安德鲁·费尔德曼谈打造 Cerebras 与芯片的未来

你每六周开一次董事会，而你能说的只是“还是做不出来”。对吧？这就是董事会？我是说，我们还能聊什么？

You got board meetings every six weeks and all you've got to say is still can't make it. Right? Is that the board meeting? I mean, what else are we going to talk about?

一次又一次还是做不出来。

Still can't make it again and again.

然后他们会说：“我该介入吗？我能帮上什么忙吗？”

And they're like, "Should I get in? Can I do anything to help?"

没错。他们会说：“我们能帮忙吗？” 回答是：“不用。” 但我们始终相信，如果我们能做出来，需求会非常大。

That's right. They're like, "Can we help?" It's like, "Nope." But we always believed that if we could make it, there would be huge demand for it.

好的。今天做这个非常兴奋。Andrew，非常感谢你来。Andrew，你是 Cerebras 的 CEO 和创始人，这显然是最重要的芯片公司之一。同时也很高兴我的合伙人 Eric 在这里，他将半主持半嘉宾，他从一开始就和 Andrew 合作。感谢你们两位抽出时间。我想从这开始：现在是 2026 年，Cerebras 在 AI 领域非常重要，但你在 2016 年创立了公司，那时 AI 还不是现在这样，所以我想当时的心态是什么，想法和洞察是什么，你当时在构建什么，思考是什么？

All right. Super excited to be doing this today. Andrew, thanks a bunch for being here. So Andrew, you're the CEO and founder of Cerebras, which is obviously one of the most important chip companies. And also really happy to have my partner Eric here, who's going to be kind of half host, half guest, who has been working with Andrew since the beginning. But thank you both for making time. What I want to start with is it's 2026 now and obviously Cerebras is very important in the sort of AI landscape but you started the company in 2016 and AI was not what it is now back then so I guess what was the kind of headspace then what was the idea the insight like what were you what were you building when you started the company and what was the thinking?

嗯，Jack，感谢邀请。和 Eric 一起总是很有趣，感谢你邀请我。我认为作为计算机架构师，当你看到一个新的工作负载出现时，你会兴奋，对吧？在成熟的计算市场赢得份额非常非常困难，所以当新事物出现时，你会兴奋，你会向前倾，你会说：“我能让它更快吗？我能造一个在这方面更好的芯片吗？”然后你问另一个问题：“这种工作足够多吗？值得造一个更好的芯片吗？”所以有两个问题：我能吗？我应该吗？我们在 AI 中看到的是一个极其计算密集型的工作负载。不像例如手机 ARM 处理器的崛起，它们不是计算密集型，而是功耗密集型。对吧？我们看到一个问题，计算上会很困难，我们看到正在使用的架构正在为此改造，我们可以造出更好的东西。这就是洞察，然后就是漫长而艰难的跋涉。

Well Jack thanks for having me it's always fun to hang out with Eric and so appreciate you having me. I think as a computer architect when you see a new workload on the horizon you get excited right it's very very difficult to win share in a mature market in compute and so when something new emerges you get excited and you lean forward and you say well can I make it faster can I build a chip that is better at this work and then you ask the other question is there enough of this work to justify building a chip that's better? So you have two questions. Can I and should I? And what we saw with AI was an extremely computationally intensive workload. Unlike for example the rise of ARM processors for cell phones that they weren't computationally intensive, they were power intensive. Right? We saw a problem that was going to be hard on compute and we saw that what was being used was an architecture that was being remodeled for it and that we could build something better. And that was sort of the insight and then it was a long hard slog.

在我们进入漫长而艰难的跋涉之前，Eric，显然对你来说，作为投资者，你知道，你当时并不在这个 AI 时代，但你看到了什么让你投资，而且我认为你以前不是芯片投资者，你做的更多是

Before we go to the long hard slug, Eric, obviously for you, like as an investor, you know, you were not, you know, also in this AI era, but you saw something that made you invest and you were, I think, you know, previously not a chips investor and you done more like

而且可能再也不会了。

and probably never will be again.

可能再也不会了。比如，你每次都会再做一次吗？有什么让你清楚，还是当你投资时发生了什么？

Probably never will be again. Like, would you do that every time again? Was there something that made it clear to you or was it like what what happened for you when you made the investment?

我认为风险投资最可怕的事情之一，我一直在想，就是你越是在公司上工作，你看到挑战，你看到有多难，你就会积累疤痕组织，然后事情就像，好吧，你应该再做一次，但你会再做一次吗？我一直在问自己，实际上你需要一些天真，你需要有这种“我能做到”的态度和这种天真的勇气来尝试，否则你就不会再做了。我肯定积累了一堆疤痕组织。我不知道如果我当时知道对他们来说会有多难，需要建立多少科学和技术，多少创新，公司会多少次走到那个点。

I think one of the scariest things about venture and I think about this all the time is the more you work on companies and you see the challenges and you see how hard it is you build up scar tissue and then the thing is like well you should do it again but like will you do it again and I ask myself that all the time like there you actually need some naivete like you need to have this like oh can do attitude and this like naive vet around it to attempt it like otherwise you just don't do it again. And I've definitely built up a bunch of scar tissue. And I don't know had I actually had any idea how hard it was going to be for them, like how much science and technology had to be built, how many innovations, how many times like the company was going to come to the point.

别告诉你的风投你的鞋子有多难。故事的寓意是告诉他们没问题。把它藏在心里。压下去。压下去。只是别告诉他们难题。一切都好，伙计。

Don't tell your VCs how hard your shoe is. And the moral of the story is tell them it's no problem. Keep it inside. Push it down. Push it down. Just don't don't tell them hard problem. All's good, man.

是的。那么，发生了什么？基本上，你创立公司，你到了流片，然后你就完成了，对吧？就像那样。

Yeah. So, what happened? So, you basically, you know, you you start the company, you get to the tape out, and then you're done, right? It's like that.

不，从来没有。

No, never.

那么，进展如何？发生了什么？

So, how's it go? So, what what happened?

我认为，嗯，我们，你知道，这不是我们的第一家芯片公司。这是我的第五家。团队，创始团队在上一次就在一起，我们对应该做什么有一些非常清晰的理念，我们的观点是，如果你要挑战歌利亚，如果市场上有一个巨人，就像当时的 Nvidia，那么稍微好一点或稍微便宜一点并不是可用的策略，对吧？他们有高利润率，所以如果你进来，即使你是三分之二的价格，他们也可以降低成本。他们可以少收费。他们可以捆绑。他们可以做一百件其他事情。这意味着你必须拿出更好的东西，快 10、100、500 倍。你必须拿出一个价值主张，即使其他人免费送，他们也无法竞争。所以这是我们思考的一个方向。另一个建立在此基础上的是，我们想做能产生这种优势的难事，对吧？我们希望所有难的东西都在内部，因为那是我们控制的。所有其他东西不在我们控制之下。但如果我们能造出如此快的东西，没人能做到，你不能免费送你的部件来达到我们能达到的。那么你就有了。在我们看来，唯一的方法就是激进创新。

I think uh we we had uh you know, this isn't our first chip company. This uh this is my fifth. and the team had been together uh the founding team had been together at the last one and we we had some really clear sort of philosophies on on on what you should do and I our view is that if if you're going to attack Goliath if if there's a a giant standing in the market like Nvidia was even at that time that being a little bit better or a little bit cheaper is is not an available strategy right they had high margins so if you come in and and Even if you're two/ird the price, uh they can just cut cost. They can just charge less. They can bundle it. They can do a hundred other things. What that means is you have to go out with something way better, 10, 100, 500 times faster. You have to come up with a product that has a value proposition that that even if the other guys give it away, all right, they can't compete. And so that that's sort of one direction of our thinking. And the other building off that is is we want to do hard things that produce that advantage, right? That we want all that hard stuff within the building because that's under our control. All the other stuff's not under our control. But if we can build something that is so fast nobody else can do it, you can't give away your parts to achieve what we can achieve. Um then you're on to something. And the only way to do that in our opinion is radical innovation.

你不能通过渐进式改进达到大幅提升。

You can't incremental your way to vastly better.

那么什么必须在你的控制之下？

So what has to be in your control?

你必须拥有设计、实现的所有方面，你必须理解制造，你必须理解每一个部分。对我们来说，这意味着芯片、板卡、系统、软件，一直到 API。那并不容易。那让它更贵。让它花更长时间。意味着有更多失败的方式。当你做激进创新时，没有供应商在等你，对吧？当你造一个晚餐盘大小的芯片时，你不能去目录里找散热器，对吧？因为没人造过比邮票更大的。所以没人有现成的东西给你。所以你最终投入大量时间构建支持你的东西，对吧？周围的组件。但结果是，你发展出这种非凡的专业知识。我们开始时不是封装的世界领导者。我们现在是世界最好的封装。我们通过一次又一次的失败，年复一年地赢得它，直到我们做到。所以你想做的是，你想思考如何在完整解决方案的各个元素中创新，并尽可能努力推进。

You have to have all aspects of the design, the implementation, you have to understand the manufacturing, you have to understand every part. And for us that meant chip, board, system, software, all the way up to the API. And uh that was not easy. That makes it more expensive. It makes it take longer. It means there are more more ways to fail there. When you do radical innovation, there are no vendors waiting for you, right? When you build a a chip the size of a dinner plate, you can't go to a catalog and find a heat sink, right? Because nobody' ever built one bigger than than a postage stamp. So nobody has stuff ready for you. And so you you end up investing an enormous amount of time in in building things that support your thing, right? The the surrounding components. But the result of that is is you develop this extraordinary expertise. We didn't start world leaders in packaging. We're right now the best in the world of packaging. Um we earned it failure after failure year after year until we got it. And so what you want to do is you you want to think about sort of h how you can innovate across the the the various elements of of a full solution and push as hard as you can.

我想 Andrew 在半导体课和其他场合教过我这一点，就是为了让这个具体化——你需要瞄准的目标。所以如果你想想任何市场里的现有霸主，你知道，他们每年大概变好两倍。在这样一个市场里，变化非常快。一家新公司要达到规模，开始达到规模，至少需要五年，假设所有人都在用 AI 工具尽可能快地推进。你一切顺利——第一次流片成功，第一次上电成功，一切都对——五年，好吧，达到规模。所以你基本上就是 2 的 5 次方。那就是 32 倍。然后你还需要至少一个倍数优势。比如 3 倍。所以你基本上就是 100 倍。所以当你听到很多这些新想法时，会发生的一件事是，他们说，‘嘿，我们会好 50%。我们会好两倍。’但他们说的是，‘一旦我们有了这个设计，它会比今天现有的好 10 倍。’但今天现有的并不是目标。因为如果你看看现在市场上的公司，这些拥有大量资源的大公司实际上在做——他们做得很好。每个人——Nvidia 每年都在实质性地推进，非常显著。

I think Andrew taught me this in semis and everything else, just to make this concrete — what you need to be shooting for. So if you think of whoever the incumbent is in whatever market, you know, they're getting, say, twice as good every year. In a market like this, it's very dynamic. For a new company to get to scale, to start to get to scale, it's going to take five years at least, let's say, with everyone moving as fast as they can with AI tools. You nail everything — first tape-out works, first bring-up works, everything goes right — five years, okay, to get to scale. And so you're basically at two to the fifth. So that puts you at 32x. And then you need at least a multiple advantage of that. So you say 3x. So you're basically at 100x. So one of the things that happens when you hear a lot about these new ideas is they're like, 'Hey, we're going to be 50% better. We're going to be two times better.' But what they are, they're like, 'Once we have this design, it's going to be even 10 times better than what exists today.' But what exists today isn't the target. Because if you're looking at the companies that are out there right now, these big companies with a lot of resources are actually doing — they're doing great work. Everyone's — Nvidia is advancing the ball materially every year, like very significantly.

