供商。所以这是一个值得注意的有趣趋势。

And we're looking forward to taking our partnership with NVIDIA further in the future. He actually mentioned NVIDIA more than he mentioned AMD, which was what the announcement was about. So there's this cautious treading around, okay, whose GPU's are we gonna take, which is, a strategy which Sam has been employing, not just in the GPU world, but with compute in general. Remember he had an over reliance on Microsoft less than a year ago, and now he's kind of, like, spread that to Oracle and a number of other providers. So just a a interesting trend to note.

是的。我喜欢屏幕上Jensen的这个引述。他真是太有竞争意识了。然后我听取了他说的很多话。他今天早上早些时候在CNBC有一个对话，我想。

Yeah. I love I love this quote from Jensen that we have on the screen. He's just so hypercompetitive. And then I listened to a lot of the things that he said. He just had a conversation on CNBC earlier this morning, I think.

嗯。那是关于NVIDIA的交易和交易的性质。他公开表现得非常积极。当他谈论这些公司时，他觉得每个人都很棒。他总是抽时间赞扬所有了不起的公司和了不起的CEO们。

Mhmm. And it was about the NVIDIA deal and about the the nature of the deals. He publicly makes it seem very positive some. When he talks about these companies, he is everybody's great. He always makes time to give credit to all of the amazing companies, the amazing CEOs.

现实是

The reality

就像他是

is is like he is

一个非常硬核的运营者。他极具竞争力，当他看到这样的交易时，我的意思是，他说的没错，AMD放弃了公司10%的股份。就像，英伟达的芯片非常厉害。我们不需要放弃任何东西。我认为这引导我们进入下一笔交易。

a very hardcore operator. He is hyper competitive, and when he seals sees deals like this, I mean, he he's right in the sense that, well, AMD gave away 10% of the company. Like, NVIDIA ships are badass. We don't have to give away anything. And I think that kind of leads us into the next deal.

我的意思是，EJ，你以英伟达10万亿美元的估值开始了节目，所以请帮我们推导一下如何达到这个目标。就像，他们这周有一笔大交易，对吧，这可能在与x AI的估值中扮演重要角色。

And I mean, EJ, as you started the show with with a $10,000,000,000,000 valuation for NVIDIA, so please back us into how we get there. Like, they had one big deal this week, right, which which probably plays a large role in that valuation between x AI.

好的。

Okay.

请所以

Please So

所以这里的头条是英伟达将持有埃隆·马斯克的x AI高达20亿美元的股权。那是埃隆·马斯克的AI模型公司，对吧？他们拥有推特。他们拥有臭名昭著的Scruc模型，该模型在所有基准测试中领先。

so the headline here is NVIDIA is to take up to $2,000,000,000 worth of equity stake in Elon Musk's x AI. That's Elon Musk's AI model company. Right? They own Twitter. They own the infamous scruc model, which is leading all benchmarks.

所以这是一个相当重大且出乎意料的公告。既然我们刚刚谈过AMD的交易，乔什，我确实想知道黄仁勋是否有点生气，就像，好吧，去你的，萨姆。我要提醒你，我并没有过度依赖或宣誓效忠于OpenAI。

And so this was a pretty major and unexpected announcement. Now that we've just spoken about the AMD deal, Josh, I do wonder if Jensen got kind of pissed and was like, well, screw you, Sam. I'm gonna remind you that I don't have an overreliance or, oath sworn to OpenAI.

在今天早上与CNBC的对话中，他非常明确地表示，我非常兴奋能参与埃隆的所有公司。我打算继续参与。是的。我唯一的，是的，遗憾就是我没有投入更多，没有在这些了不起的公司投资更多。

On the conversation that he had with CNBC this morning, he made a very explicit point to say, I'm very excited to participate in all of Elon's companies. I intend on continuing to participate. Yep. My only Yep. Regret is that I didn't spend more, that I didn't invest more in these amazing companies.

是的。所以很明显这就像是，这是一场战争。我们是

Yep. So it is very clear this is like, this is this is war. We are I

我意思是，有趣的是，这笔交易承诺为埃隆·马斯克和XAI提供大量英伟达GPU。这些GPU将专门用于埃隆·马斯克名为Colossus 2的主要数据中心，这个项目之前在这个节目中提到过。如果你还记得，当埃隆宣布要建造Colossus 2时（他之前已经建了Colossus 1），黄仁勋曾说，这是我见过建造数据中心最聪明、最有效率的人。当时他对埃隆表达了充分的信心和积极看法。现在我们看到他似乎在进一步投入，通过表示无论埃隆做什么，我都想参与其中。

I I mean, what what's interesting is, this deal promises Elon Musk and XAI a heck ton of NVIDIA GPUs. And these GPUs are gonna be specifically used for Elon Musk's major data center called Colossus 2, which was spoken about on this show before. And if you remember at the time where Elon announced that he was, going to build Colossus two, when he'd built Colossus one, Jensen had said, this is the smartest, and most effective man that I've ever seen built a data center. And he had full faith speaking his, you know, positivity towards Elon at that time. And we're seeing him kind of like lean into it, but by him saying like, whatever Elon does, I wanna be involved in.

对吧？这也让我想到，如果萨姆或AMD将他们公司10%的股份给OpenAI，你还记得特朗普宣布政府要持有英特尔10%股份的时候吗，乔什？还记得吗？

Right? It also gets me to kind of think that if Sam or AMD is giving away 10% stake, in their company to to OpenAI, do you remember when Trump announced that the government was taking a 10% stake in Intel, Josh? Do you remember?

我记得，很清楚。是的。还有之后的股价走势。

I do. Very well. Yeah. And then the price action afterwards.

之后的股价走势，就是股价飙升。但你还记得黄仁勋对那个消息的即时公开回应吗，就是当天的那种？

The price action afterwards, which was the the stock just surged. But do you remember Jensen's immediate public response to that news, like, the day itself?

我不太确定是否记得。我猜他的情绪是积极的，但不记得他具体说了什么。

I'm not sure I do. I would assume it was positive in sentiment, but I don't remember exactly what he said.

哦，不。他在一次采访中被官方记录下来了，因为他那天在一个会议上被问到这个问题，他说如果政府试图像国有化NVIDIA公司那样做（他用了国有化这个词），如果政府强迫他交出超过10%的股份，他会把他们告上最高法院。所以从一开始他就明确表示NVIDIA不出售，这让英特尔和现在的AMD在这一举动中显得有点弱。而他则反其道而行之，说，不，我要入股AI模型公司，因为我们才是这里的大佬。我们是标普指数的领头羊，我们要做到这一点。

Oh, no. He he on official record in an interview because he was at a conference that day and he was asked about it where he said, if the government tries anything like that in terms of nationalizing, and he used the word nationalizing of NVIDIA's company, if they if the government forces him to give over 10 percent, he will sue them to High Haven. So he made it very clear at the start that Nvidia is not for sale and it kind of makes Intel and now AMD look a little weak, in, in this move. And he's going the other way and saying, no, I'm gonna take a stake in the AI model company because we're the big dogs here. We're top of the S and P and we're gonna make it.

