本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
我们有一项全新的技术，这是有史以来最重要的技术，而且你需要一套全新的基础设施。
We have a whole new technology that's the most important technology ever, and you need a whole new infrastructure.
通常我们谈论基础设施时，指的是服务器、存储和网络。
Normally, we talk about the infrastructure world, we're talking about servers and the storage and the network.
但在这里，它一直延伸到矿山，铜矿。
Here, it goes all the way down to the mines copper mines.
是的。
Yeah.
这就是这项技术会普及到什么程度。
That's how widespread this thing is gonna be.
以前你建东西的时候，那是个工程问题。
It used to be when you built something, it was an engineering problem.
而现在，感觉这真的是资源限制的问题。
And here, it feels like it really is a resource limitation.
所以不管是不是代币，我们都在往系统里投入大量资金，然后这些系统会产生结果。
So whether it's tokens or not, we're pouring a ton of money into systems, and then those systems are producing a result.
而现在，我们的瓶颈在于系统本身的能力，跟不上我们投入的资源。
And right now, we're bottlenecked on the systems' ability to actually match the resource we're pouring into them.
领先的内存厂商说，他们现在的订单需求，就算满负荷生产三年也供不上。
The leading memory developer said the demand they have today, it will take them three years of capacity to supply.
如果这个基金能达到我们预期的效果，那五到十年后，世界会是什么样子？
If this fund does what we think it will do, how do we see the world in five to ten years?
美国在基础设施上会赢，真的。
America wins in the infrastructure, damn.
那就是我们了。
That would be us.
今天，a16z宣布成立“机器时代基金”，这是一个专门投资于支撑AI下一个时代的基础设施的基金。
Today, a16z is announcing the Machine Age Fund, a new fund dedicated to the infrastructure powering the next era of AI.
今天请到了本·霍洛维茨、拉古·拉古拉姆和马丁·卡萨多，来聊聊为什么我们现在要推出这个基金，以及为什么我们认为AI下一个主要的瓶颈可能不是模型本身。
I'm joined by Ben Horowitz, Raghu Raghuram, and Martin Casado to explain why we're launching it now and why we believe the next major bottleneck in AI isn't necessarily the model.
而是模型底下的一切。
It's everything underneath it.
芯片、内存、网络、电力、冷却和数据中心，这些都被推到了超出原本设计能力的极限。
Chips, memory, networking, power, cooling, and data centers are all being pushed beyond what they were originally designed to handle.
与此同时，AI正在改变技术领域的一条老规矩。
At the same time, AI is changing an old rule of technology.
以前，光靠砸钱解决工程问题，不一定能让进展更快。
Throwing money at an engineering problem didn't necessarily make it move faster.
但现在，资本越来越能直接转化为算力，而算力又能转化为更强的智能。
Increasingly, capital can be converted directly into compute and compute into more capable intelligence.
我们来聊聊这个转变意味着什么，新的基础设施公司能在哪里突破，以及为什么新一代创业者正在回归计算领域最棘手的一些难题。
We unpack what that shift means, where new infrastructure companies can break through, and why a new generation of founders is returning to some of the hardest problems in computing.
Ben、Martin、Raghu，欢迎你们。
Ben, Martin, Raghu, welcome.
谢谢。
Thank you.
好的。
Alright.
谢谢。
Thank you.
我想先引用一句名言来介绍这个新基金。
I wanna start with a marked quote to introduce this new fund.
这是我一生中最大的技术革命。
This is the biggest technological revolution of my lifetime.
这显然比互联网还要大。
This is clearly bigger than the Internet.
可以与之相比的是微处理器、蒸汽机、电力，或者可能是轮子。
The comps on this are the microprocessor, the steam engine, and electricity, or maybe the wheel.
各位，请介绍一下机器时代基金。
Guys, the Machine Age Fund, please introduce it.
本，你先来。
Ben, start us off.
嗯，基本上发生的是我们拥有了一项全新的技术，这是有史以来最重要的技术。
Well, basically what's happened is we have a whole new technology that's the most important technology ever.
每当出现一种全新的方式来利用我们所有热爱的基础设施时，会发生什么呢？
And what happens every time there's a dramatic new way of using all of the things that we love infrastructure.
你需要一套全新的基础设施，而这次的影响比以往任何时候都要大。
You need a whole new infrastructure and never has it been more high impact as it is on this one.
所以我们不仅需要新的芯片、新的系统软件，还需要新的供电方式，需要替换铜线。
So not only do we need new chips, new system software, we need new ways of doing power, we need to replace copper.
我的意思是，这几乎涉及一切。
I mean, like, it's absolutely everything.
所以这是一个非常激动人心的时刻。
So it's a very exciting time.
特别是对于这个新时代的硬件方面，我们需要一种新的方法。
So particularly for the kind of hardware aspects of this new era, we needed a new approach.
对，我同意。
Yeah, I would agree.
通常，至少在我们谈论计算领域的基础设施时，我们指的是服务器、存储和网络。
Normally when we, at least in the computing, when we talk about the infrastructure world, we're talking about the servers and the storage and the network.
这次是深入到最底层，连铜矿都要涉及。
Here it goes all the way down to the mines, copper mines.
这个影响范围就是这么广。
That's how widespread this thing is gonna be.
这是第一点。
That's number one.
第二点，我觉得过去三年我们看到的是模型能力的稳步提升，模型本身已经不再是瓶颈了。
And number two, I think what we have seen over the last three years is the steady increase of the capabilities of the models, where the model is no longer the bottleneck.
实际上，通过使用AI，这些模型进步得越来越快，越来越快。
And in fact, using AI, these models are getting better faster and faster and faster.
现在的瓶颈都在我所说的模型之下，所以我们需要在这方面下功夫。
Now the bottleneck is all what I call south of the model, and so that's why we need to work on that.
我只想快速补充一点，我们倾向于关注创始人，过去几年我们看到越来越多的优秀团队在攻克复杂难题。
The only thing I'd add very quickly is we tend to follow founders, and we've been watching over the last couple of years as the number of very strong teams going after complex harder problems has increased.
我不知道具体数字，但我周末试着估算了一下，以前顶级创始人带来的交易中大概只有5%是硬件相关的，现在我看到超过20%到30%，所以创始人群体，他们通常比风投群体更聪明，已经把这个领域视为创新的重点，并且正在积极行动。
I don't know the actual numbers but I was trying to estimate it over the weekend, so I think we'd get maybe 5% of the deals from top founders would come in would be hardware before, now I see north of 20% or 30% right now, so like the founder community which tends to be much smarter than the BC community has identified this as a very area for innovation and they're responding.
我觉得5%可能都算高了。
I think 5% is probably generous.
对。
Yeah.
非常低。
It's very low.
非常低。
It's very low.
对。
Yeah.
3%。
3%.
对。
Yeah.
说说看，是什么样的宏观条件导致了这种变化，让这么多创始人都在追求这些想法？
Explain some of the macro conditions that have led to this change in terms of the surplus of founders pursuing these ideas.
他们看到了什么，才觉得有这个机会？
Like what are they seeing that's enabling?
嗯，最明显的就是对AI的需求几乎是无限的，所以供应链的每个环节都承受着压力。
Well, mean, the obvious is the demand for AI is basically infinite, and as a result of that, every part of the supply chain is under duress.
就是说，所有东西，包括制造内存这类产品所需的材料，都面临这种情况。
I mean, everything, including like materials used to make things like memory.
还有一点很有意思，AI有个独特之处，因为需求无限、增长无限，你担心的往往是公司的利润率，也就是效率高不高。
It's also very interesting where there's something unique about AI, which because the demand is infinite and growth is infinite, what you tend to worry about is the margin of companies, which is how efficient it is.
通常你会担心增长问题。
Like, normally you worry about growth.
比如，能不能找到人买这些东西？
Can I just can get I people to buy this stuff?
但在这里你不用担心这个，问题是你能不能做到盈利。
You don't have to worry about that here, the question, can you do this in a way that's profitable?
而很多效率其实严格来说受限于硬件物理条件，所以即使是AI浪潮的商业模式，也确实给现有系统带来了很大压力，因为这些系统不是为AI设计的，也不是为这些工作负载设计的。我觉得全球都有这样一个共识，就是我们确实需要改变核心组件，才能获得那种效率，从而推动增长，提升企业价值。
And a lot of the efficiencies are actually strictly a physical limitation of hardware, and so even the business model of the AI wave is really putting a lot of stress on the existing systems, because they weren't built for AI, weren't built for those workloads, and I think there's just this global observation that we actually need to change the core components to get that efficiency to help drive the growth and to drive the value of the businesses.
对。
Yeah.
那我们怎么知道需求真的超过了供给，而不是又一轮炒作呢？
And how do we know that demand is actually outpacing supply here rather than this being another hype cycle?
嗯，我是说
Well, I mean
对。
Yeah.
有很多城市。
There are an alien number of cities.
首先，一些对需求判断最准的人正在下大额采购订单。
Firstly, it is that some of the smartest judges of demand are cutting huge purchase orders.
你看那些超大规模云服务商，对吧？
I mean, if you look at the Hyperscaler, right?
他们的资本支出一直在猛增。
Their CapEx spend has been exploding.
明年，据说大型超大规模数据中心的总支出会达到一万亿美元，而今年大约是七千亿美元。
Next year, supposedly it's gonna reach a trillion dollars collectively across the big hyperscalers this year, it's about $700,000,000,000.
对吧？
Right?
而且你想，这些超大规模数据中心在行业里的位置，他们能看到来自四面八方的需求。
And if you think about the hyperscalers position in the industry, they see demand from everywhere.
他们显然能看到前沿实验室需要算力，看到AI数据公司，看到企业客户，也看到美国本土和国际上的需求。
They see obviously the frontier labs wanting their compute, they see the AI data companies, they see the enterprise, they see The US geography, the international geography.
所以要说谁对市场有洞察力，那就是他们了。
So if anybody has visibility, it is down.
他们一直在大幅增加资本支出，因为这种需求是前所未有的。
And they've been jacking up their CapEx, because it's never been seen before.
所以这绝对是一个明确的信号。
So that's a clear, clear sign.
其次，看看我们日常接触的那些公司，它们都在飞速发展。
And secondly, if you look at the companies that we see on a day to day basis, they are all ripping.
所有应用公司的增长都太疯狂了。
All the application companies, the growth is insane.