Nvidia 和 Google 以及 Tranium，他们都在推进。

Nvidia and Google and Tranium, they're moving the ball.

所以如果你要到达，你必须瞄准好 100 倍、50 万倍。

And so you got to aim at 100, 500,000 times better if you're going to arrive.

是的，如果你要赶在他们之前切入市场。

Yeah, if you're going to intersect the market ahead of them.

所以当你意识到这一点时，你会想，‘哦，等等。这不可能是某个东西的小改动。必须有一个非常大的底层架构变化，一个大的想法。必须有某种实质性的不同，无论是晶圆级、还是别的什么，那些东西很重要。’

And so that just like when you realize that you're like, 'Oh, wait. It can't be small changes in something. There has to be a very big underlying like architectural change, a big thought. Like there has to be something materially different, whether it's wafer scale, es whatever it is, like there those things matter.'

你不可能靠一堆适度的改进到达那里。

You can't get there with a collection of modest improvements.

嗯。

Mhm.

对。你最大的竞争对手买硅比你便宜。他们买制造产能比你便宜，对吧？他们可能付更少的 EDA 工具费用。所以你不能直接冲向他们，你知道，你不能正面硬刚歌利亚。我的意思是，你必须思考如何交付某种深刻不同的东西。

Right. Your biggest competitor buys silicon for less than you. They buy manufacturing capacity for less than you, right? They probably pay less for their EDA tools. And so you can't go at them, you know, you can't run at Goliath straight on. I mean, you got to think about how you can deliver something profoundly different.

那些——在这些早期时刻之后，一路走来是什么？我知道公司有过，你知道，一些濒死体验之类的。那些——那些艰难的障碍是什么？

What were the — like along the way after some of these like early moments? I know that the company had like, you know, some near-death experiences or whatever. Like what were the — what were sort of the hard hurdles?

我有过一些濒死体验，当你无法造出你应该造的东西时，胸口疼。是的。

I had some near-death experiences, pain in my chest when you can't build the thing you're supposed to build. Yeah.

那些时期——你会如何描述那些时期，你会想我们只需要——我不知道——是需要一定的时间和一定的资本，然后就是一个死亡之谷？还是你只是不知道你是否会取得技术突破？是什么导致那些时期？

Like what are those per — like how would you frame what those periods are where you're like we just need — I don't know — is it a certain amount of time and a certain amount of capital and it's just going to be a valley of death? Is it that you just don't know whether you're going to make a technical breakthrough? Like what are the things that lead to those?

我认为在我们这个领域，我们一直有信心，如果我们能造出来，我们就能卖出去，对吧？我们的生活中有两个轴：一个是你能不能造出来，另一个轴是你能不能卖出去，对吧？我们一直相信，如果我们能造出来，就会有巨大的需求。我认为从来没有人做过晶圆级。有一大群人一直在说这永远不可能成功，我们也不确定。我的意思是，我们相信我们能做，但没有证据。有一段时间大约 18 个月，我们造不出来，我们每月花大约 800 万美元。每六周开一次董事会，你只能说还是造不出来，对吧？那就是董事会。我的意思是，我们还能谈什么？还是造不出来，一次又一次又一次。我给了巨大的——

I think in our space we were always confident if we could make it we'd sell it, right? There are sort of two axes in our life: there's can you make it, and the other axis is can you sell it, right? And we always believed that if we could make it there would be huge demand for it. I think nobody had ever done wafer scale. There was a whole sort of collection of people who were saying it could never work, and we were unsure too. I mean, we believed we could do it, but there was no evidence. And there was a period of time about 18 months and we couldn't make it and we're spending about 8 million a month. You got board meetings every six weeks and all you've got to say is still can't make it, right? That's the board meeting. I mean, what else we going to talk about? I mean, still can't make it again and again and again. And I give a huge —

对。对。

Right. Right.

他们会说，‘我该介入吗？我能做点什么帮忙吗？’

And they're like, 'Should I get in? Can I do anything to help?'

没错。他们会说，‘我们能帮忙吗？’ 就像，‘不，’

That's right. They're like, 'Can we help?' Like, 'No,'

没什么可做的。没什么可做的。但我们相信，因为我们还有想法。我们还没有用尽想法。每次我们造出来然后失败，我们都会遵循良好的工程实践。我们会做完整的失效分析。我们会理解它。我们不会再以那种方式失败。

there's nothing that can be done. Nothing to be done. But we believed because we had ideas still. We hadn't run out of ideas. And each time we built it and it failed, we'd go through sort of good engineering practice. We'd do a full failure analysis. We'd understand it. And we wouldn't fail that way again.

全是新错误。

All new mistakes.

全是新错误。所以我们有那个——那是我们的座右铭。只有新错误，对吧？只有新失败。随着时间的推移，我们能看到进步。我的意思是，一开始，我们在几秒钟内就碎裂晶圆，对吧，然后需要几分钟，然后我们有一次运行了一个小时，然后我们又几分钟内碎裂了一些，然后我们重新积累到——2019 年 7 月的一天，我们正在运行，温度平稳，我们就站在一个被改造成实验室的小办公室里，我们在墙上钻了一个洞把空气抽出去，盯着一个服务器，那和看一个油漆桶一样无聊。说，‘天哪，我们解决了 75 年来计算领域没有人解决过的问题。’那是我一生中最伟大的时刻之一。

All new mistakes. So we sort of had that — that was sort of our mantra. Only new mistakes, right? Only new failures. And over time we could see progress. I mean, in the beginning, we were shattering wafers, right, in seconds, and then it took minutes, and then we had one run for an hour, and then we shattered some in minutes again, and then we built back up to the point — we had a day in sort of July of 2019 where we were running and temperature was flat and we just stood there in a tiny little office that had been converted into a lab and we drilled a hole in the wall to suck the air out and stared at a server which is about as exciting as looking at a paint drive. Said, 'Holy crap, we've solved this problem that nobody in 75 years of compute had ever solved.' And that was one of the great minutes of my life.

那很酷。当你向前看——就像如果你——所以你知道有晶圆级，然后围绕它的一切，封装、冷却和供电，所有这些。但你知道，当你看看未来五年你拥有的研发范围，对吧？我知道有你兴奋的事情，但比如你把这个巨大的东西带向哪里？下一个晶圆级创新可能是什么？

That's cool. When you look forward — like if you — so you know there's the wafer scale and then everything around that, the packaging and cooling and powering it, all these things. But you know when you look at the R&D envelope that you have going forward over the next five years, right? I know there's like things you're excited about, but like where do you take — like where do you take this like giant thing? Like what could be the next wafer scale innovation?

我认为如果你把我们构建的东西简化到最基本的元素，计算机由三样东西构成。我们做计算，我们存储结果，然后我们把结果移动到它们有用的地方。所以我们构建一个做计算的核心。我们使用内存，我们把数据移入和移出内存，在那里我们存储结果，然后我们有 IO，这就是我们把结果运送到有用地方的方式。我认为如果你像我们一样从事 AI 计算机业务，你最好在这三个方面都努力。你最好思考如何让你的核心更快，如何让它们为 AI 调优，但又足够通用以承受 AI 中正在发生的创新。你最好思考内存以及容量，你能存储多少以及你能多快地把数据移入和移出。然后你必须思考如何把结果从你的芯片上弄到有用的地方，那就是你的 IO。所以我们正在这三个方面努力。

I think if you simplify what we build down to its most fundamental elements, a computer is built of three things. We do calculations, we store the results, and then we move the results to where they're useful. So we build a core that does the calculation. We use memory and we move data to and from memory where we store the results, and then we have IO and that's how we ship the results to somewhere where it's useful. I think if you're in the computer business like we are for AI, you better be working on all three. You better be thinking about how to make your cores faster, how to make them tuned for AI, but general enough to withstand the innovation happening in AI. You better be thinking about memory and both capacity, how much you can store and how fast you can get data on and off it. And then you got to be thinking about how you get the results off your chip and somewhere where they're useful, and that's your IO. And so we're working on all three of those.

我们已经在和美国政府合作一些大项目，思考如何堆叠内存。如何把 HBM 放到基于 SRAM 的晶圆上。这样做的目的是让 HBM 表现得像 SRAM，这正是大家想要的。你既能获得 HBM 的容量，又能获得 SRAM 的速度。我们正在研究光学晶圆堆叠。把光开关放到晶圆上，这将改变世界。Jack Dongarra 说过——他是大型计算领域的先驱之一——他说：“我们在制造浮点运算和移动浮点运算方面做得更好，那就是 IO 部分。”因此，思考如何通过将光交换紧贴计算来解决这个问题，也是我们拥有大型政府合同的项目，我们对此非常兴奋。所以，你必须更快。你必须找到存储更多数据并更快访问内存的方法。你还必须找到方法，以 10 倍、100 倍、1000 倍的速度将那些结果移动到有用的地方。

And we have programs with the US government already, big programs where we're thinking about how to stack memory. How to put HBM onto an SRAM-based wafer. And what this does is it enables HBM to behave like SRAM, which is exactly what everybody wants. You get the capacity of HBM and the speed of SRAM. We're working on optical wafer stacking. So you put an optical switch onto a wafer and this would change the world. Jack Dongarra said—and he's one of the pioneers in big computing—he said, "We've been better at making flops and moving flops, and that's the IO part." And so thinking about how to solve that problem by bringing optical switching smack up against compute is something that we also have large government contracts for and we're enormously excited about. So you've got to get faster. You've got to find ways to store more and get to memory faster. And you've got to find ways to move those results at 10, 100, a thousand times faster to where they're useful.

我想在供应链方面谈得更广一些。我认为现在大家都很清楚，AI 非常受供应限制，但可能和许多人一样，我不会说我对供应链有完美的理解。在这个播客中，我最喜欢做的事情之一就是，像你这样的成功人士在场，而且有摄像机在拍，我可以问你一个非常简单的问题，并且得到你的包容。所以，你能用相对简单的方式解释一下，我们如何从沙子变成 ChatGPT 的答案吗？

I want to go a bit broader in the supply chain. And I think it's kind of well understood right now that AI is very supply constrained, but like probably many other people, I wouldn't say I have a perfect understanding of the supply chain. And one of my favorite things doing on this podcast is you get a successful person like you and because there's cameras on, I get to ask you a really simple question and I get to be humored with it. So could you explain sort of in a somewhat simple way how do we go from like sand to a ChatGPT answer?