在我深入讨论数字之前，乔希，你有什么想法吗？

And I oh, do you have any thoughts before I I jump into the numbers here, Josh?

是的。当你成为造王者时，你就有权决定这些事情。我不确定英特尔是否真的有选择。英特尔的選擇是慢慢流血而死，或者得到政府补贴，在一个他们逐渐失去相关性的世界中成为一个重要的公司。所以这对英特尔来说可能是一笔好交易，但是，是的。

Yeah. When you're the kingmaker, you get to decide these things. I'm not sure Intel really had a choice. Like, Intel's choice was to slowly bleed out or get subsidized by the government and and become an important company in a world where they were, like, declining in relevance. So it it was probably a good deal for Intel, but for yeah.

当然，NVIDIA永远不必处理这个问题。他们有太多的筹码。如果有人对此有意见，那真是太糟糕了。没有人能制造出像他们那样好的GPU。所以我觉得这很有趣。

Of course, Nvidia would never have to deal with this. They have far too much leverage. And if they and if anyone has a problem with it, well, it's too damn bad. No one else makes GPUs as good as they do. So I found that interesting.

关于这笔交易，我发现的另一个最有趣的事情是交易的实际动态，因为这不是一个传统的“我们投资你，我们进行股权交换”的方式。它使用了特殊目的载体（SPV），用GPU作为抵押品。这笔交易有很多这种技术细节上的微妙之处，我认为这使得它独一无二，我很乐意详细讲解。

One of the other things I found most interesting about this deal was the actual dynamics of the deal because this wasn't a traditional, like, we're investing in you. We're getting an equity swap. It uses special purpose vehicles. It uses GPUs as collateral. There's a lot of these, like, weird technically nuanced things about this deal that I think makes it unique that I would love to walk through.

你们，如果有什么，是的。我想请你带我们详细看看Twitter上的这篇帖子。

You guys, if you have any, yeah. I guess walk us through this this detailed post here on Twitter.

是的。所以，所以这是我的理解。NVIDIA没有直接投资x AI，而是投资了一个SPV，即特殊目的载体，你知道，很多公司这样做是为了绕过某些组织层级结构。这个SPV总共筹集了大约200亿美元，其中125亿美元是债务，其余75亿美元是实际股权。NVIDIA的20亿美元投资投入了那75亿美元的股权部分。

Yeah. So so so here's my understanding of this. NVIDIA didn't invest directly in x AI, but they, invested into an SPV, a special purpose vehicle, which, you know, a lot of these companies do to kind of like get around certain organizational hierarchical structures. And this SPV, I think in total was raising around $20,000,000,000 $12,500,000,000 in debt and the rest $7,500,000,000 in actual equity. NVIDIA's putting the $2,000,000,000 of their investment in that $7,500,000,000 of equity.

所以这是一个纯股权的事情，不涉及债务。这笔资金完全流向XAI，这真的很有趣。再深入一层，英伟达实际上并不是直接把这笔钱交到他们手上，而是以GPU的形式提供，乔什。

So this is an equity only thing. It's not in debt. It's it's going purely to XAI, which is really interesting. Going a layer deeper, NVIDIA's not actually putting this money directly into their hands. It's coming in the form of GPUs, Josh.

如果你认出了这个模式，这正是我们刚才谈到的AMD与OpenAI合作的方式。所以这正演变成一种交易，就像这条推文所阐述的，他认为我们正在见证GPU建设的新融资模式出现。资金直接与硬件挂钩。因此，如果项目失败，贷款方可以扣押GPU，因为它们本身就具有价值。实际上，英伟达的芯片正在有效地变成一种独立的资产类别。

And so if you recognize the pattern that's happening, that's what AMD was doing with OpenAI that we just spoke about. So it's kind of becoming this deal, which is laid out by this tweet over here where he goes, I think we're watching the new financing model for GPU build outs emerge. The funding is tied straight to the hardware. So if a project fails, the lenders can seize the GPUs because they hold their own value. So in effect, NVIDIA's chips are effectively turning into their own asset class.

这难道不疯狂吗？这基本上是在说GPU是新的黄金，就像在说：去你的股权，去你的钱或任何形式的抵押品，或者你从银行借的贷款存款，我都不在乎。我想要你的GPU。如果你搞砸了，我就会收回它们。

Isn't, isn't that insane? It's basically saying GPUs are the new gold, like screw your equity, screw the money or any kind of collateral that you have or that bank lending deposit that you've, that you've taken out. I don't care. I want your GPUs. And if you screw me over, I'm gonna recall them.

是的，这太神奇了。感觉就像金本位，但针对的是GPU。它们是这些非常抢手、高度复杂的东西，在一个由AGI驱动的世界里，它们是唯一重要的东西。对我来说，这笔交易是一种非常有趣、前瞻性的看待世界的方式，用GPU而非美元来计价价值，因为，我们知道美元正处于这种通胀状态，我们的购买力在下降，但GPU却恰恰相反，用这些GPU作为抵押来构建交易是一种非常有趣和迷人的方式。

Yeah. This is amazing. It's like it feels like a gold standard, except for GPUs. They are these very highly coveted, highly complex things that are the only thing that matters in a world that is run by AGI. And to me, this deal is such a a fun forward looking way of viewing the world of denominating value in GPUs versus dollars because, I mean, we know the dollar is kind of like we're in this inflationary thing, our, like, spending purchasing power is going down, but GPUs are the inverse of that, and using these GPUs as collateral is such a fun and fascinating way of structuring the deal.

那么XAI为什么要这样做？嗯，他们需要一种方式来获得更多算力，因为显然AI正在爆炸式增长。嗯。他们的Colossus超级计算机已经有10万个GPU，这笔交易极大地扩展了这一点。所以英伟达希望保持其AI芯片的霸主地位，尤其是在与OpenAI和AMD达成那笔交易之后，而xAI需要这些GPU，并且需要更多资金来购买这些GPU。

So why did XAI do this? Well, they need a way to get more power because obviously AI is exploding. Mhmm. Their Colossus supercomputer already has 100,000 GPUs, and this deal helps extend that in a huge way. So NVIDIA wants to stay on that's the top dog of AI chips, especially after that deal with OpenAI and AMD, and xAI needs these GPUs, and they need more money to buy these GPUs.

