The SpaceX IPO, Fable 5, AI Capex Update & Market Check w/ Gavin Baker, Andrew Fox & Clark Tang | BG2

频道: BG2Pod with Brad Gerstner and Bill Gurley
视频: https://podcasters.spotify.com/pod/show/bg2pod/episodes/The-SpaceX-IPO--Fable-5--AI-Capex-Update--Market-Check-w-Gavin-Baker--Andrew-Fox--Clark-Tang--BG2-e3km4fb
原文语言: en
统计: 共 110 轮 · Brad Gerstner 51 · Gavin Baker 19 · Andrew Fox 7 · Clark Tang 8

[0:00] Brad Gerstner

I think we're all pretty AI-pilled.And if you're AI-pilled, that means we've got to build a lot more compute than the world thinks.And that these models are going to be a lot more valuable than people think.You combine that with their core business,I don't know another entrepreneur or another business that's a better bet on the future than SpaceX.And so I think for most institutional investors, it's a must-buy, a must-own.It's set it and forget it in order to have a real bet on both the space and the AI future.All right, here we go.Early morning, Silicon Valley, BG2 is back.We're chopping it up on all things tech and markets.To do that, I have none other than GB in the house, Gavin Baker from Atreides.He's brought his main guy, Andrew Fox.And of course, I had to draft Clark Tang into the mix, my partner,to talk about some of the big questions of the day.You know, how should we be thinking about the SpaceX IPO?

我们几个基本都是「AI 信徒（AI-pilled）」。而如果你信 AI，那就意味着：我们必须建出比这个世界所以为的多得多的算力，而且这些模型会比人们想象的值钱得多。把这一点跟它的核心业务叠在一起——我想不出还有哪个创业者、哪门生意，是比 SpaceX 更好的「押注未来」的标的。所以我认为，对绝大多数机构投资者来说，这是一只必买、必持的股票。买了就放着不动，就为了在太空和 AI 这两个未来上真正下一注。

好，我们开始。硅谷的清晨，BG2 又回来了。我们照例把科技和市场的事从头聊到尾。为此，我请来了大名鼎鼎的 GB——Atreides 的 Gavin Baker。他还带来了他的头号干将 Andrew Fox。当然，我也得把我的合伙人 Clark Tang 拉进阵容，一起聊聊当下的几个大问题。比如：我们该怎么看 SpaceX 的 IPO（首次公开发行）？

[1:04] Brad Gerstner

You know, what are the big levers?There are big numbers out there for what's going to happen over the course of the next few years.So let's break that down a bit, help simplify it for folks.Mythos launched yesterday.I want to talk a little bit about like who's up, who's down in the race for super intelligence.Where are we?

再比如：关键的杠杆变量有哪些？外面已经放出了一些关于未来几年会发生什么的大数字。我们把它拆开一点，帮大家简化一下。Mythos 昨天发布了。我想聊聊在这场通往超级智能（superintelligence）的竞赛里，谁在上升、谁在下滑。我们现在到哪一步了？

[1:20] Brad Gerstner

What did we learn with the Mythos launch?And Clark was in Taiwan last week with Jensen at Computex and GTC.So what was our takeaway there?What's going on with GPUs, memory?Where are the bottlenecks?And where do we go from here?

Mythos 的发布让我们学到了什么？另外，Clark 上周在台湾，跟黄仁勋（Jensen）一起跑了 Computex 和 GTC。我们从那儿带回了什么结论？GPU、内存现在是什么状况？瓶颈在哪儿？我们接下来往哪走？

[1:34] Brad Gerstner

To start everything off, you know, maybe just kick it over to you, Gavin,talking about the SpaceX IPO.The IPO is in two days.You're a big shareholder.Congratulations.We're also a shareholder.You know, we also expect to be buying in the IPO.The Wall Street Journal is reporting, you know, the Goldman Sachs are both saying $160 billion in revenue in 2028.We know that the IPO is $135 a share, $1.77 trillion.So when we think about kind of what the big levers are, there are so many moving parts in this IPO.Nobody's better than you at just breaking it down, simplifying it.What are the key levers that we ought to be thinking about that you're thinking about over the course of the next few years?

作为开场，Gavin，先把话筒交给你，聊 SpaceX 的 IPO。IPO 就在两天后。你是大股东，恭喜。我们也是股东。我们也预计会在 IPO 里买入。《华尔街日报》在报道，高盛他们也都在讲 2028 年 1600 亿美元收入。我们知道 IPO 定价是每股 135 美元，对应 1.77 万亿美元。所以，当我们思考关键杠杆是什么的时候——这次 IPO 里活动的零件实在太多了。没有人比你更擅长把它拆开、简化。未来几年里，我们应该想清楚、你自己正在想的关键杠杆是哪些？

[2:20] Gavin Baker

Sure.So great to be here.Thank you for having me.I thought we were going to call it BGGB, but we can stick with BG2.I'm in your house.Hey, hey, hey.All subject to revision.That's okay.That's okay.So I think there's two big levers or variables that I think people should focus on.And, you know, I'm not going to comment on where I think those variables go.But one is you guys have this chart.Did you post this on X?

好的。首先，很高兴来这儿，谢谢你邀请我。我本来还以为我们要把节目改叫 BGGB 的，不过还是叫 BG2 吧。我在你家嘛。哎哎哎。一切都还可以改。没事，没事。

我认为有两个大杠杆、或者说两个大变量，是大家应该盯住的。而且我不打算评论这些变量会走向哪里。第一个是——你们做了一张图。这张图你在 X 上发过吗？

[2:46] Gavin Baker

I did before.And then we also included a new addition with XAI's new deals as well.Yeah.So Clark, who I've known for many years, did a great analysis here.And he shows that XAI's deal with Google for cloud computing generates more operating profit per gigawatt than Anthropic, than Meta, than Google, than OpenAI.Their deal, actually, with Anthropic also generates probably more operating profit than anyone but Anthropic.And so, you know, your colleague at Altimeter, Freida, also, she calculated a 55% IRR on Colossus I.You know, if you can borrow money at 6%, 7%, 8% and invest in something with a 55% IRR, I'm not the most sophisticated thinker, but that math, maths.Right.And so I think the most important variable, one of the two most important, is how quickly they bring on terrestrial data centers.We do know from Jensen that Elon brings data centers up faster than anyone, 122 days.Speed is literally cost.Right.Because every day you're paying electricians and plumbers, that's cost.And they're now monetizing them at arguably the highest rate.And so I think, you know, everybody should run their own math on that, but that is a massive variable.Yes.Truly massive variable.The second thing is, you know, we have a chart, and it's wildly out of date now.

我之前发过。后来我们又出了一版新的，把 xAI 的新交易也加了进去。

对。Clark 我认识很多年了，他这个分析做得非常好。他这张图说明：xAI 跟谷歌签的云算力交易，每吉瓦（GW）产生的营业利润比 Anthropic、比 Meta、比谷歌、比 OpenAI 都高。而他们跟 Anthropic 那笔交易，每吉瓦营业利润大概也只输给 Anthropic 自己。另外，你们 Altimeter 的同事 Freida 还算过，Colossus I 的内部收益率（IRR）是 55%。如果你能以 6%、7%、8% 的成本借到钱，然后投进一个 IRR 55% 的东西——我不是最精于计算的人，但这笔账算得过来。对。

所以我认为最重要的变量之一，是他们把地面数据中心拉起来的速度有多快。我们从黄仁勋那儿知道，马斯克建数据中心比任何人都快，122 天。速度就是成本。因为你每多拖一天，就得多付电工和管道工的工钱，那就是成本。而他们现在的变现单价，可以说是全行业最高的。所以我觉得每个人都该自己去算这笔账，但这是一个巨大的变量。是的，真的巨大。

第二件事是，我们还有另一张图，现在已经严重过时了。

[4:18] Gavin Baker

It's kind of freaking amazing.This chart is, I think, is this chart from 10 days ago?But in like the 10 or 12 days since this chart, since we made this chart, which shows the Pareto curves for Opus 4.7 for coding, for Codex from OpenAI.And now we've had Opus 4.8.It was already out of date.And now we have Fable.Totally.And Mythos, which is freaking wild.In 10 days, like we would have had to update the chart twice.Right.But what the Pareto curve shows is how much intelligence you can get for a given amount of cost.And I do think being all revenue will accrue to the Pareto curve.All at least kind of frontier model revenue will accrue to the Pareto curve.And this is Pareto curve for coding.And what I think is so impressive is that you can see in the chart that Composer 2 was Pareto dominant, or, you know, at the lowest level of intelligence with very little training.Right.And this just reflects, and I know you know Cursor well.I think you know Cursor a shitload better than I do.A vast amount better than I do.But my understanding is that Cursor and Anthropic have more tokens of proprietary coding data than anyone else, and they each have more tokens of proprietary coding data than exist on the public internet.

这事儿简直离谱。这张图……是十天前做的吧？但就在做完之后这十来天里——这张图画的是编程能力上的帕累托曲线（Pareto curve），上面有 Opus 4.7，有 OpenAI 的 Codex。然后我们又有了 Opus 4.8，图就已经过时了。现在又出了 Fable。完全是。还有 Mythos，简直疯了。十天里，我们得把这张图更新两次。对。

帕累托曲线画的是：花一定的钱，你能买到多少智能。我确实认为，所有收入都会归到帕累托曲线上——至少所有前沿模型的收入都会归到帕累托曲线上。这张是编程的帕累托曲线。让我觉得最震撼的是，你能在图上看到 Composer 2 是帕累托占优的（Pareto dominant）——在智能水平最低的那一档上，而且训练量非常小。这就反映出……我知道你很了解 Cursor，你比我了解 Cursor 多得多，多太多了。但据我理解，Cursor 和 Anthropic 手里的私有编程数据 token 比任何人都多；而且他们各自手里的私有编程数据，都比整个公开互联网上存在的还要多。

[5:33] Gavin Baker

And so they fed, Cursor fed, used KimiK.25, used their own private data, did some RL, some supervised fine-tuning, and they got a really good model.And then they spent three weeks in the Colossus 2 cluster, and they got a model that 12 days ago was Pareto dominant with Composer 2.5.Now, it's on their own benchmark, Cursor bench, so maybe take it with a grain of salt.But I think this just suggests that the Cursor data is very valuable for coding.And when it is trained, you know, Chinchilla optimal or beyond Chinchilla optimal with reinforcement learning, you know, I think it suggests that XAI and SpaceX AI has a shot of being a real player in coding.I mean, I think one of the interesting things is, you know, the way you answered the question, right, we didn't talk about launch, right?

所以 Cursor 就拿 Kimi K2.5、用自己的私有数据喂进去，做了一些强化学习（RL）、一些监督微调，做出了一个相当好的模型。然后他们在 Colossus 2 集群上跑了三周，做出了十二天前处于帕累托占优的 Composer 2.5。当然，这是在他们自己的基准 Cursor Bench 上，所以可以打点折扣。但我觉得这至少说明：Cursor 的数据对编程非常有价值。而当它按 Chinchilla 最优、甚至超过 Chinchilla 最优的量去训练，再加上强化学习，我认为这说明 xAI 和 SpaceX AI 有机会在编程上成为一个真正的玩家。

我觉得有意思的一点是，你回答这个问题的方式——我们压根没聊发射，对吧？

[6:24] Brad Gerstner

We didn't talk about Starlink or communications.Those, up until really six months ago, were the business.Yeah.Right?And, you know, and then we merged in X.AI, and we merged in Cursor, and then we announced these deals where it was very clear he was kind of building AWS right under our nose, you know, in terms of this.But what I want to do is go to Fox.Give us the breakdown.Three big lines of business, right?

我们没聊 Starlink，没聊通信。而直到大概六个月前，那些才是这家公司的业务本体。对吧？后来我们把 xAI 并了进来，把 Cursor 并了进来，然后又宣布了那几笔交易——那时候大家才看清，他就在我们眼皮底下建了一个 AWS。

不过我想把话筒转给 Fox。你来给我们拆一拆。三大业务线，对吧？

[6:49] Brad Gerstner

We've got the communication Starlink launch business.We've got the, you know, AI compute business.And then I want to come back to X.AI that you were just clicking on.But if we just go to the core business, what do we have to assume goes right in the core business, both with launch and with Starlink, in order to achieve the numbers that are out there?

我们有通信业务，Starlink 加发射；有 AI 算力业务；然后我想再回到你刚才点到的 xAI。但如果我们先只看核心业务：要达到外面流传的那些数字，在发射和 Starlink 这两块上，我们必须假设哪些事情做成了？

[7:08] Andrew Fox

Yeah, sure.So, look, I think the thing that's foundational to everything is the launch business.Right.Right.This is the kind of crown jewel of SpaceX.It's something that no one else really has, notably reusability.Right.And soon, rapid reusability.Right.This is, I think, what you need to believe in to get to the economics in AI that make orbital compute something that's very economically attractive.Right.Outside of the idea that we are in shortage for power, shortage for chips.Right.So, I think rapid reusability is the main thing that we're watching for and I think most people should watch for.You know, Elon talks about it a lot, but getting these rockets to fly at a cadence that's comparable to an airline.Right.And Gavin has used this analogy before, but the old rocket industry was kind of like, imagine boarding a plane, flying to California, getting off the plane, the plane explodes.Right.Right.Right.Right.So, I think what SpaceX are ultimately trying to achieve is have a Starship fly both stages, not just the booster, 30, 40, 50 times before you have to retrofit that ship.And when you do that, you're amortizing the cost of the vehicle over many flights.Right.And that's what brings the cost down significantly.

好的。我觉得一切的地基是发射业务。这是 SpaceX 的皇冠明珠，是别人真的没有的东西，尤其是可重复使用（reusability）。而且很快会变成快速重复使用（rapid reusability）。我认为，你必须相信这一点，才能推出那套让轨道算力（orbital compute）在经济上极具吸引力的账——这还是在「我们缺电、缺芯片」这个大前提之外。

所以我觉得快速复用是我们、也是大多数人应该盯的核心指标。马斯克经常讲这件事：要让这些火箭飞出接近航空公司班次的频率。Gavin 之前用过一个比喻：老的火箭工业就像是，你登上一架飞机，飞到加州，下了飞机，然后飞机爆炸了。对，对。所以 SpaceX 最终想做到的，是让星舰（Starship）的两级——不只是助推器——在需要返厂大修之前能飞 30 次、40 次、50 次。做到那一步，你就把飞行器的成本摊到了很多次飞行上，这才是让成本大幅下降的关键。

[8:29] Andrew Fox

But that's a really hard problem to solve.It's extremely difficult.And look, I think the company, you know, have been loud and clear.They're going to attempt to bring back the second stage of Starship later this year.Right.And then make it reusable, you know, refly the second stage next year.And from there, ramp up the cadence.But at the end of the day, driving down the cost of launch is what enables all of these other businesses and is what makes them so attractive relative to incumbents.So, how many times have Starship just launched?

但这是个非常难解的问题，极其困难。而且公司说得很清楚：他们今年晚些时候会尝试把星舰的第二级带回来，然后让它可复用，明年把第二级重新飞一次，之后再把频次拉上去。但说到底，把发射成本压下来，才是让其他所有业务成立、并且相对在位者如此有吸引力的原因。

那星舰到目前为止一共发射了多少次？

[9:05] Brad Gerstner

Starship 3, you know, just launched.How many launches are, you know, do you think kind of the consensus out there is assuming, you know, two or three years from now?Like, what is the launch cadence?Are we launching one of these every day?

星舰 3 刚刚发射。那你觉得外面的共识预期是，比方说两三年后，发射频次会到哪儿？我们是每天发一枚吗？

[9:17] Andrew Fox

Are we launching one of these every week or every month?Like, where are we in terms of expectations?Yeah.So, look, I think expectations for now, you know, we're going from, you know, call it 160, 165 launches last year up into the high hundreds of launches in the next several years.And, you know, getting into the thousands of launches probably in the next three years thereafter.Okay.I think the company of aspirations.Thousands of launches.You're launching.You're doing two or three launches a day.Right.Right.And then talk to us a little bit.What does this enable?