老兄，就是台积电。就这样。有一个黑箱。你称之为台积电，通过另一个叫 ASML 的黑箱。然后另一端就出来了很棒的东西。

Dude, it's just TSMC. That's it. There's a black box. You call that TSMC by another black box called ASML. And out the other end comes comes great.

我们受供应限制。但我不知道——你能教我吗？供应——它是如何运作的？

We're supply constrained. But I don't know what the—can you teach me what the supply—how it works?

是的。我认为首先要考虑的是，一个晶圆厂，尤其是建造尖端几何尺寸的晶圆厂，是一座现代金字塔。它是人类制造的最伟大的东西之一。

Yeah. I think the first thing to think about is that a fab, and especially a fab that builds at cutting edge geometries, is a modern pyramid. It's one of the greatest things humans make.

它是什么？

What is it?

它是一组机器，取一块硅，使用光刻工艺，将晶体管蚀刻到那块硅中，这样当你供电时，它们就会进行计算。它是一组——我的意思是，这是一个工厂，只是一个合理的方式——一个真正的工厂要花费 400 或 500 亿美元建造，寿命为四到五年。

It is a collection of machines that take a chunk of silicon and use a photolithographic process in which they etch transistors into that piece of silicon such that when you deliver power to it they do calculations. And it is a collection—I mean this is a factory, it's just a reasonable way—a real factory costs 40 or 50 billion to make, it has a four or five year lifetime.

它有多大？

How big is it?

足球场那么大。

Football fields.

好的。

Okay.

当三星在得克萨斯州建厂时，他们开始建造一座发电厂。发电厂用来制造混凝土。他们运行混凝土卡车，数百辆混凝土卡车，每周 7 天，每天 24 小时，持续数年，浇筑足够的混凝土来建造放置晶圆厂的地基，对吧？这些是难以置信的复杂事物。

When Samsung was building a factory in Texas, they began building a power plant. The power plant was used to make concrete. They ran concrete trucks, hundreds of concrete trucks, 7 by 24 for years to pour enough concrete to build the foundation on which to put the fab, right? These are unbelievably complicated things.

所以，ASML 制造机器。ASML 制造进行光刻的机器。那台机器是什么样的？

So, ASML makes the machines. ASML makes the machine that does the photolithography. What's that machine like?

它的尺寸——每台机器大约 50 或 60 英尺长，20 英尺高。

It's the size of—each machine is what, 50 or 60 feet long and 20 feet high.

成本大约 5 亿美元？

Costs what, half a billion?

是的，它们很贵。有趣的是，他们把同样的机器卖给不同的晶圆厂，晶圆厂以不同的方式使用它们，并能够做不同的事情，台积电能够实现别人无法实现的事情。

Yeah, they're expensive. What's interesting is that they sell the same machines to different fabs and fabs use them in different ways and are able to do different things with TSMC able to achieve things that others can't.

还有其他人能制造这些机器吗？真的只有 ASML 能制造这些机器吗？

And can anyone else make these machines? Is it really just ASML that can make these machines?

目前，只有 ASML。

Right now, it's only ASML.

在世界上。

In the world.

在世界上。

In the world.

这很奇怪，不是吗？

That's weird, isn't it?

是的。我的意思是，很少有东西——这是一个真正的垄断，不是像戴比尔斯那样试图控制供应。他们拥有其他人无法复制的技术。

Yeah. I mean there are very few things that—this is a true monopoly not born of what De Beers did, which was try and control supply. They have technology that others haven't been able to replicate.

我们在光刻机上受供应限制吗？

Are we supply constrained on lithography machines?

可能不是。我认为我们行业的挑战在于需求变化极快。如果你看看像英伟达这样的伟大公司的需求，它是指数级的。但你不能那么快地建造晶圆厂。如果一个晶圆厂需要 5 年才能建成，而需求却在增长，没有人会预料到芯片需求会像现在这样飙升，他们总是落后。

Probably not. I think the challenge in our business is that demand moves extremely quickly. And if you've looked at a great company like Nvidia's demand, it's exponential. But you can't build fabs that fast. If a fab takes 5 years to build, and you see demand increasing, there's no way anybody would have expected demand to shoot up the way it has for chips, they're always behind.

是的，他们正在对这些设施的建设投入巨额资本，而需求正超过做出这些巨大的 5 年赌注的能力，我们现在在数据中心方面也有同样的问题，对吧？AI 以软件的速度发展，而数据中心以房地产的速度发展，这就是为什么我们落后。就是这样。

Yeah, and they're investing huge capital blocks in the construction of these facilities and demand is racing past the ability to make these huge 5-year bets and we have the same problem right now with data centers, right? AI is moving at the speed of software and data centers are moving at the speed of real estate and that's why we're behind. That's just it.

良好的表达很好。好的，所以台积电购买光刻机，对吧？台积电购买光刻机。他们有数十年的经验。他们把它整合在一起。他们取一个硅锭，是一管硅。它被切成我们称为晶圆的圆盘。晶圆被送入流程。流程通过光刻工艺将晶体管蚀刻到硅中。它通过流程移动。它们被切割——我们称之为切割——或者在我们的情况下，它们不被切割。另一端出来的是芯片。令人惊奇的是，你把你见过的一些最聪明的人放在一个项目上两年，然后把结果发送给台积电，他们通过一个花费 400 或 500 亿美元建造、耗时数年建成的工厂运行它，另一端出来的是一个 22 美元的芯片。当你想到这一点时，它会让你大吃一惊。每一个都需要各自领域最聪明的人。

Good articulation is good. Okay, so TSMC buys the lithography machine, right? So TSMC buys lithography machines. They have decades of experience. They wrap it together. They take an ingot which is a tube of silicon. It's sliced into discs we call wafers. The wafers are sent through the process. The process etches through a lithographic process transistors into the silicon. It moves through the process. They're cut—we call it diced—or in our case they're not cut. Out the other end comes chips. And what's amazing about this is you take some of the smartest people you've ever met and you put them on a project for two years and you send the results to TSMC and they run it through a factory that cost 40 or 50 billion to make and took years to make and out the other end comes a $22 chip. It blows your mind when you think about it. Each one of these took some of the smartest people in their field.

这是人类做过的最令人印象深刻的事情，对吧？这是最令人印象深刻的事情之一。

It's the most impressive thing humans have ever done, right? It's among the most impressive things.

而一个墨西哥卷饼是 23 美元。

And a burrito is $23.

没错。对。正是。对。你在这附近的 Mission District 街上走，花 17 美元买一个墨西哥卷饼。没错。对。所以，

That's right. Right. Exactly. Right. And you go down the street here in the Mission District and you pay $17 for a burrito. That's right. Right. And so,

一旦你有了芯片，你就完成了。对。然后你就好了。

and once you have the chip, you're done. Right. Then you're good.

不。一旦你有了芯片，你解决了大约四分之一的问题。

No. Once you have the chip, you've solved about a fourth of your problem.

好的。但仍然停留在台积电。

Okay. But still stay TSMC.

是的，他们操作——他们接收设计，操作机器，给你芯片。

Yes, they operate the—they receive a design, they operate the machine, they give you the chip.

是的。

Yes.

但这本身似乎不可复制。就像我们在美国没有这样的晶圆厂。

But that itself seems like not replicable. Like we don't have fabs like this in the US.

我们有由台积电和三星运营的晶圆厂。

We have them run by TSMC and by Samsung.

那么——

So what—

Global Foundries 也有一个。而且

and Global Foundries have got one too. And

瓶颈有多少在台积电层面？

how much of the bottleneck is at the TSMC level?

很大一部分。

Huge amount.

好的。为什么是很大一部分？因为建造晶圆厂的速度。

Okay. Why huge amount is that? Because of the speed of building fabs.

因为需求超过了建造晶圆厂的能力。

It's because the demand has outpaced the ability to build fabs.

这不像公寓楼，有数百个建筑商可供选择来建造你的公寓楼。实际制造晶圆厂是只有台积电才拥有的技能。即使在台积电，他们也无法同时做 12 个，对吧？制造这些东西的技能如此稀有，掌握在如此少的人手中，你无法轻易搞定。这不是千篇一律的，这就是为什么我们现在没有足够的晶圆厂。事实上，在美国，三十年的糟糕政策把晶圆厂推走了。当晶圆厂离开时，工具供应商，那些为他们提供服务的供应商集合，他们都离开了。工艺的下一步叫做封装，那些是像 Amkor 和 ASE 这样的公司。他们也离开了。

It's not like an apartment building where there are hundreds of builders you could choose to build your apartment building. The actual making of the fab is skills that only TSMC has. And even there they can't do 12 at once, right? The skills to make these things are so rare and in so few hands that you can't just knock them out. It's not a cookie cutter, and that's why we don't have enough of them right now. And the fact is in the US, three decades of bad policy pushed the fabs away. When the fabs left, the tool vendors, the collection of vendors who provided services to them, they all left. The next step in the process called packaging, those are companies like Amkor and ASE. They left.

我们就这样放弃了一个战略产业，我们必须做得更好。那并不明智。

And we just punted a strategic industry and we got to do better. That was not smart.

是的。那么，你认为那里应该发生什么？

Yeah. So, what do you think should happen there?

嗯，我认为我们应该与 GlobalFoundries、台积电和三星坐下来，有一个 20 年的时期，我们免除所有当地法令，允许他们建造晶圆厂。我认为这是关于建立美国国内晶圆厂产能。

Well, I think we should sit down with GlobalFoundries and TSMC and Samsung and have a 20-year period where we wave all local ordinances to allow them to build fabs. I think this is about building US domestic fab capacity.

是的。我的意思是，看，当一艘大船试图在美国运河平行停泊时，对吧，我们在芯片上被延误了，我们买不到洗衣机。

Yeah. I mean, look, when a big ship tried to parallel park in the US canal, right, we were delayed in chips and we couldn't buy washing machines.

对。我的意思是，如果我们失去了芯片产能，对我们的行业将是灾难性的。不仅对我们的行业，而且对国家，对经济。

Right. I mean, if we lost our chip capacity it would be catastrophic for our industry. Not just for our industry but for the country, for the economy.

埃里克，作为董事会成员，当你边做边学这些东西时，你显然仍然，你知道，我不认为安德鲁在骗我。我认为你设法非常 helpful。比如什么？

Eric, how did you as a board member when you were learning this stuff as you went, you still obviously, you know, I don't think Andrew's been lying to me. I think you've managed to be very helpful. Like what?

埃里克非常 helpful。

Eric was extremely helpful.

那么你是如何 approach 这个的？

So like how did you approach this?