那么，向谁筹集所有这些资金最好呢？嗯，就是那个制造所有GPU和（视频相关产品）的人。如果这笔交易成功，如果真如黄仁勋、埃隆和整个XAI团队所描述的那样发展，嗯，他们很快将不仅趋向于建成世界上第一个千兆瓦级集群，而且是世界上第一个拥有数百个相干GPU的多千兆瓦级集群，这些事情正在非常快速地发生。根据我今天早上听到的黄仁勋的谈话，很明显他对XAI团队非常看好。而且他

So who is the best person to raise all this money from? Well, the person who's making all of the GPUs and video. And if this deal works, and if this actually plays out as as described by Jensen and Elon and the whole XAI team, well, they're very quickly trending towards not only the world's first gigawatt cluster, but the world's first multi gigawatt cluster with a couple of 100 coherent GPUs, and and these things are happening very quickly. And it was clear based on the conversation that I heard Jensen have this morning that he is very bullish on the XAI team. And he

哦，是的。

Oh, yeah.

非常坚信他们拥有所有资源来实现这一目标，并且愿意提供他们所需的一切来使之成为现实。因为下游效应，一是黄仁勋会卖出大量GPU，二是XAI团队充分展示了英伟达公司的全部能力。这就像如果你真的有能力制造这些数据中心并快速建设，那么使用我们的产品就能取得如此惊人的成果。所以这对所有人都是双赢。这是一笔非常酷的交易。

Very strongly believes that they have all the resources to do so and is willing to give them basically whatever they need in order to make it a reality. Because the downstream effects, one, are that, well, Jensen just sells a ton of GPUs, but two, the XAI team is very much showcasing the full capacity of NVIDIA, the company. It's like if you actually have all of the ability to manufacture these data centers and build them quickly, well, this is how amazing the results could be if you use our product. So it's a win win for everyone. It was a really cool deal.

我认为这是一个令人兴奋的新范式，我们正在以GPU计价，因此对英伟达非常看好，伙计。就像，在整个过程中，他们似乎处于最佳位置。他们只有增长。不存在英伟达市值不持续上涨的世界。就像，我我非常理解那一万亿美元的市值。

I think this is this is an exciting new paradigm where we are denominating in GPUs, and as a result, just very bullish on NVIDIA, man. Like, through all of this, they seem to be in the best position. They have nothing but growth. There is no world in which the NVIDIA market cap does not keep going up. Like, I I do I very much understand the $10,000,000,000,000 market cap.

就像，当然，也许存在一些支出泡沫，但其长尾效应是巨大的。这是一个在很长、很长时间内都不会改变的趋势。

Like, sure, maybe there is some spending bubble, but long tail effects of this are huge. This is a trend that is not changing for a very, very long time.

我只想指出，这是对所有正在听这个的悲观主义者们的公共服务公告。每个人都在投资这个东西。每个人都在掷骰子。每个人都在下这个大赌注，而你看到这个行业的领导者，比如我们这条来自OpenAI总裁兼联合创始人格雷格·布罗克曼的推文，说，我们正在努力在未来几年内尽可能快地构建尽可能多的计算能力。他接着提出了一个观点，我认为每个人都低估了AI对计算能力的需求。

I just wanna point out this is a public service announcement for all the doomers out there that are listening to this. Everyone is investing in this thing. Everyone's rolling the dice. Everyone's taking this big bet, and you're seeing the leaders of this industry, like we have on this tweet up here from Greg Brockman, president and co founder of OpenAI, saying, we're working to build as much compute as we can as quickly as we can over the next few years. And he goes on to make the point that I think everyone is underestimating the demand for compute that AI will have.

所以，不是所有那些悲观主义者说，嘿，我们处于资本支出泡沫中，你知道，AGI还没有出现。而是这些领跑者，整个行业的领导者，通过他们在计算硬件上的重大投资和他们达成的交易告诉我们，计算能力是必要的，他们将掷骰子赌这个行业。这里还有一个更广泛的观点，乔什，我不知道你是否同意，那就是，此时此刻这对美国来说是成败攸关。我昨天在推特上谈到了这一点，因为我只是在思考整个美国与中国的事情。而且，你知道，我们明天将推出一期关于机器人的节目，在我看来，中国在那方面处于领先地位。

So instead of all the doomers, that are saying, hey, we're in a CapEx bubble, you know, AGI hasn't emerged yet. The front runners, the leaders of this entire show is telling us through these major investments in compute hardware and the deals that they're making that compute is necessary, and they're gonna roll the dice betting on this industry. And there's a wider point here, Josh, which I I'm wondering if you agree with, which is, it's make or break for America at this point. I I tweeted about this yesterday because I was just kind of, like, thinking about this whole, like, America versus China thing. And, you know, we're gonna have a robot episode coming out tomorrow, and China's, like, in my opinion, leading, ahead, over there.

但这就像，美国如何在这里获胜？他们需要下这些大赌注。他们也许需要政府持有10%的股份，并将所有制造业带回国内，增加机器人投资，并推动标普指数每周都创下历史新高。因为如果他们不这样做，我们就有可能失去一切。我认为这对美国来说完全是启动模式，我很兴奋地想看看它会走向何方。

But it's like, how does America win here? Like, they need to take these big bets. They need to perhaps take a 10% stake for the government and kind of like bring all manufacturing onshore and ramp up their robotics investments and push S and P to all time highs every single week. Because if they don't, we risk losing it all. And I think that that is just full on startup mode for America, and I'm excited to see where it goes.

我喜欢这个想法，就是对中国视而不见，不去真正关注，只把焦点放在我们自己内部。再回到你那一万亿美元的估值，我最近才开始意识到的一件事是，基本上整个世界都运行在CPU架构上。就像，一切都运行在CPU上。这就是我们使用计算机的方式。这就是你看到的全世界所有计算机的运行方式。

I like the idea of just just turning blinders to China and not really paying attention and just keeping the focus on us in house. To to back in again to your $10,000,000,000,000 valuation, one of the things that I've just began to realize recently is the fact that, like, basically the entire run world runs on CPU architecture. Like, everything is run on CPUs. It's like that's how we use computers. That is how all the computers that you look at across the world run.