还是每周一枚、每月一枚？预期大概在什么位置？

是这样，我觉得目前的预期是：去年大概 160、165 次发射，未来几年会涨到几百次的高位；再往后三年，大概会进到上千次。好。这是公司的目标。上千次发射。那就是一天两三次发射了。对，对。

然后跟我们讲讲，这能解锁什么？

[9:49] Brad Gerstner

Obviously, you know, I'm here in Silicon Valley.I can't even keep a call on Sand Hill Road two decades into the mobile revolution.I mean, it's the craziest thing.It's like a third world.It's a major business problem when you're freaking out here.It's crazy.It's crazy.Right by the Starwood, dead zone.And I'm like, how can this possibly be?

你看，我人在硅谷。移动革命都过去二十年了，我在 Sand Hill Road 上连一通电话都打不完整。这太离谱了，简直像第三世界。你在这儿信号断掉，这是很实际的生意问题。太疯狂了。就在那家酒店旁边，一个信号死角。我就想，这怎么可能呢？

[10:09]

It's almost like it's a joke.It's the epicenter of technology in America.And you can't maintain a call.Okay.So, we're all going to switch to Starlink Mobile when it comes along because I don't want to lose that call on Sand Hill Road.So, walk me through a little bit just, again, high level.It's a big portion of the revenue growth expected in the business over the course of the next two to three years.My hunch is a lot of this is driven by direct-to-sell connectivity.Walk me through a little bit of those economics.Yeah.So, look.It's actually interesting.The broadband business is still very early stage when you think about the percent of households that have actually been penetrated to date.You look at the percent of global households with Starlink.It's less than 1%.And that's the broadband.You know, you kind of have a base terminal at your house, on your car, on your boat, and in airlines now as well.So, I actually think broadband can scale to hundreds of millions of terminals.Okay.Hundreds of millions of users.And today, the subscriber base.Hundreds of millions if they get rapid reusability of Starship, which is really hard.You know, if there's not competition.Hundreds of millions, it's possible.

简直像个笑话。这里是美国科技的震中，你却打不完一通电话。好吧。所以等 Starlink 手机业务出来，我们都会转过去，因为我不想在 Sand Hill Road 上掉线。

那还是从高层面讲一下。未来两三年，这块占了公司预期收入增长的很大一部分。我猜其中很多是直连手机（Direct-to-Cell）连接驱动的。给我讲讲这块的经济账。

好。其实挺有意思的。宽带业务从渗透率看还处在非常早期。全球装了 Starlink 的家庭占比不到 1%。这说的是宽带——你家里、车上、船上装一个基础终端，现在飞机上也有了。所以我其实认为宽带能扩到几亿个终端。好。几亿用户。而今天的订户基数……

——前提是他们做出星舰的快速复用，这非常难。前提还是没有竞争。几亿，是有可能的。

[11:23] Brad Gerstner

But maybe.I always say around here, it's funny.I love seeing PM and kind of analysts in this situation.It's exactly what I do with Clark.Clark will say something.I'll say, the future is a distribution of unknown probabilities.It's either more likely or less likely.So, give me the distribution.Are we talking 20%, 30%?

但也可能不。我在这儿老爱说一句，挺好玩的。我特别喜欢看投资经理（PM）和分析师处在这种情境里。我跟 Clark 就是这么干的。Clark 说一个判断，我就说：未来是一个概率未知的分布。它要么更可能、要么更不可能。那你给我一个分布。我们说的是 20% 还是 30%？

[11:40]

It's hilarious.Well, no, 100% the same thing.And, like, I've watched Elon do many hard things.And this is a really hard thing.So, I think it's reasonable to think that they're going to succeed with rapid reusability.But I just think it's important to acknowledge that, like, orbital compute, you know, Starlink, you know, Starlink V3, Starlink Direct-to-Sell.We need first reusability for Starship V3.And then rapid reusability unlocks a lot of this.Right.When I see the models that the banks are putting out there, right, and Wall Street, everybody's reported on these.These things have been widely leaked.They largely have the revenue on connectivity.So, let's call it Starlink, Direct-to-Sell, et cetera, going from, you know, let's call it $10 billion to $50 billion by 2028.And so, I'm not asking you guys to react to, you know, to tell me your specific numbers.But when I'm talking to Clark, all I'm trying to size up is order of magnitude.Do we think we can 5X the business over the course of the next three years?

太逗了。不不，百分之百是同一回事。而且我看着马斯克做成过很多难事。这也是一件很难的事。所以认为他们能把快速复用做成，是合理的。但我觉得重要的是要承认：轨道算力、Starlink V3、Starlink 直连手机——我们得先有星舰 V3 的可复用，然后快速复用才能把这里面的很多东西解锁出来。对。

当我看那些投行放出来的模型——华尔街的，大家都报道过，这些东西早就被广泛泄露了——它们大体上把连接业务的收入（就算 Starlink、直连手机等等）从大概 100 亿美元做到 2028 年的 500 亿美元。我不是要你们对具体数字表态。但我跟 Clark 聊的时候，我想掂量的只是数量级：我们是不是认为，这门生意未来三年能做到 5 倍？

[12:41]

Is there enough TAM, both in terms of broadband and direct-to-consumer?And I think the answer to that is yes.Here's what I just say very simply is I have – I travel with Starlink.I'm a big video gamer and very consistently, wherever I am in the world, Starlink is the best connection.Yes.It's the fastest.It's the lowest latency.And I do think once they get to rapid reusability, it's also going to be they're going to have the cheapest cost per gigabyte or megabyte delivered.And better, faster, cheaper has been a winning formula.And so, $50 billion, that's, you know, 0.3% penetration of the global telecom market.Now, maybe there's some deflation with Starlink pricing.But that's the way I'd frame it up.Yeah.I like betting on better, faster, cheaper.Clark, I would say probably the biggest surprise of the last six weeks is that Elon – you know, we talked about it on All In Podcasts.We called it EWS, Elon Web Services, right, that he struck these huge deals with Anthropic and Google.I don't even think people were thinking about SpaceX in the AI compute game, right?

从宽带和直连消费者两边看，可触达市场（TAM）够不够大？我觉得答案是够。

我就很简单地说：我出差随身带着 Starlink。我是个重度游戏玩家，非常一致地，不管我在世界哪个角落，Starlink 都是最好的连接。是的。最快，延迟最低。而且我确实认为，一旦他们做到快速复用，他们每 GB、每 MB 的交付成本也会是最便宜的。而更好、更快、更便宜，一直是制胜公式。所以 500 亿美元，也就是全球电信市场 0.3% 的渗透率。当然，Starlink 的定价可能会有些通缩。但我会这么框定它。对。我喜欢押注更好、更快、更便宜。

Clark，我觉得过去六周最大的意外，是马斯克——我们在 All-In 播客上聊过，我们管它叫 EWS，Elon Web Services——他跟 Anthropic 和谷歌签了这几笔巨额交易。我觉得之前根本没人把 SpaceX 放进 AI 算力这盘棋里，对吧？

[13:50] Brad Gerstner

If you looked at the models as of a few months ago, it was connectivity, so Starlink, and then it was X.AI, the model.But this whole category of taking all of this compute, which he's uniquely good at standing up, right, and then reselling it in a way that's highly profitable was not in a lot of people's forecasts.Now it's a major component of the forecast.You know, you and I did this podcast with Jensen where Jensen said Elon is an N of 1.What they achieved is singular.Never been done before.Just to put in perspective, 100,000 GPUs, that's, you know, easily the fastest supercomputer on the planet.That's one cluster.A supercomputer that you would build would take normally three years to plan.Right.And then they deliver the equipment, and it takes one year to get it all working.Yes.We're talking about 19 days.Wow.N of 1 is right.Elon is an N of 1.And his ability to secure supply, stand up the supply, you know, deploy it in a way that's, you know, coherent and effective for both himself and I guess now for others.So walk us through kind of that.It looks to me again like this is a major component of the revenue story.Totally.I mean, so we were all at the macro hard data center, and it was just very evident the amount of engineering that had gone into building these sites.

如果你看几个月前的模型，那就是连接（Starlink），加上 xAI 这个模型。但「把这么多算力拿下来——这正是他独一无二擅长的事——再以极高利润转售出去」这一整个品类，压根不在很多人的预测里。现在它成了预测里的一个主要构成。

你和我跟黄仁勋录过一期播客，他说马斯克是「独一份（N of 1）」。他们做到的事情是绝无仅有的，前无古人。给个参照：10 万块 GPU，那轻轻松松就是地球上最快的超级计算机。而这只是一个集群。你要建这样一台超算，正常要花三年做规划。然后设备交付、把所有东西调通要一年。而我们说的是 19 天。哇。独一份，没错。马斯克就是独一份。

还有他锁定供应、把供应搭起来、并以对自己（现在也对别人）连贯有效的方式部署出去的能力。所以给我们讲讲这块。在我看来，这又是收入故事里的一个主要构成。

完全是。我们当时都在 Macrohard 那个数据中心，那里投进去的工程量一眼就能看出来。

[15:21] Clark Tang

You know, people always talk about Google and their ability to build a TPU and sell the TPU to Anthropic to generate revenues for AI.I think it's a pretty similar dynamic here with Elon able to secure power, build these sites faster than anyone else, and also be able now to monetize it to this massive AI market that's ahead of us.If you look at the relationships that he's forged with a lot of his suppliers, you know, be it Jensen, be it, you know, all of these different sites that actually want XAI as a tenant, his ability to finance these deals at very attractive financing rates relative to a lot of the other players in the space.You know, these are advantages that compound over time, and when you've built the credibility to stand up these sites and monetize at these levels, you know, it's actually a very attractive proposition for a lot of folks involved.And actually, you know, if you look at these deals in particular, Gavin, you pointed out, but, you know, they're actually monetizing, you know, perhaps better than other players in the space by selling this infrastructure.A lot higher.Google is obviously paying SpaceX a huge premium for this compute.Fox, you said something that I thought was really important, which is, you know, it may very well be that in order to get, you know, first in line on space compute, which Google certainly wants to do, that they're willing to pay a premium for their terrestrial compute.

大家总在讲谷歌做 TPU、把 TPU 卖给 Anthropic 来创造 AI 收入的能力。我觉得这里是很相似的动力学：马斯克能拿到电、能比任何人更快把这些站点建起来，而且现在还能把它变现给我们面前这个巨大的 AI 市场。

你看他跟很多供应商建立的关系——不管是黄仁勋，还是那些真心想让 xAI 来当租户的各个站点——还有他以远优于同业的融资成本给这些交易融资的能力。这些优势是会随时间复利的。当你已经建立起「能把这些站点拉起来、并且能变现到这个水平」的信誉，对很多参与方来说，这就是一个非常有吸引力的方案。

而且，Gavin 你刚才也点到了，如果你具体看这几笔交易，他们靠卖这套基础设施的变现，可能确实比这个领域里其他玩家都好。高很多。谷歌显然为这批算力向 SpaceX 付了一大笔溢价。

Fox，你说过一句我觉得很重要的话：很可能是为了在太空算力上排到第一顺位——谷歌显然想这么做——他们才愿意为地面算力多付一笔溢价。

[17:00]

And so, to me, that's how you kind of square the circle as to why the premium.Any thoughts?Yeah, look, I think there's some of that embedded there.But, look, at the end of the day, SpaceX can stand up compute quickly.They can stand it up coherently.And they can stand up a lot of it in one place and have it readily available.So, look, I think that's most of the premium.But outside of that, certainly, people are going to space over time.Pay a little call option to get first in line for space.There you go.We've all been investing in the neocloud space.So, like, there's a fundamental belief around this table that we lack the compute needed to continue to push the frontier on intelligence.So, we have to build a lot of compute.Okay?

所以在我看来，这就把「为什么会有溢价」这件事说圆了。有什么想法？

是的，我觉得里面确实有这个成分。但说到底，SpaceX 能快速把算力拉起来，能拉得很连贯，还能在一个地方拉起很大的量、并且随时可用。所以我觉得溢价的大部分来自这个。除此之外，人们确实会随时间往太空去，付一点期权费换太空的优先排队权，也说得通。就是这样。

我们大家都投过新型云（neocloud）这块。所以这张桌上有一个基本信念：我们缺算力，缺到没法继续把智能的前沿往前推。所以我们必须建很多很多算力。对吧？

[17:39] Brad Gerstner

Now, there's competition going on.On one end, you have the hyperscalers who are building out that capability.Then we have AI-dedicated clouds that are building out that capability.And now, literally, in a matter of weeks, right, we have a giant that's emerged in this category, which is SpaceX.The question to you, Gavin, is can they consolidate this market, right?

现在竞争正在发生。一头是超大规模云厂商（hyperscaler）在建这个能力。然后是 AI 专用云在建这个能力。而现在，真的就在几周之内，这个品类里冒出了一个巨人，就是 SpaceX。

Gavin，问题是：他们能不能整合掉这个市场？

[18:02] Brad Gerstner

Because if I think about a marketplace, Elon has a unique ability to get the supply.He has a unique ability to cut deals on the other side.And nobody can stand it up like he can stand it up.So, I think there might be a real consolidation in the AI compute market where you have the hyperscalers on the one hand.And on the other hand, you know, he may emerge as the largest, strongest player in the AI compute market.Yeah.So, I think they are – are they the number four or number five hyperscaler today after the Google deal?

因为如果我把它想成一个市场：马斯克在拿供给上有独特能力，在另一头谈交易上也有独特能力，而且没人能像他那样把东西拉起来。所以我觉得 AI 算力市场可能会出现一次真正的整合——一边是超大规模云厂商，另一边，他可能会崛起成为 AI 算力市场里最大、最强的玩家。

对。所以我觉得他们现在——在谷歌那笔交易之后，他们是第四还是第五大超大规模云？

[18:32] Brad Gerstner

It will be number four.Kind of wild.Right.In 30 days, we went from not being an AI hyperscaler to being number four.And we passed a lot of companies, including Oracle.CoreWeave is a huge business, right?

会是第四。挺疯狂的。对。30 天里，我们从「不是 AI 超大规模云」变成了第四。而且我们超过了一堆公司，包括甲骨文。CoreWeave 是一门很大的生意，对吧？

[18:47] Gavin Baker

We're investors in, you know, and have been investors in, right?But there are a lot of other players, the Nebbias's of the world, the Irons of the world.And I would say that there are probably 50 Neolabs being funded in Silicon Valley right now as we speak because of the shortage in compute.Absolutely.So, that's kind of crazy in 30 days.That's just extraordinary.What I would say is that I think there is a belief that these data centers are commodities.And I do not share that belief.I don't think anybody around this table shares that belief.And in the same way that Elon was able to re-engineer a rocket from first principles and make it reusable, he engineered an electric car from first principles.You know, everyone else was trying to, you know, make an electric car like an internal combustion engine car.And he thought about it differently.And I think he looked at data center design from first principles.And he designed something fundamentally different.And I did actually ask the team.I said, hey, guys, maybe be a little less public about things that are very obvious to you about how to design a data center, but are revelations to other people.Because I think what you're doing is maybe more differentiated than you perhaps realize because what you're doing is so logical to you, but maybe not logical to everyone else.

我们是投资人，一直都是，对吧？但还有很多别的玩家，Nebius 之类的，IREN 之类的。而且我要说，此刻硅谷大概有 50 家新云公司正在被融资，就因为算力短缺。绝对的。所以 30 天里发生这些，挺疯狂的。真是不同凡响。

我想说的是，我觉得外界有一种信念，认为这些数据中心是大宗商品。我不认同这个信念。我认为这张桌上没人认同。就像马斯克能从第一性原理重新设计一枚火箭、让它可复用，他也从第一性原理设计了一辆电动车——别人都在试着把电动车做成内燃机车的样子，而他换了个想法。我认为他也从第一性原理去看数据中心设计，然后设计出了一个根本不同的东西。

我其实还跟他们团队说过：伙计们，有些在你们看来非常显然的数据中心设计的事，对别人来说是启示，也许可以少公开一点。因为我觉得你们做的东西也许比你们自己意识到的更有差异化——正因为它在你们看来太合乎逻辑了，但对别人未必。

[20:07] Clark Tang

And that's how he was able to do it 122 days.Yeah.I mean, to that point, Brad, yesterday we were meeting one of our portfolio companies and we were talking about behind the meter and we're, you know, really thinking about it.There's only maybe two or three players now that can actually reliably engineer a behind the meter data center.And, you know, there's real engineering work that goes into all of this.So if you think about this, if you're a gas combustion, if you're Vernova and you say we only have a certain number of gas combustion engines, now we can sell them to X.ai or we can sell them to one of these startup neoclouds.Who are you going to sell them to?