我的能力圈非常小。不是他谈论的任何东西。如果你让我现在解释什么是封装，我做不到。我当然无法以适当的方式做到。安德鲁一直向我解释事情。所以我认为重要的是要像这样，嘿，这是你能做的，这是你不能做的。在任何公司都有这样的时期，特别是如果你在做技术创新，你只需要让工程师去工程，让科学家做他们的事情，不要挡道，并保持他们的资金。也许保持他们的资金非常重要，我想。

I have a very small circle of competence. It is not any of the stuff that he talked about. If you asked me to explain what packaging is right now, I could not do it. I certainly could not do it in an adequate way. And Andrew explains things to me all the time. So I think it's important just to be like, hey, this is what you can do and this is what you can't do. And there are periods of time in any company, particularly if you're doing technological innovation, where you just have to let the engineers engineer and the scientists do their thing and stay out of the way and keep them financed. Maybe keeping them financed was very important I guess.

我认为有几件事让埃里克成为一个好的董事会成员。我认为这些是成为好成员的基础，对吧？你并不了解一切，在你真正专业的领域分享并让我们变得更好，不要谈论其他领域，对吧？有时会有很多话，我们很幸运。我们有一个非常好的董事会，每个人都知道自己擅长什么。

I think there are a couple things that made Eric a good board member. And I think that are foundational in being a good member, right? You don't know about everything and share and make us better in those domains where you're a real expert and don't talk about those other domains, right? That sometimes there are a lot of words and we were lucky. We had a really good board and everybody knew what they were good at.

他们在他们真正专业的领域帮助我们，对吧？对。我的意思是，风险投资世界的优势之一是你可以看到整个行业，对吧？当我们深入其中时，我们是在深入。他们可以看到广泛，对吧？但董事会没有试图解决技术问题。他们没有那个，那不是他们的专长。谈论我们如何融资公司，思考各种其他事情。他们非常 helpful，他们很有耐心，我认为他们问了深思熟虑的问题，我们的董事会成员，我们的董事会会议让我们变得更好。

And they helped us in those domains in which they had real expertise, right? Right. I mean, one of the advantages of the venture world is you can see across an industry, right? When we're deep in it, we're going deep. They can see wide, right? But there weren't efforts by the board to try and solve technical problems. They didn't have that, that wasn't their expertise. Talking about how we might finance the company, thinking about all sorts of other things. They were enormously helpful and they were patient and I think they asked thoughtful questions and our board members, our board meetings made us better.

就像在硬件公司中，董事会一般应该做什么，当你知道你预期在产品上贡献会少得多？是融资和招聘吗？

It's like what should a board do in a hardware company in general when you know you expect that there's going to be much less to contribute on the product? Is it financing and recruiting?

我认为当你有一个长期艰难的项目时，问题就在于此，对吧？这与 SaaS 相反，对吧？你筹集资金，然后花两三年时间构建一个，然后你才真正知道客户会说什么。你去和客户交谈，他们说，‘是的，是的，听起来很棒。’因为谁不会说，‘是的，是的，听起来很棒。’对吧？所以产品管理非常困难。你调查客户，因为它是无成本的，很容易说，‘是的，是的，这很棒。’没有人想扼杀别人的想法，对吧？他们说，‘是的，是的，那很棒。’你需要三年到两年半的时间才能达到可以带给客户的程度。我认为作为投资者，你能做的是理解那个轨迹。理解你的第一个芯片很少是好的，而是第二个或第三个。我的意思是，即使是像谷歌 TPU 团队这样真正强大的团队，第四个才是好的。

I think the question when you've got a long hard project, right? And this is the opposite of SaaS, right? You raise money and then you spend two or three years to build one before you have any real idea what the customer is going to say. You go and you talk to customers and they say, 'Yeah, yeah, that sounds great.' Because who's not going to say, 'Yeah, yeah, that sounds great.' Right? And so the product management is unbelievably difficult. You survey customers and because it's no cost and it's easy to say, 'Yeah, yeah, this is great.' Nobody wants to put their foot on the throat of somebody else's idea, right? They say, 'Yeah, yeah, that's great.' It takes you three years to two and a half years to get to the point where you can bring it to a customer. I think what you can do as investors is understand that trajectory. Understand that your first chip is very rarely a good one and it's a second or third one. I mean, even really strong teams like Google's TPU team, the fourth one was good.

前两个不好，第四个或第五个才是真正好的部分。这需要数年时间。你必须在开始时就知道这一点。你必须知道这是一场长期游戏，对吧？问与长期游戏相关的问题，对吧？你在招聘合适的人吗？你是否在考虑这个——你在专业化和灵活性之间做出了正确的决定吗，对吧？这些是真正能帮助锐化我们思维的问题。但在设计中使用这种技术还是另一种技术，你使用一个工具供应商还是另一个，你必须让团队选择。

The first two were not good, the fourth or fifth were really good parts. It takes years. You have to know that going in. You have to know it's a long game, right? Ask questions related to the long game, right? Are you hiring the right people? Are you thinking about this in the—have you made the right decision between specialization and flexibility, right? These are questions that can really help sharpen our thinking. But whether to use this one technique or this other technique in design, whether you use one tool vendor or the other, you got to let the team pick that.

对你来说，与基础设施角色相比，感觉真的很不同吗？

Did it feel really different for you versus like an infrastructure role?

是的，完全不同。就是真的不同。我的意思是你的收入时间要长得多。收入像巨大的块状到来。最初的客户，你知道我们有美国政府客户，然后我们有巨大的主权云和 G42，你知道，然后现在显然有 OpenAI，所以你有这些东西，所以你的客户集中度更高——

Yeah, totally. It's just really different. I mean your time to revenue is so much longer. The revenue comes in like giant chunks. The initial customers, you know we had US government customers, then we had huge sovereign cloud and G42, you know, then you have now today obviously OpenAI and so you have these things, so your customer concentration is more—

就像每个动态都不同，进入市场的投资要少得多，所以我认为感觉真的非常不同。顺便说一下，相同的事情是：你需要来自各个领域的真正优秀的人聚集在一起，在很长一段时间内保持动力。那部分实际上是一样的。然后市场非常动态。就像，我的意思是，想想看，在 Transformer 之前你创办了公司，在 Transformer 之前。

Just like every dynamic is different, the investment in go-to-market is so much less, so I think it feels really, really different. The things that are the same, by the way, are: you need really good people from a whole bunch of domains coming together who stay motivated over a really long period of time. That part's actually the same. And then the market is very dynamic. Like, I mean, think about it, pre-Transformer you started the company, pre-Transformer.

我们确实如此。

We did.

就在那时，TensorFlow 是主导的，所以第一个——

Right at that time, TensorFlow was dominant, so the first—

TensorFlow 是主导的，ResNet 仍然存在。

TensorFlow was dominant, ResNet was still a thing.

对。与我们今天所处的位置相比，这些是非常小、非常简单的网络。是的，实际的机器学习，人工智能看起来与今天完全不同。

Right. These were very small, very simple networks compared to where we are today. Yeah, the actual ML, the AI looked nothing like it does today.

嗯，实际上我觉得这对收听这期节目的创业者来说真的很有意义，公司最初是作为一个训练系统起步的。你当初推销的愿景、你解释的一部分内容是，由于反向传播，训练是一个更难的问题。你认为它最终会运行在设备和计算机上，并且真正分布式，这或许仍会发生。快进到 2022 年、2023 年，也许是 24 年，你们把公司转向了——或者我不知道“转向”这个词是否准确——但转向了推理。那么，我们是如何、为什么以及在哪些地方走了运，又在哪些地方真正有先见之明？

Well, actually I think it's really, you know, for entrepreneurs who listen to this, it was started as a training system. Part of the vision that you pitched and part of what you explained was that training is a harder problem because of backprop. You thought it was going to be on devices and computers and really distributed, which may still end up happening. And fast forward to 2022, 2023, maybe '24, you pivoted the company—or I don't know if pivot is the right word—but to inference. So how and why and where did we get lucky and where did you have real foresight on that?

我认为做决策和思考有两种方式。一种，我想想老式电话线路，对吧？他们建立一条专用线路一直连到你纽约的兄弟，这是一种做愿景的方式。这是一种你有一个关于未来会是什么样子的想法，远远在那里的方式。那是我们想到的电路，对吧？另一种方式是路由器与互联网的工作方式，你一跳一跳地走。你到克利夫兰，然后决定是否最好走下一跳。这就是我们工作的方式，我们知道那里有一罐金子，但我们知道通往它的路径是不可知的。所以每次你翻过一座新山，你环顾四周，重新决定。所以这是这种一跳一跳的思维。那么我们能建造它吗？能。我们能把它用于路由吗？能。路由现在是所有人关注的重点吗？那就是他们关注的。我们能成为一个快速路由器吗？能。但现在我们看到了地形。我们看到了 AI 的展开，智能的快速增长。那么，谁会使用它？我们通过推理使用 AI。所以我们处于一个位置——我们参与对话，我们的产品被使用——这让我们看到了一些新东西，那个新东西就是：天哪，AI 的轨迹将使我们足够聪明，以至于每个人都想使用它。如果每个人都想使用它，推理将压垮算力基础设施。我们最好在那里。所以每次我们取得成就，它就把我们推上一座山或一座丘。它给了我们新的景观视野。我们可以做出一些新决定，我们总是朝着同一个方向前进，但我们不知道如何到达那里。所以决策一跳——你得到一个观点，仔细思考，赢得下一个观点。

I think there are two ways to make decisions and to think about it. One, I think about an old telephone circuit, right? They set up a dedicated line all the way to your brother in New York, and that's one way to do vision. That's one way that you've got this idea of the way the future's going to look way out there. And that's a circuit we think of, right? The other way to do it is the way routers work with the internet, where you go hop to hop. You go to Cleveland and then there's a decision whether it's best to take the next hop. And that's sort of the way we worked, where we knew there was a pot of gold out there, but the path to it we knew was unknowable. And so each time you get over a new mountain, you look around and you redecide. So it's this sort of hop-to-hop thinking. So can we build it? Yes. Can we make it work in routing? Yes. Is routing now where everybody is focusing? That's where they're focusing. Can we be a fast router? Yes. But now we're seeing the lay of the land. We're seeing the unfolding of AI, the rapid growth of intelligence. Well, who's going to use it? We use AI through inference. So we were in a position—we were engaged in conversations, we had product being used—where it allowed us to see something new, and that new thing was: holy cow, the trajectory of AI will make us smart enough that everybody will want to use it. If everybody wants to use it, inference is going to crush the compute infrastructure. We better be there. And so each time we achieve something, it moved us up a mountain or a hill. It gave us a new view of the landscape. We could make some new decisions, and we were always moving in the same direction, but we didn't know how to get there. And so decision hop—you get to a point of view, think carefully, earn the next viewpoint.

是的。

Yeah.

赢得下一个。

Earn the next.

嗯，这很有趣，因为你知道，对于软件，你显然既可以灵活思考，又可以在第二天改变一切。

Well, it's interesting because you know with software you obviously can both think nimbly and change everything the next day.

对。

Right.

硬件你不能那样做。

Hardware you can't do it like that.