但随着人工智能的出现，尤其是最近的强化学习和推理技术，为了给出更好的响应，需要生成或必须生成的令牌数量大幅增加，这就像是两股力量在起作用：一方面，人工智能创新的自然趋势导致需要生成的令牌数量呈指数级增长；另一方面，所有这些令牌都是在GPU上生成的。因此，如果你相信人工智能将渗透到全球所有计算中这一理念，那么你也不得不相信整个世界的CPU架构都需要转向GPU。如果这个设想成真，那么所有内置处理器的设备都将被颠覆。这将是

But with the advent of AI, and most recently, reinforcement learning and inference, where much more tokens are generated or required to be generated in order to give a better response, there are like these two forces at play where, one, while the natural flow of this AI innovation is leading to exponentially more tokens needed to be generated, and two, all of those tokens are generated on a GPU. So if you believe in the idea that AI will be ambient throughout all the computing in the world, then you also kind of have to believe that the entire CPU architecture of the world needs to flip over to GPUs. And in the world that that's true, that displaces everything with a processor inside of it. It is

是啊。

a Yeah.

一个巨大的产业，规模远超当今任何现有行业。所以，如果说像英伟达这样的公司还有2到2.5倍的成长空间，基于这个前提似乎非常合理——只要你相信人工智能是未来，相信强化学习和推理训练，

Ginormous industry, bigger than anything that we have currently today by a long shot. So to say that a company like NVIDIA still has a two to 2.5 x left in it seems very reasonable in that basis, where if you believe that AI is the future, if you believe reinforcement learning and inference training,

它似乎有点被低估了。

it kind of seems undervalued.

确实如此。短期来看，没人知道到底会发生什么。但长期而言，这是一个非常稳固的趋势，而且有一家公司显然处于最前沿。所以我看好英伟达，看好GPU，看好所有这些交易。

It does. On a short time scale, no one has any idea what the hell is gonna happen. But over a long period of time, this is very much a sticky trend, and there's one company very clearly at the forefront. So bullish on NVIDIA, bullish on GPUs, bullish on all these deals.

别忘了英伟达最初是一家游戏GPU或游戏硬件公司。我们就是这样认识他们的。

Don't forget NVIDIA started off as a gaming GPU or gaming hardware company. That's how we know about them.

因为是啊。老兄，我超爱玩游戏的。我

Because Yeah. Dude, I'm a freaking I loved gaming. I

组装了一台定制电脑。法律。告诉我们法律，乔什。

built a custom PC. The law. Tell us the law, Josh.

我嘛，我以前一直玩游戏，还组装了一台定制电脑来玩游戏，那时候买的就是英伟达的GPU，我总是买最新的英伟达GPU，然后后来，嗯，以太坊出现后，它们开始被用于以太坊挖矿，可以买些GPU来做这个，然后后来它又转回了游戏领域，这些东西会瞬间售罄。所以我以前做过运动鞋转卖和转卖机器人。你会试着买GPU来转卖，因为它们非常稀有且难以获得。现在呢，它就像是从游戏玩家升级到矿工，再到电脑极客，再到世界上最大的超级大国，现在每个人都想要GPU。所以听着，我不是说，其实我是说这些书呆子游戏玩家，他们周末做的事情，你真的得关注一下，因为没错。他们走在潮流的前面。

I well, I used to I gamed all the time, and I built a custom PC to play games on, and the thing that you bought was the NVIDIA GPUs, and I would always just buy the newest NVIDIA GPU, and then eventually, well, when Ethereum came around, they started being used for for, like the Ethereum mining, could buy some GPUs for that, and then eventually it kind of pivoted to back to gaming, where these things would sell out instantly. So I used to do sneaker reselling and reselling bots. You would try to buy the GPUs to resell them because they were so rare and so difficult to get, And now it's like it's kind of leveled up from like the gamers to the miners to like the computer nerds to like the largest superpowers in the world, and now everyone wants the GPU. So listen, I'm not saying, actually, am saying that these nerdy gamers, like what they do on the weekends, you really gotta pay attention to it because Yep. They are ahead of the trend.

有时候早到看起来和愚蠢的玩具没什么区别，但他们非常正确。我用英伟达GPU已经十多年了，我可能应该把这当作信号多买点股票，因为天哪，多么不可思议的公司。

Sometimes so early that it seems indistinguishable from like a stupid toy, but they are very much right. I have had NVIDIA GPUs for well over a decade, and I probably should have used that as signal to buy some more stock because my god, what an incredible company.

黄仁勋正在上演传奇，电脑极客们赢了。这是什么世界啊，如果我告诉我13岁的自己现在这样，他绝对不会相信。嗯。好吧。那么，乔什，我们还有几件事要聊。

Jensen Huang is on a legendary run, and the computer nerds are winning. What what a world, my 13 year old self would never believe in, if I told him, this now. Mhmm. Okay. So, Josh, we have a few more items to cover.

我们开始吧。

Let's get to it.

一个小消息是，OpenAI这周相当重要。当我说重要的一周，我实际上是指重大的一周。而且最近OpenAI哪一周不重要呢？看看这些数据。他们宣布有8亿周活跃ChatGPT用户。

One tiny item is, OpenAI had a pretty big week. And when I say big week, I I actually mean major week. And what other week is not major for OpenAI recently? Check out these stats. They announced 800,000,000 weekly chat GPT users.

乔什，四个月前这还是5亿。

Josh, this was 500,000,000 four months ago.

等等，各位，我整理了一些有趣的数据：这8亿周活跃用户意味着地球上每10个人中就有一个人每周都在使用他们的产品。如果用户组成一个国家，它将是地球上仅次于中国和印度的第三大国。你每周可以用你的观众填满8000个超级碗体育场，外面还有人排队等候，而且每分钟有79000人在使用它。这相当于每60秒就有一个完整的足球场的人数登录。所以这是一个惊人的周活跃用户数量，而且周活跃用户不应与月活跃用户混淆，我确信月活跃用户数量要多得多，他们正在逐渐接近世界上最大的用户群体之一。

Wait, you guys, I pulled some fun stats about about this 800,000,000 weekly So active users that means one in every 10 people on earth is using their product every week. If users were a country, it would be the third largest nation on Earth behind only China and India. You can fill 8,000 Super Bowls every week with your audience and still have people waiting outside, and every minute 79,000 people are using it. That's more than an entire football stadium logging in every sixty seconds. So this is like a outrageous amount of weekly active users, and weekly should not be confused with monthly active users, which I'm sure is far more, where they they have to be gradually approaching among the largest user bases in the world.

我想，可能仅次于Meta，我不确定还有谁在那个类别里，但这是一个巨大的数字，而且在不算太长的时间内就达到了。

I guess, maybe behind Meta, I'm not even sure who else would be up in that category, but this is a tremendously large number to do so in not that long of a time period.