这也正是他能做到 122 天的原因。

说到这一点，Brad，昨天我们跟一家投资组合公司开会，聊到表后供电（behind the meter，即绕开电网、自建电源直接供给数据中心），我们认真想了想：现在真正能可靠地把一个表后数据中心工程化落地的玩家，可能只有两三家。这里面是有真工程活儿的。

所以你想想：如果你是做燃气发电的，如果你是 GE Vernova，你说我们手上只有这么多台燃气轮机，现在我可以卖给 xAI，也可以卖给某家新云初创公司。你会卖给谁？

[20:43]

Well, and there's another dynamic.Everyone starts making more money when the GPUs get energized and sold faster.So literally speed is money for all of the suppliers.Power, land, turbines.So I think it's what we'll see.Right.Hey, Brad, man.But this is just we're just talking terrestrial.Terrestrial.I do want to hit on and then you can flip it back on me.Talk to me.OK, so let's assume, right, that they continue to build out the terrestrial landscape.They continue to find buyers for that.Walk us through, you know, what this unlocks, you know, and how this is related to space data centers.Because I think, you know, once you start talking terrafab capacity and beyond.So we're talking a thousand gigs.Right.And this year, what we're doing, 25 or 30 gigs just to put it all in perspective.Yeah.Right.2025 gigs.OK.So once we start scaling up, walk us through, do we have to have space data centers in order to get excited about buying the IPO?

而且还有另一个动力学：GPU 通电、卖出去得越快，所有人赚的钱都越多。所以对所有供应商——电力、土地、涡轮机——速度就是钱。所以我觉得我们会看到这个。对。

不过 Brad，这些都还只是在讲地面。地面。我想讲讲这个，然后你可以再把问题抛回给我。你说。

好，那我们假设他们继续把地面这块建出去，也继续给它找到买家。给我们讲讲这解锁了什么，以及这跟太空数据中心是什么关系。因为我觉得，一旦你开始谈太瓦级产能以上……我们说的是一千吉瓦。而今年我们做的大概是 25 到 30 吉瓦，给个参照。对。

所以一旦我们开始往上扩，跟我们讲讲：我们是不是必须有太空数据中心，才有理由对买这次 IPO 感到兴奋？

[21:43] Brad Gerstner

Right.And then there's obviously this debate in the world.I heard Jeff Bezos say, you know, I think it's more like six years, but Elon's going to say three because if he says six, then it will take even longer.So say three and we may get it four or five.But are space data centers integral and essential to, you know, the IPO?

而且世界上显然有这个争论。我听贝索斯说，他觉得更像是六年，但马斯克会说三年——因为如果他说六年，那实际会拖得更久。所以说三年，我们可能拿到四五年。但太空数据中心对这次 IPO 来说，是不是不可或缺、必须成立的？

[22:02] Gavin Baker

And what do you think the timeline is to either of you guys?So I don't think, I think if you think about those variables around what cursor could mean for XAI.Yeah.And we do have an existence proof that once you really get on that Pareto frontier, revenue can scale rapidly and it's called anthropic.And there does seem to be an exhaust, there seems to be a lot of demand for coding.And I do think Amjad Massad posted something very interesting.The founder of Replit.The founder of Replit.Yeah.He called it bitter lesson adjacent that coding may be the fastest path to AGI and ASI.Because if you're really good at coding, you can write code, if a model's good at coding, to do anything.Correct.So I think that's a profound point and I think coding is going to continue to be very important.So I think if you think about that variable, if you think about Starlink Direct to Cell, enabled by Starlink V3,and you think about how quickly they can or cannot bring on terrestrial compute,I don't think orbital compute is necessary for the IPO valuation.But it's certainly important.And it's...Well, maybe another way to say it is you may think we're going to get to ASI faster than we're going to get to orbital compute.

你们两位觉得时间表是什么？

我不这么认为。我觉得如果你去想 Cursor 对 xAI 意味着什么这些变量……对。而且我们有一个存在性证明：一旦你真的站上帕累托前沿，收入可以极快地放大，这个证明叫 Anthropic。而且编程的需求看起来确实非常旺、像有一股尾气在往外冒。

我觉得 Replit 的创始人 Amjad Masad 发过一条很有意思的东西。Replit 的创始人。对。他管这叫「与苦涩教训相邻（bitter lesson adjacent）」——编程可能是通往 AGI 和 ASI 最快的路径。因为如果一个模型真的很擅长写代码，它就能写代码去做任何事。没错。

所以我觉得这是个很深刻的点，编程会继续非常重要。所以如果你把这个变量、把由 Starlink V3 支撑的直连手机、以及他们能多快（或多慢）把地面算力拉起来这几件事放在一起——我不认为轨道算力对这次 IPO 的估值是必需的。但它当然很重要。而且它……

换个说法可能是：你可能认为我们会先到达 ASI，再到达轨道算力。

[23:17]

That may take us from 300 IQ to 400 IQ, 500 IQ and beyond, and the ability to scale it up to consume 10% of global GDP.But maybe that's where we should move next.No, no.I think on orbital compute, I think Foxy would be great, Art Clark, to lay out the math for first principles.Clark has this great chart on the gigawatts it costs, the dollars per gigawatt.Walk us through the economic case.Yeah.Yeah.So, I mean, on this point of is orbital key to investing here.I don't think it is.And the first point I'll make is what are the implied monetization rates based on expectations today for the AI business?

那可能会把我们从 300 的智商推到 400、500 甚至更高，并且有能力把它扩到消耗全球 GDP 的 10%。不过也许我们该往下一个话题走了。

不不。关于轨道算力，我觉得让 Foxy 或者 Clark 从第一性原理把这笔账摆出来会很好。Clark 有一张很棒的图，讲每吉瓦要花多少钱。给我们讲讲这个经济学论证。

好。那么，关于「轨道算力是不是投资这家公司的关键」这个点——我不认为它是。我要说的第一点是：按今天的预期，AI 业务隐含的变现率是多少？

[23:59] Andrew Fox

You know, I think you threw out the $160 billion number that's been leaked out there that people are talking about.The implied monetization rate on that number is something like $14 billion per gigawatt per year for the AI business.They just signed Anthropic at 22 to 23.They just signed Google at 50.Right.Right.So, I think you can invest behind the AI business terrestrially and still be excited about it.But with orbital...I think it's an important point.Excited about it if they can get the land and the power.Right.But, I mean, I think for most investors, right, they have an easier time getting their head around how SpaceX wins terrestrially.Like, can they go get land power and chips?

你刚才抛出了外面泄露的 1600 亿美元这个数字，大家都在谈。按这个数字倒推，AI 业务隐含的变现率大概是每吉瓦每年 140 亿美元。而他们刚跟 Anthropic 签的是 220 到 230 亿，刚跟谷歌签的是 500 亿。对，对。

所以我认为，你完全可以只押注地面的 AI 业务，就依然对它感到兴奋。但如果加上轨道……

我觉得这是个重要的点。前提是他们能拿到地和电。对。但我的意思是，对多数投资者来说，「SpaceX 怎么在地面上赢」是更容易想明白的：他们能不能去搞到地、电和芯片？

[24:37] Andrew Fox

The answer to that is high probability yes.Yeah.Okay.And what we're saying is at the rate they're monetizing that, that gets you to the numbers that are being leaked out there before you even have to take the leap of faith that they're going to extend the lead with orbital data centers.But take us there on that, too.Yeah.Yeah.So, look, with orbital, I think the key thing is two-stage reusability.Yeah.And beyond that, rapid two-stage reusability.Yeah.So, today with Starship, they've shown that they can successfully re-land the booster.Mm-hmm.The second stage, we'll see what happens later this year.I think they're attempting to bring that back and then make it reusable by next year.But the thing that's important about two-stage reusability when it comes to the economics for orbital compute, right, is the cost per kg comes down significantly.You know, we're talking about going from $1,500 per kg on Falcon, somewhere in that range, to $250 per kg, something lower.And the more that you can reuse the rocket, the more that price comes down.Right.Right?

这个问题的答案是高概率能。对。好。而我们说的是，按他们现在的变现速率，你根本不需要先做出「他们会靠轨道数据中心继续扩大领先」这个信仰之跃，就已经能算到外面流传的那些数字了。

不过也带我们讲讲轨道那部分。好。关于轨道，我觉得关键在于两级复用。对。再进一步是快速两级复用。对。

今天在星舰上，他们已经证明能成功回收助推器。嗯。第二级今年晚些时候看结果。我觉得他们会尝试把它带回来，然后明年让它可复用。

但两级复用之所以对轨道算力的经济账重要，是因为每公斤成本会大幅下降。我们说的是从猎鹰（Falcon）上大概每公斤 1500 美元左右，降到每公斤 250 美元，甚至更低。而火箭复用次数越多，这个价格降得越多。对，对吧？

[25:40] Andrew Fox

Because you're just depreciating the cost of the launch.And eventually, you asymptote to the cost of the fuel.Right.Right.Right.So, assuming you can use a rocket for forever.Yes.Right?Which will take a very long time for us to really achieve that.But, and at that point, we're talking about something well south of 250 per kg.So, then you look at the specs of these AI satellites.You know, Elon did a great…

因为你是在折旧这次发射的成本。最终你会渐近到只剩燃料成本。对，对，对。当然，前提是这枚火箭可以永远用下去。是的。对吧？真要做到那一步，还得花非常久。但到了那时候，我们说的就是远低于每公斤 250 美元了。

然后你再看这些 AI 卫星的规格。马斯克那期做得很好……

[26:03]

Yeah, that pod was incredible that he laid out the other day, the specs on the satellites.It was really great because I think they are finally showing people, here's how you could viably design one of these satellites.And how heavy is the satellite?

对，他前几天那期播客太棒了，把卫星的规格都摊开讲了。真的很好，因为我觉得他们终于在给大家看：这样一颗卫星，可以怎么可行地设计出来。那卫星有多重？

[26:19] Andrew Fox

How many could you fit into a starship launch?And when you back into the numbers, you get to something like 5 megawatts of capacity per starship launch.Right.There's 100 metric tons in one of those starships.So, you can back into the math of how much will it cost per gigawatt to launch these satellites into space.Right.Launch this compute into space.And the math that you get to before you account for things like bad GPUs, bad satellites, right?

一次星舰发射能塞进去多少颗？把数字倒推一下，你会得到大概每次星舰发射 5 兆瓦（MW）的算力容量。对。一艘星舰能装 100 公吨。所以你可以倒推出：把这些卫星送上天，每吉瓦要花多少钱。对，把这些算力送上天。

而你算出来的数字——这还没算坏 GPU、坏卫星这些因素，对吧？

[26:50]

These will all be things that happen.But the math you get to is it's about $5 billion per gigawatt of CapEx to put these in space.Right.For comparison, terrestrially, talk about the switch gears, the generators, the transformers, the shell, getting the power.That today is about $20 to $25 billion per gigawatt.So, we're talking about a 5x reduction in cost on half of your bill of materials for the data center.Right.Which is a huge number.Just very simply, I mean, just to say it, it costs $60 billion to put a gigawatt on the ground today.And we'll call it 35 of that are the GPUs and the silicon that's doing the training and the inference.And $25 billion is the land, the shell, the power, and the cooling.I would hypothesize that those elements are probably going to be inflationary.So, that $25 billion may not go down.And because space, power, cooling are effectively free in space.And when I say space, I mean land.Yes.You know.Yes.There's no land in space.But there is space in space.There's a lot of space in space.You're talking about putting a gigawatt into space for $30 billion and having lower operating costs.Now, the first $60 billion that's inflationary and that $30 billion, that $5, may be deflationary over time.

这些当然都会发生。但你算出来的结果是：把这些东西放到太空里，大约是每吉瓦 50 亿美元的资本开支（capex）。对。作为对比，在地面上，你要算配电开关柜、发电机、变压器、厂房外壳、还有拿电——今天这些大概是每吉瓦 200 亿到 250 亿美元。

所以我们说的是，数据中心物料清单里有一半的成本，降到了原来的五分之一。对。这是个巨大的数字。

很简单地说：今天在地面上放一吉瓦，要花 600 亿美元。其中我们就算 350 亿是 GPU 和做训练与推理的硅片，250 亿是土地、厂房外壳、电力和冷却。我倾向于假设，这后面这几项大概率是通胀性的。所以那 250 亿可能降不下来。而在太空里，空间、电力、冷却基本是免费的。我说「空间（space）」的时候，我指的是土地。是的。你懂的。是的。太空里没有土地。但太空里有空间。太空里空间多得是。

所以你说的是，用 300 亿美元把一吉瓦送上太空，而且运营成本更低。而地面那第一笔 600 亿是通胀性的，而那 300 亿——那个 50 亿——随时间可能是通缩性的。

[28:16]

But what we need to consider is, you know, the reliability and the maintenance.And so, as long as, you know, everybody can do the math.But as long as these satellites in space aren't failing at an astronomical rate, the math, maths.And by the way, we know GPUs melt and lasers fail.We know this happens in data centers, particularly during big training runs.Yeah.I mean, GPUs melt.So, as long as the reliability and maintenance is not dramatically lower, the math is there once we have reusability and then rapid reusability for Starship V3.When you look at this, okay, so we went through Starlink.And we said, okay, like, it just stands to reason that we're going to have direct to sell on Starlink.Like, the assumptions there are, you know, again, seem like you can get your head around.Then when it comes to building terrestrial data centers, again, not a hard one to think that based on these couple deals that Elon's going to build a much bigger, Starlink's going to build, or SpaceX is going to build a much bigger business there.And then you have this call option on space that would drop the price even further.The one thing we haven't talked about is their model, right?

但我们需要考虑的是可靠性和维护。所以，只要这些天上的卫星不是以一个天文级别的速率在失效，这笔账就算得过来。每个人都能自己算。而且顺便说，我们知道 GPU 会烧掉、激光器会失效。我们知道这些事在数据中心里就会发生，尤其是在大规模训练跑批的时候。对。GPU 会烧掉。所以只要可靠性和维护不是急剧变差，一旦我们有了星舰 V3 的可复用、再到快速复用，这笔账就成立。

你这么看：好，我们过了一遍 Starlink，我们说，直连手机会做出来，这个基本合理，假设你能想明白。然后到建地面数据中心，凭那几笔交易，认为马斯克、Starlink、或者说 SpaceX 会在这块建出一门大得多的生意，也不难想。然后你还有一个太空的看涨期权，能把价格进一步压低。

我们唯一还没聊的，是他们的模型，对吧？

[29:29] Brad Gerstner

And I find this surprising, right?Six months ago, X.ai was competing.They were doing pretty well.But they've done something dramatic over the course of the past couple months, which is they bought Cursor, right?

而我觉得这件事很出人意料，对吧？六个月前，xAI 还在竞争之中，他们做得还不错。但过去这几个月他们做了一件很戏剧化的事，就是把 Cursor 买了，对吧？

[29:41] Brad Gerstner

Cursor is 700, 800 people, was already doing incredibly well from a revenue perspective.Our own projections were that they could exit this year at up to $10 billion of revenue.So they were growing very fast, one of the leading coding agents.But they also had this incredible team with the potential, right, to really build a frontier-level model.But they were compute constrained.So all of a sudden, they get bought by X.X has massive compute that they can now train on.And when I think about the revenue in AI, like if I look at that line item in the models, having it go from $10 billion to $150 billion,yes, a lot of that will be the core weave type business that they have.But the question is, how much of that is going to be the core X.ai business that's really powered by the new team from Cursor?

Cursor 有七八百人，从收入角度看已经做得非常好了。我们自己的预测是，他们今年可以做到最高 100 亿美元的年化收入。所以他们增长非常快，是领先的编程 agent 之一。但他们同时还有一支了不起的团队，具备真正做出前沿级模型的潜力——只是他们受算力约束。结果突然之间，他们被 X 收购了。X 有海量算力，他们现在可以拿来训练了。

而当我看 AI 这块的收入，比如我看模型里那一行，要从 100 亿美元长到 1500 亿美元——是的，其中很大一部分会是他们那种类似 CoreWeave 的业务。但问题是：其中有多少会是 xAI 核心业务本身，靠 Cursor 来的新团队驱动的？

[30:30] Gavin Baker

So any thoughts on that, Kevin?Right now, so Composer 2.5 was Pareto dominant 12 days ago.It was trained on the Kimi K2.5 base model.Now, what's happening is the Grok 4.3, 1.5 trillion parameter model is training.One would hypothesize, based on scaling laws, that that might be a better base model.And then the cursor data is being injected into the pre-training process, not just reinforcement learning.And we'll see.And I think that is going to be a very important data point when that comes out.And I just think everyone should keep in mind that once you are at multiple places on that Pareto curve, if you have compute, you can scale really rapidly.And I think Michael and the team at Cursor as well.