没错。我们的，我们有更大的离散决策，对吧？这就是为什么在芯片中选择专用化还是灵活性如此重要。我们在第一个架构中做了几个非常好的决定，我们决定不嵌入会加速卷积网络的技术。相反，我们说我们不知道那些能持续多久。如果我们在那之下工作，加速我们所知道的所有 AI 所依赖的代数，那是一个非常好的决定，因为当 Transformer 出现时，我们在那两个上是最快的，尽管我们从未见过它们，从未听说过它们，而且当我们设定架构时它们还没有被发明。所以做出几个好的决定是有帮助的。

That's right. Ours, we have bigger discrete decisions, right? And that's why the choice in your chip of specialization versus flexibility is so important. We made a couple really good decisions in our first architecture where we decided not to embed technology that would accelerate convolutional networks. Instead, we said we don't know how long those will last. If we work underneath that and accelerate the algebra that underpins all AI we knew about, that was a really good decision because when Transformers came out, we were the fastest at those two, even though we'd never seen them and never heard of them and they hadn't been invented when we set the architecture. And so it helps to make a few good decisions.

是的，这很有趣，就像在 SpaceX、Anduril、Cerebras、Palantir 等公司的背上，也许还有其他公司，但显然硬件现在热得不能再热，每个人都想投资硬件。但——只是跟我们讲讲——这疯狂地难。

Yeah, it's interesting like on the backs of companies like SpaceX, Anduril, Cerebras, Palantir, maybe others, but obviously hardware is now hotter than hot and everyone wants to fund hardware. But it's—just talking us through it—it is crazy hard.

真的很难。

It's really hard.

是的。

Yeah.

很难，它需要巨大的内在毅力，它需要——历史上，成功一直由该领域的一些经验所预测。

It's hard and it requires enormous internal fortitude and it requires—success has historically been predicted by some experience in the field.

是的。

Yeah.

不同于 AI，你能看到优势。我认为在 AI 和社交网络中，很多创始人和领导者是在为自己、同学和朋友构建工具，他们有独特的洞察力。我的意思是，如果你看看 Cognition，如果你看看 Cursor，这些是世界上最好的工程师、软件工程师，在为自己构建工具，对吧？那是真的。那不同于——

Different than AI where you see the advantage. I think in AI and social networking, a lot of the founders and the leaders were building tools for themselves and their classmates and their friends, and they had unique insight. I mean, if you look at Cognition, if you look at Cursor, these are some of the best engineers, software engineers in the world, building tools for themselves, right? That's true. And that's different from—

我也想知道这最后一个话题是否可能在其中扮演重要角色，那就是当你可以第二天改变你的观点，而你昨天错了也没关系。速度和决策速度胜过——相比之下，如果你只是，你知道，你做了一件事，然后你必须忍受它一年、几年。

I also wonder if this last topic probably plays a big role in it, which is when you can change your opinion the next day and it's fine that you were wrong yesterday. Speed and decision-making speed trumps—versus if you just, you know, you do something and then you have to live with that for a year, years.

几年。

Years.

你不能只是错了然后明天修复它。

You can't just be wrong and fix it tomorrow.

我认为在芯片中，我们量三次，切一次，对吧？它真的毫不留情。

I think in chips we measure three times and cut once, right? It is really unforgiving.

是的。

Yeah.

对于大错误。所以你必须非常非常确定你把大项目做对了。

To big mistakes. And so you have to be very, very sure that you get the big items right.

是的。我认为关于你如何建立公司，有一件事很有趣，那就是，你知道，你提到很多早期团队一起做过多个公司。就像你们很多人一起工作了很长时间。你们更年长，更有经验。

Yeah. One of the things that I think is interesting about how you built the company, which is, you know, you mentioned a lot of the early team had done multiple companies together. Like a lot of you guys had worked together for a long time. You're older and more experienced.

我觉得你没那么老。

I think you're not that old.

但有人叫我婴儿潮一代。我就想，我爸爸才是婴儿潮一代。婴儿潮一代。

But somebody called me a boomer. I'm just like, my dad's a boomer. A boomer.

你知道，在某个时候你抬头看，那只是——都一样。当你那么低的时候，你抬头看。就像你八岁时，你觉得 27 岁真的很老。

You know, at some point you're looking up and that's just—it's all the same. When you're that low, you're looking up. It's just like when you were eight, you really old, like 27.

嗯，但你知道实际上有趣的是，如果我看你们组织中的产品负责人。

Well, but you know what's interesting actually is if I look at the product leaders in your organization.

如果我看一些市场推广领导层，

If I look at some of the go-to-market leadership,

是的，

yeah,

这些人实际上非常年轻。

these are they're actually very young.

他们是的。我只是感兴趣，比如这背后是否有意为之？显然不一定是年龄歧视，但就像你如何将年轻前沿产品负责人的视角与经验丰富的硬件工程师结合起来。

They are. And I'm just interested in like how was there intentionality around that? Not necessarily ageist, obviously, but just like how you combined the perspective of young at the cutting edge product leaders with seasoned hardware engineers.

我认为我们非常努力地思考经验在哪里重要，以及极致的智慧在哪里重要。你知道，如果你看看我们的产品组织，聪明得难以置信。我的意思是，一些我见过的最好的产品人员。都很年轻，都是从内部提拔的。你知道，我们没有大公司规则。

I think we tried to think really hard about where experience mattered and where blistering intelligence matters. You know, if you look at our product organization, unbelievably smart. I mean, some of the best product people I've ever seen. All young, all promoted from within. You know, we don't have big company rules.

你得在一个职位上待够一定时间才能升职，对吧？我的意思是，如果你非常出色，我们会给你越来越多的责任，如果你做得好，我们能带你去的地方没有上限。我的联合创始人 Sean，我们五个人之一，在我上一家公司，我们招他时他是一名个人贡献者，四年后我们被收购时，他 2007 年加入我们时才 25 或 26 岁，对吧？当我们被 AMD 收购时，我们让他成为了公司院士，对吧？就是没有上限。现在他是一家上市公司的创始人兼联合 CTO。

You got to be in a job for this amount of time before you can get promoted, right? I mean, if you're extraordinary, we're going to give you more and more responsibility and more and more and if you do a great job with it, there's no end to where we will take you. My co-founder Sean, one of the five of us, in my last company, we hired him as an individual contributor and when we were acquired four years later, the guy was 25 or 26 when we hired him in 2007, right? When we were acquired by AMD, we made him a corporate fellow, right? There was just no end. And now he is a founder and he's co-CTO of a public company.

我们在这方面相当果断。有些领域要出类拔萃需要大量经验。有些领域则不需要任何经验。我们在理解客户对 AI 的需求方面没有很长的历史。所以他们的方法、聪明才智和洞察力胜过经验。在其他领域，比如芯片制造，我的经验是，之前造过若干年芯片是能否交付卓越芯片的最佳预测指标。这在机械方面、系统方面也是如此。

And we do that sort of ruthlessly. And there are some areas where to be exceptional requires a tremendous amount of experience. There are some areas where it doesn't require any experience. We don't have a long history of understanding what customers want in AI. So their methodology, smarts, insight trump experience. In other areas, in the making of chips, it's been my experience that some odd years of previously building chips is the best predictor of whether you're going to deliver exceptional chips. And that's also true on the mechanical side, on the system side.

在大学或研究生院，真正去造硅、真正去造机器的机会很少。它们太贵了。太难造了。耗时太长，即使在博士项目中，你也没机会流片、交付、启动、放到板子上、通电、为它写软件。所以这需要一些时间。

There's just not a lot of chance in college or in graduate school to actually build silicon, to actually build a machine. They're so expensive. They're so hard to build. They take so long that even in a doctoral program, you don't get a chance to tape out a chip, to deliver it, to bring it up, to put it on a board, to power it, to write the software for it. And so it takes some time.

这些关系是至关重要的，还是年轻人可以很快赢得？

Are the relationships something that are critical or can those be earned quickly by young people?

我认为是这样的。首先，五个创始人毫无疑问太多了，除非我们之前一起工作过，所以每个人都知道其他人擅长什么，对吧？所以根本没有太多冲突。我的意思是，我们之前都一起工作过。我们都非常尊重彼此能做的事，也对自己不能做的事保持谦逊。所以我们能够做到。你的具体问题是建立信任需要很多时间吗？我认为和某人一起工作六到八周内，你就能看出他们是否非凡。

I think the following. First, five is too many founders without question except that we'd worked together before and so everybody knew what the other folks were good at, right? And so there wasn't a lot of headbutting at all. I mean, we'd all worked together previously. We all had tremendous respect for what the others could do and some humility about what we couldn't do. And so we were able to do that. Your specific question is does it take a lot of time to build trust? I think within the six or eight weeks of working with someone, you can tell if they're extraordinary.

嗯。

Mhm.

对。我的意思是，非凡的人发来的第一封邮件，你就会想，“哇，这正是我需要的。”

Right. I mean extraordinary people in the first email they send you, you go, "Whoa, that's exactly what I needed."

是的。

Yeah.

对。每个列表都按重要性降序排列。没有太多废话。信号很强。你会想，“哇。”然后你又看到了。然后你观察他们主持会议的方式，你会想，“哇。”然后你看着他们交付东西。他们激励周围的一群人。他们能跨组织工作。你说，“嗯，这是我下一个重要项目需要的人。”

Right. Every list is in descending order of importance. There's not a lot of fluff. There's high signal. You go, "Whoa." And then you see that again. And then you watch the way they run a meeting and you go, "Whoa." And then you watch them deliver something. They motivate a bunch of people around them. They can work across the organization. You say, "Well, that's a person I need on the next important project."

是的。

Yeah.

然后他们就是能啃下工作。

And then they just chew through work.

我一直认为招聘时，如果我能学到真正的东西，并且我想再进行一次面试或会议，只是因为我想，我会学到更多东西。我就想，那是我最好的预测。

I always thought with recruiting, it's like if I came away like learning real things and I wanted to have another interview or meeting just because I was like, I'm going to learn more stuff. I was like, that's my best prediction.

最好的。

The best.

太好了。

It's so good.

最好的。

The best.

那外部关系呢？就像，你知道，你得和公司外的很多人合作。你知道，你得在台湾。你得和，你知道，

What about the external relationships? It's like, you know, you got to work with a lot of people outside your company. You know, you got to be in Taiwan. You got to work with, you know,

你知道，把一些关系带过来会有帮助，对吧？所以，你知道，我们和台积电一起造芯片几十年了。我们和我们的代工厂合作几十年了。所以，尤其是在竞争激烈的时期，那些关系已经存在多年，你一直信守承诺，不是一次，不是两次，不只是在顺境，而是在顺境和逆境中都信守承诺。那真的非常非常重要。我认为你以完全相同的方式赢得与新合作伙伴的关系。你信守承诺。你尽早给他们信息。你写——我的意思是，你知道，写一封感谢信。不，真的。我的意思是，做你妈妈说的，对吧？做个好人。写感谢信。做你说过要做的事。

you know, it helps to bring those with you a little bit, some of them, right? So, you know, we'd been building chips with TSMC for decades. We'd been working with our contract manufacturers for decades. And so, especially in a time of contention, that those relationships had been in place for years, that you'd been good to your word, not once, not twice, not just in good times, but you'd been good to your word over good and bad times. That was really, really important. I think you earn relationships with new partners in exactly the same way. You're good to your word. You get them information early. You write—I mean it's, you know, write a thank you note. No, really. I mean, do what your mother said, right? Be a good person. Write a thank you note. Do what you say you're going to do.