我认为另一个需要注意的重要数字是400万开发者。有一点要警惕的是，你的产品只有用户说好才算好，但你的产品能变得多好，取决于开发者有多少机会接触它，有多少意愿在上面构建。有400万开发者专注于一个单一产品，基本上就是ChatGPT，直到这次公告——我们在屏幕上看到的，来自本周OpenAI开发者日——这简直太疯狂了。我们有一整期节目，Josh和我在直播结束后立即就报道了这件事。

And I think the other major number to pay attention to here is the 4,000,000 developers. Now one thing to be wary of is your product is only as good as your users say it is, but your product only gets better as much as your developers have access to it, as much as your developers want to build on it. Having 4,000,000, developers focusing on one singular product, which has pretty much been ChatGPT up until, this announcement, which we're seeing on the screen, which is from their OpenAI Dev Day this week is absolutely insane. And we have an entire episode. Josh and I covered this immediately after the stream.

我们是最早发布相关节目的品牌之一。快去看看吧，在我们的YouTube页面、Spotify，或者你常听播客的地方都有。但Josh，我最喜欢的亮点之一是应用SDK。你还记得这个吗？

We were one of the first, brands to put out an episode on this. Go and check it out. It's it's on our YouTube page, Spotify, wherever you listen to stuff. But one of my favorite takeaways from this Josh was the, apps SDK. Do you remember this?

这基本上是一个新的升级，可以让你在ChatGPT中与应用程序聊天。虽然我们已经讨论过这个，我也分享了我的想法，但昨天我和女朋友有一次非常有趣的对话，她问，'你能和应用程序聊天是什么意思？'我向她解释了整个过程。我说，记得我们正在考虑可能搬到美国或世界其他地方的新地方吗？你知道你通常输入提示，比如'嘿，我想住在哪里？'

This is basically, a new, upgrade, which will allow you to chat with apps in your chat GPT. And whilst we've spoken about this already, and I've given you my thoughts, I had a really interesting conversation, with my girlfriend yesterday, which was, she was like, what do mean you could chat to apps? And I walked her through the process. I said like, well, remember how we're kind of like looking at potentially new places to move to, in America or around the world? And you know how you usually type in a prompt and you're like, Hey, like, where do I wanna live?

比如，告诉我这个城市的顶级地点或社区。然后她说，是的，现在你可以说，'那里有哪些顶级的出租或购买房屋列表？'它会接入Zillow。然后你可以说，'我想为这些地方安排看房，我下周要去那里。'

Like, tell me the top places or neighborhoods in this city. And then she's like, yeah, well now you can say, what are some of the top listings for a home to rent or buy there? And it taps into Zillow. And then you could say, well, I wanna set up viewings for these, places. I'm gonna visit there next week.

然后它就像，'酷，需要我为你预订航班吗？需要我为你预订这些列表的看房吗？让我访问你的日历。'它会为你完成所有这些事情。

And they're like, cool. Do you want me to book a flight for you? Do you want me to book, the bookings for the listings for you? Let me give access to your calendar. And it does all of these things for you.

它让你的生活更轻松。但我想到的一点是，乔希，这对应用程序构建者来说是一种全新的方式，让他们与这一新范式互动。我的意思是，你的应用现在不再服务于人类，而是服务于这些AI模型，服务于这些AI智能体。

It makes your life easier. But the thought that I had here, Josh, was this is a completely new way for the app builders themselves to interact with this new paradigm. What I mean by that is your app now no longer caters to humans. It caters to these AI models. It caters to these AI agents.

我认为这与品牌从实体店转型的过程并无太大不同——以前他们需要关心地理位置、客流量以及产品在店内的摆放位置，而现在转向在线电商，突然之间人们不得不担心SEO、在谷歌搜索页面和排名中争第一。所以在我看来，这就像是又一次飞跃，跳向一个不同的媒介，而人们正在实时摸索中。

And I think that that is not too dissimilar from, brands going from brick and mortar physical stores, where they had to care about location. They had to care about foot traffic. They had to care about where they positioned their products when people walked in, to online e commerce where suddenly people had to worry about SEO and becoming number one in Google's search page and search rankings. So in my opinion, this is like another jump. It's another jump to a different medium, which people are figuring out in real time.

而且我认为他们还没有做好准备。这就是我的看法，我觉得必须分享出来，因为我觉得，天啊，这完全是范式转变。

And I don't think they've prepped for. And that was just my takeaway that I had to share because I was like, oh my God, this is completely, like, paradigm shifting.

是的。他们正在吞噬世界。我认为我们看到OpenAI和NVIDIA之间有类似的动态——他们拥有每个人都想要的东西，一边是GPU，另一边是用户。如果你有8亿周活跃用户，并且有机会接入这个用户流，真正为用户体验增值，那简直是显而易见的选择。很明显，OpenAI现在正试图把所有人都吸引过来，充当外部世界的聚合者。

Yeah. They're eating the world. I think we see similar dynamics between OpenAI and NVIDIA in the sense that they have what everyone wants, which is one case is GPUs, but the other case is users. If you have 800,000,000 weekly active users and you have an opportunity to plug in to that user flow and actually add value to the user experience, that's a no brainer. It is very clear that OpenAI is now just trying to bring everyone to them and serve as an aggregator to the outside world.

我认为这很棒，因为他们绝对是最有能力做到这一点的。所以，是的，这是本周一个有趣的小插曲。再次提醒那些不知道或没跟上进度的人，我们昨天确实发布了那一期节目，快去看看吧。你想在这里谈谈Sora吗？Sora更新第一弹。

And I I think that's great because they're definitely the best equipped to actually do that. So, yeah, that that was an interesting little thing this week. Again, for people who don't know, or who aren't up to date, we did drop the episode yesterday, so check it out. Do you wanna talk about Sora here? Sora update number one.

我看到Sam发了博客文章。我很乐意聊。所以是Sora更新第一弹，但它也正式成为App Store排名第一的应用，乔希。就在本周或一周前，OpenAI宣布了他们的新文本转视频模型，但他们有个转折——他们还推出了一个全新的社交媒体平台，有点像TikTok，但每个视频都是AI生成的，乔希，我有点吐槽它。

I see a Sam on blog post. I I would love to. So Sora update number one, but it is also officially the number one app in the App Store, Josh. So this this week or or a week ago, OpenAI announced their new text to video model, but they did it with a twist. They also announced a brand new social media platform, which is kind of like TikTok, but every single video is AI generated Josh and I kind of hated on it.

我在节目结尾大致总结说，也许这会成功，也许不会。但我得说，从那以后我反而爱上了它，我创建了一个… 抱歉，我会让你…

And I kind of rounded off the episode with, maybe this might work, maybe it won't. And I have to say I've ended up loving it since, I've, I've created a, sorry, I'll have you

你得告诉我为什么。

know You gotta tell me why.