对此有什么想法，Gavin？

现在的情况是：Composer 2.5 在十二天前是帕累托占优的。它是在 Kimi K2.5 的基座模型上训练的。而现在正在发生的是，Grok 4.3——1.5 万亿参数的模型——正在训练。按缩放定律（scaling laws）推断，那可能会是一个更好的基座模型。然后 Cursor 的数据正在被注入到预训练过程里，而不只是强化学习阶段。我们拭目以待。我认为那个结果出来的时候，会是一个非常重要的数据点。

我只想让大家记住：一旦你在那条帕累托曲线上占住了多个位置，如果你有算力，你就能扩得非常快。我对 Michael 和 Cursor 团队也是这么看的。

[31:45] Brad Gerstner

This is an extraordinary team that he just downloaded, right, into SpaceX.SpaceX was already building good models.And what they have is they have this way to monetize compute that gives you this call option that you can pull all that compute in-house, right, to train a model and then to run the model.I suspect if there's an upside surprise, if we went around the table, I'd say this is the place that's getting the least amount of attention and could have the biggest upside surprise.Any thoughts, Clark, on what you think is being overlooked or areas that you think are misunderstood about the business today?

这是一支了不起的团队，他刚把它整个「下载」进了 SpaceX，对吧？SpaceX 本来就在做不错的模型。而他们拥有的，是这样一条把算力变现的路子——这给了你一个看涨期权：你可以随时把那些算力全部收回自用，拿来训练模型、再跑模型。

如果说哪里会有超预期的惊喜，我们绕桌子问一圈的话，我会说这是最少被关注、却可能有最大上行意外的地方。

Clark，你觉得今天这门生意里，有什么被忽视了、或者被误解了？

[32:20] Clark Tang

I would say what the last few weeks have proven is that Elon, their team can stand up all this compute.Actually, if you just went back one and a half years, they were behind in the race to stand up compute.They didn't have that many H100s.They brought in Colossus.Then they brought in Colossus 2 at a scale much larger than anyone else.And now, as we gear for Verirubin, from a lot of my conversations, it looks like they've secured maybe up to 20% of Verirubin capacity.Especially in the early days of when these chips are very scarce, that they're going to have a lead on all of this.Because people think that they can stand up this compute better.So I think what the last few weeks have actually shown is that Elon will take a shot at hitting the frontier.But if for whatever reason, they have over-procured some capacity, this is a very scarce asset.That they have shown that they can monetize at actually best-in-class margins and payback periods.The irony is, you and I have been doing this long enough to know, that's why Bezos built AWS.He had to build capacity for Black Friday.But then the rest of the year, he sat on all this capacity they had to build.And he figured out a really incredible way to monetize this.

我会说，过去几周证明的是：马斯克和他的团队能把这么多算力拉起来。其实你只要回看一年半前，他们在「谁能先把算力拉起来」这场竞赛里是落后的，他们没有那么多 H100。后来他们上了 Colossus，接着又上了规模远超所有人的 Colossus 2。

而现在，随着我们迈向 Vera Rubin（英伟达下一代平台），从我做的很多访谈来看，他们似乎已经锁定了最多约 20% 的 Vera Rubin 产能。尤其在这些芯片极度稀缺的早期，他们会在这上面拥有领先。因为大家认为他们能把这些算力拉得更好。

所以我觉得过去几周真正显示出来的是：马斯克会去搏一把冲击前沿；但万一出于什么原因他们采购过量了，这也是一项极度稀缺的资产——而他们已经证明，他们能以实实在在的行业最佳毛利和回本周期把它变现。

讽刺的是，你我干这行够久了，会知道——这正是贝索斯建 AWS 的原因。他必须为黑色星期五准备容量，但一年剩下的时间，那些不得不建的容量就闲在那儿。然后他想出了一个极其精彩的变现方式。

[33:52]

And by the way, investors at the time, 2009, 2010, when he was building out the capability around AWS, hated it.Because he was consuming all that free cash flow.Meanwhile, he was digging the biggest gold mine in the history of the world.One of the biggest.Among them.Among them.At the time was probably the biggest.Yeah, Google search might want to have a discussion.By the way, I do think it is important.Grok 4.3.I think the cursor, if they acquire it, that may end up being very important.But Grok 4.3 was on the Pareto frontier.And as of 10 or 12 days ago, these things move fast.Most intelligent 500 billion parameter model in the world.And they were on the frontier.And there are four companies on the frontier.XAI.SpaceX AI.Google won with Gemini 3.1 Pro.And then the rest of it was dominated by Anthropica and OpenAI.But they were on the Pareto frontier.Now we'll see what they do with Cursor.I want to come back to that in a second.By the way, man, I want to ask you some questions.What do you think?

顺便说，当时的投资者——2009、2010 年他在建 AWS 那套能力的时候——是讨厌这件事的。因为他把自由现金流全烧进去了。与此同时，他其实是在挖世界历史上最大的一座金矿。之一。之一。当时可能确实是最大的。是啊，谷歌搜索可能有话要说。

顺便，我确实觉得有一点很重要。Grok 4.3。我觉得 Cursor，如果他们真把它收进来，最后可能会非常重要。但 Grok 4.3 本身就在帕累托前沿上。而且就在十天、十二天前——这些东西变得太快了——它是世界上最聪明的 5000 亿参数模型。他们当时就在前沿上。而前沿上有四家公司：xAI、SpaceX AI、谷歌靠 Gemini 3.1 Pro 拿下一段，剩下的部分被 Anthropic 和 OpenAI 主宰。但他们确实在帕累托前沿上。

现在就看他们拿 Cursor 做出什么了。我一会儿想回到这个点上。顺便，老兄，我想问你几个问题。你怎么看？

[34:51] Brad Gerstner

So you think the biggest source of potential upside is the model.Yes.What do you think?I think that's the thing that's least talked about.Least talked about.Right?And so, listen.When I look at the bull bear case on the IPO, right?

所以你认为最大的潜在上行来源是模型。是的。你怎么看？我觉得那是最少被谈论的。最少被谈论。对吧？

那么，听着。当我看这次 IPO 的多空两面……

[35:05] Brad Gerstner

The bears are looking at last year's revenue.Say it was $18 billion.And they're looking at the forecast from the banks of $160 billion, you know, three years from now.And they're saying, listen.Not many companies in the history of the world have basically 8x their revenue over three to four years.Right?

空方看的是去年的收入。就算是 180 亿美元。然后他们看投行给的三年后 1600 亿美元的预测，然后说：听着，人类历史上没几家公司能在三四年里把收入翻 8 倍。对吧？

[35:20] Brad Gerstner

So that's where, you know, I think people get nervous about the valuation.When I look at this, again, when you break it down as an analyst's first principles part by part, which is what I tried to do here.Right?

所以我觉得，这就是大家对估值紧张的地方。而我看这件事的方式，是像分析师那样从第一性原理一块一块拆开——这也正是我在这儿试着做的。对吧？

[35:33] Brad Gerstner

When you look at Starlink, it looks totally doable.When I look at what they're building in AI compute terrestrially, looks totally doable over the course of the next three years.When I look at the model itself after the acquisition of Cursor, you know, combining those things around the compute they have, that looks to me like it could be an upside surprise.So I would say that I think that, you know, in the IPO, but I think when you look back three years from now, there's a decent chance that everybody's like, oh, my God, that was super obvious.Right?

看 Starlink，完全做得到。看他们在地面 AI 算力上建的东西，未来三年完全做得到。看收购 Cursor 之后的模型本身，把这些东西和他们手上的算力结合起来，在我看来这可能是一个上行意外。

所以我会说，在这次 IPO 里……我觉得当你三年后回头看，有相当的概率大家会说：天哪，这事当初也太显然了。对吧？

[36:01] Brad Gerstner

Even though today, all of these things have risk associated.Back to where we started.I'm not, you know, none of us are here to pump the IPO at 1.77 trillion.It's really to just break it down as we do inside our shop and to say, what is that distribution of future probabilities?

尽管在今天，所有这些事情都是带风险的。

回到我们开头说的：我们谁都不是来给这个 1.77 万亿美元的 IPO 抬轿子的。这只是像我们在自己公司内部做的那样，把它拆开来问：未来概率的分布是什么样的？

[36:16] Brad Gerstner

What's the probability that is higher from here?What's the prob?And I think we're all pretty AI pilled.And if you're AI pilled, that means we got to build a lot more compute than the world thinks.And that these models are going to be a lot more valuable than people think.You combine that with their core business, I don't know another entrepreneur or another business that's a better bet on the future, right, than SpaceX.And so I think for most institutional investors, it's a must buy, a must own.It's set it and forget it, right, in order to have a real bet on both the space and the AI future.From your lips to God's ears.I mean, listen, again, I think that you're going to have to wait.But, you know, we had this chart last week, right, that came out everybody was sending around Twitter, conveniently timed.And, you know, it's like shows the average max drawdown post IPO for like 20 companies from Facebook, Twitter, Alibaba, Shopify is, you know, over 50%.And so maybe that, again, we'll end this section here.You know, Gavin, you and I have been doing this a long time.We know it's going to be bouncy around the IPO.You know, how do you as a manager try to manage that?

从这儿往上的概率有多大？往下的概率有多大？

我觉得我们几个基本都是 AI 信徒。而如果你信 AI，那就意味着我们必须建出比这个世界所以为的多得多的算力，而且这些模型会比人们想象的值钱得多。把这一点跟它的核心业务叠在一起——我想不出还有哪个创业者、哪门生意，是比 SpaceX 更好的「押注未来」的标的。所以我认为，对绝大多数机构投资者来说，这是必买、必持的一只股票。买了就放着不动，就为了在太空和 AI 这两个未来上真正下一注。

借你吉言。

我是说，听着，我还是觉得你得等一等。上周有那张图，大家都在推特上转，时机还挺凑巧——它显示 Facebook、Twitter、阿里巴巴、Shopify 这 20 家公司 IPO 之后的平均最大回撤是超过 50% 的。

所以也许……我们就在这一段收尾吧。Gavin，你和我干这行很久了，我们都知道 IPO 前后会很颠簸。作为一个基金经理，你怎么应对这种颠簸？

[37:26] Brad Gerstner

Do you try to trade around the IPO?Do you set it kind of and forget it?I would say from an altimeter perspective, what we tend to do is we take a base position that we set and forget, right?And then we may size up or size down depending upon how the market reacts in, you know, in a particular moment.But any thoughts on this chart or, you know, how people, you guys are thinking about it in particular?

你会围着 IPO 做交易吗？还是买了就放着不动？从 Altimeter 的角度，我们倾向于先建一个「买了就放着」的底仓，对吧？然后我们可能会根据市场在某个特定时刻的反应加仓或减仓。但对这张图、或者对你们具体怎么想的，有什么看法？

[37:52] Gavin Baker

You obviously own a lot going into it.First, agree with absolutely everything you said.And I actually think about it the same way.Set it and forget it.You've talked about you have ballast.You move around and you move the ballast to one side of the ship.When you want the ship to lean into the wind to go faster and you move it to the other side, we don't want the ship to tip over.I think that's a great analogy.Think about all important companies in the portfolio the same way.So 100% agree.I mean, this chart is a bummer.What I would say is, you know, there's data on IPOs.But what I would just say is this is a really unprecedented situation.We've never had an IPO this big.We've never had an IPO that's going to go into an index this quickly.We simply do not know how much selling there will be from investors.I would hazard a guess.I mean, I don't know.But Elon, I don't think he needs liquidity.And I think he owns, what does he own, Foxy?

你显然是带着很大的仓位进去的。

首先，你说的每一句我都完全同意。而且我想的方式跟你一模一样：买了就放着。你讲过「压舱物」这个说法——你把压舱物挪到船的一侧，是想让船顶着风倾斜、跑得更快；然后再挪到另一侧，是不想让船翻掉。我觉得这个比喻很好。对组合里所有重要公司，我都是这么想的。所以百分之百同意。

我是说，这张图挺让人扫兴的。我想说的是，IPO 是有历史数据的。但我要说的是，这是一个真正前所未有的情况。我们从来没有过这么大的 IPO。我们从来没有过一个会这么快被纳入指数的 IPO。我们完全不知道会有多少投资者卖出。

我可以斗胆猜一下。我不知道，但马斯克，我觉得他不需要流动性。而且他持有……他持有多少，Foxy？

[38:50] Gavin Baker

50%-ish.50% of the company.And by the way, he's locked up for 365 days.Yes, exactly.So we know he's not selling.So I just think it's an unprecedented situation.And the right answer is, I don't know what's going to happen in the short term.And the right answer that I would just, you know, encourage every investor making their own decisionis to just think exactly the way you articulated it.We have these different levers.We have these different variables.Think about each one of them from first principles.Make your own decision.Do your own due diligence.Be thoughtful.But there are a lot of variables here.And then it is a little funny to me that, you know, it was 100 times trailing TTM revenue.Well, after the deals they signed, I think it's at 39 times.That can change fast.So they added $29 billion in a month.Yes.No.By the way, have you ever seen that happen?

50% 左右。公司的 50%。顺便说，他有 365 天的锁定期。是的，没错。所以我们知道他不会卖。

所以我只是觉得这是一个前所未有的局面。正确的答案是：我不知道短期内会发生什么。而我想鼓励每一位自己做决定的投资者的正确答案，就是完全按你刚才表述的方式去想：我们有这些不同的杠杆，我们有这些不同的变量，从第一性原理逐个去想每一个，自己做决定，自己做尽调，认真思考。但这里的变量真的很多。

还有一点我觉得挺好笑的：这曾经是过去十二个月（TTM）收入的 100 倍。而在他们签了那些交易之后，我觉得现在是 39 倍。这个东西可以变得非常快。所以他们一个月里加了 290 亿美元。是的。是啊。顺便说，你见过这种事吗？

[39:39] Brad Gerstner

Never.Never.And, you know, it just goes to show, first, Elon is not only a great engineer.He and Gwen and the team are great at business.And Brett.They understand what needs to be done to raise the capital, to get to the next phase.They have a long-term mission in the business.And so to me, again, what we saw in the course of the last few weeks with Cursor, what we saw with these deals that they cut,I don't know that any of the Mag 7 could have moved that quickly to adjust the business that they did.It's exceptionally entrepreneurial at scale, which we very rarely see in businesses.Two other things I would just say.Can I give you a hug, Brad?

从来没有。从来没有。这件事本身就说明：首先，马斯克不只是个伟大的工程师，他和 Gwynne（葛温妮·肖特维尔）以及团队做生意也非常厉害。还有 Bret。他们明白要走到下一个阶段、要把资本募到位，必须做什么。他们的生意里有一个长期使命。

所以对我来说，我们过去几周看到的 Cursor 那件事、看到他们谈成的那些交易——我不觉得七巨头（Mag 7）里有哪一家能这么快地把生意调整成那样。这是在超大规模上依然保持极强的创业精神，而这在大公司里非常罕见。

还有另外两件事我想说。我能给你个拥抱吗，Brad？

[40:18] Brad Gerstner

Two other things I would just say.Number one is people talk a lot about the total amount of capital being raised.If you add up the capital here, right, for Anthropic, what they may raise, what OpenAI may raise, what, you know, SpaceX may raise.Let's call it $250 billion.That's 1% of the Mag 7.Okay, it's 1% of the Mag 7.Yeah.And we will as well.You know, like that to me is like a bet on the future that we all believe in.And so if I said, where are we out of consensus?