对我来说，这是最大的学习之一，特别是在新冠时期等等。如果我想到软件公司，甚至我们合作的软件基础设施公司，真正重要的供应商是 AWS，也许现在重要的供应商是 AWS 和一家基础模型公司之类的。好像只有两个供应商重要。这就像我不知道供应商名单是什么，但它

This to me was one of the big learnings, like particularly during the COVID era and everything. If I think of the software companies, even the software infrastructure companies we work on, it's really like their vendor that matters is AWS and like maybe now their vendor that matters is AWS and you know a foundational model company or something like that. There's like two vendors that matter. This is like I don't know what the vendor list is but it

这太疯狂了。是几十个，对吧？系统里所有的这些组件。所有那些制造冷却板和水系统部件的专业制造商，所以你必须让所有这些事情结合起来，我们只是不习惯，为了交付产品需要那么多依赖，或者至少

it's insane. It's dozens, right? It is all of these components that go into the system. All of these specialty manufacturers that build the cooling plate and part of the water system and like and so you have all of these things that have to kind of come together and and we just aren't used to that like that those me that many dependencies in order to deliver our product or at least

我不习惯，那是个大

I wasn't that was like a big

事情，你必须和他们合作

thing and you have to kind of work with them on

软件人开始理解供应链。

software guy coming to grips with a supply chain.

是的。是的。是的。

Yeah. Yeah. Yeah.

是的。没错。

Yeah. Exactly.

他们叫它链是有原因的。

They call it a chain for a reason.

而且它真的是。你会想，“等等。天哪。所有这些。”当你指数级增长时，一切变得更加复杂，对吧？因为就像你在投入，我从未想过这个。你知道，就像 AWS，我们想要更多容量。你上网增加容量。就像你想要更多晶圆数量，那是一个 15 个月交货期的决定

And and and it really is. And you you're like, "Wait a minute. Holy cow. This all of this stuff." And when you're growing exponentially, like then everything becomes even more complicated, right? Cuz it's like you're putting in I had never thought about this. You know, it's like AWS, we want more capacity. You know, you go online and you add capacity. like you want more capacity on in terms of like number of wafers like that's a 15month lead time kind of decision

而且需要巨额资本，你的供应商也在做巨大的分配决定，所以他们也必须同意，所以这对我来说是一个很大的学习，就复杂性而言

and a huge amount of capital and your vendor is also making a huge allocation decision and so they have to buy into it too and so I I that that was a big learning for me in just terms of like how much complexity there is in that

我们内部一直在讨论的一件事是，感觉无论你认为它有多大，都很难理解正在发生的 AI 建设规模。

one of the things we've been talking about a lot internally is that it feels like no matter how big you think it is, it it is hard to wrap your head around the size of the AI buildout that's happening.

是的。

Yeah.

以及投入的资本支出，你知道，我今天早上实际上看到一张图表，大约是 GDP 的 1%

And the capex going in, you know, I actually saw a chart this morning that was like 1% of GDP

每年花费在

was spent per year on

高速公路和电信上，

highway and telecom,

对吧？

right?

然后铁路是 2%

And then it was 2% for railroads

然后这个 AI 建设大约是 GDP 的 3.5%。

and then this AI buildout is like 3 and a half% of GDP.

就像

It's like

只有一个人做对了，

only one person got it right,

那就是

which was

你的兄弟。

your brother.

不，那是真的。

No, that's true.

是的。我是说，唯一一个人——大家都觉得他疯了。Sam 真正擅长、而我认为极其困难的一点是，他看到了指数增长，而且不害怕。

Yeah. I mean, the only person — and everybody thought he was out of his mind. What Sam's really good at, and I think is so hard, is he saw an exponential and wasn't afraid.

对吧？

Right?

你把那个指数外推两三年、四五年，你会惊呼天哪。

You take that exponential out two or three or four or five years, you go holy crap.

对吧？

Right?

他不害怕。其他所有人都害怕。时不时有人说会放缓。他说，不。

He wasn't afraid. Everybody else was afraid. Every now and then it's going to slow down. He was like, no.

记得那个 Stargate 吗，好像 7 万亿还是多少，

Remember the Stargate, like 7 trillion or whatever,

对吧？我是说现在最初的 Stargate 已经小得可怜了，

right? I mean now the initial Stargate is sadly small,

对吧？而且当时那数字令人难以置信。他们都在笑。

right? And it was mindboggling. They were laughing.

所以除了他，我们所有人都错了。我认为我们错了。我认为台积电错了。我认为英伟达错了。所有人，内存厂商也错了。除了 Sam，我们都错了。

And so all of us got it wrong except him. I think we got it wrong. I think TSMC got it wrong. I think Nvidia got it wrong. Everybody, the memory guys got it wrong. We all got it wrong except Sam.

是的。所以，你知道，现在有这种疯狂的建设。而对推理的需求似乎增长得比我们能建设的任何东西都快，所以就有很多问题，比如这对算力价格意味着什么，对可用量和 token 支出等等意味着什么。但我很好奇，你日常在数据中心这边看到什么，你的体验，现在对你来说是什么感觉？

Yeah. So, you know, you've got this insane buildout going. And the demand for inference seems like it's going even faster than no matter what we're going to be able to build, you know, and so then there's a lot of questions about, you know, what will that mean for the price of compute and what will it mean in terms of how much will be available and token spend and all of this, but I'm curious like day-to-day what you're seeing on the data center side and your experience, you know, now what what is that like for you?

数据中心是通过云交付算力的供应链中的一环。事实上，无论你是通过云交付，还是我们所说的本地部署，两种情况都受影响。一种情况是你租用容量，把你的设备放进去。另一种情况是你的客户租用数据中心。所以这是当前市场的一个限制。它暴露了美国的许多弱点。我们的电网很糟糕，基本上还是 1940 或 50 年代的技术。我们停止了一些非常有趣、结果极其清洁的技术的研发，比如核能。我是说，如果最终要靠 AI 把我们带回核能，那岂不是很讽刺，对吧？

The data center is one of the links in the supply chain to deliver compute by the cloud. And the truth is whether you deliver it by the cloud or what we call on prem, it impacts both cases. In one case you rent the capacity and you put your equipment in it. In the other case your customer rents the data center. And so it is a limitation in the market right now. It exposes a whole bunch of weaknesses in the US. Our grid is pathetic and sort of built on 1940s or 50s technology. We stopped doing work in very interesting technologies that turn out to be extremely clean like nuclear. I mean wouldn't it be ironic if it took AI to bring us back to doing nuclear, right?

我是说，没错。相当对。

I mean that's right. Pretty right.

它表明，在这些数据中心里，我们用发电机作为备用电源，这些要么是柴油，要么是 LNG，要么是某种燃气，对吧？这些来自 GE Vernova 或卡特彼勒。几十年来，发电机和柴油发电机组一直没有创新。现在突然有了创新。Boom 的人想用他们为喷气机设计的东西来给数据中心供电。我们在电池备份方面看到各种有趣的东西，还有 Bloom Energy 的有趣燃料电池。我是说，因为需求驱动，出现了巨大的创新。我们没有足够的数据中心。它们上线太慢。电网老旧。所以这是一个创新巨大的领域。我认为这个行业一开始做了一些蠢事，对自己没好处。试图把一些成本转嫁给当地社区。试图占地方政府的便宜。数据中心没有理由不自己买单。没有理由用很多水。我们用的是闭环系统，对吧？美国所有数据中心用的水都比加州杏仁种植者少。不是少 1 倍、2 倍或 4 倍，而是四到七倍，杏仁种植者用得更多。所以，我是说，

It showed that in these data centers we use generators as backups and these are either diesel or LNG or one form of gas, right? And these come from GE Vernova or they come from Caterpillar. There've been no innovation in generator and diesel gen sets for decades. Now suddenly there's innovation. The guys at Boom want to use what they designed for jets to power data centers. We're seeing all sorts of interesting things in battery backup and like from Bloom Energy, you're seeing interesting fuel cells. I mean, there's just this enormous innovation because there's necessity driving it. We don't have enough data centers. They're coming on too slowly. The grid is old. And so it's an area of tremendous innovation. I think the industry did itself no favors by doing some dumb stuff at the beginning. Tried to pawn off some costs on local communities. It tried to take advantage of local municipalities. There is no reason a data center shouldn't pay its way. There's no reason why it should use very much water. We use closed loop systems, right? All the data centers in the US use less than California almond growers. Not by 1x or 2x or 4x, but between four and seven times, the almond growers use more. So, I mean,

用水这件事根本不算个事。

the water thing is just not a thing.

用水这件事根本不算个事。是的。但作为一个行业，我们没有做好与当地社区的沟通，没有获得他们的支持，没有向他们展示我们会带来成千上万、有时是数万个高薪建筑工作岗位。我们会持续提供就业，他们的税基应该会随时间下降。我们没做好这件事。现在我们正在付出代价。

The water thing's not just a thing. Yeah. But as a community, we didn't do a good job of communicating with local communities, getting their buy-in, showing them that we're going to bring thousands and in some cases tens of thousands of high-paying construction jobs. We're going to pay ongoing jobs, and their tax base ought to go down over time. And we didn't do a good job of that. And now we're paying the price.

我是说，这跟科技行业在 AI 传播上整体做得很差这个主题是一致的。

I mean, it kind of goes with the whole theme of tech doing a terrible job communicating about AI in general.

糟透了。我们做得糟透了。

Horrible job. We're doing a horrible job.

是的。

Yeah.

我觉得数据中心这件事，另一个有趣的部分是，如果我，比如说 2018 年、2020 年随便什么时候问你，数据中心成为我们关键因素的概率有多大？

I think the data center thing, the other part of the data center thing is interesting to me is like if you if I asked you, I don't know, 2018, 2020, whatever. What's the probability that data centers were going to be a critical factor in our

又一件我会搞错的事，

Another thing I'd have gotten wrong,

对吧？是的。我们都会搞错。

right? Yeah. We'd all got it wrong.

搞错了。

Got it wrong.

而且它只是——那就像——这是件新事。就像，好吧，现在你必须从头建到底并交付。但

And it just — and that was just like a — it's a new thing. It's like, okay, now you have to build it all the way through and deliver it. But

从铲子开始要多久？我知道你开玩笑说过铲子。我其实想问你那个，但从头到尾建一个数据中心要多久？

how long does it take to go from shovel? I know you joked about shovel. I actually want to ask you about that, but how long does it take to go from start to finish on a data center?