好吧，我来告诉你原因。我来告诉你原因。而且，我不知道我是不是太骄傲了才不愿意宣布原因，但无论如何我还是要说出来。我喜欢看到自己出现在我能想出的各种随机幻想或点子中，但更重要的是，我喜欢和朋友们一起这么做。

Okay. I'll tell you why. I'll tell you why. And, I don't know if I'm too proud to announce why, but, I'll, I'll say it anyway. I like seeing myself feature in all these random fantasies or ideas that I can come up with, But, but, but more so I like doing it with my friends.

我可以客串，大卫·霍夫曼。我可以在一个有趣的小视频里让你客串一下。如果我有一个有趣的小点子，我会和你分享。而且我必须说，我这么做并不是带着那种‘乔希，看这多真实啊，是不是很酷？’的想法。

I can cameo, David Hoffman. I can cameo you in a fun little video. If I have a fun little idea and share it with you. And I have to say, I'm not doing it with the idea of like, Josh, look how real this is. Wouldn't it be cool?

我不是在给你发暗示说‘我希望我们能做这个’。更像是‘这不是一个有趣的梗吗？’就像，看我们在开玩笑。我们，你和我，永远不会这么做。所以我想这种新奇感对我来说还没有消失，对很多人来说也没有消失。在应用商店登顶可不是闹着玩的。

I'm not like sending you a hint saying like, I wish we could do this. It's more like, isn't this a funny meme? Like, look at us joking around. We, you and I would never do this. And so I guess that novelty hasn't worn off for me and hasn't worn off for a heck ton of other people Reaching number one in the app store is is no joke.

是的。我想对我来说，我甚至已经两三天没打开这个应用了。所以它对我来说已经失去了粘性，但我想我分享你的热情，因为再次强调，这是一家拥有8亿活跃用户的公司，他们一时兴起就推出了一款社交媒体应用只是为了测试。我真的很钦佩萨姆和OpenAI团队的一点是他们愿意实验，并且越来越打破边界。我的意思是，版权问题就是其中之一，你可以用宝可梦或海绵宝宝这些受版权保护的角色制作视频，他们愿意在更快发布和更快创新方面稍微变通规则，我真的很佩服。

Yeah. I think I mean, for me, I haven't even opened up the app in, two or three days. So it it has lost the stickiness for me, but I think I share your enthusiasm in the sense that, again, this is a company with 800,000,000 active users that just dropped a social media application on a whim just to test things. And I really I I admire that about Sam and the OpenAI team is their willingness to experiment and and to increasingly break down borders. I mean, the copyright thing was one of them where you can make your videos with, like, Pokemon or SpongeBob or these copyrighted characters, and their willingness to kind of bend the rules a little bit in terms of shipping faster and innovating faster, I really admire.

所以我预计这将是OpenAI的一个方向性趋势，我们会看到很多新技术应用、产品被推出，只是看哪些能留下来，哪些不能，但把它们作为这些平台来完成开发循环，新技术与产品配对，发布产品，然后看会发生什么。而Sora取得了压倒性的成功。我的意思是，在App Store排名第一，所有这些，就像我在Instagram和TikTok上看到的所有这些病毒视频，这真的是一件大事。所以向

So I expect this to be a directional trend with OpenAI where we see a lot of new technology applications, products just being launched, and just seeing which ones stick, which ones don't, but using them as these platforms to kind of complete the development circle where new technology, pair it with a product, release the product, and just see what happens. And Sora, overwhelming success. I mean, number one in the App Store, all of this, like, all of these viral videos that I'm seeing on on Instagram and TikTok and everything, like, it it's a really big deal. So props to the

唯一的答案。不过大部分病毒式传播并不是发生在Sora上。对吧？它是发生在像Instagram和TikTok这样的平台上。我见过数百万个Sora视频获得了数百万的点赞和数亿的观看量，但都不是在真正的Sora应用上。

only answer. Most of the virality is not happening on Sora though. Right? It's it's happening on on like Instagram and TikTok. I've seen a million Sora videos that have got like millions of likes and hundreds of millions of views, not on the actual Sora app.

所以这大概就是他们需要解决的一个循环。但我必须稍微反驳你一下，乔什，因为尽管他们当初很勇敢地采取了‘嘿，我们会侵犯一堆版权，你们可以起诉我们或者选择退出’的策略，但这周他们其实也在某种程度上退缩了。他们基本上已经移除了大量受版权保护的内容，结果导致用户在尝试创建视频时——我自己也遇到过——只会收到一条错误信息说：‘抱歉，我们无法创建该视频，已被禁止。’

So that's probably like a loop that they need to tie in. But I I have to push back on you, a little bit, Josh, because whilst they were brave with going down the, Hey, we're gonna infringe on a bunch of copyright and you can sue us or opt out or whatever. They've kind of been backtracking on that this week as well. They've basically been removing a bunch of copyrighted material and it's resulted in an experience where when people are trying to create videos, and I've seen this myself, you just get an error message saying, sorry, we we can't create that video. It's been prohibited.

这种事情发生两三次之后，我就有点失望，也没什么动力去制作视频了，因为我会想，那我还能做什么呢？我来这里的全部意义就是和超级马里奥兄弟一起在游戏里奔跑并制作视频，对吧？但现在我做不到了，所以这还有什么意义呢？

And after that happens two to three times, I'm kind of disenchanted and disincentivized to create a video because I'm like, well, what can I create? The whole point of me coming here is me creating a video with Super Mario Brothers and running in the game with them. Right? And now I can't do that. So so so what's the point?

是的。我得说，在第五天做这件事，而不是在第零天做，其好处是——没错——这有着天壤之别，因为我们确实获得了那五天的病毒式传播。它确实登顶了排行榜第一。在我还对这款应用感兴趣的那段时间里，它确实非常棒，因为它没有任何限制。

Yeah. I I will say the benefit of doing that five days in versus on day zero is Yes. This, like, huge astronomical difference because we did get those five days of virality. It did get to number one. I did see in in the time that I was interested in the application, it was fantastic because it had none of the guardrails.

所以，尽管他们是事后才补上这些限制，但我确实欣赏他们是在第五天而不是第零天采取了行动。我认为这可能就是登顶App Store榜首和一个从一开始就很‘拉胯’、像Meta那样的氛围实验之间的区别。至少你尝到了那种‘无拘无束’版本的滋味。另外，给那些坚持听完我们上期节目的朋友们一个小备注：我确实给评论的朋友们扔了几个邀请码。

So even though they are adding them retroactively, I do like the fact that they were able to do that on day five instead of day zero, and I think that may have been the difference between number one in the App Store and just like a vibes like meta experiment where it's just it's just lame off the bat. Like, you never even get a chance to try the unhinged version. At least you got a little taste of the unhinged version. And then also, just one little comment for the people who made it to the end of our last episode. I did drop some sort of two codes for the people who commented.