前沿实验室的增长也疯狂，这已经有记录。
The frontier labs, the growth is insane, it's been documented.
所以我觉得从需求端来看，信号从来没有这么明确过，这不是什么泡沫。
So I would say on the demand side, the signals have never been fair, that this is not a hybrid.
而且价格还在上涨。
And prices are going up.
我们从来没见过芯片价格这样涨。
We've never seen prices go like up on chips.
TP的价格倒是降了，
The TP prices went down,
但通常都是降的。
always go down.
对。
Yeah.
价格总是会降的。
It always goes down.
你看价格曲线，先是这样走，然后又这样回落。
If you look at the price curve, it went like this, and then it went back like this.
而且我们知道，目前只有大约10%的市场是真正在交易的。
And we know only like 10% of that is sort of market to stop today.
我的意思是，如果你看整个市场的供应，基本上都已经预订到2028年了。
I mean, if you look at the supply across the board, it's basically all booked out to 2028.
情况真的很紧张，我们甚至看到过几千块GPU的拍卖持续了好几天。
I mean, it's so bad, we've actually seen multi day auctions for a few thousand GPUs.
当然，另一面是需求，正如Raghu所说，我们看到了行业历史上增长最快的公司。
The other side of that of course is demand, and as Raghu said, we've seen the fastest growing companies we've seen in the history of the industry.
而且，AI能完成的工作单位，其价值在不断增加，但背后消耗的token数量却呈数量级增长，如果是聊天或代理用100个token，现在就是几千个token，对吧？
And also the unit of work that AI can do, value of that unit of work keeps increasing, but underneath the covers, the number of tokens that are consumed is going by orders of magnitude, If it's 100 tokens for chat or agent, it's thousands of tokens, right?
所以需求的两端都在扩张。
So you got expansion on both sides of demand.
一方面是，每个工作单元消耗的算力越来越多，另一方面，因此受益的人数也在增加。
One is the unit of work is becoming more and more consumptive of Tompkins, and then secondly, the number of people therefore that are gonna be benefited.
不仅仅是开发者，所有知识工作者都会受益，再往后，范围会更广。
It's not just the developers, it's gonna be all knowledge workers, and then all beyond that.
这就是我们看到的情况。
So that's what we see.
你说供应端的关键组件已经卖到了2027年，可能到2028年。
You said the key components in supply are sold out till 2027, maybe in '28.
对，一直到2028年。
Till 2028, yeah.
整个行业提前这么久就售罄了，这意味着什么？
What does it mean for an entire industry to be sold out that far?
我不知道以前有没有发生过这种情况。
I don't know if this has ever happened before.
你们还记得吗？
Do you guys recall?
我是说，还记得互联网时代吗？那时候我们大规模建设，但大部分铺设的东西都是投机性的，而且都是暗光纤。
I mean, remember in the Internet days, when we were doing massive build out, the majority that was actually being put in the ground was speculative and was dark.
还记得暗光纤吗？现在基本上每一块GPU生产出来就已经被预定了。
Remember the dark fiber in here, basically, every GPU that's being created is already presold.
所以，我们
And so We
还没到那个程度。
weren't quite there.
我的意思是，在98、99年那会儿，带宽确实短缺，但实际需求没那么大，因为上网的人没那么多。
Mean, there was a lack of bandwidth, like in the 9899 timeframe, but there wasn't that much real demand for it because there just weren't that many people on the internet.
所以这是个两方面的问题：公司都急着要更多带宽，理论上需要，但另一边没有足够的用户来消耗这些带宽。
So it was a two sided thing and the companies were all rushing there and needed more bandwidth theoretically, but there weren't the users on the other side to consume it necessarily.
而且真要消耗大量带宽，得靠视频这类高带宽应用，但当时这些应用因为种种原因还不成熟，跟数据中心有多少带宽没关系。
And then to really consume a lot of bandwidth, you have to do high bandwidth things like video, which weren't really viable for a number of reasons that had nothing to do with how much bandwidth was in the data center.
所以看起来有点像，但本质上不是一回事。
So it smelled similar, but it wasn't this.
现在的情况是，我们全力以赴，人们把GPU以四倍的价格转卖，诸如此类的事情。
This is like we're flat out and people are reselling GPUs for four times what they bought them for and this kind of thing.
而且我们的电力和冷却资源也已经用完了。
And then we're also out of power and cooling.
再加上，现在建设真的很难，因为面临巨大的政治阻力。
And then on top of that, it's really hard to build because there's these incredible political headwinds going into it.
所以，在我职业生涯中，真的从未见过这样的情况。
So it's really unprecedented in my career that we've had anything like this.
确实。
No.
我想给你讲个小故事。
I want give you a quick anecdote.
我之前和一家大型上市公司的首席财务官聊天，他以前一直对上云很抵触，他们有很多服务器。
So I was talking to a CFO of a large company, a large public company, who had historically been very resistant about going into the cloud, they had a lot of servers.
他们做了一次库存盘点，发现服务器里的内存价值增长了很多，足以支付整个上云迁移的费用。
And they were doing an inventory check and they realized that the memory in their servers had increased so much, it could fund the entire migration to the cloud.
所以我就是觉得我们现在的情况非常特殊。
So I just feel like we're in a very unusual situation.
对。
Yeah.
没错，确实是这样。
Well, that's right.
我们很多东西都缺。
We're out of many things.
电力、冷却、内存、GPU，你能想到的都有。
Power, cooling, memory, GPUs, you name it.
我们都缺
We're out
这些。
of it.
对。
Yeah.
行业里的旗舰会议叫Hot Chips，正在斯坦福举行。
So the flagship conference for the industry is the one called Hot Chips, is going on in Stanford.
领先的内存供应商说，他们现在的需求，要三年的产能才能供应上。
And the leading memory vendor said the demand they have today, it'll take them three years of capacity to supply.
这还只是现在的需求。
It is just today.
还不算未来的需求。
It's not even future demand.
那么，说所有东西都同时短缺，是因为人们低估了模型会有多好、多有用，他们根本没法预见到这种需求吗？
So in terms of being about everything simultaneously, is it because people just underestimated how good models would be, how useful they would be, they just couldn't have foreseen the demand?
我觉得甚至不是这个原因。
Well, I don't even think it's that.
我是说，这些东西是突然冒出来的。
I mean, this stuff came out of nowhere.
对吧？
Right?
我们才刚开始四年。
We were only four years into this.
所以即使我们一开始就有完美的预知能力，我也不觉得
So even if we had a perfect oracle once it started working, I don't
我们能建出那么大的产能。
think We couldn't have built the capacity.
我们根本建不出那么大的产能。
We couldn't have built the capacity.
绝对不可能。
There's no way.
而且我们说的是芯片周期，通常要三到四年。
And we're talking about, like, chip cycles, which tend to be three to four years.
我们说的是破土动工建数据中心，这你知道的，要四到五年。
We're talking about breaking ground and building data centers, which is, you know, four to five years.
而且还要连接、拆除、建造，还要有电力供应。
And and connecting breaking down and building them and having a power source.
所以，你要么得自己发电，要么通常两者都得做。
So, like, you either have to build your own power or, like, usually both.
你得自己发电，还得有电源，这并不容易。
You've gotta build your own power and have a power source, which is not easy.
对。
Yeah.
如果一个行业以30%的速度增长，那已经是很不错的增长率了。
If you have a industry that's used if it's growing at 30%, it's a great growth rate.
对吧？
Right?
而它对接的AI软件行业，基础增长率是三位数。
And it's being connected to an AI software industry that's like triple digits as the base.
所以你能看到这种脱节，对吧？
So you can see the disconnect, right?
所以差距只会越来越大。
So it's just widening.
那为什么这个基金五年前或者七年前不存在呢？或者说，为什么那时候它不是一个值得投资的好领域，布拉德？
And so why didn't this fund exist, you know, five years ago, or seven years ago, or why was it not a great category to invest in the same way, Brad?
嗯，我觉得我们可能算是刚好赶上，但要是能早个几年开始就更好了。
Well, I would say we're probably I'd like to think we're just in time, but we probably would've been well suited to have at least a couple years ago.
我想说，基本上每个时代都会出现一家独立的公司，对吧？
I will say you can actually point on basically every epoch to an independent company that came up, right?
很明显，从大型机到客户服务器的转变，我们看到了一堆公司冒出来。
Clearly the move from a mainframe to the client server, we saw a bunch of companies come up.
互联网的兴起，就有了思科和瞻博网络。即使在超大规模数据中心，顺便说一句，这主要是由现有的云服务商垂直整合推动的，你也看到了Arista的崛起。所以一直都有投资硅片和硬件的可能性，但相对较小，因为变化相对较小，比如一家芯片公司、一家交换机公司。但在这里，一切都变了，所以我同意本的观点，可能可以早点开始，但现在变化太大了，这显然是个该做的事。
The move to the Internet, that's why we've got Cisco and Juniper, even in the mega data centers, which by the way was largely driven by the incumbent cloud providers verticalizing, you saw the arising of Arista, so there has been the ability to invest silicon on hardware, but it's been relatively minor because the change has been relatively minor, like one chip company, one switch company, where here everything is, and so I agree with Ben, probably could have started a little bit earlier, but the amount of change is so high now that it's just an obvious thing to do.
另外，对智能的需求非常垂直，而且看不到尽头。
And the other thing is the demand for intelligence is so vertical, with really no end in sight.
我的意思是，因为每个采用它的公司，使用量都在快速增长。
I mean, because every company that's adopted it is growing very fast in its usage.
而且大多数公司还没有大规模采用它。
And then most companies haven't adopted it to a high degree.
而且消费者才刚刚开始使用。
And then consumers are just getting started.
所以，对代币的需求可能会以每年接近百分之千的速度增长，而供应量根本不可能增长得那么快。
And so it's gonna probably, the demand for tokens is probably gonna grow close to a thousand percent a year, which you cannot grow supply that fat.
我们不是，要让我们能够以那样的速度扩展基础设施，我们需要做的各种工作是非常庞大的。
We're not, the amount of just work we're gonna have to do across the board to get to the point where we can grow like infra at that kind of rate is pretty vast.
所以，我觉得在这条路上有很多投资机会。
So, I think there's a lot of investing opportunity on the way.
顺便说一句，另外一点是，所有硬件系统的架构都是为完全不同的计算时代设计的。
And by the way, the other thing is like all the architectures of the hardware systems were built for a whole different era of computing.