所以“铲子就绪”是你在数据中心圈子里听到的说法。意思是还没做

So shovel ready is an expression you hear in the data center world. It means we haven't done

但我们准备好了，对吧？

but we're ready, right?

我们想卖给你一堆土，准备好让你的铲子开始干活。它

We want to sell you a pile of dirt, ready for your shovels to start doing something. It

听起来确实更好。确实比“我有块生地”或“我在俄克拉荷马有块土”好听，对吧？

does sound better. It does sound better than I've got raw land or I've got dirt in Oklahoma, right?

它难道不是意味着你有许可证和其他重要的东西吗？

Doesn't it mean you have permits and other things that are important?

有时候是，有时候不是。

Sometimes, sometimes not.

嗯，我们从未想过会走到今天这一步。

Um, we would never have expected to get where we are today.

是的。

Yeah.

如果你有好的建筑商，从生地加许可证到建成 50 兆瓦，这算一个合理的区块，通常甚至大站点也是以 50 兆瓦区块展开的，大概需要——

From if you have a good builder, it takes probably from raw land and permits to stand up 50 megawatts, which is a reasonable block, and often even the big sites unfold in 50 megawatt blocks.

50 兆瓦区块现在是 18 个月，如果那里有现成建筑而且已经有电网供电，我们称之为棕地。所以现在很多情况是人们去锈带买旧工厂、造纸厂，呃，因为它们有

50 megawatt blocks is 18 months now if there's an existing building there and there's grid power already we call those brownfield sites so a lot of what's going on right now is people are going to the rust belt and they're buying old factories, paper mills, uh, because they had

大量电力，

a lot of power,

大量电力，对吧？它们已经有电力许可了

a lot of power, right? That they had the permits for power already

然后改造它们。

and retrofitting them.

那么，供应链瓶颈中，能源、芯片、建设、许可等等各占多少？

So, how much of their kind of supply chain bottleneck is energy versus chips versus the construction versus permitting versus whatever else?

供应链的每一块都有自己的供应链。所以，数据中心需要——混凝土很充足。大多数地方有劳动力。尽管在怀俄明州，为了争夺所有这些数据中心选址，他们从远至丹佛招来电工。发电机和电力传输开关目前是长交付周期项目，很难搞到。所以这些通常是长杆项。通常你可以先建起所谓的冷壳。你可以相当快地建起混凝土预制建筑或金属建筑。但之后你必须装修，把它变成算力的高功率酒店。

Each chunk of the supply chain has its own supply chain. So, data centers need — there's plenty of concrete. In most places there's labor. Although in Wyoming where there was a battle for all these data center sites, they were bringing in electricians from as far away as Denver. Generators and electrical transmission switches are long lead time items right now and those are hard to come by. And so those are the long poles usually. Usually you can get up what's called a cold shell. You can build a concrete tilt-up building or a metal building fairly quickly. But then you have to fit it out and turn it into a high-powered hotel for compute.

是的。

Yeah.

对。那需要电工、冷却、冷水机组，所有这些其他东西。

Right. And that takes electricians, takes cooling, chillers, all these other things.

这些现在都是长交付周期。我昨晚离开孟菲斯时正好飞过 Macrohard 和 Macroharder。那规模真是惊人。

Those all are long lead time right now. I happened to fly over, leaving Memphis last night. I happened to fly over Macrohard and Macroharder. It was astonishing, like the size.

没人能像埃隆那样建造。太不可思议了。

Nobody can build like Elon. It's unbelievable.

那些建筑的规模、附近的施工等等。我当时就想，哇，太庞大了。这就引出一个有趣的问题：我可以想象一个世界，要么数据中心永远不够，未来十年一直不够；要么由于交付周期的动态和当前的情况，可能会出现过度建设。押注你认为哪个世界会成真，对谁来说可能都有一些超额收益。

The size of buildings and the construction nearby and everything. I was like, whoa, that is tremendous. Which leads to an interesting question now, which is, I could imagine a world where either it'll just never be enough data centers, you know, and for 10 more years it's always going to not be enough, or because of the dynamics of the lead time and what's going on right now, you could imagine like an overbuild. And betting on which of those worlds you think is going to come out, there's probably some alpha there for whoever.

我们没有足够的数据中心。我们肯定不够。我的意思是，现在要预见我们领域十年后的情况真的很难。对吧。十年前，Transformer 还没跑起来。人们跑的是很小的模型，试图识别人脸、猫、椅子之类的。我是说，2016 年不是 AI 的大年，对吧？但五年后我们仍会追逐数据中心，仍会追逐芯片。而那些用 HBM 的人，仍会追逐 HBM。

We don't have enough data centers. We won't certainly. I mean, seeing out 10 years in our space right now is really hard. Right. 10 years ago, transformers weren't running. People were running tiny little models and trying to identify faces or cats and chairs or whatever. I mean, 2016 was not a big AI year, right? But five years we will still be chasing data centers and we'll still be chasing chips. And for those who use HBM, they'll still be chasing HBM.

是啊。政治如何介入这一切也会很有趣，因为现在政府对科技、尤其是 AI 的兴趣和参与度显然比以往任何时候都高。

Yeah. It'll be interesting also how politics plays into all this because like now you got probably, you know, more government interest and involvement in tech and AI specifically than we've ever had obviously.

不，我们需要的是更多不懂行的人来做决策。这显然是我们需要的。我是说，我觉得需要更多完全没头绪的人来做重要决策。你知道，我认为在这个领域，真的应该就什么对美国有利展开深思熟虑的讨论。

No, what we need is more people who don't understand making decisions. That's clearly what we need. I mean, I think more people with absolutely no clue making important decisions. You know, I think this is an area where there are really thoughtful discussions to be had about what's good for the US.

是啊。

Yeah.

以及对不同市政、不同农村地区有利。而这些讨论政客们并没有进行。

And good for different municipalities, different rural areas. And those aren't being had by the politicians.

是啊。

Yeah.

对。他们是条件反射式的。他们有点——

Right. They're knee-jerk. They're sort of—

让我就这一点问你个问题，暂时抛开政治。假设我们确实同意这个，比如算力这件事，算力在某种形式上受到限制，无论是内存、芯片还是数据中心空间，随便什么。存在某种约束，而需求是压倒性的。那是不是说明唯一重要的就是每瓦 token 数？基本上我们会有有限的分母，所以无论有多少，不管是每瓦 token、每设施 token 还是每芯片 token，或者别的什么，但感觉我们需要在有限的分母上最大化分子，也许就是每瓦任务。

Let me ask you a question on this, moving away from politics for a second. Just let's say we do agree on this, like compute thing, that compute is constrained in some form or another, whether it's memory or chips or data center space, whatever. There's some constraint there and the demand is overwhelming. Doesn't that argue then the only thing that matters is like tokens per watt? Like basically we're going to have a limited number of denominator and so however much, whether there's tokens per watt or tokens per facility or tokens per chip or I don't know something like that, but it feels like we need the most of the numerator for the limited denominator task per watt maybe.

所以如果情况如此，那么 Cerebras 有速度优势，巨大的速度优势。

And so like if that's the case, then you have Cerebras has a speed advantage, a huge speed advantage.

但如果我们快进来看，实际的约束到底是什么，你怎么看待这种动态？

But like how do you think about that dynamic if we fast forward in terms of what the constraint actually is?

我认为有几点。首先，速度，这是我们选择专注的，历史上创造了市场。

I think there are a couple things. First, speed, which is what we've chosen to focus on, has historically created markets.

嗯。

Mhm.

对。你想想，慢速搜索没有市场，对吧？拨号上网没有市场。你的孩子多大了，14 和 12？如果你想惩罚他们，别拿走他们的手机。把速度调回拨号上网。

Right. And if you think about it, right, there is no market for slow search, right? There's no market for dialup. Your kids are what, 14 and 12. If you want to punish them, right, don't take away their phone. Ratchet it back to dialup speed.

对。对。这才是真正的惩罚。让他用一周拨号速度，慢慢地。

Right. Right. This is a real punishment. Let him use it for a week at dialup speed slowly.

对。为什么我们笑了，然后又说 AI 慢一点没关系。我是说想想看。当互联网慢的时候，Netflix 递送 DVD 和信封。

Right. That why is we laugh and then we say it's okay for AI to be slow. I mean think about it. When the internet was slow Netflix delivered DVDs and envelopes.

是啊。

Yeah.

而当互联网变快时，他们变成了电影制片厂。对。那不是——他们并没有在原来的事情上变得更好。他们变成了全新的东西。我认为我们在 GPT-5.6 Soul 发布时看到的是，对吧？上周它以有限可用性推出，人们正在思考全新的应用。你瞬间获得前沿智能，这开启了各种新机会。所以这是速度的一个作用。另一件事是，你试着思考如何提高吞吐量，如何制造更多 token。我们正在做的一种方式是通过合作，叫做解聚合，我们与 AMD 合作，与 AWS 合作，在推理工作中，有一部分可以由别人完成，有一部分由你完成，从而使结果具有更高的吞吐量。你知道，与 AMD 合作，我们看到额外 5 倍的吞吐量，我是说在保持速度不变的情况下吞吐量提高五倍。我们在 AWS 上也看到类似的数字。由于我们的架构，我们有机会在整个 GPU 生态中做到这一点。我们可以全面做到。所以你知道，目前我们类别中有四大芯片制造商。显然是 Nvidia，我们很想与他们合作。还有 AMD、Google TPU，以及 AWS 的训练部件，我们已经与其中两家合作。所以，这是我们深思熟虑过的事情，也是我们极其感兴趣追逐的方向。

And when the internet got fast they became a movie studio. Right. That's not—they didn't get better at their other thing. They became something entirely new. And I think what we're seeing with the launch of GPT-5.6 Soul, right? It's put out in limited availability last week, that people are thinking of whole new applications. You've got frontier intelligence instantly and that opens up all sorts of new opportunities. So that's one thing you do with speed. The other thing you do is you try and think about how you can drive up throughput, how you can make more tokens. And one of the ways we're doing that is by partnering in something called disaggregation and we're doing it with AMD, we're doing it with AWS, where you think about in the work of inference, is there a part that can be done by somebody else and is there a part that can be done by you such that the result is higher throughput. And you know with AMD we're seeing 5x additional throughput, I mean five times as much throughput while keeping the speed the same. And we're seeing similar numbers with AWS. We have an opportunity because of our architecture to do that with the entire GPU landscape. We could do it across the board. And so you know there are four major chip makers right now in our category. Obviously Nvidia, we'd love to partner with them. There's AMD, there's the Google TPU, and there's AWS with their training parts, and we're already working with two. So, that's something we've thought a great deal about and is a vector we are extremely interested in chasing down.