我把我的全给出去了，一个不剩了。所以，Aegis，如果你还有的话，请给这些可爱的朋友们再扔一些。不过谢谢大家的收听和回复。我的全被领走了。

I gave you all of them. Have none left. So, Aegis, if you have any more, please drop drop them to the nice people. But thank you for for listening and replying. All of mine got taken.

我之前还跟我朋友们搞了个小‘庞氏骗局’，我邀请一个人，然后拿到四个邀请名额，这样我就能发更多。但现在我也弹尽粮绝了，不过还是谢谢大家的评论。我们会试着再去搞一些，我们会尝试补充库存，因为我真想给收听节目的朋友们一点小奖励，尤其是那些坚持听到最后的人。

I was, like, doing a little Ponzi scheme with my friends where I would invite one and then take the four invites so I could drop more. I'm out. I got nothing left, but thank you for commenting anyway. And we'll try to get more. We'll try to refill them because I want to give the people who listen a little reward, especially for those that make it to the end.

所以谢谢

So thank

又是你啊。是的。不过确实。嗯，是的。正如乔什提到的，我们正在与OpenAI合作，为无限听众们争取一大堆代码。

you again. Yeah. But yeah. Well, yeah. As Josh mentioned, we are working with OpenAI to get you a bunch of codes for limitless listeners.

你们一直是最热情的，所以我们当然不会忘了你们。好了，乔什。苹果可能要搁置他们更便宜、更轻便的Vision Pro升级版——Vision Air的开发，因为他们可能想推出一系列眼镜，你知道，就是那种能做AI功能的眼镜，类似于Meta的眼镜，甚至可能和OpenAI即将发布的类似。对此你有什么看法？因为我脸上现在可是一副相当得意的表情，因为我曾以为VisionPro才是未来。

You guys have been the most enthusiastic ever, so we we have got you there. Alright, Josh. Apple is potentially shelving the development of their cheaper, lighter vision pro revamp called the vision air because they might wanna create a bunch of spectacles, you know, some glasses that do AI stuff similar to meta glasses, and maybe even similar to what OpenAI might release. What do you have to say to that? Because I got a pretty smug face on my I've got a pretty smug look on my face right now because I thought VisionPro was the future.

我以为他们会创造出一些酷炫的新东西，结果他们现在却要去做眼镜了。

I thought they were gonna create something cool new, and now they're gonna be making glasses.

我对此有几点看法。第一点我想说的是市场上已有的眼镜，特别是你几周前订购的那副。它们在哪？那些Meta Quest眼镜呢？它们根本没到货。

I I have a couple points to this. The first point I wanna make is about the glasses that exist in the market, particularly regarding yours that you ordered a couple weeks ago. Where are they? Where are those Meta Quest glasses? They're not here.

它们没到货。是的。为什么没到？他们连这该死的东西都还没发货。根本没人能拿到这玩意儿。

They're not here. Yeah. Why are they not here? They haven't even shipped the damn product. No one could get the freaking thing.

它们根本不好用。烂透了。他们不发货。买不到。连配镜服务都没有。

They don't work. They suck. They're not delivering them. They're not available. There's no prescriptions available.

如果你戴眼镜，你根本没法戴它们，因为它们不支持很多度数。这次发布简直就是一场灾难。这根本不是个真正的问题。

If you have glasses, you literally cannot wear them because they do not support a lot of prescription. It's just like this disaster of a release. It's not a real problem.

我现在正在看我的邮件，上面还是显示它甚至还没发货。天哪，我被坑了。

I'm just looking at my email right now, and it still says hasn't been it's not even outfit delivery. God. I'm being screwed.

哦天哪。看吧？听着，我我讨厌说“我早告诉过你”，但你知道，也许我确实告诉过你。不管怎样，总有一天会的。是的。

Oh my gosh. See? Listen, I I hate saying I told you so, but, you know, maybe I told you so. Anyways One day. Yep.

目前眼镜市场上没有真正的重量级玩家。所以这是原因之一。我认为苹果内部有很多事情可能导致了这种情况。其中最重要的一个实际上是现在正在发生的领导层变动。苹果高管层有很多高管已经任职了几十年，甚至几十年以上，其中很多人要么离职，要么重组，要么改变他们的职位。

There is no serious player in the glasses market right now. So that is one reason why. I think there's a lot of things happening at Apple that are probably contributing to this. One of them one of the most important ones actually is the leadership shakeup that we're seeing now. There's been a lot of executives in the Apple c suite that have been around for decades, multiple decades plus, and a lot of them are either leaving or restructuring or changing where they're gonna be.

蒂姆·库克明确表示，他开始逐步退出CEO的位置。将会有一位新的——我想他是产品负责人，首席产品工程师，

Tim Cook is making it clear that he is starting to transition out of the CEO spot. There will be the new I think he's the chief of product, the chief product engineer,

约翰，哇。

John wow.

是的。特努斯。约翰·特努斯。是的。太棒了。

Yeah. Terminus. John Terminus. Yeah. Awesome.

他们最近开始大力宣传他。他很可能会接任CEO职位。我希望他在公司的战略变革中发挥作用，因为OpenAI非常——哦天哪。OpenAI。蒂姆·库克更偏向于运营者。

They've recently started giving him a lot of publicity. He's likely to step into the CEO slot. I'm hoping he has a role in strategic changes in the company because OpenAI is very much or oh my god. OpenAI. Tim Cook is very much an operator.

他是一位企业经营者。他将苹果公司从一个微不足道的市值运营到今天这个天文数字般的巨头，但结果却是产品创新受到了影响。因此，随着一位新的产品负责人离开公司，我满怀希望地认为他们将开始更加认真地对待他们的产品，也许眼镜就是向这个方向的一个转变。我认为VisionPRO，而且我认为很多人也这么想，是通往眼镜形态的一个垫脚石。从改进VisionPRO直接转向研发眼镜，我愿意相信这意味着他们的创新速度比最初预期的要快一些。

He's a business operator. He ran Apple from whatever small market cap they were to this astronomical giant they are today, but as a result, product innovation suffered. So with a new product leader at the forefront leaving this company, I'm hopeful that they will start to take their products much more seriously, and perhaps the glasses are a pivot to that. I think the VisionPRO, and I think a lot of people kind of think that, is a stepping stone towards the glasses form factor. Switching from improving the VisionPRO to just going right for the glasses, I would like to believe means that they have just kind of innovated a little faster than they initially intended.