所以，不仅仅是需要更多的容量，我们还需要建设容量的能力。
And so more than just we need more capacity, we need capacity to build.
有很多机会去建设不同类型的基础设施。
There's lots of opportunities to build different kinds of infrastructure.
对，它们都达到了设计时的物理极限。
Yeah, they're all reaching their the physics limits for what they were designed.
对吧？
Right?
就像Ben刚才提到的，铜之类的材料，诸如此类。
Like what Ben was talking about, copper, and so on and so forth.
而且你可以逐一审视这些类别，然后发现，好吧，这类技术已经到极限了。
And you could go across every one of these categories, and you could find, okay, this is the limit of this type of technology.
所以现在你得有技术突破才能进入下一个阶段。
So now you gotta get some technical breakthroughs to get to the next one.
对。
Yeah.
我想花点时间深入聊聊需求端。
I wanna dive deeper on the demand side for a second.
随着我们从聊天机器人发展到推理模型，再到智能体，再到多智能体，每一步都让单个任务消耗的token数量成倍增加，而且是数量级的增加。
As we've moved from chatbots to reasoning to agents to to multi agents, each step has multiplied the number of tokens a single task takes up by orders of increasing orders of
没有谁比AI自己更喜欢用AI了。
Nobody likes to use AI more than AI.
对。
Right.
简单。
Easy.
所以
So
为什么这种情况会持续发生，而不是趋于平稳？你觉得这种情况会无限期地持续下去吗？
why does that keep happening instead of a leveling off, and do you just see that happening, you know, indefinitely, just continuing to?
嗯，这个问题有几个答案。
Well, so there's a couple answers to that.
首先，可以肯定的是，现在如果你看看我们实现扩展的方式，我们主要是通过大量的推理来实现的，也就是通过大量的token，对吧？
The first one is is for sure right now, if you look at, like, the the way we're achieving scaling, the way we're doing it is through a lot of inference, so through a lot of token, right?
如果你想想强化学习是什么，你知道，它需要大量的推理。
If you think about what RL is, you know, so it's a lot of inference.
如果你想想思维链，它也需要大量的推理。
If you think about chain of thought, it's a lot of inference.
如果想想长期运行的代理，当然，这需要大量的推理。
If think about long running agents, course, it's a lot of inference.
所以这基本上就是我们一直在用来扩展的方法之一。
And so that's just basically been one of the approaches that we've been using to to scaling.
我觉得如果你想退一步看，这里的宏观趋势是什么？
I think if you want to step back and say, kind of what is the macro trend here?
以前你构建东西时，那是一个工程问题。
It used to be when you built something, it was an engineering problem.
投入一堆工程师，但这无法扩展，而且会有工程物理的自然法则，这就是《人月神话》的来源。
Throw a bunch of engineers at it, and that doesn't scale, and that would have a natural law of engineering physics, which is what the Mythical Man Month came from.
而在这里，感觉这真的是资源限制，不管是代币还是什么，我们投入大量资金到系统中，然后这些系统产生结果。
And here it feels like it really is a resource limitation, so whether it's tokens or not, we're pouring a ton of money into systems, and then those systems are producing a result.
而现在，我们受限于这些系统将资源投入转化为实际产出的能力。
And right now, we're bottlenecked on those position those systems' ability to actually match the resource reporting into them.
所以我觉得，代币现在可能就是我们处于扩展曲线上的位置，但我们没有像以前工程那样的自然调节器。
And so I think, like, tokens right now is probably where we're on the scaling curve, but we don't have a natural regulator like engineering like we did before.
所以我觉得这种情况会持续下去，我们必须建立相应的供应来支持它。
So I think we should expect this to continue, and we have to build a supply to support it.
对，换个角度看，任何问题只要有足够的基础设施和GPU算力就能解决，所以只要问题没完，需求就不会断。
Yeah, and the similar way to think about it is any problem that you have can be solved with enough infrastructure, And GPUs power and so until we run out of problems, we're not going to run out of demand.
这就是挑战所在。
And that's the challenge.
AI要变得更好，答案就是使用更多的AI。
AI's answer to getting better and better is to use more AI.
推理是它反复使用的一个基本构建块，所以这些token才会成倍增长。
Inference is one basic building block that it keeps using over and over and over again, and so that's why these tokens multiply.
是的，甚至还有自催化效应，比如用AI来创造更多AI，就像创建GPU内核，这本身就是用更多AI作为过程的一部分。
Yeah, even the auto catalytic effect, so even the idea of using AI to create more AI, like creating GPU kernel of course is just using more AI as part of the process.
所以再说一次，我们思考这个问题的一种方式是，过去钱投进来，你面对的是工程问题，我们知道这需要两年时间，通常还会失败，这就像是一个国家级的制约因素，然后你才能得到产品。但现在，钱投进去和硬件产生智能之间没有别的环节，所以我们只受限于我们创造供应的能力，所以这非常非常不同。
So again, one way that we think about it is in the past money would come in, you have an engineering problem, we know that it takes two years, normally fails, know, it's a national governor and then you get the product on the other end, There's nothing between the money going in and then the hardware creating intelligence, and so now we're just limited by our ability to create supply, so very, very different
对，只要给我更多GPU，太阳能就没了。
Yeah, just giving me more GPUs and our solar is gone.
对。
Yeah.
没错。
That's right.
所以这就是一个循环。
So that's the cycle.
只要你有钱、有GPU、有数据，在可预见的未来，你就能不断扩展这些东西。
As long as you have the money, the GPUs, and the data, you know, for the foreseeable future, you'll be able to scale these things.
这很有意思，因为你知道，过去十年里，感觉有太多人，而且普遍的情绪是，有太多钱涌向了初创公司。
And it's fascinating because I've, you know, over the last decade it feels like there are so many, you know, people and the pervasive sentiment was there's too much money going to start ups.
我们给这些初创公司投的钱太多了。
We're we're over funding these start ups.
风险投资里的钱太多了。
There's too much money in in in in venture capital.
Ben，你多说说这意味着什么，因为以前人们会怀疑，你往这个行业投的钱越多，结果就越大，而现在我们说的是，从好的方面看，某种程度上，市场的大小取决于我们大家共同投入多少。
Say more, Ben, about what that means because there there used to be this skepticism that the more money you put into the industry that there would be bigger outcomes, and now we're saying at the upside that there's, to some degree, market is as big as we collectively contribute to it.
对。
Yeah.
所以你看。
So this is look.
在创业圈里，我们都明白一件事：如果我在你前面领先两年，你想靠雇一千个工程师来追上我，那你会把公司搞垮。
The one thing we all knew in startup world is that if I have a two year lead on you, and you try and cast me by hiring a thousand engineers, you're gonna wreck your company.
这从来就行不通。
Like, that never works.
这就是所谓的人月神话。
It's a mythical man month.
九个女人一个月也生不出孩子。
Nine women can't have a baby in a month.
就是这样。
That's it.
那永远行不通。
That never works.
好，现在这个办法行得通。
Okay, now that works.
但这不是雇十万个工程师，而是拿三十亿美元去点亮一个巨大的计算集群。
But it's not hiring 100,000 engineers, it's taking $3,000,000,000 and lighting up a magnificent cluster.
然后突然之间，不管什么Grock都能凭空冒出来，一下子就成真了，Kimmy，或者别的什么。
And then all of a sudden, whatever Grock can come out of nowhere and all of a sudden it's real, Kimmy, or what have you.
就像这些领先优势，你可以用钱砸向问题，而且几乎任何问题都能用钱解决，这招管用。
It's just like these leads, you can throw money at the problem and you can throw money at almost any problem and that works.
所以这跟我们经历过的任何情况都完全不同。
And so that is just completely different than anything we've ever lived through.
顺便说一句，我们都在心理上适应这个变化。
So we're by the way, we're all psychologically adjusting to this.
ChatGPT应用每周活跃用户有十亿。
The ChatGPT app has a billion weekly actives.
大约有三千万开发者，他们使用的计算需求占了相当大的比例。
There's about 30,000,000 developers who are using relatively a big portion of compute demands.
我们现在怎么看待计算需求，以及未来，考虑到人们实际在用AI做什么？
How do we think about compute demand needs now and in the future in light of what people are actually doing with AI?
这就是发展的过程。
That's the progression.
对吧？
Right?
所以Chargeic Video是个休闲应用，而编程是给专业人士用的。
So Chargeic Video was a casual app, and it's coding for professionals.
对吧？
Right?
现在，通过编程，我们已经为知识工作者打造了很棒的工具。
Now using coding, have now built amazing tools for knowledge workers.
所以这就是下一个前沿。
So that's the next frontier.
而现在，有超过十亿的知识工作者在等着呢。
And now there are over a billion knowledge workers in the line.
对吧？
Right?
而且要达到那个需求，还有很长的路要走。
And for that, it's a long ways to go for that demand.
顺便说一句，他们做的工作，所有这些流程和自动化等等，然后他们到了后台，那里全是智能体。
And by the way, the work that they do, all this workflow and automation and so on, and then they get to the back office which is all the agents.
所以每一步都会解锁，我觉得，一个数量级更多的需求。
So progressively each of these things unlocks, I would say, an order of magnitude more demand.
我的意思是，这还只是开始。
I mean, just at the start of this.
嗯，现在你有了GrockBot，你知道，就像编码领域发生的那样，现在通过GrockBot，所有计算机的使用都在发生类似的事情。
Well, and now you have GrockBot, which is kind of, you know, what happened with coding is kind of happening with all use of computer via GrockBot.
所以我们正处在另一波需求浪潮中，而且肯定还会有更多。
And so we're in a whole another wave of demand and most certainly there's going to be more to come.
所以目前看起来确实是无止境的，而且我们还没涉及到具身智能或机器人，那将是另一个
So it does seem quite unlimited at the moment, and we haven't even gotten into embodied AI or robots, which are gonna be another
需求来源。
source of demand.
马丁是专家，但据我了解，Rockport 用的是计算机使用技术，就像有人坐在电脑前不停地打字一样。
Martin is the expert, but my understanding is that Rockport uses computer use, which is just like human beings sitting inside the computer typing away.
我周末真的用了它来更新信用卡信息，处理了一堆我一直懒得弄的服务，还取消了一些订阅。
I literally used it over the weekend to update my credit card with a bunch of services that I'd been lazy to do, and cancel a bunch of subscriptions.