还有另外两件事我想说。第一件是，大家很爱讨论正在被募集的资本总量。如果你把这里的资本加起来——Anthropic 可能募的、OpenAI 可能募的、SpaceX 可能募的——就算 2500 亿美元。那是七巨头市值的 1%。好吧，是七巨头的 1%。对。我们也会参与。

对我来说，这就是一个我们都相信的、对未来的押注。所以如果你问我，我们在哪里不在共识里？

[40:48] Brad Gerstner

What is our variant perception?We actually think it's going to be bigger, faster.And we've thought that for a couple of years.So first, it's only 1% of the Mag 7 market cap.And then you referenced it, the amount of selling.I've got a chart we'll post here.This is, you know, the Dribbble's share release for SpaceX shareholders.You know, so there's not a lot that can be released up until after the first earnings.We saw this in the Cerebris IPO.There's a version of it here in this IPO.And so, again, I think the banks have been thoughtful here, knowing that this is a very large IPO.And I'm not saying that won't trade down.Like, there's a possibility, you know, these things trade down.But again, for me, telescope out.Is there any company better positioned as a bet on the future?

我们的异见认知是什么？我们其实认为它会更大、更快。而且我们已经这么认为好几年了。

所以第一，它只占七巨头市值的 1%。第二，你刚提到的卖压。我这儿有一张图，我们会贴出来。这是 SpaceX 股东的股份「涓滴式」解禁安排。在第一次财报之前，能被解禁出来的量并不大。我们在 Cerebras 的 IPO 上见过这一幕，这次 IPO 里也有一个类似的版本。

所以我还是觉得投行在这件事上是想过的，他们知道这是一次非常大的 IPO。我不是说它就一定不会跌破发行价——这种事完全有可能。但对我来说，把镜头拉远：还有哪家公司比它更适合作为「押注未来」的标的？

[41:33]

I think what they've shown over the course of the last five weeks, they're probably number one.But let's move on.No, no, can I just say one thing about the employees?I think another thing that's unprecedented here is the employees and, to a large degree, the investors here, have had liquidity every six months for like the last 10 years.So if you're a SpaceX employee or former employee and you wanted to sell, you've had, whatever that is, close to 20 chances.And it is a matter of historical record that large investors have been able to sell.So I would think a lot of the people.Great point.They've chosen to own it.Now there's a new valuation.We'll see what they do.But just this is utterly unprecedented.And we'll see.Yeah, no, it's a great point.We've, in fact, called these companies quasi-public.You and I both know that SpaceX, and I put Anthropic in this category as well, Databricks in this category.These things, in many ways, have been more liquid over the course of the past three years than some public biotech companies we know.Absolutely.Right?

我觉得从过去五周他们展示的东西看，他们大概是第一。不过我们往下走吧。

不不，我能就员工这件事再说一句吗？我觉得这里还有一件前所未有的事：这里的员工，以及很大程度上这里的投资人，在过去大概十年里每六个月就有一次流动性机会。所以如果你是 SpaceX 的员工或前员工，你想卖，你已经有过差不多 20 次机会了。而且大额投资者能够卖出，这是有历史记录的事实。所以我会觉得，很多人……很好的一点。他们是主动选择了继续持有。当然现在有了一个新的估值，我们看看他们会怎么做。但这真的是前所未有。我们走着瞧。

对，这是很好的一点。我们其实把这类公司叫做「准上市公司（quasi-public）」。你我都知道，SpaceX——我也把 Anthropic、Databricks 放进这一类——这些东西在过去三年里，很多方面比我们知道的一些上市生物科技公司还要有流动性。绝对是。对吧？

[42:35] Brad Gerstner

And so there's a continuum of liquidity here.We treat it as a binary, private versus public.But it's really about this continuum.You know, let's keep going on models.You know, Anthropic Launch Fable 5, which you referenced yesterday, which is basically mythos with some classifiers and safeguards around cyber and biology, chemistry, and distillation.When those things get triggered, it fails back to Opus 4.8.You know, there was a Kaparthi tweet about this yesterday.He said, you know, it's SOTA on all the benchmarks.But what really makes it special is long-running tasks.Okay?

所以这里其实是一条流动性的连续光谱。我们习惯把它当成二元的——私有 vs 公开。但真实情况是一条连续谱。

我们继续聊模型吧。Anthropic 发布了 Fable 5，你昨天提到过——它基本上就是 Mythos，外面套了一些分类器和针对网络安全、生物、化学以及蒸馏方面的防护。当这些防护被触发时，它会回落到 Opus 4.8。

昨天 Karpathy 发了一条推。他说它在所有基准上都是最优（SOTA）。但真正让它特别的，是长时运行任务（long-running tasks）。对吧？

[43:13] Brad Gerstner

You retweeted our good friend, you know, Noam Brown.You know, ChatGPT 5.5 also exhibited these capabilities.You know, it led Noam, right, to suggest that it's not very relevant to do these snapshot benchmarks anymore.Like, the x-axis has to be time or tokens or compute.Because we can solve most problems now if we just let these frontier models for a very long point in time.So, Gavin, what is this new class of model, right?

你转发了我们的好朋友 Noam Brown。ChatGPT 5.5 也展现出了这些能力。这让 Noam 提出：再去做这种「快照式」的基准测试已经没什么意义了——横轴必须换成时间、token 或者算力。因为现在只要我们让这些前沿模型跑足够长的时间，大部分问题都能解出来。

所以 Gavin，这一类新的模型是什么？

[43:44] Gavin Baker

Fable 5, ChatGPT 5.5.What does it mean for the race in superintelligence?Who's up?Who's down?Who's still on the frontier?Give us your thoughts.I mean, it's hard to say that Anthropic's not up.Yeah.Like, after the revenue numbers they put up, after the Fable 5 release, and Mythos is evidently even better.But I just think that Noam Brown post from yesterday, polynomial, is so profound.Yeah.And just the idea that we do not know how smart these models are.Okay.And we made...Say more about that.Why don't we know how smart they are?

Fable 5、ChatGPT 5.5。它对这场超级智能竞赛意味着什么？谁在上升？谁在下滑？谁还在前沿上？说说你的看法。

我是说，很难讲 Anthropic 不是在上升。对。看他们交出来的收入数字、看 Fable 5 的发布，而且 Mythos 显然还更好。

但我就是觉得，Noam Brown 昨天那条关于多项式（polynomial）的帖子太深刻了。对。就是这个想法：我们并不知道这些模型到底有多聪明。好吧。而且我们……

展开说说。为什么我们不知道它们有多聪明？

[44:22] Gavin Baker

Because nobody has run Mythos for a year continuously.And we may never know how smart each generation of models actually is or was.But because we don't have time to appropriately evaluate their intelligence before the next model comes out.I mean, this is a profound statement.And just imagine, okay?

因为没有人把 Mythos 连续跑过一年。而我们可能永远都不会知道，每一代模型实际上到底有多聪明、或者曾经有多聪明。因为在下一个模型出来之前，我们根本没有时间去恰当地评估它的智能。

这是一个很深刻的判断。你想象一下，好吗？

[44:42] Gavin Baker

So I always say, like, when you think about FSD, just imagine a human being who never gets distracted, never gets tired, never talks on the phone in the car, never drinks and drives, never yells at their kids, never has to go to the back seat to give their baby a bottle.And, like, of course you would think that over time that is superior to humans who are distracted.I don't know how long...How long can you think deeply about one topic, Brad?

我老爱这么说：想想全自动驾驶（FSD）。想象一个人类司机，他从不分心、从不疲劳、开车时从不打电话、从不酒驾、从不冲孩子吼、从不为了给宝宝喂奶瓶而回头往后座伸手。那你当然会觉得，随着时间推移，他会比那些会分心的人类更强。

我不知道你能……Brad，你能对一个题目连续深度思考多久？

[45:08] Gavin Baker

It can be an hour.It can be an hour.It's like an hour.Oh, man.Yeah.That makes me feel terrible.Because I think I can think deeply about one topic continuously before having a stray thought enter my mind for, like, maybe five minutes.Now I can come back to that.Imagine if Albert Einstein had been able, instead of, you know, maybe he could think for three hours at a time.Right, exactly.Clearly an exceptional intellect.Yes.But imagine Albert Einstein had just thought about fundamental physics 24 hours a day.Yeah.He doesn't have to eat.He doesn't have to sleep.He doesn't have to relax.He doesn't drink.And never gets old.Never gets old.Never has diminutive intelligence.And he thought for one year.I mean, we might already, you know, have solved a lot of these intractable problems.So I just think that's an extraordinary thought.And just my takeaway was, however bullish I was on compute before then, I'm just a lot more bullish.Right, right, right.So that is a, you know, we saw when, that was probably what really unlocked Opus 4.6.It was the first really long-running model that could maintain that context, maintain that memory, solve some of these longer-running problems.Right.For us, the signal was in January.

可以有一个小时。可以一小时。差不多一小时。

哦，兄弟。那我可太难受了。因为我觉得我对一个题目连续深度思考、直到有杂念冒出来之前，大概只有五分钟。当然我可以再回到那个题目上。

想象一下，如果爱因斯坦能够——他也许一次能想三个小时。对，没错。显然是超凡的智力。是的。但想象一下爱因斯坦一天 24 小时都在想基础物理。对。他不用吃饭。不用睡觉。不用放松。不喝酒。而且永远不会变老。永远不老。智力永远不会衰退。然后他这样想了一年。

我是说，我们可能早就已经解决了很多难以攻克的问题。所以我觉得这是一个了不起的念头。而我的结论就是：不管我在这之前对算力有多看多，我现在都更看多了。

对对对。所以这就是——我们当时看到的，那大概正是真正解锁了 Opus 4.6 的东西。它是第一个真正能长时运行、能维持上下文、维持记忆、解决那些更长周期问题的模型。对。对我们来说，信号是在一月份出现的。

[46:26] Brad Gerstner

We knew, we felt like that was a big moment.But then when you started to see the revenue go up, we knew that lots of people were voting independently, that that was a profound moment, that they became much, much more useful.So, but one of the things that, the consensus going into this year, right, so the big question going into this year was, was the AI revenue going to show up?

我们当时就知道，我们感觉那是一个大时刻。但当你开始看到收入往上走，我们才知道有大量的人在独立地投票，证明那确实是一个深刻的时刻——这些模型变得有用得多得多了。

不过，进入今年时的共识是……今年最大的问题是：AI 的收入会不会真的出现？

[46:50] Brad Gerstner

Were we going to get to these thresholds of intelligence that caused enterprises and consumers to use them more?And I think the consensus at the time, at least on this podcast, the debate with Bill, was that open source models, cheap tokens, were catching up on the frontier.That perhaps these models were beginning to asymptote.That people wouldn't really pay for premium tokens.And it seems to me that the evidence on the field six months into the year is just the opposite, right?

我们会不会达到那些让企业和消费者更多使用它的智能门槛？我觉得当时的共识——至少在这档播客上、在跟 Bill 的辩论里——是：开源模型、便宜的 token 正在追上前沿；这些模型也许开始进入渐近平台期了；人们不会真的为高价 token 付钱。

而在我看来，今年过了六个月，场上的证据恰恰相反，对吧？

[47:23] Brad Gerstner

That frontier tokens are capturing the vast majority of all the revenues.And that, in fact, if you believe in the long-running capabilities and more compute allows you to do that, they may actually be extending their lead, right, on some of these models that were built on distillation.So I just opened it up to anyone around the table.What are your thoughts on whether or not, you know, have we challenged this thesis that cheap open source tokens are going to always, you know, close the gap on these frontier models?

前沿 token 正在拿走绝大部分收入。而且事实上，如果你相信长时运行能力、相信更多算力能让你做到这一点，那它们可能实际上正在拉开领先——相对那些靠蒸馏（distillation）做出来的模型。

所以我把这个问题抛给在座各位：我们是不是已经挑战了「便宜的开源 token 总能追平前沿模型」这个论点？

[47:53] Clark Tang

Or are they extending their leads?I think this debate, like this same debate has existed since the beginning of, since we started training these models to begin with, which was, hey, we're always kind of three, six months behind the frontier.But empirically, like you can just see all of the revenue has actually just accrued at the frontier.And I think that's because every time we release the frontier, a whole new, like slew of use cases that previously we could have never tackled before.Like coding.But also just, you know, we've just been locked at our desks for the last day just, you know, hammering Claude because, you know, it's just fascinating the things that now we can do with Fable 5 that we just couldn't do with Opus 4.8 just a day before.So what are some of those things, man?

还是说前沿正在扩大领先？

我觉得这场辩论，从我们开始训练这些模型的第一天起就一直存在：嘿，我们永远落后前沿三到六个月。但从经验事实看，你只要看就知道，所有收入其实都归到了前沿。

我觉得这是因为，每次我们把前沿往前推一次，就会有一整批以前根本没法碰的新用例冒出来。比如编程。但也不只是编程——我们这一整天就锁在办公桌前疯狂用 Claude，因为现在能用 Fable 5 做到的事情，一天前用 Opus 4.8 还做不到，实在太迷人了。

那都有哪些事，老兄？

[48:42] Clark Tang

I'm curious.So I think it's really, really good at multi-agent orchestration now.So Anthropic released a blog post about like different agent, six different agent like orchestration patterns that, you know, they've talked about.But really, like once you start being able to manage all these agents, the harness and the model itself is being RL'd with one another.They're actually being, you know, fused closer and closer together.But the model can understand the, you know, the extent of your work.So, you know, one of the things, for instance, is I just threw in like seven of our models and just said, okay, like I want to create a master view of like my beliefs given all of these assumptions of all these companies, TSMC capacity, like, and then produce me a report on all this stuff.And, you know, the model is able to reason through all of our assumptions.Like, actually, if you believe this, this thing is inconsistent with this.What are the contradictions?

我挺好奇的。

我觉得它现在在多智能体编排（multi-agent orchestration）上真的非常非常强。Anthropic 发了一篇博客，讲了六种不同的 agent 编排模式。但真正的关键是，一旦你开始能同时管理这么多 agent，脚手架（harness）和模型本身是被放在一起做强化学习的，它们其实在被越来越紧密地融合到一起。而且模型能理解你这份工作的全貌。

举个例子：我把我们的七个财务模型一股脑扔进去，然后说，好，我要基于所有这些公司的假设、TSMC 的产能等等，做一个「我的信念总览」，然后给我出一份报告。而这个模型能够把我们所有的假设推理一遍：其实，如果你相信这一条，那这一条就跟那一条不自洽。矛盾在哪里？

[49:42]

Yeah.It was fascinating.It's really interesting.And, you know, before we never do that.But now, you know, I think we're just step one into multi-agent orchestration.We're going to do this even further.And that's one example.I've also dumped all my notes into it.And it's reasoned across all my notes from the last three years and said, you know, here are some of your ideas that were consistent.Here are like, you know, the sources that were actually the highest signal to what actually played out.You know, and then it was actually just super fascinating what you could do.And we've just blown through our limits.It's unlocking all this stuff.I mean, like, they gave examples yesterday in the release.Anthropic did, you know, 50 million line Ruby code base.That stripe that was, you know, refactored in a day versus many weeks with many people.You think about where this is impacting biology and life sciences just across the spectrum.And to me, it really gets back to this fundamental point.Number one, if you believe this to be true about long-running agents, then we're going to produce and consume more tokens in the future as far as the eye can see.So the world, this gets me back to, you know, TerraFab and Space Orbital and all this because we may, in fact, unlock real thresholds of intelligence.

对。太迷人了。真的很有意思。而以前我们根本不会去做这种事。但现在，我觉得我们才刚迈进多智能体编排的第一步，我们还会往前走得更远。这只是一个例子。

我还把我所有的笔记都倒了进去。它把我过去三年的所有笔记做了一遍推理，然后说：这里是你一直保持一致的一些观点；这里是那些对最终真实发生的事情信号最强的信息源。你能做的事情实在太让人着迷了。而我们已经把用量额度用爆了。它把这些东西全解锁了。

我是说，他们昨天在发布里就给了例子。Anthropic 讲了 Stripe 那个 5000 万行的 Ruby 代码库，一天之内完成了重构，而以前要很多人干很多周。你再想想这在生物学、生命科学上的影响，各行各业都是。

对我来说，这又回到那个根本判断：第一，如果你相信长时运行 agent 这件事是真的，那我们在可见的未来会生产和消耗越来越多的 token。所以这个世界——这又把我带回太瓦级晶圆产能、太空轨道这些话题——因为我们可能真的会解锁一些实实在在的智能门槛。

[50:56] Gavin Baker

But we're going to have to let these horses run for a long time in order to get there.Yeah, I'll just say two things can be true.The majority of economic value may continue to accrue to the frontier.And man, has it ever accrued to the frontier thus far?