我们见证了这项技术为iPhone Air实现了微型化。我认为这是一个非常好的概念验证。我们看到了非常小的显示屏，所有的计算都留在了那个平台上。这是一个非常优雅、漂亮的设备。他们把它们塞进了眼镜里。

We saw the technology get micro sized for the iPhone Air. I think that's a really good proof of concept. We saw the really small displays, all the compute being left in the plateau. It's this, like, really elegant beautiful device. They stick them in glasses.

他们有一款杀手级产品。我认为这是正确的举措。我有点失望，因为，天哪，VisionPros真是太棒了。我的意思是，看看这张照片。它们太惊艳了。

They have a killer product. I think this is the right move. I'm a little disappointed because, man, the VisionPros are fantastic. I mean, look at this this photo. They're they're amazing.

你有一副吗？嗯，我没有。没有。因为它们要3500美元。这是一

Do you have a so well. I don't have a pair. No. Because they were 3,500. It's a

大笔

lot of

钱，很难证明这笔开销是合理的，就像我不会经常使用的东西一样。是的。因为再说一次，没有生态系统开发者支持，但作为一款产品，它们非常出色。所以，想到我们在那个方面没有得到大量的创新，我感到难过，但我希望这意味着眼镜会更快面世。而且当苹果推出眼镜时，我向你保证，它们一定会非常出色。

money to justify them like something I won't use that much. Yeah. Because there's just again, there's no ecosystem developer support, but as a product, they're amazing. So it makes me sad that we're not getting a ton of innovation on that front, but I'm hopeful this means that the glasses will come sooner. And when Apple delivers glasses, it shows, I promise you, they will be actually good.

而且他们会准时交付，他们会配备高分辨率显示屏，并且在你实际参与的生态系统中具有很高的实用性，而不是这种奇怪的孤立

And they will deliver them on time, and they will have high resolution displays, and they will have a lot of utility in an ecosystem that you actually participate in instead of this weird siloed

元宇宙那玩意儿。而且它可能要到2030年或2035年才会问世。是的。我们只能再等十年了。

meta thing. And and it'll just come out in 2030 or 2035, maybe. Yeah. We'll just have to wait a decade.

那会是很久以后的事。所以别指望很快就能看到智能眼镜。但与此同时，我们将有OpenAI的硬件可以实际享用，因为在人工智能方面会有一些不错的硬件，希望能让我们撑到超酷眼镜发布的那天。嗯。但我想大概就这些了。

It'll be it'll be a long time. So don't expect glasses anytime soon. But in the meantime, we will have hardware from OpenAI that we can actually enjoy, because we'll have some good hardware on the AI front, which I think will hopefully be able to hold us over until our super cool glasses get released. Mhmm. But I think that's probably everything.

对吧？除非你还有更多不满。除非你还想再争论一下苹果的事。

Right? Unless you want unless you got more beef. Unless you wanna fight some more about Apple.

不。不。不。我我我大体上同意你的观点。我只是对他们一直以来的交付方式持悲观态度。

No. No. No. I I I largely agree with your your point of view. I just have been bearish at the way that they've been delivering.

我希望这次领导层变动真的能发生，Josh。如果他们让一个硬件专家掌舵，那将是非常乐观的信号，因为这意味着他会推出新的硬件产品，而且很可能不会是另一款iPhone。很可能会是某种重大产品。他想站稳脚跟，留下自己的遗产印记。所以我对这个很期待，但本周的综述节目就到这里了，各位。

I hope this new leadership change actually happens, Josh. And if they're putting a hardware guy at the helm, that is extremely bullish because that means he's gonna come out with a new hardware product and he's probably not gonna be another it's not probably not gonna be another iPhone. It's probably gonna be something major. He wants to put his foot down, stamp his, his kind of like legacy. So I'm excited about that, but that is it for this week's Roundup episode folks.

正如Josh提到的，我们要送出的东西是Sora的访问码。Josh现在可能没有很多，但大约两天后当我收到OpenAI团队的回复时，他就会有大量的码了，Josh。顺便说一下，这是最新消息。

As Josh mentioned, the thing that we're giving out are Sora codes. Josh may not have many right now, but he will have plenty in about two days time when I get a response from the OpenAI folks, Josh. That that is the latest update, by the way.

好的。行吧。因为我本来想说，别承诺我们给不了的东西，但我们确实有内部消息。

Okay. Alright. Because I was gonna say, don't promise the people we can't give, but we do have insider info.

我们确实有内部消息。所以我们

We do have insider info. So We

有关系。

got the connect.

这周，Josh，我们应该激励听众做什么来获得邀请码？这周你来选。

This week, Josh, what should we incentivize listeners to do to get a to get an invite code? You get to pick this week.

嗯，他们可以说他们坚持到了最后。比如，你想给他们一个关键词之类的吗？如果你还在看这期节目，你值得拥有所有。你居然和我们一起待了将近四十分钟，不管我能给什么，请收下吧。

Well, they could just say that they made it to the end. Like, do you wanna give them a keyword or something? Like, if you're still watching this episode, you deserve everything. The fact that you're stuck with us for, like, almost forty minutes, whatever I could give, please take it.

那就评论黄仁勋。天啊。他是那个人。评论黄仁勋，我们会给你一个Sora码。

So Comment Jensen Huang. Oh god. He's the guy. Comment Jensen Huang, and we'll get you a Sora code.

好的。EJazz会帮你搞定Sora，他可是深入AI世界的。他是这周早些时候和人聊天的那位。如果还没看，快去查看那期节目。非常精彩。

Okay. EJazz is gonna get you the Sora he's, like, plugged into this whole of an AI world. He was the one that chatted with people earlier in this week. Check out the episode if you haven't. It's fantastic.

全是关于记忆和脉冲的。我们会努力帮你拿到那些代码，这样你就可以试试Sora，然后你可以告诉我们你是喜欢还是讨厌它，因为我可是讨厌派。EJazz是喜欢派，但我想让你有机会尝试。如果还没做，别忘了点赞视频、订阅、分享给你的朋友，所有好事。明天我们回来，带来一期关于机器人的精彩节目。

All about memory and pulse. We'll work on getting you those codes so you could try Sora, and then you could let us know if you like it or hate it because I'm very much in the hater basket. EJazz is a lover, but I want you to have the opportunity to try. And if you haven't, please don't forget to like the video, subscribe, share with your friends, all the good things. And we're back tomorrow with an exciting episode about robots.

有一个新的机器人公告。目前还没有人知道，但它即将到来，就在明天。我们会详细讨论它，到时候见。感谢观看。祝好。

There's a new robot announcement. No one knows it yet, but it's coming, and it's gonna be tomorrow. We're gonna talk all it, and I will see you guys then. Thank you for watching. Peace.

哦，对了，Cameo也出现在Sora上了。耶。

Oh, and Cameo was on Sora. Yay.