我的意思是，这不是编程什么的，这是真正的计算机使用。
I mean, this is not coding or whatever, this is true computer use.
突然之间，就创造了可能性，比如知识工作的案例，只不过它们都坐在电脑里干活。
All of a sudden, create possibility like knowledge work cases, except they're all sitting inside of the computer doing work.
我确实认为
I do think
马克·安德森说得对，这里正确的类比是蒸汽机或电力，原因如下。
that Mark Andreessen is right, it's like the right analog here is like the steam engine or electricity, in the following way.
就像引入了这种新东西，你可以把它变成工作，现在有一些非常明显的应用，但可能还有三四十年的时间来把各种计算机问题，任何有明确回报信号的东西，都扔进去，而我们才刚刚开始，就像我们有了语言和代码，仅此而已，计算机时代才刚刚开始。
Like introduced this new thing that you can turn to work, and there's some very obvious applications now, but there's probably thirty, forty years of throwing computer problems, anything with a clear reward signal, and we're just starting, like we've got language and code, that's it, just starting computer years.
但除此之外，我们还在看什么？
But like what else are we looking at?
从科学、材料、生物学的角度来看，当然创造力也是一个巨大的应用领域。所以，我们正处于一段漫长旅程的非常非常早期的阶段，而且我们已经消除了一个关键瓶颈，那就是传统的软件工程。
We're looking at in terms of science, materials, biology, I mean of course creativity is a massive use, and so listen, we're on the very very early part of a very long journey, and we've removed this key bottleneck, which is traditional software engineering.
当然，瓶颈会转移，其他地方会出现更多的复杂性，但我认为我们正处于一个非常漫长的、不断解决计算机问题的过程的早期阶段，所以让我们预期这种计算需求会持续几十年。
Now of course, bottlenecks will move and there'll be kind of more complexity elsewhere, but I think we're very early in a very long run of throwing computer problems, so let's expect this compute need to persist for decades.
既然你提到了，马蒂，说说Gragbot吧，因为我们在异地会议上讨论过它给你留下的印象。
Because you mentioned it, Marty, talk about Gragbot, because we were talking at the off-site about how what struck you about it.
显然，我们以各种可能的方式参与其中，但你觉得它有什么特别有趣的地方？
Obviously, we're involved in every possible way you could involved, but what did you find so interesting about?
我觉得我们整个行业已经经历了好几个阶段，来理解AI如何进入我们的生活，对吧？
So I think we've, as an industry, gone through kind of multiple realizations for how AI enters our lives, right?
最开始，我们觉得，好吧，你给产品加上AI，就像加个搜索栏一样，然后你可以和它聊天，它也会回应，这就是传统的方式。
And very early on, we're like, okay, well, you add AI to a product and it's like a, whatever, it's a search bar, and then you kind of, you do chat with it and the chat's back, so that's kind of the traditional way to do it.
然后Open Clock出现了，那是今年早些时候的事。看到Open Clock，我想，也许它不仅仅是比谷歌更好，也许那还不是它的全部形态。不如让它成为一个独立的东西，但它是你的延伸，它共享你的密钥，知道你的密码，然后它就能做你会做的事情，对吧？
And then Open Clock kind of showed up, and that was earlier in the year, and with Open Clock I said okay, maybe like it just being like Google but better, maybe that's not the full embodiment of it, how about we'll have it be a standalone thing, but it'll be an extension of you, and it'll share your keys and it'll know your passwords and it'll just kinda do stuff that you would do, right?
所以它有点像你的延伸，但更像是一个人，你的延伸。
So it's kind of an extension of you, but it's more like a human, an extension of you.
然后我觉得拉古拉姆真正做对的是，不，不如把它当成一个真正的员工。
And then when I think Raghuram got really right is, no, how about it's just actually an employee?
所以现在你有了一个实体，它不需要特别访问你的密钥之类的，它有自己的电脑和浏览器，因为它们是世界上最聪明的模型，它能做任何员工能做的事。
So now you have this thing that's an entity, and it doesn't have like special access to your keys or whatever, it has its own computer and it has its own browser and because these are the smartest models in the world, it can do whatever an employee can do.
这挺有意思的，因为现在如果我想做点什么，我首先想到的是，高拉夫能帮我做吗？
And it's kind of interesting because now actually if I if I want something done, first thing I I think is like, can Gaurav do it for me?
而且答案往往是肯定的，即使是你想不到的事情。
And often the answer is yes, even if it's something you wouldn't expect it.
明显的例子就像，管理我的日历，安排会议。
So the obvious ones are like, whatever, like manage my calendar, like book a meeting.
还有一些不那么明显的。
There's also non obvious ones as well.
比如，我会让它读我的邮件并做分类。
Like, so for example, I'll have it read through my email and do triage.
我不会告诉它具体怎么做，但它会知道在真正处理之前先跟我确认一下。
And I don't tell it how to do that, but it will know to check with me before actually doing the triage.
所以这些东西已经足够智能，你可以交给它相对高层的任务，它也能因此完成一些复杂的事情。
So like these things are sophisticated enough that you can give it relatively high level tasks and it'll do kind of, sophisticated things as a result.
Ben，我知道你在这方面想了很多，也思考了这在组织里如何运作。
Ben, I know you're thinking a lot about this, and how this works in the organization.
你当然对文化也有很多思考。
You think a lot about culture, course.
你对此有什么看法？
What are your thoughts here?
嗯，我觉得你看看我们，就像是有了一种新类型的员工，而且会有很多这样的员工。
Well, I I think if you just look at us, it's like having a new kind of employee and there's going to be a lot of them.
我们得，你知道，我们花了很多很多年才学会怎么和普通的人类员工合作。
We have to just like, you know, with our we spent many, many, many years figuring out how to work with our kind of regular human employees.
现在有了这些其他类型的员工，和他们合作也需要一个学习过程。
And now we've got these other kinds of employees and there is a learning curve with them.
所以它们可能会消耗大量令牌，花很多钱，却什么正事都没干成。
So they can burn a lot of tokens and spend a lot of money and get nothing productive done.
它们可能会忘事，可能会编造内容，表现有好有坏，还可能带来安全问题。
They can forget stuff, they can make stuff up, they can have good behavior, they can have bad behavior, they can create security problems.
所以，这些方面都存在，但它们也能超级高效地工作。
So like, there's all those aspects to it, but they can also be super duper productive.
所以，我觉得如何把它们整合进来，让它们和真正的人类同事好好配合，这些都是我们在学习的东西。
So, I think figuring out how to integrate them in, have them work nicely with the people that they're working with, the actual humans, is all something that we're learning how to do.
我不是说我已经完全搞明白了。
I mean, don't want to sit up here and say, I've cracked the code.
我们有了这个神奇的系统，整个公司现在完全自动化了，然后我就要慢慢把所有人类都裁掉，因为我能这么做。
We've got this marvelous loop and the whole firm is just completely automated now, and I'm going to slowly get rid of all the humans because I can.
这根本不是我们现在的情况。
That's not at all where we are.
我们更多是在想，怎么让我们的人类员工都变成超人，同时又不把公司搞乱，因为机器人可能会失控。
We're much more going like, okay, how do we make all our humans superhuman without wrecking the place because the bots get out of control?
对，这挺有意思的，因为我们试过几种不同的方式，想看看怎么把智能体最好地引入系统，最后是马丁的想法，就把它们当人一样对待，然后搞定就行了。
Yeah, and it's interesting because we've we've tried a couple of different ways as to how best to get agents into the system, if you will, and eventually, it was Martin's insights, just treat them as people and get it done.
我们现在就是这么做的，而且这被证明是在组织内部推动这件事最稳妥的办法。
And that's what we're doing, and that's turned out to be the most durable way of getting this thing going inside of an organization.
我想回到供给侧，更深入地聊聊瓶颈问题。
I wanna go back to the supply side and go deeper into the bottlenecks.
我们之前谈到，数据中心、芯片架构、系统软件、设施本身，这些都不是为人工智能设计的。
We were talking about how, you know, in terms of the data centers, chip architecture, system software, facilities themselves, that none of them were designed with AI in mind.
如果它们是为人工智能设计的，会是什么样子？
What would it look like for them to be designed with AI?
思考这可能意味着什么，有什么样的思维模式？
What is sort of the mental model for thinking about what that could mean?
对，所以，如果你从一个说法开始，就是说，嘿，基于实体模型的基础设施原始模式必须改变。
Yeah, so, I mean, if you start with a statement that you just say, hey, original model of infrastructure on entities models has to change.
你可以一个类别一个类别地看，看到各种不同的砖块，对吧，然后你开始逐一解锁这些瓶颈。
You can go category by calling and see varied bricks, right, And then you start unlocking the bottlenecks in each one of these things.
所以最终你得达到一个系统，你看推理引擎做什么，对吧，它占用大量内存，同时生成新令牌和进行计算。
So eventually you have to get to a system where if you look at what an inference engine does, right, it takes up a lot of memory, It generates new tokens along with the compute.
所以你可以想想怎么优化这一切？
And so you can just think about how do I optimize all of this?
内存需要什么样的？
What does the memory need to be?
计算需要什么样的？
What does the compute need to be?
它们之间怎么通信？
How do they need to talk to each other?
每个需要多少电力？
How much power does each of them need?
如果它们需要这么多电力，你怎么给每个供电？
And if they need all of this power, how do you call each of these?
然后你怎么把这些东西组合在一起？
And then how do you put the collections of these things together?
这正是目前整个行业里很多创始人在做的事情。
That is the exercise that's underway in the industry right now with a lot of the founders.
他们把问题拆解成最基本的组成部分，然后问自己：这里面的计算到底是什么性质的？
So they're breaking down the problem into the fundamental components and saying, what is the exact nature of the compute that's getting done?
好，既然主要是矩阵乘法，那我该怎么针对这种场景优化我的计算机？
Okay, it's how it's gonna be matrix multiplications, how do I optimize my computer and that kind of a scenario?
然后，生成这些token的过程中，它们都需要逐步增加内存。
And then they all need memory progressively to generate these tokens.
那么，怎么分层组织这些内存才是最优的呢？
What is the best way of hierarchically arranging this memory?
对吧？
Right?
还有，功耗方面又该怎么处理？
And then how does the power consume?
我的意思是，接下来你得把这些东西连接起来。
I mean, then you gotta connect it together.
在同一块芯片上、跨芯片以及跨数据中心，有哪些连接方式？
What are the ways of connecting it on the same chip across chips and across data centers?