但我们得让这些马跑很长很长的时间才能到那儿。

对，我就说两件事可以同时为真。大部分经济价值可能会继续归到前沿。而且到目前为止，它归到前沿的程度简直了。

[51:11] Gavin Baker

And for sure the first six months this year.But the majority of tokens consumed in the world may be open source.And they are today.Yes.And I think that this current state is likely to persist.Harvey had a great blog post that they put out on X.And they used, and it's just amazing how everything gets out of date like in five days, you know.But they use their own proprietary legal data to do reinforcement learning and supervise fine-tuning.With fireworks on an open source model.And then they used a router.And a router being something that picks which model you send which query to.And which model you use to check which model.And they got better outcomes than an Opus 4, either 4.7 or 4.8 at a lower cost.Yes.And I think that is the future.And the reality is they were still consuming a lot of Opus.But a majority of the tokens they were processing probably were in their own open source model.We hear the same thing.We did an enterprise survey that we'll post of 300 companies.Which ones were optimizing?

今年前六个月绝对是这样。但世界上被消耗的大部分 token，可能是开源的。而且今天就是。是的。我觉得目前这个状态很可能会持续。

Harvey（法律 AI 公司）在 X 上发过一篇很棒的博客。而且太神奇了，什么东西五天就过时。他们用自己专有的法律数据，在一个开源模型上、跟 Fireworks 一起做强化学习和监督微调。然后他们上了一个路由器（router）——路由器就是决定把哪个查询发给哪个模型、以及用哪个模型去检查哪个模型的东西。结果他们拿到了比 Opus 4（4.7 或者 4.8）更好的结果，而且成本更低。是的。

我认为那就是未来。而现实是，他们仍然在消耗大量的 Opus。但他们处理的 token 里，大多数很可能跑在他们自己的开源模型上。

我们听到的是一样的。我们做了一份 300 家公司的企业调研，会贴出来。哪些公司在做优化？

[52:16] Brad Gerstner

So these are folks who are kind of looking at model routing and saying we're going to send certain tokens over here.Which ones are thinking about optimizing?Which ones aren't optimizing yet?And then what is their expected use of frontier model tokens?

也就是那些在研究模型路由、说「我们要把某一类 token 发到这边去」的人。哪些公司在考虑优化？哪些还没开始优化？然后他们预期自己会用多少前沿模型的 token？

[52:28] Brad Gerstner

Right?And they're all expecting to consume a lot more even though they're already in the process of optimizing.Think of it in the context of JP Morgan.If they're doing some back of the house stuff, right?

对吧？结果是，尽管他们已经在优化的过程中了，他们全都预期自己会消耗多得多的 token。

拿摩根大通（JP Morgan）举例。如果他们在做一些后台的事，对吧？

[52:39] Brad Gerstner

On customer service or whatever.They may very well use an open source model.Now I think they're loathe to use Chinese open source models.So they're waiting on kind of US open source models to be able to really deliver the bang that they need.But my hunch is for these enterprises, a lot of that back of the house stuff will get rooted there.That will probably be a majority of the tokens.But I think the really high value stuff, you know, coding as an example, they don't want to write second tier code.I think the vast majority of that will continue to be on the frontier.You don't need Albert Einstein to book you a trip.You don't need Albert Einstein to do KYC.But this is the debate we had literally at this table two years ago.However, if you just look at the revenue curves, right?

比如客服之类的，他们完全可能用一个开源模型。当然我觉得他们很不愿意用中国的开源模型，所以他们在等美国的开源模型能真正给出他们需要的性能。

但我的直觉是，对这些企业来说，很多后台的事情会被路由到那边去，那大概率会占掉大多数 token。但我觉得真正高价值的东西——比如编程——他们不会想要二流的代码。我认为其中绝大部分会继续留在前沿上。

你不需要用爱因斯坦来帮你订一趟行程。你不需要用爱因斯坦来做身份核验（KYC）。

但这正是我们两年前就在这张桌子上辩过的。然而，你只要看看收入曲线，对吧？

[53:22]

What folks concluded when they said that, they said, therefore, the frontier models will not accrue most of the revenue.And what we're seeing right now, it's 90% of the revenue.That has been decisively wrong.Probably more than 90%.And it may continue to be decisively wrong.Frontier might be 90% of the economic value.Open source might be 80% of tokens.Something that I think is very important on open source is that, you know,I think there's this belief that it's bearish for AI.It's actually, it may be bearish for the frontier models.There's that bear case you talked about.It's actually really bullish for compute and hardware.Because if the frontier models are capturing less of the margin, then you're going to spend more on compute.So the better open source does, the better it is for compute providers.I will say, there is a very, I would say between spending time in the heart of, like, the West, Silicon Valley, and also spending time in Asia,there is, like, a very big, like, a deep-seated belief in one versus the other, which is, like, if you spend a lot of time here,it's, like, all closed source, cloud, every, all traffic is going to go, you know, by way of this direction.And then you spend time in Asia, you know, the overwhelming belief is that we're going to find the right model to the right workload,

当时说这话的人得出的结论是：因此，前沿模型不会拿到大部分收入。而我们现在看到的是，前沿拿走了 90% 的收入。那个结论已经被决定性地证伪了。可能还不止 90%。而且它可能会继续被决定性地证伪。前沿可能占 90% 的经济价值，开源可能占 80% 的 token。

关于开源，我觉得有一点非常重要：外界有一种看法认为开源对 AI 是利空。它其实——它可能对前沿模型公司是利空，就是你说的那个空头论点。但它对算力和硬件其实是非常大的利多。因为如果前沿模型拿走的毛利更少，那你就会在算力上花更多钱。所以开源做得越好，对算力提供方越好。

我要说，在西方核心的硅谷待过、又在亚洲待过之后，会发现两边有一种很深的、非此即彼的信念差异。如果你在这边（硅谷）待久了，那就是：全都闭源、全都上云，所有流量都会走这个方向。而你去亚洲待一阵，那边压倒性的信念是：我们会给对的工作负载找对的模型，

[54:41] Clark Tang

and we're not going to overspend.And I think, you know, I would say the next year is probably going to be the most indicative of which way this falls.Because I think the reason why closed source models have captured so much of the value is because the models actually get the intentionand actually carry through the work.And this was the first year where we actually had agents that actually carried out user intentionfrom just answering a chatbot request to actually producing useful work.Right.Now the level of this intelligent has scaled so rapidly, and we continue to push against, like, the most economically valuable tasks,which are coding and finance and all these, like, knowledge work tasks.But, like, for the long tail of tasks, if open source continues to maintain a six-month lag,we might actually see a lot more open source used for, you know, our everyday tasks.That we might actually...And that's basically Jensen's argument, right?

而且我们不会多花冤枉钱。

我觉得接下来这一年，很可能最能说明这件事最终倒向哪边。因为我认为，闭源模型之所以拿走了这么多价值，是因为这些模型真的能领会意图，并且真的能把活干完。而今年是第一次，我们真的有了能执行用户意图的 agent——从只是回答一个聊天请求，变成真的产出有用的工作成果。对。

现在这个智能水平扩张得如此之快，我们持续在冲击那些经济价值最高的任务：编程、金融，以及所有这些知识工作任务。但对于长尾任务，如果开源继续保持只落后六个月的节奏，我们可能真的会看到越来越多的日常任务用开源来做。我们可能真的会……

这基本上就是黄仁勋的论点，对吧？

[55:43] Brad Gerstner

Jensen's argument is you're going to have model routing, and we're just in a moment in timewhere the frontier models, game to the advantage, can do long-running tasks.The open source models couldn't do it very well, and so they're accruing all of the value.But as soon as the open source models can do the long-running tasks as well, which is not far away,that they, too, will grab a bunch of this revenue.Are you investors in reflection?

黄仁勋的论点是：你会走向模型路由；我们只是恰好处在一个时间点上，前沿模型凭着优势能做长时运行任务，而开源模型做不好，所以价值全归了前沿。但只要开源模型也能把长时运行任务做好——而那并不遥远——它们也会拿走一大块收入。

你们投了 Reflection 吗？

[56:06]

I'm not.Okay.Nor are we.But I'm very impressed by Misha and the team and what they're doing.I very much want a frontier open source U.S. lab to win.We know that, you know, I heard you say recently, and I believe it to be true,NVIDIA, any day that they really wanted to, right, they already have some great open source models.They could absolutely build a frontier open source model whenever they chose to do it.And so it's not a question in my mind as to whether or not the U.S. is going to have a frontier open source model.It's just a question about timing.And then, like, at that point in time, is the, you know, let's assume they get these long-running capabilities.Have the frontier labs now achieve something yet again that allows them to keep the stranglehold on the revenues?

我没有。好。我们也没有。但我对 Misha 和他的团队、以及他们在做的事非常佩服。我非常希望有一个前沿级的美国开源实验室能赢。

我们都知道——我最近听你说过，我也认为是真的——英伟达只要它真心想做，任何一天都可以。他们本来就已经有一些很好的开源模型了。他们绝对可以在任何他们选择的时候做出一个前沿级开源模型。

所以在我心里，问题不是美国会不会有一个前沿开源模型，而只是时间问题。然后到了那个时间点，假设它们也拿到了长时运行能力——那时候前沿实验室是不是又一次做出了某种新东西，让它们继续掐住收入的咽喉？

[56:52] Gavin Baker

Yeah, and I just think it's, if you're, wow, that's a cute ASIC you've built there.That is so cute.Right.How would you like open source to join the frontier?Right.How would you like that?How do you like the baffles?

对，我就觉得，那画面是：哇，你自己造的这块专用芯片（ASIC）挺可爱的呀。真是可爱。对。那你希不希望开源也加入前沿呢？对。你觉得怎么样？喜欢这个风门设计吗？

[57:06]

So, I mean, I'm not sure that's the explicit calculation, but I do think Jensen is...Say more.Just double-click on that for everybody at home.Yeah.If they were to put an open source model out there, how does that impact the ASIC landscape?

我是说，我不确定这就是他明确在算的账，但我确实觉得黄仁勋……

展开说说。给在家听的朋友们再点开一层。

对。如果他们真的放一个开源模型出来，这对专用芯片（ASIC）的格局有什么影响？

[57:18] Gavin Baker

Well, you might not have the revenue to fund that, the revenue or the margins to fund that ASIC.And I do think NVIDIA is highly likely to be the world's dominant provider of open source AI.And I do think Jensen will bring open source.You know, right now it's, whatever, six months behind the frontier.Yeah.We might see it creep closer and closer and closer.And I do think Jensen has a big business decision.I see this, you know, chart here.So let's, you know, chop it up about NVIDIA, as you say.But if all of his customers are going to compete with him, then why not compete with his customers?

那你可能就没有足够的收入、或者足够的毛利，去养得起那颗自研芯片了。

而且我确实认为英伟达极有可能成为全世界最主要的开源 AI 提供方。我也确实认为黄仁勋会把开源往前带。现在它大概落后前沿六个月。对。我们可能会看到它一点点、一点点地逼近。

我确实认为黄仁勋面临一个很大的商业决策。我看到这儿有张图。那我们就顺着你说的，聊聊英伟达。但如果他所有的客户都要来跟他竞争，那他为什么不去跟他的客户竞争呢？

[58:01]

And we have all these neoclouds.So that's a cloud computing business that can compete with all these cloud computing businesses.He has his own models that are really, really good.Nematron 3 or 3.1 was actually really, really cool from a compute efficiency perspective.And he's always careful to release small models so as to not tread on anthropic open AI, Google's toes.But I do think that is a choice he is making.And just, you know, if the economics change, I think NVIDIA can join the frontier and become one of the world's largest cloud computing companies much faster than people think.Interesting.Interesting.Clark, walk us through this chart.Yeah.So I think one of the takeaways from spending time in Taiwan was there is certainly a lot of excitement around the next wave of ASICs.But I think it's like a very clear moment now where NVIDIA, it used to be an argument of NVIDIA versus ASICs, one or the other, and, you know, total domination, one or the other.Now, I think it increasingly, every year, everyone assumed that NVIDIA was going to lose share dramatically on a revenue scale, on a gigawatt scale, on a unit scale.And actually, if you actually look at the last few years, you know, they've actually maintained their share very, very handsomely.

而且我们有这么多新型云。所以那是一门可以跟所有这些云计算生意竞争的云计算生意。他自己的模型也真的非常非常好。Nemotron 3 或者 3.1 从算力效率角度看真的很酷。而且他总是很小心地只放出小模型，以免踩到 Anthropic、OpenAI、谷歌的脚。但我确实认为那是他在做的一个选择。

所以，如果经济账变了，我认为英伟达能够加入前沿、并成为全世界最大的云计算公司之一，而且比人们想的快得多。有意思。有意思。

Clark，给我们讲讲这张图。

好。我觉得在台湾待那段时间的一个收获是：大家对下一波专用芯片（ASIC）确实非常兴奋。但我觉得现在很明显到了一个新阶段：以前的讨论是英伟达 vs 专用芯片，非此即彼、一方全面统治。而现在，每一年大家都假设英伟达会在收入口径、吉瓦口径、出货量口径上大幅丢份额。但你要真去看过去这几年，他们其实把份额守得非常非常漂亮。

[59:21] Clark Tang

Actually, if you accounted for the fact that Anthropic was not really using NVIDIA, they probably actually gained share against, if not for in 2526.So I think what was very interesting, though, was a new class of accelerators or ASICs, MediaTek with their new V8T versus, you know, Broadcom's V8I for TPUs, actually was a big topic of discussion.And, you know, I think for ASICs, the argument now is that more and more will look custom to the actual workload.And that is like one vector that people are moving in versus NVIDIA now has kind of shown itself as the predominant provider of compute to a lot of the world.And for, you know, internal workloads, perhaps it will go more and more custom and more and more down the stack.And I remember just, you know, one year ago when it was kind of a Broadcom or NVIDIA battle.It seems there's a lot more nuance now to, you know, what type of accelerators will fit which workloads and fit which customers and fit which business models.And, yeah, I thought that was a new topic.New realization, though.I think we all kind of shared this view for a long time.Yeah, I was just shocked.I mean, I'm out here.I did a board meeting with one of our companies.And just, you know, their biggest one thing they emphasized is we thought the world would be consuming less NVIDIA than it is.

事实上，如果把 Anthropic 基本没在用英伟达这一点也算进去，他们在 2025、2026 年可能其实是在涨份额的。

不过我觉得非常有意思的是，出现了一类新的加速器或专用芯片：联发科（MediaTek）的新 v8t，对上博通（Broadcom）给 TPU 做的 v8i，成了一个很大的讨论话题。

我觉得对专用芯片来说，现在的论点是：它们会越来越贴着具体工作负载去做定制。这是大家在走的一个方向。而英伟达现在已经把自己确立为全世界很大一部分算力的主要提供方。至于内部的自用工作负载，也许会越来越定制、越来越往下沉到栈的底层。

我还记得就在一年前，这还是一场博通 vs 英伟达的对决。而现在，关于「什么类型的加速器适配什么工作负载、适配什么客户、适配什么商业模式」，似乎有了多得多的层次和细节。对，我觉得这是一个新话题、一个新的认知。我觉得我们其实早就大致持这个看法了。

对，我当时就是很震惊。我出来这边跟我们一家公司开了个董事会。他们最强调的一件事就是：我们本来以为世界消耗的英伟达会比现在少。

[1:01:00] Brad Gerstner

And if anything, NVIDIA is accelerating and they just continue to out-execute their competitors.And I think a lot of people are indexing to this OpenAI gigawatt and, you know, NVIDIA has 10.Broadcom has 10.Who has six?

而事实上，英伟达在加速，他们持续在执行上碾压竞争对手。

我觉得很多人是拿 OpenAI 那个吉瓦分配表来对标的：英伟达 10 吉瓦，博通 10 吉瓦，谁是 6？

[1:01:18]

AMD.AMD has six and they have warrants.And then Cerebrus, our shared portfolio company, has a gigawatt.And I just, that is what's on paper.What actually gets deployed, let's see.I will be very surprised if, you know, that 10 out of 27.What's that math?