Nvidia 到底是什么让他们有了这样疯狂的走势，你知道，就像我——

What has it been about Nvidia that has made them have the just crazy run that they've had, you know, like I've—

我认为大多数人对 Nvidia 为何伟大都看错了。首先，Nvidia 在本世纪头十年，毫无疑问是伟大的公司，对吧？我认为人们关注一堆东西。他们关注 CUDA。我不认为那是 CUDA。他们关注芯片架构。我不认为那是芯片架构。而是那种令人难以置信的毅力和强度，源于作为上市公司十年没有成功的经历。我认为如果你看他们的股价图，大约 2003 到 2013 年，2004 到 2014 年，十年间，对吧，他们的股价表现糟糕。你是一家上市公司，你拼命奋斗，没人在听你，你卖不出多少，那种坚持不懈和毅力是了不起的。而十年后从中走出来成为最有价值的公司，对吧，世界上最有价值的公司。在我看来，对于他们这种规模的公司，那种强度和毅力使他们成为历史上最伟大的公司之一。不是那些别的东西。那些只是人们说的。但那种作为上市公司十年、战斗、战斗、战斗的经历，会进入你的 DNA，那太了不起了。

I think most people are wrong about what makes Nvidia great. First, Nvidia's, you know, in the first quarter of this century, they're the great company without any question, right? And I think people look to a bunch of things. They look to CUDA. I don't think it's CUDA. They look to their chip architecture. I don't think it's a chip architecture. It's this unbelievable grit and intensity that was born of a decade of not having success as a public company. I think if you look at their stock chart between about 2003 and 2013, 2004 and 2014, for a decade, right, they traded horribly. And you're a public company and you're fighting tooth and nail and no one's listening to you and you can't sell very much, and the relentlessness and the grit that that takes is awesome. And to come out of that as the most valuable company a decade later, right, the most valuable company in the world. That sort of intensity and grit for a company of their size in my view makes them one of the great companies in history. It's not these other things. These are just things people say. But that sort of a decade of being a public company and fighting and fighting and fighting, that gets in your DNA and that's awesome.

是啊。我的意思是，你看到黄仁勋仍然在活动上——

Yeah. I mean like you see Jensen like still at events—

拼命战斗。

Fighting tooth and nail.

是啊。

Yeah.

对。对。

Right. Right.

我是说，你能想象出万亿美元市值的公司吗？你能想象在黄仁勋之前的老一代科技领袖做到这一点吗？

I mean, where could you have imagined trillion-dollar companies? Could you imagine old tech leaders before Jensen doing that?

不能。

No.

对。对。有那种斗志。是的。

Right. Right. Have that kind of fight. Yeah.

在我看来，那才是最了不起的。那种斗志。其他东西很酷、很好，但当我思考自己能在哪些方面做得更好，当我思考作为 CEO 应该考虑什么时，我关注的就是这些。

I mean, that in my view is what's awesome. That level of fight. I think the other stuff is cool, good, but when I look at what I can do better, when I look at what I ought to be thinking about as a CEO, those are the things I look to.

你觉得你的濒死经历是否给公司注入了这种基因？

Do you think your near-death experiences gave your company some of that DNA?

你知道，我们确实是，而且我觉得黄仁勋身上有意思的一点也是，他仍然把自己看作弱者。现在他是巨头了，但你知道，这是我的第五家创业公司。之前我们卖掉了三家，把一家带上市，现在又把这家带上市了。我是一个职业的大卫，在与歌利亚战斗，我每天醒来都带着这种心态。我们现在更大了，我们有更大的竞争对手，对吧？我们有更大的挑战。有更多人在向我们扔石头。所以我希望我们能保持这一点。我个人每天醒来都带着那种热情和干劲。我是说，每天我都想，我们刚起步时他们说这永远行不通。你做不到晶圆级，我们做到了晶圆级。然后他们说，好吧，你做到了晶圆级，但你没法量产良率，然后我们实现了量产良率。他们说，好吧，你能量产良率了，但你没法做量产封装，然后我们做到了量产封装。他们说，好吧，现在你做到了量产封装，但你只有政府客户，我们说，好，然后我们拿下了主权云。然后他们说，你没有超大规模云厂商，你没有前沿实验室，然后我们拿下了 OpenAI。他们说，好吧，你有政府、有主权云、有前沿实验室，但你没有超大规模云厂商。然后我们拿下了一家超大规模云厂商。对吧？每次他们都说，然后他们说你有这些很酷的客户，但你做不了大模型。现在我们正在服务 GPT。

You know, we are, and I think this is what's interesting about Jensen too, is he sees himself still as the underdog. Now he's the big dog, but you know, this is my fifth startup. We sold three and took one public previously, and we've now taken this one public. I'm a professional David in the battle with Goliath, and I wake up every day with that mentality. We're now bigger, we have bigger competitors, right? We have bigger challenges. We have more people throwing stones at us. And so I hope we take that. I personally wake up every day with that passion and that drive. I mean, every day I think to myself, when we started they said it would never work. You can't do wafer for scale, and we made wafer for scale. Then they said, all right, you did wafer for scale, but you can't yield it in volume, and then we yielded it in volume. They said, okay, you can yield it in volume, but you can't package it in volume production, and then we packaged it in volume production. They said, okay, now you've packaged it in volume production, you only have a government customer, and we said, okay, then we won a sovereign cloud. Then they said, you don't have a hyperscaler, you don't have a frontier lab, and then we won OpenAI. And they said, okay, you've got government, you've got sovereign clouds, you've got a frontier lab, you don't have a hyperscaler. Then we won a hyperscaler. Right? Each time they said, then they said you've got all these cool customers, but you couldn't do big models. Now we're serving GPT.

那他们现在说什么？现在又是什么说法？他们现在说什么？

What are they saying? What's that now? Like what are they saying?

CEO 是个婴儿潮一代的老古董。这挺好。我是说，但每一次——

CEO is a boomer. That's good. I mean, but each time—

我变成了千禧一代，然后我闭嘴了，你知道——

I became a millennial and I shut down, and you know—

对，所以我觉得当你算是个职业的大卫，当你骨子里是个创业者时，对吧，每一次这种事都会让你热血沸腾。对吧？我们只对解决别人解决不了的问题感兴趣。我们只对做别人做不了的事感兴趣。这就是我们每天早上起床的原因。不是为了钱。不是——我是说，也不是为了名声。是因为我们热爱打造酷东西，而且我们真的很喜欢打造那些难到别人做不出来的酷东西。

Right, and so I think when you're sort of a professional David, when you are an entrepreneur at heart, right, each one of those fires you up. Right? We're only interested in solving problems that other people can't solve. We're only interested in doing things that other people can't do. That's why we get up every morning. It's not the money. It's not—I mean, notoriety. It's because we love building cool things, and we really like building cool things that are so hard that other people can't build them.

Eric 告诉我，你的童年相当酷、相当独特。我不知道那是否对这一切有影响，但你是怎么长大的？

Eric told me you had a pretty cool and unique childhood. I don't know if that fed into this at all, but like how'd you grow up?

我在斯坦福校园长大。我父母是教职员工，那里有个小社区，你的邻居全是其他教授。我的邻居是威廉·肖克利，晶体管的发明者。

I grew up on the Stanford campus. My parents were faculty, and there's a little neighborhood where all your neighbors are other professors. My neighbor was William Shockley, the inventor of the transistor.

太疯狂了。

That's crazy.

对。所以我们对他所知道的，你知道，我们才十岁八岁，就是他妻子在万圣节发全尺寸的糖果棒，对吧？那位发明家曾在贝尔实验室。他带来了硅谷，对吧？他通过把肖克利搬到西海岸，为硅谷奠定了基础。而我们想的是，他发大号的 Three Musketeers 巧克力棒，对吧？但我觉得有几件事很辉煌。首先，唯一的硬通货是智力，对吧？没人在乎你有没有钱。没人在乎你有没有创办过公司。那是 70 年代。没人在乎。他们在乎的是，那家伙真聪明。他做的工作很棒。我另一个邻居是阿尔蒙，他和 David Conorman 一起拿了诺贝尔经济学奖。我爸爸的网球局，有六到八个人轮换。他们周六周日打双打。在我二十多岁某时，我意识到其中三个有诺贝尔奖，还有一个有菲尔兹奖。

Right. So all we knew about him, you know, we're 10 or eight, is that his wife gave out full-size candy bars at Halloween, right? The inventor was at Bell Labs. He brought Silicon Valley, right? He, by the movement of Shockley to the West Coast, created the foundation for Silicon Valley. And we're thinking he gives big Three Musketeers bars, right? But I think there were a couple things that were glorious. First, the only currency was intellectual horsepower, right? Nobody cared if you're rich. Nobody cared if you'd started a company. This was the 70s. Nobody cared. What they cared about was, well, that dude's really smart. He does good work. My other neighbor was Almon, for his work with David Conorman, they got the Nobel Prize in economics. My dad's tennis match, there were six or eight guys in rotation. They played doubles on Saturday and Sunday. Somewhere in my mid-20s I realized three had Nobel prizes and one had a Fields medal.

对吧？

Right?

而他双打很烂。

And he sucks at doubles.

没错。不，我跟你说，那就是老年人的网球。我是说，他们的发球很糟糕。

That's right. No, let me tell you, it was some old man tennis. I mean, their serves were grim.

他们的物理很好。

Their physics was good.

那是一种疯狂的成长方式。

It's a crazy way to grow up.

那是一种疯狂的成长方式。你知道，那个社区很安全。我们骑上自行车，整个夏天到处跑，天黑才回来。

It's a crazy way to grow up. And you know, the neighborhood was safe. We'd get on our bikes and we'd just go all summer and come back when it was dark.

那你当时觉得自己会成为学者吗？

And so did you think you'd be an academic?

是的。我其实当时在读博士。我有点厌倦了。我在完成博士资格考试的同时去了斯坦福商学院。我有点误打误撞被卷进了这一行。你知道，我爸爸现在还问我，他说，你会读完博士吧，Andrew？我说，爸，我的教授们都去世了。没人还在了。

I did. I actually was working on a PhD. I got a little bored. I went to the business school of Stanford while I completed my qualifying exams for my PhD. And I sort of got sucked into this by mistake. And you know, my dad still asks me, he says, you're going to finish your PhD, Andrew? Like, dad, all my professors are dead. Nobody left.

是的，太棒了。好的。Andrew，这次聊得很开心。非常感谢你和我们做这期节目。显然，你知道，你打造了极其特别的东西，能通过 Eric 间接地观察和学习，真的很酷。所以，感谢一切。

Yeah, amazing. All right. Well, Andrew, this was a blast. Thanks a ton for doing this with us. And obviously, you know, you've built something extremely special, and it's been cool to just watch and learn vicariously through Eric. So, thanks for everything.

很高兴能来这里和你们聊天。你知道，Benchmark 是一位非凡的合作伙伴。我是说，我觉得如果你做硬件，你就要和你的投资方绑在一起十年，所以要挑好的，我很自豪我们挑对了。谢谢。

It's a pleasure to be here and chat with you guys. And you know, Benchmark was an extraordinary partner. I mean, I think if you do hardware, you're going to be in bed with your backers for a decade, and pick good ones, and I'm proud we did. Thank you.