每次数据传输要消耗多少电力？
How much power does each of these data transmission take?
所以你必须逐步拆解，然后从这些基本构件重新搭建起来。
So you have to progressively break it all down and rebuild it from these fundamental building blocks.
这就是我们正在看到的趋势，也是我们看到的机遇所在。
And that's what we see underway, that's where we see the opportunity.
我给你一个有趣的思维模型，来想想这个格局是怎么变化的。
Let me give you an interesting mental model to think about, like how the landscape's changed.
现在要训练一个前沿模型，大概要花30到50亿美元，对吧？
So today to build a frontier model costs, let's say, 3 to $5,000,000,000, right?
训练要花这么多钱，推理当然至少得赚回这个数，才能让这一切可行，对吧，大概得翻倍。
And that's to train it, and so the inference has to pay back at least that, of course, right, in order for any of this stuff to be viable, let's say two times that.
所以现在推理得赚到100亿美元。
So let's say that now inference has to make $10,000,000,000.
所以如果你能在这方面节省20%的效率，那就是20亿美元，而你很容易就能用20亿美元造一个ASIC，对吧？
So if you can save 20% of efficiency on that, that's $2,000,000,000, and you can easily build an ASIC for $2,000,000,000, right?
所以我们在行业里确实到了一个有意思的节点，为一个模型专门造一个ASIC是划算的，因为那个模型上的资本投入太大了。而且不像传统软件，传统软件有很多状态，而且非常动态，这些模型是固定的，模型权重是固定的。所以我们不知道世界会不会走向每个模型一个ASIC，但这给了你一个很好的思维模型，让你明白怎么把架构演进得更定制化，来匹配我们正在做的这些巨额资本投入。
So we've actually gotten to this interesting point in the industry where it actually makes sense to build an ASIC per model, just because the amount of capital investment in that model, and then unlike traditional software, traditional software has a lot of state and a lot of, you know, it's very dynamic, these models are fixed, the model weights are fixed, and so we don't know if the world goes to per model ASICs, but it gives you a great mental model of how you would evolve the architecture to be far more bespoke for these massive capital investments we're doing.
我觉得在行业历史上，我们从来没有创造过一个数字产品，直接投入了50亿美元进去。
Like I don't think in the history of the industry we've ever created a digital artifact with something like $5,000,000,000 that went directly into that artifact.
所以我觉得这会给硬件带来前所未有的巨大需求。
And so, you know, like, this, I think, is gonna put the greatest demands on hardware that we've ever seen.
哇。
Wow.
为此，机架功率需求从大约5到10千瓦，上升到100到250千瓦。
To to that end, rack power requirements are moving from roughly five to 10 kilowatts to a 100 to two fifty kilowatts.
计算密度大概提升了70倍。
Compute density is climbing something like 70 x.
冷却方式也从风冷变成了液冷，这是必须的。
Cooling is moving from air to liquid as a requirement.
展开剩余字幕（还有 298 条）
那这样一来，投资机会在哪里？
What are the investment opportunities as a result of this?
首先，当每个机架的功率达到那个水平时，交流电就不管用了。
Well, first of all, when you get to that level of power per rack, AC power doesn't work anymore.
这还真是挺夸张的。
So like, that's a pretty wild thing.
所以现在得用直流电，而且直流电本身也需要冷却。
So now you're into DC power, which by the way, requires its own cooling.
而且直流电还挺危险的，这挺讽刺的，因为爱迪生当年推广直流电的时候，还拿交流电有多危险来说事儿，甚至用电击动物来演示。
And is like, by the way, fucking dangerous, which is kind of ironic because Edison promoted DC power by claiming how dangerous AC power was in demonstrating it by electrocuting animals and things.
对，那匹马。
The horse, yeah.
对，那匹马。
The horse, yeah.
但是
But
他说得对，但那是针对他自己那种电力，那种电力非常强大，这是好消息。
he was right, but around his own kind of power, which is extremely powerful, it's the good news.
所以单从电力说起，就会变得非常非常不同。
So just starting with power, that's gonna be very, very different.
再想想冷却，这就涉及到，是的，我们要从风冷转向液冷。
Think with cooling, and this gets into, so yes, we're going air cooling to liquid cooling.
我觉得对于任何城市或数据中心来说，液冷已经是板上钉钉的事了。
I think we're already at liquid cooling for any That's city or data already kind of a done thing.
但问题来了，考虑到当前的政治环境等等，光靠液冷还不够。
But it gets into, okay, given the political environment and so forth, liquid cooling isn't enough.
必须是环保的液冷，直流电本身还不够。
It's got to be eco friendly liquid cooling of DC power is not enough.
电力必须为社会做贡献，而不是消耗社会资源。
It's got to be power that contributes to the power of society, not takes away from it.
所以有些数据中心表现不好，其实比例不大，大概10%，浪费了大量水资源。
And so you have data centers who have been behaving badly, small percentage actually, probably 10%, wasting a lot of water.
虽然没有网上人们说的开心果或杏仁那么夸张，但他们在这方面还可以做得更高效。
Not as much as pistachios or almonds and so forth as people demonstrate on the internet, but they could be a lot more efficient with that.
还有一些数据中心就像电力寄生虫，不把电力回馈给社会。
And then there are ones that kind of are parasites of power and don't contribute power back.
我觉得这一切都会结束。
I think all that's gonna end.
这必须结束，因为我们已经过了那个不可逆转的关口。
It's gonna have to end just because we've kind of gone through a one way door on that.
所以这需要很多公司还没投入的工程水平，所以这即将到来。
So that requires like a level of engineering that many haven't invested in yet, so that's coming.
而且，如果机架密度这么高，还有其他问题，比如地板的设计必须能承受那样的重量。
And then, if racks are that dense, there are other things, like the way the floors are designed, have to support that kind of weight.
这种事情是真实存在的。
That kind of thing is actually for real.
而且，我觉得，你需要的东西真的很多。
And then, I think that there's, you you just need a lot of everything.
而且，你知道，这些设备也会非常吵。
And so there's going to be, you know, kind of also by the things are really loud.
所以你得把数据中心的墙建得更厚，不然会打扰到邻居的安宁，这肯定不行。
So you have to build the data center with thicker walls, or you're going to disturb the peace in the neighborhood, which is not going to be acceptable.
我觉得没有哪个州会允许这种情况。
Like I don't think any state's going to allow that.
所以，很多人以前设计和建造建筑的方式已经彻底过时了。
And so a lot of the ways people have architected and designed the buildings themselves are already completely obsolete.
一旦我们达到费曼那个水平，现在能用的数据中心就会少很多。
Like once we get to Feynman, a much smaller percentage of the data centers that we have today work.
其实，大家都说内存价格涨，但涨得最快的领域之一其实是钢筋混凝土，真是讽刺。
In fact, I mean, everybody talks about memory prices, but one of the fastest areas where prices increasing is reinforced concrete for their sins.
这些数据中心给机架送800伏电压，首先变得非常危险。
The other thing that happens when these data centers are sending 800 volts to the rack is it's become so dangerous, number one.
其次，我们也没有足够的电气承包商，他们得有处理数据中心里800伏电压的专业知识，因为这是高压电。
But secondly, we don't have enough electrical contractors that have the expertise to deal with the 800 volts inside the data center because this is high voltage.
只有百分之二
Only two percent
的美国电工获得了直流电认证。
of electricians in The US have been certified on DC power.
所以这大概能让你有个概念。
So like that gives you an idea.
现在Meta有一个完整的培训项目来培训人员，这很棒。
Now Meta's got a whole program to train people up and so forth, which is great.
这就像一个新工作，他们免费培训人们来做这个工作。
It's like a new job where they train people for free to do this job.
但有趣的是，AI正在取代所有的工作。
But it's funny, AI is taking all the jobs.
AI将会培养出很多新的电工。
AI is gonna create a lot of new electricians.
是的，我觉得
Yeah, I think
我们只是在太空中做点什么。
we're just doing something in space
那些拥有大型云数据中心的大公司，都在拼命试验机器人，用来组装服务器或者把服务器放进数据中心等等。
I big guys that own the big cloud data big data centers, they all are furiously experimenting with robots, right, to do the work of assembling or putting servers into the data center, etcetera.
所以你会看到，随着AI的发展，这种情况会越来越多。
And so you will see that increasing as a result of the evolution in AI.
顺便说一下，关于我们正在筹集的基金，我们的重点是计算机科学基础设施。
By the way, to be to be clear on the on the actual fund that we're raising, our focus is on computer science infrastructure.
所以任何模型运行所需的东西，都属于计算机科学。
So anything a model runs on, that's computer science.
对吧？
Right?
所以想想看，芯片、网络、互联、存储，可能一直到电力。
So think, you know, chips, network, interconnect, storage, all the way down probably to the electricity.
对。
Yeah.
再多说说机器人手臂这块，我们会做什么，跟美国活力那边有什么区别，或者该怎么理解这个。
And and say more about the robotics arm in terms of what we'll be doing versus maybe American dynamism or how how to think about that.
对。
Yeah.
对。
Yeah.
当然。
For sure.
所以，你知道，我们再次认为，AI运行的任何平台，比如，AI带来的一个重大突破就是让计算机能够与物理世界互动，对吧？
So, you know, again, we we think that any platform that that AI will run on, like, one of the the the great breakthroughs that AI does is it allows computers to interact with the physical world, right?
它能看、能听、能说，对吧？
It can see, it can hear, it can talk, right?
这意味着新的平台，对吧？
And this means new platforms, right?
人们常说的边缘设备，但这其实没什么具体含义，对吧？
And the simplest way people say edge device, but that doesn't really mean anything, right?
我的意思是，它可以是移动设备，可以是CDN，也可以是笔记本电脑，但也可能是能到处走的实体设备。所以，作为专注于基础设施的投资者，我们不碰重监管行业或更垂直化的行业，但任何能推动AI进一步发展的计算机科学平台，我们都非常感兴趣。
I mean it could be a mobile device, it could be a CDN, it could be a laptop, but it also could be an embodied device that goes around, and so again, we as infrastructure focused investors don't do heavy regulated industries, or more verticalized industries, but any sort of computer science platform that's gonna push AI further out, we're quite interested in.
对。
Yeah.
回到数据中心的话题。
Going going back to the data centers.