AMD。AMD 6 吉瓦，而且他们还拿了认股权证。然后 Cerebras——我们共同投资的组合公司——有 1 吉瓦。

我就想说，那只是纸面上的。实际会部署出来多少，我们走着瞧。如果最后真是 27 里面的 10，我会非常意外。那是多少来着？

[1:01:37] Gavin Baker

Let's see who's best at math.What percentage market share is that?30%.Yeah.I'll be very surprised if that is where they land.I think that is an extremely unlikely outcome.And especially as long as we are in a watt-constrained world.If you can get more tokens per watt, which is literally revenue, with NVIDIA than a lot of alternatives.Just if you build your factory with another chip, you may save some money, but you're going to have less revenue.And the margins may be lower.And that's a point that Jensen keeps hammering and I think is really important.And by the way, credit where credit is due.The most important, one of the most surprising things to me in this ASIC landscape, I would say Meta and Microsoft have been probably disappointing.Yes.You know who made a good ASIC?

看看谁数学好。那是多少市场份额？30%。对。

如果最后真落在那儿，我会非常意外。我觉得那是一个极不可能的结果。尤其是只要我们还处在一个受电力（瓦特）约束的世界里。如果用英伟达，你每瓦能拿到的 token 比很多替代方案更多——而每瓦 token 就等于收入。你要是用别家芯片建工厂，你可能省了点钱，但你的收入会更少，而且毛利可能更低。这是黄仁勋一直在反复敲的一个点，我觉得非常重要。

顺便，该给的信用要给。在专用芯片这个格局里，最让我意外的事情之一是：我会说 Meta 和微软大概是让人失望的。是的。你知道谁做出了一颗好芯片吗？

[1:02:23]

Yes.Well, I know you know.Yes.Jalapeno.Yeah, exactly.From OpenAI.Yeah.They made a great chip.Yes.Now, unfortunately, it needs to run at a much lower temperature than the NVIDIA GPUs, which means you need to spend more money on cooling and that consumes more power.They made a great chip.I mean, I think the question there and the question for everybody is going to be, is that the highest and best use of your time?

是的。你肯定知道。是的。Jalapeño。对，没错。OpenAI 的。对。他们做了一颗很棒的芯片。是的。

不过很遗憾，它需要在比英伟达 GPU 低得多的温度下运行，这意味着你得在冷却上多花钱，而冷却又要多耗电。

他们做了一颗很棒的芯片。但我觉得那里的问题、也是每个人都要面对的问题是：这是不是你时间的最高最优用途？

[1:02:43] Brad Gerstner

Right?Like, you know, I tend to think that the frontier companies, like, there's this belief that they've got to be vertically integrated.But if you believe, like I do, that the race to superintelligence, particularly as we get these recursive loops working, may be over in the next two to three years.Then I think focus, focus, focus, focus.You exist to build the best intelligence in the world and to deliver the best intelligence in the world.And that means you have to have all the revenue.Because if you want to build out the compute that's going to be required to continue to push the frontier, you have to have the revenue in order to support it.So I think that, you know, subject to the focus question, I think they certainly did.This all brings me back to kind of a reality check, though.You know, we just got done talking about test time compute, inference time compute, long-running agents.This is really the thing that's unlocked the revenue this year.It all pushes us in the direction of more CapEx.Google just raised $80 billion, right?

对吧？我倾向于认为，那些前沿公司——外界有一种信念，觉得它们必须纵向一体化。但如果你像我一样相信，通往超级智能的竞赛（尤其是当这些递归自我改进的回路跑通之后）可能在未来两三年内就见分晓，那我觉得答案就是：聚焦、聚焦、聚焦、聚焦。

你存在的意义是造出世界上最好的智能、并把世界上最好的智能交付出去。而这意味着你必须拿到全部的收入。因为如果你想建出继续推进前沿所需要的算力，你就必须有收入来支撑它。

所以我觉得，在「是否够聚焦」这个前提之下，他们那颗芯片确实做得不错。

不过这一切又把我拉回到一次现实检查。我们刚聊完测试时算力（test-time compute）、推理时算力、长时运行 agent——这些才是今年真正解锁收入的东西。而它们全都把我们往「更多资本开支（capex）」的方向推。谷歌刚刚募了 800 亿美元，对吧？

[1:03:39] Brad Gerstner

We've now taken the MAG-5 or MAG-7 free cash flow, you know, down dramatically 80% from just a few years ago.And Morgan Stanley, you've got this chart in front of you, up to their 2027 CapEx forecast from $950 billion to $1.1 trillion.I mean, we were talking about this with Jensen.That was his forecast two years ago.So, you know, obviously, this doesn't even include SpaceX, CoreWeave, etc.So I think the number on 2027 is likely closer to $1.5 trillion.And if we compare this to the total incremental inference revenue, so the thing that the market gets worried about, you know, back to my Sam Altman podcast, you know, in October of last year,can we really afford to spend $1.5 trillion of CapEx a year if we're only generating X amount in inference revenue?

我们现在已经把五巨头、七巨头的自由现金流，从几年前的水平大幅砍掉了 80%。而摩根士丹利——你面前有这张图——把 2027 年的资本开支预测从 9500 亿美元上调到了 1.1 万亿美元。

我们当时跟黄仁勋聊过，这正是他两年前的预测。而且显然，这甚至还没包括 SpaceX、CoreWeave 等等。所以我觉得 2027 年的实际数字很可能更接近 1.5 万亿美元。

如果我们把这个跟总的增量推理收入去比——这正是市场担心的地方，回到我去年十月跟 Sam Altman 那期播客——如果我们只创造出 X 的推理收入，我们真的负担得起一年 1.5 万亿美元的资本开支吗？

[1:04:28] Brad Gerstner

The thing I think that lit the fuse this year was Anthropics showed up in a major way with revenue, right?And so we have, you know, the AI lab revenue, everybody combined at around $300 billion next year, right?

我觉得今年真正点燃引信的，是 Anthropic 以一种非常显著的方式交出了收入，对吧？所以我们现在有——所有 AI 实验室加起来，明年收入大概在 3000 亿美元左右，对吧？

[1:04:43] Brad Gerstner

So can't, you know, roll that out to 2027, or that is 2027, $300 billion.So we're spending $1.5 trillion of CapEx on $300 billion of inference revenue.Does that math math for you?And what would cause you, you know, to get more nervous again about our ability to continue to make these investments?

所以，把这个推到 2027 年，或者说那就是 2027 年，3000 亿美元。那我们是在用 1.5 万亿美元的资本开支，去对 3000 亿美元的推理收入。这笔账在你那儿算得过来吗？还有什么会让你重新对「我们能否继续做这些投资」感到紧张？

[1:05:04]

Because the second we get nervous about it, the entire semi-complex is going to come down a lot.Well, what do you think the gross margins are on that $300 billion?Yeah, let's call it 50%.I would guess they're probably a little bit higher than that.I might say 60 or 70.But I mean, that math starts to math.And what I would just say is I think that $300 billion is low, man.Yeah.I just think it's low.From your mouth.Yeah, exactly.I think we end this year well over $200 billion in inference revenue, well over.And so I think the math really maths.And I do think we have to give Jensen, our friend, some credit because he said some thingsthat seemed outlandish.Right.And he was conservative.He was low.He said a trillion two years ago.And I mean, he was really low.Right.And so like, let's give the guy some credit and think about what he is saying right now.For sure.For sure.And listen, I would say consistently, Elon's been taking the over.Sundar's been taking the over.Sam, Dario.You know, Dario did the podcast with Dworkish when he was talking about country geniusesin a data center.He said that will be here by 2028.He said revenues will go into the low hundreds of billions by 2028.So let's call that, you know, three, 400 billion of revenue by 2028.

因为我们一旦对这件事紧张起来，整个半导体板块就会跌一大截。

那你觉得这 3000 亿美元的毛利率是多少？就算 50% 吧。我猜可能比这更高一点，我会说 60% 或 70%。但我的意思是，这笔账开始算得过来了。而我要说的是，这 3000 亿是低估了，老兄。对。我就是觉得低了。借你吉言。是啊，没错。

我认为今年年底我们的推理收入会远超 2000 亿美元，远超。所以我觉得这笔账真的算得过来。

而且我确实觉得，我们得给我们的朋友黄仁勋一些信用，因为他说过一些当时听起来很离谱的话。对。而他其实是保守的。他说低了。他两年前说的是一万亿。而他真的说低了。对。所以，给这位老兄一些信用，然后认真想想他现在正在说什么。

没错，没错。而且听着，我要说一贯以来：马斯克一直在赌「超过」，Sundar 一直在赌「超过」，Sam、Dario 也是。Dario 在 Dwarkesh 的播客上讲「数据中心里的天才国度」，他说那会在 2028 年到来，他说到 2028 年收入会进到几千亿美元的低位区间。那我们就算 2028 年三四千亿美元的收入。

[1:06:23] Brad Gerstner

And he said that a while ago now.So he may even be revising up his number.And he said, it's hard for me to see that there won't be trillions of dollars in revenuebefore 2030.And if you're on that revenue trajectory, if we're on a trajectory to 200 by the end ofthis year, let's call it four or 500 by next year and a path to trillion plus by2029, then the math, maths.And we got to keep in mind that half of the spending is there to, you know, for training,maybe a little less than half.What is it, Foxy?

而他说这话已经有一阵子了，所以他甚至可能正在上调他的数字。他还说过：在 2030 年之前，他很难想象收入不会达到数万亿美元。

如果你在这条收入轨迹上——如果我们今年年底能到 2000 亿，明年就算四五千亿，然后到 2029 年有一条通往一万亿以上的路径，那这笔账就算得过来。

而且我们要记住，其中有一半的支出是用来做训练的，也许略少于一半。是多少来着，Foxy？

[1:06:53]

It's probably, it depends on the lab, but I would say it's increasingly less than half.Yeah.Okay.So we'll call it 35% is spending that's not revenue generating, but is going to kind ofmake the next model.So I think the math, maths.Right.And there's still this prisoner's dilemma where if you opt out, that may be an existentialdecision.For sure.And I think like coming into this year, going back to this kind of what narratives were violated,you know, I think into this year, everyone expected token pricing, the price of compute,it's all deflationary and it will be kind of a smooth line deflationary over time.But I think this year what we've seen is the opposite.And, you know, it's all comes back to supply demand.The demand side of the equation seems to be far outstripping the supply.Right.And I think you look at the deals signed by SpaceX and others, the monetization ratesper watt are increasing.And look, that is on a pretty nascent small base of users, right?

大概……这取决于哪家实验室，但我会说它正在越来越低于一半。对。

好，那我们就算 35% 的支出是不产生收入的、是拿去做下一个模型的。所以我觉得这笔账算得过来。对。而且这里还有一个囚徒困境：如果你选择退出，那可能是一个生死攸关的决定。绝对是。

我觉得，回到「今年哪些叙事被打破了」这个话题：进入今年时，所有人都预期 token 价格、算力价格是通缩的，而且会是一条平滑的通缩曲线。但今年我们看到的恰恰相反。

而这一切都回到供需。需求端似乎在大幅超过供给端。对。你看 SpaceX 和其他公司签的那些交易，每瓦的变现率是在上升的。

而且要注意，这还是建立在一个相当初生、相当小的用户基数上，对吧？

[1:07:50]

Like Alex at Whale Rock has this great way to frame it.Less than 0.2% of people on earth are actually using AI in an agentic way.Right.Right.Like I'm not a technical person, but I'm consuming 500 CPU cores in a VM instance, fiveGPUs 24-7.I mean, if you draw that out to any meaningful percentage of the population, I mean, we'regoing to be in, you know, this kind of shortage environment maybe for some time.So I think that is all positive for this ROI question.Man, Foxy, 100 to 1 CPU to GPU ratio.What kind of a jitter workflow?

Whale Rock 的 Alex 有一个很好的框定方式：地球上真正在以 agent 的方式使用 AI 的人，还不到 0.2%。对，对。

我不是技术出身的人，但我现在一个虚拟机实例里就在跑 500 个 CPU 核心、5 块 GPU，7×24 小时不停。你把这个外推到人口中任何有意义的比例，我是说，我们大概会在这种短缺状态里待上相当长一段时间。所以我觉得这对那个投资回报（ROI）的问题来说全是正面的。

老兄，Foxy，100 比 1 的 CPU 对 GPU 比例。你这是什么 Jupyter 工作流啊？

[1:08:29] Clark Tang

Of course.You said course.Fine.Fine.Yes.I'm being smart with my spend.Good.Good.Good.Excellent.I will say also that ratio of 300 to 1 point, you know, call it 1.2, 1.5.There is also a rate that now physically we can only expand how much we can produce andhow much we can actually increase that spend by.Whereas we're seeing the opposite right now on the willingness to pay for these tokens.Right.And actually, like, when the willingness to pay for these, when the monetization per gigawattis actually increasing from, you know, call it like 20 billion in the best of cases forat the beginning of the year to now like 30 to even pushing 40.Per gigawatt.Per gigawatt.All of that is a very heavy fixed cost base.But all of that is like pure margin flow through now.Right.And you're actually, you know, as we scale like the willingness to pay for all of this.And now all of this stipulated by like, you know, everything we're talking about of likehow much is open source versus not and all of these different flows.But really, like as we're climbing this curve, you know, the revenue is might actually outstripour fixed cost base by a significant amount.And I think that's why all the labs are pushing, you know, the gas of the pedals because they

当然咯。——你说的是「核（cores）」。行吧，行吧。是的，我花钱花得很精明。好，好，好。太棒了。

我还想说，那个 300 比 1 的比例，就算 1.2、1.5 吧。还有一件事：现在物理上，我们能扩产的速度、能真正把这笔支出往上加的速度，是有天花板的。而在为这些 token 付费的意愿上，我们看到的却是相反的方向。对。

实际上，当为这些 token 付费的意愿——当每吉瓦的变现率从年初最好情况下的约 200 亿美元，涨到现在的 300 亿、甚至逼近 400 亿。每吉瓦。每吉瓦。

这背后是非常重的固定成本基础。但现在这些增量基本是纯毛利穿透下来的。对。而且随着我们把付费意愿整体扩上去——当然这一切都还受制于我们刚才聊的开源占比多少这些不同的流向——但真的，随着我们爬上这条曲线，收入可能会大幅超过我们的固定成本基础。

我觉得这就是为什么所有实验室都在把油门踩到底，因为他们都看到：如果这条曲线继续这样走，三年之内，

[1:09:54] Brad Gerstner

all see like within if we continue this curve within like three years, you know, we're justgoing to be so short on all the compute.Go ahead.This is a great, I'm sorry.But I mean, like, it's a great point.Like you thought you were getting, when you made these decisions in November of 2025, youthought you were getting a certain return.Yeah.You may be getting triple that return today.At Tropic, no way, no way did they think they were going to be anywhere close to breakeven.Yeah.Right.In this part of the curve.And the reason, like I called it accidental profitability that, you know, people have been talkingabout that because they want to spend a lot more money on compute.They've just had a hard time doing it.Now, maybe with SpaceX, you know, they could take some of those dollars and go spend themother places.But that to me is, you know, a fundamental change.The first argument against the frontier labs was they'll never generate revenue.Okay.And then that got blown up.Then it was like, even if they generate revenue, it'll be really shitty gross margins and they'llnever be able to make money.And then kind of that's blown up.And, you know, I think now, you know, people are falling back and they're saying, well,

我们的算力会短缺到不行。你说。

这是个很棒的——抱歉。我是说，这是个很棒的点。就像你在 2025 年 11 月做这些决策的时候，你以为你会拿到某个特定的回报率。对。而今天你拿到的可能是那个回报的三倍。

Anthropic 当初绝对、绝对没想到自己会离盈亏平衡这么近。对。在曲线的这个位置上。

而原因——我把它叫做「意外的盈利（accidental profitability）」，大家一直在谈这个——是因为他们本来想在算力上花多得多的钱，只是一直很难花出去。当然，现在有了 SpaceX，他们也许可以把其中一些钱拿到别的地方去花。

但对我来说，这是一个根本性的转变。反对前沿实验室的第一个论点是：它们永远赚不到收入。好，那个论点被炸掉了。然后变成：就算它们有收入，毛利也会烂得一塌糊涂，永远赚不到钱。这个论点差不多也被炸掉了。而现在，我觉得大家在往后退一步，说：好吧，

[1:10:59] Brad Gerstner

they're overcharging.This is token maxing.My good friend, you know, Tamatha said there's no ROI on any of this spend.It's all this token maxing.I mean, my best evidence for why we all know, of course, when somebody puts on this muchspend like at Altimeter, we're not optimally spending every single dollar.But the question is, why are millions of independent businesses, small, medium, and large, why aremillions of consumers all choosing to do the same thing?

它们是在超额收费。这是在「刷 token 量（token maxing）」。我的好朋友 Chamath 说，这些支出根本没有任何投资回报，全是在刷 token 量。

我能给出的最好的反证是——我们当然都知道，当一家机构像 Altimeter 这样投入这么多支出时，我们不可能每一块钱都花在最优处。但问题是：为什么成百上千万家彼此独立的企业——小的、中的、大的——为什么成百上千万的消费者，都在同时选择做同一件事？

[1:11:25] Brad Gerstner

They're not dumb.These are, you know, rational economic actors that are all simultaneously saying, I want todo this because it makes my life better.It makes my business better, et cetera.To me, that is the best evidence as to why I think this revenue can continue.Yeah.And Clark, I think like the point you made is dead on because, I mean, you want to ownasset heavy businesses and inflationary environments and token pricing is going up and supply demandis tightening.So totally agree.You know, as we begin to find our way to the exit ramp and wrap here, one of the things I,you know, you and I've been doing this for a long time, Gavin, a couple of decades.You may even sketch longer than me, even though I'm a little bit older than you.You know, we have, I always like to do a market check because I find a lot of time that analystscome on these things and they talk their, you know, talk their book.And, you know, there are a lot of people who listen to these things, retail investors andothers.And it's just kind of like, what do we really think?

他们不傻。这些都是理性的经济主体，他们在同时说：我想这么做，因为它让我的生活更好，让我的生意更好，等等。对我来说，这就是我认为这份收入能够持续下去的最好证据。

对。Clark，我觉得你说的那个点太到位了——因为在通胀环境里你想持有重资产的生意，而 token 定价在涨、供需在收紧。所以完全同意。

那么，在我们找到出口匝道、准备收尾之前——有件事，Gavin，你和我干这行很久了，几十年了。你入行可能比我还早，虽然我年纪比你稍微大一点。我总喜欢做一次「市场体检」，因为我发现很多时候分析师上这种节目，都是在给自己的持仓喊话。而听这些节目的人有很多，有散户投资者，也有别的人。所以就想问一句：我们到底真实地怎么想？

[1:12:22] Brad Gerstner

And so I always characterize as kind of small, medium and large.Like, what am I doing?Do I have small exposure on?Do I have medium exposure on?Do I have large exposure on?You know, and if you look at what's happened in the markets, semis ripped this year.I mean, like you've been doing this a long time.I don't, I've never seen it before.Right.I've never seen, you know, the doubles and the triples across the board like we saw,but there's been huge dispersion right in the market.Internet's down 16 percent, software's down 8 percent on the year, you know, SPY and NASDAQare up, but really up because of their components that are related to AI and compute.And so the market itself has kind of struggled.Meanwhile, if you were in the stuff that we were invested in, we've all done pretty well.I think, you know, I've said it a couple of times.I think if the anthropic revenue had not shown up this year because that was the overhang onthe market, I think the whole market could be down this year.Right.But that showed up.You know, we just had these huge months in April and May.For us, you know, because prices came up so much, because I have some worry about, youknow, geopolitics, the macro backdrop with, you know, with what's going on with inflation

我一直用「小、中、大」来刻画：我在干什么？我的敞口是小仓、中仓还是大仓？

你看今年市场发生的事：半导体今年一路狂飙。你干这行很久了，我从没见过这种场面。对。我从没见过这样全线翻倍、翻三倍。但市场的分化也非常大：互联网板块今年跌了 16%，软件跌了 8%，标普 ETF（SPY）和纳斯达克是涨的，但真正涨上去是因为其中跟 AI 和算力相关的成分股。所以市场本身其实过得挺挣扎的。

与此同时，如果你在我们投的那些东西里，我们大家都做得不错。

我说过好几次了：我觉得如果今年 Anthropic 的收入没有兑现——因为那是压在市场上的那块石头——我觉得整个市场今年可能是跌的。对。但它兑现了。我们刚刚经历了四月和五月这两个大月。

对我们来说，因为价格涨了这么多，因为我对地缘政治有些担忧，对宏观背景、对短期通胀在发生的事情有些担忧，

[1:13:31] Brad Gerstner

in the short run.And just like, you know, needing a little consolidation in this market to answer someof these questions because now expectations are higher.You know, we dialed back from what I would call large for altimeter to something kind oflike medium small.Again, it's never all or nothing for us.It's like, what is the risk reward at a given price?

而且市场也需要一点整固，来回答其中一些问题——因为现在预期更高了——所以我们从我称之为「大仓」的位置，回调到了「中偏小」的位置。对我们来说，从来不是全有或全无。而是问：在给定价格下，风险收益比是多少？

[1:13:51] Gavin Baker

And so we think this is a, you know, maybe going to be a period of consolidation on wayto much higher highs.Curious just how you run the book, how you think about it like a portfolio manager.Very similarly, man.I always think stocks, the markets, I imagine them as runners.Yeah.Okay.And like in 22, that rudder had gone downhill.It had a lot of energy.Painful.Painful.It was painful.It wasn't fun.Yeah.But coming out of that, there was a lot of kind of pent up upside in the market.And, you know, the market, particularly the last two months, it has run up a very steephill.And a lot of companies, semiconductor companies in particular, you know, ironically, you know,NVIDIA and Broadcom, they have been laggards.Total.And so, but a lot of these, like I do see a lot on X about finding the next bottleneck.I think that was the last game.Yes.Yes.That game is over.Yes.You've had a lot of stocks that forget climbing a mountain or a hill.Yes.They've gone straight up a cliff.Yes.Okay.Yes.They're tired.Yeah.They need to rest.Yeah.And we'll see.Do they just rest at the top of that cliff they climbed?

所以我们认为，这可能是一段整固期，之后会走向高得多的新高。

我很好奇你怎么管这个组合、你作为一个投资经理怎么想。

非常相似，老兄。我一直把股票、把市场想象成跑者。对。比如在 2022 年，那个跑者是在下坡。他积攒了很多势能。很痛苦。很痛苦。是很痛苦，不好玩。对。

但从那里出来之后，市场里积压了很多向上的动能。而市场，尤其是最近这两个月，它跑的是一段非常陡的上坡。而且很多公司，尤其是半导体公司——讽刺的是，英伟达和博通反而是落后的。完全是。

我在 X 上看到很多人在找「下一个瓶颈」。我觉得那是上一局的玩法。是的。是的。那一局已经结束了。是的。

现在你有一大堆股票，别说爬山、爬坡了，它们是直着爬上了一面悬崖。是的。好吧。是的。它们累了。对。它们需要休息。对。我们走着瞧：它们是就在爬上去的那面悬崖顶上休息，

[1:14:56]

Yeah.Do they hang out in their harness for a while?And we've seen the last week.We've seen, you know.We've seen some.Some retracement.Or do they need to go downhill for a bit?Yeah.We'll see.But I'm thinking very similarly to you.But it is.And I think there's, you know, the market is seasonal.Yes.I think there's real concerns around inflation and rates.What was CPI this morning?

对。还是挂在安全带上晃一阵？我们看到过去一周……我们看到了一些。一些。一些回撤。还是说它们需要往下走一段？对。我们走着瞧。

但我的想法跟你非常相似。而且我觉得市场是有季节性的。是的。我觉得围绕通胀和利率的担忧是真实的。今早的 CPI 是多少？

[1:15:16] Brad Gerstner

It was 4.2.I think we added core came in at like 0.2 versus 0.3.So a little bit better.But, you know, clearly we're above 4 again.Yeah.And there's short-term pressure on, you know, core PCE, et cetera.And we have some unknown unknowns.But the market, I mean, if I had told you the fact pattern for this year,that we're going to be in a war with Iran, that, you know, oil was going to be at $100,that CPI was going to be creeping back up, that internet was going to be down 15%,software was going to be down 8%, you would have said, I want nothing to do with that market.Right?

是 4.2。我记得核心通胀是 0.2，预期是 0.3，所以稍微好一点。但很明显我们又回到 4 以上了。对。而且核心 PCE 等等也有短期压力。我们还有一些未知的未知。

但市场——我是说，如果我提前告诉你今年的事实图景：我们会跟伊朗打仗，油价会到 100 美元，CPI 会重新往上爬，互联网板块会跌 15%，软件会跌 8%——你会说，这种市场我碰都不想碰。对吧？

[1:15:50]

Yeah.And here we are, the market's done pretty good in the stuff that we traffic in becausethe world underestimated AI revenues and underestimated the amount of compute that was going to beneeded.And so I'll just say, you know, we're heading into a seasonally weak period with all thesefears.Right.AI has actually been seasonal for the last three summers.That's interesting.Yeah.Token consumption has kind of plateaued, slowed down.And that's because, you know, college kids are big AI consumers and they don't use as muchAI.You know, hopefully they're all using it to learn and not cheat.But that may happen.It may not happen because of a gigantic AI.My 15-year-old is building swarms of agents, building a SpaceX model.He's going to the SpaceX IPO with me at the exchange on Friday.Awesome.But he had to build an AI model, using AI agents.He had to build a model, a DCF, before we go to the exchange.He is mesmerized.He is absolutely, and it's extraordinary what he's building.So he's one kid who's not using less compute in the summer.He's burning it.He's burning it.Yeah.But, you know, if token consumption plateaus, if open source takes some share, there'sa Silicon Data Index that has showed, which is an index of kind of consumption and pricing.

对。而结果我们在这儿，在我们涉足的那些领域里，市场表现得相当不错，因为这个世界低估了 AI 的收入，也低估了将会需要的算力总量。

所以我要说，我们正带着这一堆担忧，走进一个季节性偏弱的时段。对。

AI 其实已经连续三个夏天呈现季节性了。这挺有意思。对。token 消耗量会走平、放缓。因为大学生是 AI 的消费大户，而他们暑假不怎么用 AI。当然，希望他们都是用它来学习，不是用来作弊。

不过这次可能会发生，也可能不发生。我 15 岁的儿子在造 agent 集群，在建一个 SpaceX 的模型。他周五要跟我一起去交易所参加 SpaceX 的 IPO。太棒了。但我要求他在我们去交易所之前，先用 AI agent 建出一个模型，一个现金流折现模型（DCF）。他完全着迷了，他造出来的东西非常了不起。所以他就是那个暑假不会少用算力的孩子。他在猛烧。他在猛烧。对。

不过，如果 token 消耗走平，如果开源拿走一些份额——有一个 Silicon Data 指数，是追踪算力消耗和定价的，

[1:16:57]

I think there may have been a little bit of a shift over the last two weeks to open sourcetokens that are cheaper.I think people looking at that data as bearish or not understanding it.But nonetheless, like, I just think there's reasons, you know, to look around, be careful,be thoughtful.I always assume a bullet is coming for me, head on a swivel.It's the bullet you don't see that gets you.So I'm trying to spin as fast as I can.But yeah, it's, the market may need to take a breather.But man, when I think about what Noam Brown said, and when I see the capabilities of Fable,it's just hard for me to get too bearish.I mean, like, to me, we got two, I think, of the most extraordinary guys of, you know,the next generation, you know, sitting in the room.We have, at Altimeter, we have deep admiration for the work that you guys do.I always appreciate when you send me a note about the work that we do and we publish.But for the guys who are newer to the business, they might think this is the way that it kindof always was, right?

我觉得过去两周可能确实有一点点向更便宜的开源 token 的转移。我觉得看那份数据的人要么把它读成利空，要么就是没读懂。

但不管怎样，我就是觉得有理由四处看看、保持小心、保持审慎。我总是假设有一颗子弹正朝我飞来，脑袋要像装在万向轴上一样四处转。真正打中你的，是你没看见的那颗子弹。所以我在尽可能快地转。

但是的，市场可能需要喘口气。不过老兄，当我想到 Noam Brown 说的那些话，当我看到 Fable 的能力，我实在很难变得太悲观。

我是说，对我来说，我们这屋里坐着的是下一代里最了不起的两个人。在 Altimeter，我们对你们做的工作深怀敬意。我一直很感激你在我们发表东西时给我发来的反馈。

但对那些刚入行的人来说，他们可能会以为这行一直都是这样的，对吧？

[1:17:57] Brad Gerstner

And like this line, the steepening of the line of creative destruction, the steepening ofthe line of, you know, scale advantages.I always believed it was going to be true.I never thought it would be true at this rate.I went back last night.In the last seven years, we've added $1 trillion of revenue to the Mag7 in the last seven years,okay?

而这条线——创造性破坏这条曲线的陡峭化，规模优势这条曲线的陡峭化——我一直相信它会成真，但我从没想过它会以这个速率成真。

我昨晚回去查了一下：过去七年里，七巨头（Mag 7）新增了 1 万亿美元的收入，就在过去七年里，好吧？

[1:18:21] Brad Gerstner

To get to a trillion, to get to the first trillion, you know, took over 20 years.In the last seven, we added another trillion.And that added $17 trillion in market cap, that trillion dollars, okay?The forecast now that we're going to add another trillion of revenue in just three companies,SpaceX, Anthropic, and OpenAI, over the next four to five years, okay?

而要做到第一个 1 万亿美元收入，花了二十多年。过去七年，我们又加了一个万亿。而那一个万亿美元的收入，带来了 17 万亿美元的市值增长，好吧？

现在的预测是，我们会在未来四到五年里，仅靠三家公司——SpaceX、Anthropic 和 OpenAI——再加上一个万亿美元的收入，好吧？

[1:18:47]

Like, not seven companies, three companies, and in half the time, right?And so I would say that, you know, we are going to have bumps in the road.I know that it's going to be like this, but we're going to higher highs because the sizeof the prize, this is going to transform 5%, 10%, 15% of global GDP.There is no doubt in my mind, and 10% of global GDP is $10 trillion.It's an exciting future to be a part of.It's fun to do it with you guys.I think we're going to have to do our work to do the things to make sure America winsand that we evolve the social contract, keep everybody, you know, lift the floor, takeeverybody with us on this ride.But it's a really exciting time to be doing what we're doing.It's fun to be doing it with you guys.Yeah, I just want to say, Brad, thanks for having us, and thank you for what you've donewith the Trump accounts.I actually think it's super important for America, for the world, to give people an equitystake at a very young age.They will see it compound over their lifetimes.This is a great thing you've done for the world, so thank you.I'd echo all your comments, like deep admiration for you, your team, gratitude for the collegialityand friendship between our firms.

不是七家公司，是三家公司，而且是在一半的时间里，对吧？

所以我要说，我们路上会有颠簸，我知道过程会是这样的，但我们会走向更高的新高，因为奖赏的体量摆在那里：这件事会改造全球 GDP 的 5%、10%、15%。我心里毫无疑问。而全球 GDP 的 10% 就是 10 万亿美元。

这是一个令人兴奋的未来，能参与其中很棒。跟你们一起做这件事很有意思。我觉得我们还得做我们该做的功课，去确保美国能赢，去让社会契约跟着演进，把地板抬起来，把所有人一起带上这趟车。但这真是一个非常令人兴奋的时代，能做我们正在做的事。跟你们一起做很开心。

对，我就想说，Brad，谢谢你请我们来，也谢谢你在「特朗普账户（Trump accounts）」上做的事。我真心觉得这对美国、对世界都极其重要——让人们在很小的年纪就拥有一份股权，他们会看着它在一生中复利增长。这是你为这个世界做的一件大好事，所以谢谢你。

我也附和你所有的话，对你和你的团队深怀敬意，也感谢我们两家机构之间的同侪情谊和友谊。

[1:19:53] Brad Gerstner

I know Clark and Foxy, they hang out like all the time.That's what people think that, you know, and there are people in our business who don'twant to share anything.Our view is, like, we open source it, but there are very few people who we actually calland ask their opinion, because there are very few people who do the thousands of hours ofwork that we do, you know, that are adding to that.And you do it, and we appreciate that, and you do as well, Gavin, we appreciate that.So with that love fest, let's call it a wrap.Thanks for being here.Thank you.Thank you.

我知道 Clark 和 Foxy 一直混在一起。大家都以为——我们这行里有些人什么都不愿意分享。而我们的观点是，我们把它「开源」出去；但真正会打电话去问意见的人非常少，因为真正像我们这样投入成千上万小时做功课、并且能给你增量的人非常少。你们就是这样的人，我们很感激。Gavin，你也是，我们很感激。

那就在这场彼此吹捧的爱意里收尾吧。谢谢你们来。谢谢。谢谢。