到2028年，新建的数据中心将需要大约44吉瓦的额外电力，而预计电网新增容量可能只有25吉瓦。
By 2028, new data centers are gonna need something like 44 gigawatts of additional power against maybe 25 gigawatts of expected grid additions.
哦，等等，我们用了“吉瓦”这个词。
Oh, hold We we use that word gigawatt.
不。
No.
就像，哦，我们会有100吉瓦。
It's like it's like, oh, we'll have a 100 gigawatt.
马丁，什么是吉瓦？
Martin, what's a gigawatt?
我是说，那有多大？
I mean, how big is it?
相当于好几个足球场那么大。
It's multiple football fields.
我是说，非常庞大。
I mean, it's massive.
相当于五万人的规模。
It's 50,000 people.
你是什么意思？
What do mean?
它到底能提供多少电力？
What is it power?
大概相当于五万户家庭的用电量？
Like The equivalent is like 50,000 houses?
差不多是一座城市了。
To city.
五万。
50,000.
五万户家庭。
50,000 homes.
五万户家庭。
50,000 homes.
我在亚利桑那州的弗拉格斯塔夫长大，那是一个四五万人的小镇，具体人数看大学开学情况。
I grew up in Flagstaff, Arizona, which is a town of 40 to 60,000 people, depending on the universities.
我们整个镇的用电量还不到一千兆瓦。
We have less than a gigawatt of power consumption.
所以基本上，一千兆瓦就能让你的整个镇子亮起来，还能开空调。
So you can basically light up and air condition your entire town for a gigawatt.
对。
Yeah.
我是说，他这是在随便说
Mean, this is He's just throwing
到处乱扔。
them around.
不。
No.
不过话说回来，大家都在谈论千兆瓦。
But by the way, everybody talks about the gigawatt.
实际上真正运营的千兆瓦级数据中心非常少。
There's very few gigawatt data centers that are actually up.
我的意思是，我们还有很长的路要走。
I mean, we've got a long way to go.
要建起来。
To put in.
为什么公用事业公司和超大规模数据中心不能建得更快呢？
Why can't utilities and hyperscalers just build faster?
哦，原因太多了。
Oh, there's so many things.
首先，现在需要人来建造它们。
Well, there's first of all, right now, need humans to build them.
所以，就是普通的建筑施工。
So there there's just, like, the regular construction.
但更重要的是，你需要许可证。
But much more than that, you need permits.
你需要能接入的电力。
You need access to power that you can plug in.
所以你要么是组合方式，你得获得电力，这本身就是一个巨大的监管和竞标难题。
So you're either, you're doing a combination of, you've gotta get access to power, which is a massive kind of regulatory bidding struggle.
能接入的天然气、电网之类的资源非常有限。
There's very limited amounts and things you can tap into in terms of natural gas, power grids, what have you.
但你还得自己发电。
But then you also have to build your own power.
你猜怎么着？
And guess what?
我们现在变压器、涡轮机以及相关的各种设备都短缺。
We've got shortages of transformers and turbines and everything that goes into that.
所以，你得把所有这些都搞定。
So, it's just you've got to get all that stuff.
这可不是软件问题。
This is not a software problem.
不是光靠一堆工程师加加班就能解决的。
It's not just like a bunch of engineers, can't you work weekends and that type of stuff.
虽然加班也没什么用，但这里确实有真正的瓶颈。
Not that that works anyway, but there are real bottlenecks in this.
而且这些交货周期可没那么容易缩短。
And these lead times are not that easy to compress.
你看，全世界最聪明的人都在想办法缩短它，但这真的不容易。
Look, we have the best minds in the world trying to figure out how to compress them, but it's not not easy.
而且需求并没有放缓。
And the demand is not slowing down.
所以我们已经落后了。
So we're already behind.
现在需求每年增长十倍，而供应根本跟不上。
The demand is growing 10x a year right now, and the supply just can't
增长不了那么快。
grow that fast.
顺便说一下，
By the way,
现在情况糟透了，如果有新公司要买GPU，他们往往去墨西哥、澳大利亚或者其他国家，就是因为在美国太难了。
it is so bad that right now, if we have new companies going for GPUs, it's often in Mexico or Australia or in other country just because it is so difficult in The United States.
对，我们因为在这里禁止数据中心，反而在其他国家创造了大量就业和长期经济机会。
Yeah, we're creating huge job, both job and long term economic opportunity in other countries by banning data centers here.
我觉得正确的做法是设定一个标准，让数据中心回馈社区。
I think the right answer would be to set a standard where a data center contributes back to the community.
电力供应更好，没有噪音，没有水资源问题，还能增加就业。
Power gets better, there's no noise, there's no water issue, and it's adding jobs.
这应该成为标准。
That ought to be the standard.
然后每个人都应该遵守这个标准。
And then everybody ought to be just held to that standard.
顺便说一句，现在就有数据中心这么做的。
By the way, there are data centers that do that now.
这不是什么未来梦想之类的。
That's not a futuristic dream or something.
他们在那里的时候，能源价格每年都在下降。
Energy rates have gone down every year they're there.
原因就是他们自己供电。
And the reason is they provide their own power.
白天他们把电卖给州里。
They give power to the state during the day.
晚上州里不需要电的时候，他们再从州里借电。
And then at night they borrow power from the state when the state doesn't need it.
因为发电厂的运作方式就是始终按峰值容量发电。
Because the way power plants work is you're always generating peak capacity.
而数据中心白天黑夜的用电量都很稳定，城市则是白天用电猛增、晚上骤降，所以这正好形成一种互补关系。
And since a data center has steady capacity during day and night, and a city goes way up in the day and way down at night, that's a symbiotic relationship.
往大了说，我们为什么觉得——你知道，我们之前讨论了一会儿这个名字。
Zooming out, why do we think the you know, we we were batting around the name for for a little bit.
我们为什么觉得“机器时代”这个词能很好地概括我们现在做的事情呢？
Why do we think machine age is a is a compelling term for for what we're doing
在这里？
here?
对。
Mhmm.
好，听我说。
Well, listen.
让我先来说吧。
Let let me let me take it first.
首先，我觉得本说得完全对，人工智能这个词用错了，我们不应该叫它人工智能，应该叫机器智能。
So the first one is, I think Ben's absolutely right, artificial intelligence was the wrong word, like we shouldn't have called it our it's machine intelligence.
展开讲讲，为什么这么说？
Say more about that, why is that?
因为它不一定像人类那样思考，对吧？
Because it's not how humans think necessarily, right?
我的意思是，它其实是人类思考的缓存，是人类思想的集合，但今天我们还不知道如何让一个没有知识的人工智能去接触世界，然后让它自己重建语言，对吧？
I mean, it is a cache of how humans thought, it's a collection of humans thoughts, but like, today we don't know how to take an AI with no knowledge and put it out in the world and have it reconstruct language, right?
我们并没有做到这一点，对吧？
Like that's not what we've done, right?
我们构建的东西，是能够从我们已经学到的一切中学习，然后以有效的方式利用它。
We've built something that I can learn off of everything we've already learned, and then use that in a productive way.
而且，人工智能是一个通用术语，在计算机科学中正式使用已有七十年历史，适用于很多不同的东西，当然，它也背负了很多包袱，无论是来自科幻小说，还是来自尼克·博斯特罗姆的著作，或者其他什么。
And listen, AI is a general term that goes back seventy years in computer science formally that applies to many different things, and of course, it's got a lot of baggage either from science fiction or from, you know, Nick Bostrom who wrote about it or whatever.
所以，第一点就是承认，这确实是机器智能。
And so so the first one is just an acknowledgment, like, really is machine intelligence.
然后你想强调其中的机器部分，这里有个很深的讽刺，就是那些说软件正在吞噬世界的人，你知道，你已经到了一个地步，你往里面投钱，然后却被底下的实际机器所限制，所以我觉得这有点像在暗示硬件组件在这一波浪潮中非常重要，我们想承认这一点。
And then you want to emphasize the machine part of it, and there's kind of this deep irony, and this is from the, you know, the software's eating the world people that, you know, you've really come to a place where you pour money into something, and then you're limited by the actual machines below it, and so I think it is kind of a nod to like the hardware component is so significant in this wave, and we want to acknowledge that.
对，我觉得那会带来下一个突破。
Yeah, I think that's what's gonna create the next breakthroughs.
底层机器的质量决定了整个体系，所以这基本上就是名字的由来。
Tenantologies is the quality of the machines underneath, so that's basically the reason for the name.
这名字也挺酷的。
It's also a cool name.
机器时代。
Machine Age.
听起来不错。
It sounds good.
很有未来感。
Futuristic.
是的。
Yeah.
考虑到到目前为止在AI基础设施上已经投入了这么多资金，而且这些业务的资本支出强度可能很高，我们是不是已经过了新公司还能在实质性层面上打入市场的阶段？
The given how much has been spent on AI infrastructure to date to date and how CapEx intensive these businesses can be, are we past the point where new companies can break in at sort of material levels?
为什么像NVIDIA、Quariv这样的现有巨头不能直接拿走这些市场的大部分份额呢？
Why not incumbents like NVIDIA, Quariv, etcetera, just take the lion's share of these markets?
他们都做得很好。
They're all doing well.
这一点毫无疑问。
There's no question about it.
对吧？
Right?
但回到我们之前的讨论，你需要根本性的新创新来保持增长持续，或者改进速度持续，无论是每秒token数、每美元token数、每瓦特token数、每机架token数，还是电力，随便哪个指标。
But to our discussion earlier, when you're get you need fundamentally new innovations to keep the growth continuing, or the pace of improvement continuing, whether it's tokens per second, per dollar, tokens per watt, tokens per rack, right, or power, you take any metric.
如果你想让这些指标提升10倍，就必须有新的创新。
If you wanna have a 10 x on those metrics, you gotta have new innovation.
而新的创新通常来自那些才华横溢的创始人，他们从第一性原理出发，用不同的方式思考如何解决问题。
And new innovation traditionally comes from brilliant founders thinking about solving the problem from first principles in a different way.
对吧？
Right?
而这就是下一轮创新突破所需要的。
And that's what's needed here for the next jump in innovation.
我的意思是，这就是市场规律，对吧？
I mean, is the law of markets, right?
你看，假设现有的芯片巨头市值都是数万亿美元，这绝对是事实。
I mean, let's assume that the existing Silicon incumbents are multi trillion dollars in market cap, which is absolutely the case.
哪怕只是其中的5%，那也是一家巨大的私营公司，对吧，我们说的是，你可能会说，英伟达也能做到，他们确实能，但他们为什么要做呢？
Even 5% of that is a massive private company, massive private company, right, we're talking and you could say, well but Nvidia could do that, they could, but why would they?
如果他们专注于那90%的业务，而这些业务同样能带来增长，你总是会问这些问题，我们在云计算时代就问过，比如亚马逊为什么不做这个，在微软时代也问过，微软为什么不做这个？
If they're focused on things that are in the 90%, which is also driving the same amount of growth, and you always ask these questions, we ask these questions during the cloud days, like why wouldn't Amazon do this, would ask these questions during the Microsoft days, why wouldn't Microsoft do this?
市场有一个很自然的规律，就是一旦你达到一定规模，边缘地带就会有巨大的创新机会。
There's a very natural law of markets, is that once you get to a certain scale, there's tremendous opportunity for innovation at the margins.
对，我们合伙人亚历克斯·兰佩尔说过一句有趣的话，他当时有个初创公司叫TrialPay，他想把服务卖给Meta，也就是当时的Facebook。
Yeah, there's a funny quote from our partner, Alex Rampell, where he had this startup called TrialPay, and he was trying to sell it to, or sell his services to Meta, then Facebook.
当时丹·罗斯是公司发展部门的负责人，亚历克斯，他说这很棒。
And Dan Rose, who was the head of corp dev at the time, Alex, that's great.
听起来你能收集很多银砖，但我说，我有太多金砖了，根本捡不过来。
It sounds like you can collect a lot of silver bricks, but I'm like, I have so many gold bricks, I can't even pick them all up.
所以我最后想的就是看银砖，我觉得英伟达也处于那种情况。
So, the last thing I'm doing is looking at a silver brick, and I think NVIDIA's in that position too.
完全同意，是的。
100%, yeah.
我们之前聊到，关于模型提供商，如果你正好处于OpenAI或Anthropic能做的核心领域，那可能是比较难的位置，但除了那三到五个领域之外，其他领域可能……
We were talking about, as it relates to model providers, that if you're in the sweet spot of what OpenAI or Anthropic can do, one of their main interest areas that might be a tough place to be, but anything outside of those three to five areas might
市场扩大时，就会分化。
As markets expand, they fragment.
这种情况经常发生。
And it happens all the time.
记住，在福特早期，有鲁日河工厂。
And remember, in the early days of Ford, there was the Rouge River Plant.
真的是，把水、煤和橡胶树运进去，车就出来了。
Literally, made cars in went water, coal, and rubber trees, and out came cars.
顺便说一句，他在亚马逊丛林里买下了一整片橡胶种植园。
By the way, he bought a whole rubber plantation the Amazon Jungle.
还有一本很棒的书叫《福特兰迪亚》，因为他想完全掌控整个产业链，就在亚马逊丛林里建了一座叫福特兰迪亚的城市，那里完全美国化了，有音乐台、冰淇淋什么的。
And then there's a great book called Fordlandia, because he wanted to own the complete vertical thing, where he created this city called Fordlandia in the Amazon Jungle, which was all Americanized, bandstands and ice cream and all this kind of stuff.
一开始还真行，直到他要求大家准时上班，然后大家就不干了。
And it actually worked for a while until he made people show up to things on time, and then they were like,
去他的，滚蛋吧。所以现在你看汽车行业，当然有多级供应商，还有很多公司，这种事总是发生。
Screw this, get the fuck So now if you look at the car industry, of course, there's multiple levels of supplier, and then there's a bunch of companies, and this always happens.
所以市场扩大就会分散，一旦增长放缓，就会开始整合。
So as markets expand, they fragment, and then once that growth slows down, they tend to consolidate.
整合可能是收购，也可能是新挑战者崛起，这就是私募市场永恒的循环。
The consolidation can either be acquisition, or it can be like new challengers rise up, and that is the everlasting cycle of private markets.
对，因为应用场景在成倍增加，就算你是最大的公司，也不可能覆盖所有场景，但场景实在太多了。
Yeah, because the use cases are multiplying, and there's no way, like if you're the biggest company, you can get to the biggest use cases, but there's so many use cases.
而且，就像马丁说的，这些都是非常有价值的应用场景，只是很难做到特别好。
And all, as Martin was saying, very valuable use cases that it's just very hard to get to in a great way.
对，以前推理还只是简单的单一架构，对吧？
Yeah, even inference used to be one simple architecture, right?
现在可不一样了，变得特别复杂。
It no longer is just like, it's so complex now.
所以，你肯定能用不同的方式来优化。
So it's inevitable that you can optimize things in a different way.
顺便说一句，有个很有意思的点，很多人没意识到，就是利润率在传统的软件技术模式里其实挺高的，对吧？
By the way, here's a very interesting thing, like people don't often don't understand that like margins kind of fell out of the standard way of doing the technology with software, right?
这其实不算技术问题，一旦业务跑通了，利润率通常都不错，因为软件就是这样，尤其是你交付之后，哪怕作为服务也是。
Like it wasn't really a technology problem, like once you got the business working, you tended to have pretty good margins because that's just kind of how software works, certainly when you shipped it, but even as a service.
但AI就不一定了，我们可能真的进入了一个新时代，硬件优化对业务增长的影响会非常大，这在以前是没见过的。
And that's not necessarily the case with AIs, we may actually be entering an era where the optimization in the hardware is absolutely meaningful to the upside of the business in a way that we haven't seen in the past.
所以有很多
So there's a lot
这里有很多机会。
of opportunity here.
我们深入聊聊要投资的公司类型吧。
Let's get deeper in talking about the types of companies we'll be investing in.
也许我们可以先说说细分领域，或者如果可能的话，聊聊我们已经做的一两个投资。
Maybe we could start by either illustrating the the subsectors, or if we can, talk about a few or a couple investments that we've made.
我知道有些还没公布，但Raghu，你想带我们看看吗？
I know there's some that haven't been announced yet, but Raghu, do wanna take us down?
好的。
Yeah.
我是说，这些细分领域，就像我们一直说的，每个类别都是机会，对吧？
I mean, the subsectors, as we've been talking all along, is every one of these categories, right?
最明显的就是计算芯片。
The obvious ones are compute chips.
但现在光造芯片已经不够了，你得打造整个系统，对吧？
But these days, it's not enough to build a chip, you need to build a full system, right?
那么，系统里到底需要什么呢？
And then therefore, what goes into the system?
可能包括内存创新、网络创新、电源芯片等等。
There's potentially memory innovation, there's potentially networking innovation, there's potentially power chips, and so on and so forth.
所以这些类别里，每一个都可能出现像上市公司那样的公司，这些都是我们在关注的。
So each one of these categories are categories where you can see public company style companies emerging, and those are all things that we are looking into.
然后当你把所有东西整合起来，还需要一层软件来自动化所有这些，管理这些集群等等。
And then once you put it all together, there's a layer of software around it to automate all of these things, to manage these fleets, and so on and so forth.
所以这也是一个重要的领域。
So that is another important area.
这些环节相互叠加，但每一个类别都很重要。
So these things keep building on each other, but every one of these categories is important.
说说这些公司和普通公司有什么不同。
Talk about what's different about these kinds of companies from the usual company.
我的意思是，从我们公布的公司来看，他们的首轮融资规模都很大。
I mean, one thing you can tell from the companies we announced is their first rounds have been massive.
你知道，好几亿美元。
You know, hundreds of millions.
是不是不同类型的创始人？
Is it a different kind of founder?
或者，跟我们传统的软件相比，在建设和投资这类业务时，还有什么其他不同之处？
Or what else is different as we think about just the practice of, know, building and investing in these kinds of businesses relative to our traditional software?
嗯，我觉得你提到了一个关键点，就是很多钱在产品出来之前就投入了。
Well, I I think the big thing is you hit on one of the big things, is a lot of money goes in before they get to a product.
这就是它的本质。
And that's just kind of the nature of it.
大模型也是这样，但我觉得那条路更明确一些，而这个领域风险更大，投入也更多，比我们之前做的其他事情都要多。
Now that's true on big models too, but I would say that's a little more of a known path, whereas this has got a little more risk and a little more money than some of the other things that we've done.
你看，很多芯片行业的创始人都是过去那批人。
Know, look, a lot of the chip founders are here from the past.
比如那些懂内存制造的人，他们年纪都不小了。
You know, like the guys who know how to make memory, they're not young.
所以，这部分也不一样，但还挺让人兴奋的。
So, that part is different too, but it's kind of exciting.
对。
Yeah.
这些创始人还有一个共同点，他们都必须是系统型创始人。
The other thing about these founders, they all got to be systems founders.
我的意思是，你不能只是个研究员或者优秀的计算机科学家。
So what I mean by that is you can't just be a researcher or a great computer scientist.
对吧？
Right?
你得有能力去架构和设计芯片或整个系统，不管是什么。
Gotta be able to architect and design the chip or the system, whatever it is.
然后你还得考虑这个东西实际上怎么生产出来。
Then you gotta think about how is this thing gonna actually get manufactured.
谁会来供应这些？
Who's gonna be supplying this?
还有一大堆下游的事情，通常做软件的话，这些都不用考虑。
And a whole bunch of these downstream things, which normally if you're building software, don't have to think about all of these things.
所以真正最优秀的创始人，当然，黄仁勋就是这行的迈克尔·乔丹，对吧，他们在开始设计芯片之前，就已经把整个生态系统想清楚了，对吧？
So really the best founders, of course, Jensen is the Michael Jordan of this, right, They think the entire ecosystem right from the get go, before they start designing the chip, right?
因为瓶颈的本质，还有所有这些必须凑在一起的事情。
Because of the nature of the bottlenecks and all these things that have to come together.
所以这是很大的一个不同点。
So that's a big characteristic that is different.
还有两个环境因素也很重要。
There's two environmental factors that are important too.
第一个是实验室非常急切，他们愿意跟初创公司合作，所以我们其实很早就能得到不少信号，因为实验室在硬件还没出来之前就跟公司签协议了，这跟五年前相比是个很大的变化，对吧？
The first one is the labs are so desperate that they will engage with startups, and so like we actually have quite a bit of signal early on because the labs are inking deals with companies before they actually have hardware available, and that's a big big shift than, you know, five years ago, right?
以前你根本不可能拿着你那不太靠谱的硬件去卖给谷歌什么的，所以这确实是个变化。
Like, you just didn't go and, you know, sell your kind of janky hardware thing to Google or whatever, so that that's a shift.
第二个是资金的可获得性大大放宽了。
The second one is the the capital availability has loosened up a lot.
我觉得大家普遍认为，现在是重塑这个行业的时候了。
I think there's general consensus that that, you know, it is the time that to reshape this stuff.
所以后续融资轮次，市场上资金很充裕，当然，你肯定想投资那些有钱可投的领域。
And so follow on rounds, there's a lot of capital available, which, know, of course, you wanna be investing into areas where there's capital available.
所以整个氛围也不一样了。
And so the atmospherics are also just different.
帕特里克·卡尔森，你知道，几年前说过一句话。
Patrick Carlson, you know, remarked a few years ago.
他说，嘿。
He said, hey.
感觉现在年轻的创始人少了，不像当年扎克伯格还在上大学就创办了Facebook，或者盖茨那样，同样地创办了微软。
It feels like there's less younger founders today in in the way that, you know, Zuck, you know, in college building the next Facebook or or or Gates, you know, in the same way with Microsoft.
当然，像迈克尔·特罗尔这样的人，还是有一些年轻创始人在打造标志性的公司，但确实，像你说的，现在更多是年纪大一些的创始人在做这些公司，二十多岁的年轻人少了。
And, of course, you know, the Michael Trolls of the world, there there's still, you know, some young founders building iconic companies, but it does seem, you know, your point, that there's more older founders building these companies or less 20 year olds.
我很好奇，这是不是真的，以及为什么。
I'm curious if it resonates and why.
我觉得拉古说得对，如果你要做的项目供应链很复杂，需要制造东西，而且技术难度高，那有些经验确实有帮助。
Well, I think it's Raghu's point that if you're building something that has a very complicated supply chain, has to manufacture things, and is technically complicated, that some experience helps.
你看埃隆或者特拉维斯·卡兰尼克，他们年轻的时候做的都是软件公司。
You know, And if you look at Elon or Travis Kalanick, their companies when they were young were software companies.
即使是他们这些最厉害的人，也是先积累了一些建立公司、开发技术的经验，然后才去涉足那些更复杂、更庞大的领域。
It wasn't till they got like a lot, even those guys, the best guys needed some experience in building a company's building technology and so forth to kind of graduate to the much more kind of complicated or I would say elaborate domains.
这些领域里涉及的环节太多了。
There's just much more, there are many more moving parts in these things.
所以你看，当你学习怎么建立一家公司时，如果你完全理解产品，那已经够难的了。
And so, look, when you're learning how to build a company, it's hard enough if you completely understand the product.
如果你不完全理解产品，还得一边建公司一边学，这对一个全新的创业者来说，学习曲线太陡峭了。
If you don't completely understand the product and have to learn it while you build the company, That's just such a steep learning curve for a brand new entrepreneur.
所以我觉得我们现在看到的是，一方面像迈克尔，他非常年轻，很聪明，但他做的是纯软件和人工智能的东西。
So I think that what we're seeing is you see Michael on the one hand, who is a very young guy, brilliant, but what he built was kind of a pure software AI thing.
另一方面，像埃隆或者特拉维斯，他们有足够的经验。
And then on the other end, you have like an Elon or a Travis who can who's got enough experience.
我觉得迈克尔大概十年后能做到这一点。
I think Michael could probably do that, you know, ten years from now.
但放在今天，那会很难。
But today, that would have been hard.
重要的是要记住，过去二十年里，整个行业和学术界都淡化了这一点。
It's important to remember, it's been defocused by the entire industry and academia for the last twenty years.
它不再像以前那样是一个同等机会了。
It just hasn't been the same opportunity.
这块市场一直存在，但从来不是增长的重点领域。
It's been there, but it's never been a growth area.
真正增长的是软件、网络这些方向。所以我觉得，高校里或者大企业里，有相关经验的人才实在太少了。
The growth areas have been software, networking, things like that, and so I also think we just have a paucity of people coming out of the universities, or having experience at large companies that have done this, and there's just not that many.
毕竟没人会去实习时专门造芯片。不过现在情况正在改变，我们会培养出一批从这些新公司走出来的创始人，他们懂得怎么做硬件。而且你会发现，很多年轻人才会被招进来。我认为埃隆在这方面留下的最大遗产之一，就是他创立了那些伟大的公司，但更重要的是，从SpaceX走出来的一大批创业者正在重塑整个工业体系，这份遗产甚至比公司本身还要重要。我相信计算机和硬件领域也会迎来同样的变化。
Like you don't go intern and build a chip, So but a lot of that's changing now, like listen, we're gonna create a whole generation of founders that come from these new companies that will know how to do this, and you know, they'll be hired in much more junior, and like, I would say actually one of the greatest legacies of Elon towards this is of course, he's created these great companies, but the amount of entrepreneurs that have come out of SpaceX that are changing the entire industrial complex, maybe even greater legacy than the companies themselves, and I think we're gonna see the same thing for computer science and hardware.
确实如此。
Yeah.
事实上，我们投资的一家公司，是由两位二十多岁的创始人创办的，但如果你去他们的办公室看看，也会看到有经验的人，所以这是一个理想的组合。
As a matter of fact, one of our investments, was started by two founders in their twenties, but if you go watch one of their offices, you see the experienced people as well, so it's an ideal combination here.
对。
Yeah.
不一定非得是创始人自己有经验，但创始人必须能够利用那些经验，对。
It doesn't necessarily have to be the founder with experience, but that founder better be able to tap into that experience Yeah.
而且，还要能和他们合作，他们也得是优秀的人，还有诸如此类的事情。
And a real Well, and then be able to work with them and they have to be good and and all these kinds of things.
这很复杂。
It's complicated.
说到经验，这是我们正在推出的一只大型新基金，而且没有新的普通合伙人。
Speaking of experience, this is a a big new fund we're we're we're launching, and there's no new GPs.
对。
Yeah.
我们算是正在募集。
We're sort of collecting.
这是因为你们团队其他人在这个领域有很多经验，你知道，这个领域之前一直比较沉寂。
It's because you guys have a lot of experience in the rest of the group, you know, in this field that has been kinda latent and dormant.
对。
Yeah.
嗯，说起来挺有意思的。
Well, it's kinda funny.
我觉得我们几乎需要被提醒不要这样做，因为我们的背景……这很难说。
I think we almost had to be warned against it almost, just because our backgrounds are from it's hard.
我觉得我们需要提醒的原因是，这一切都太……
And I think the reason that we needed a reminder is because all of this has been so much
在我们的职业生涯中，比如系统和硬件，我们很容易被吸引过去。
in our careers, like systems and hardware, we're kinda drawn to that.
所以你看，这些年来我们明显在硬件上投入了很多。
And so listen, we've been clearly invested in hardware over the years.
对吧？
Right?
我们是在SpaceX。
We're in SpaceX.
我们参与了。
We're in.
这些都是非常早期的投资。
These are very early checks.
参与了。
In.
你知道吗？
You know?
所以即使在早期阶段，我们也做了不少这样的投资。
So we even even early hours, we did a number of those investments.
但是，你知道，这已经深深印在我们的骨子里了，所以我
But, like, you know, this is because it's so much in our DNA, and so I
觉得没有必要扩大团队或提升能力。
don't think this is necessary to increase the team or competencies.
这只是额外的收获。
It's just bonus.
如果这个基金能达到我们预期的效果，那五到十年后，我们会看到世界变成什么样呢？
If this fund does what we think it will do, how do we see the world changing or look looking like in in five to ten years?
嗯，你知道，希望美国能在基础设施方面胜出。
Well, you know, hopefully, America wins in the infrastructure ding.
而且我们会有很多超级环保、高效的数据中心，还有大量的芯片、内存和电力。
And we have lots of, like, super eco friendly efficient data centers out there and lots and lots of an abundance of chips and abundance of memory and abundance of power.
那真是太棒了。
And, you know, that would be awesome.
而且我觉得，这又回到了我们真的认为美国是个特别的地方。
And I think we like we, it goes back to like, we really think America is a special place.
我们不仅对这里的每个人重要，也对世界上任何想要做出贡献、做点超越自己的事情的人重要。
And we're important not only to everybody here, but anybody in the world who wants to kind of make a contribution and do something bigger than themselves.
这里是那种你一无所有却能做出非凡成就的最佳之地。
It's kind of the best place to come with nothing and do something profound.
所以我们希望继续保持这种势头。
So we'd like to keep that going.
我觉得如果我们在技术上失去领先地位，这种势头就难以持续。
I think that doesn't continue to go if we lose our lead in technology.
我想我们会进入另一个时代，由另一个国家主导，而他们可能有着不同的价值观。
I think we'll be in another era, and it'll be another country, and maybe they have a different set of values around them.
感谢你今天的总结。
Thanks for this wrap.
很好。
Great.
感谢Streaming Us基金团队的Ben和Raghuram。
For Streaming Us Fund, our team, Ben, Raghuram.
谢谢。
Thank you.
谢谢。
Thank you.
本也谢谢。
Ben too.
谢谢。
Thank you.
感谢收听本期a16z播客。
Thanks for listening to this episode of the a 16 z podcast.
如果你喜欢这期节目，记得点赞、评论、订阅，给我们评分或写个评论，并且分享给你的朋友和家人。
If you like this episode, be sure to like, comment, subscribe, leave us a rating or a review, and share it with your friends and family.
想听更多节目，可以去YouTube、Apple Podcast和Spotify。
For more episodes, go to YouTube, Apple Podcast, and Spotify.
在X上关注我们@a16z，并订阅我们的Substack：a16z.substack.com。
Follow us on x at a sixteen z, and subscribe to our Substack at a16z.substack.com.
再次感谢收听，我们下期见。
Thanks again for listening, and I'll see you in the next episode.
提醒一下，这里的内容仅供参考，不应视为法律、商业、税务或投资建议，也不应用于评估任何投资或证券，且不针对任何a16z基金的投资者或潜在投资者。
As a reminder, the content here is for informational purposes only, should not be taken as legal business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a16z fund.
请注意，a16z及其关联公司可能也持有本播客中讨论到的公司的投资。
Please note that a16z and its affiliates may also maintain investments in the companies discussed in this podcast.
更多详情，包括我们投资的链接，请访问a16z.com/disclosures。
For more details, including a link to our investments, please see a16z.com forward slash disclosures.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。