Cerebras CEO 谈巨型晶圆与 AI 芯片

大家好，欢迎收听另一期 Odd Lots 播客。我是 Joe Weisenthal。

Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Weisenthal.

我是 Tracy Alloway。

And I'm Tracy Alloway.

Tracy，我不得不说，不幸的是，我没有 AI 精神病。我很确定。

Tracy, I have to say, unfortunately, I don't have AI psychosis. I'm certain of that.

有争议。

Debatable.

我很确定我没有 AI 精神病。但我不得不说，不幸的是，现在感觉 AI 相关的问题——而且有很多——正在吞噬我脑海中的其他想法，无论是关于哪个模型最好以及为什么、推理的经济学、每个模型的预训练与后训练比例。它就像一个不断增长的团块，占据了我越来越多的思绪。

I'm pretty sure I don't have AI psychosis. I do have to say, unfortunately, the amount of time now where it feels like AI-related questions, and there are many of them, are swallowing up the other thoughts in my head, whether it's questions about which model's best and why, what are the economics of inference, how much training is pre-training versus post-training for each model. It's just this blob that's growing, taking up more and more of my thoughts.

你对 AI 精神病的定义是什么？因为有人可能会说，一直想着 AI 本身就是一种精神病。

What is your definition of AI psychosis? Because one could argue that thinking about AI literally all the time would be a form of psychosis.

嗯，这么说吧，首先我不是那种认为 AI 是朋友的人。我没有爱上 AI 模型。我不认为通过与 ChatGPT 合作，我就能偶然发现物理统一理论之类的东西。

Well, let's just say I'm not the type who thinks the AI is a friend, for one thing. I'm not in love with the AI models. I don't think that in collaboration with ChatGPT I'm stumbling on a unified theory of physics and things like that.

但你确实花了很多时间输入指令、按下按钮，然后看结果。

But you do spend a lot of time inputting instructions, pressing the button, and seeing what comes out.

我只是说，我认为我意识到我在和机器对话，我们并没有作为合作者、伙伴和朋友取得任何重大突破。

I'm just saying I think I'm aware that I'm talking to a machine and that we're not establishing any great breakthroughs as collaborators, partners, and friends.

认识到你有问题是治愈的第一步，Joe。不过说真的，越来越多地思考 AI 是有充分理由的，因为不仅是市场，实体经济的大部分现在都围绕着 AI 运转。

Recognizing you have a problem is the first step towards healing, Joe. Seriously, though, there's a good reason to think about AI more and more, which is that a huge chunk of not just the market, but the real economy is now revolving around AI.

所以，在 AI 对话中，有很多子类别。其中一个恰好是 Odd Lots 另一个最喜欢的话题——芯片。芯片以多种不同方式使用，在 AI 供应链的不同部分。不同类型的芯片有不同的角色，所以我们需要了解更多。

So, anyway, within the AI conversation, there are a lot of subcategories. One of them happens to be another Odd Lots favorite topic, which is chips. Chips are used in multiple different ways, in different parts of the AI supply chain. Different types of chips have different roles, and so we have to learn more.

我们需要了解更多。我不得不说，我对我们即将采访的公司特别感兴趣，部分原因是我知道他们的两件事：第一，他们刚刚进行了大规模 IPO，筹集了大约 55 亿美元，估值倍数高得离谱。我甚至无法计算市盈率，因为他们还没有盈利，但按销售额计算，大约是 67 倍远期收益，非常热门。第二，我知道这家公司制造巨大的晶圆。

We have to learn more. And I have to say, I'm particularly interested in the company we're about to speak to, partly because the two things I know about them are: number one, they just had a huge IPO, raising something like $5.5 billion at an insane multiple. I can't even do a price-to-earnings multiple because they're not profitable yet, but on a sales basis, it was like 67 times forward earnings, which is pretty hot. And the second thing I know about the company is they make giant wafers.

是的。

Yes.

这只是一个有趣的画面。

Which is just a fun image to have in your head.

没错。所以，如果你在想，‘好吧，这个领域有一个热门新进入者。他们的差异化是什么？’嗯，关于他们的一点是，他们的芯片非常巨大，大约有餐盘那么大。你可能会以为你在读《洋葱新闻》，但这是真的，而且显然有一些真正的技术优势。

That's right. So, if you were thinking, 'Okay, there is a hot entrant in this space. What is their differentiator?' Well, one fact about them is their chips are just enormous, about the size of a dinner plate. One might think you're reading an Onion article, but it's real, and apparently it has some real technical advantages.

而且这与其他人所做的不同。其他人都在做模块化网络的事情，把一堆芯片连接在一起以获得更多算力、更多内存、更多功率。但这家公司通过巨型晶圆做了不同的事情。

And it's different to what everyone else is doing. Everyone else is doing the modular networking thing, where you get a bunch of chips and connect them together to get more compute, more memory, more power. But this company has done something different in the form of the giant wafer.

巨型晶圆。如果你认为为了获得最大性能，你想减少事物之间的距离，那么就把所有东西放在一个晶圆上。总之，我们将了解更多。我非常兴奋地宣布，我们邀请到了 Cerebras 的创始人兼 CEO Andrew Feldman 做客播客，他真是完美的嘉宾。Andrew，非常感谢你在 IPO 当周来参加播客。

The giant wafer. If you figure that to get maximum performance, you want to lessen the distance between things, then put it all on one wafer. Anyway, we're going to learn a lot more. I'm very excited to say we have the founder and CEO of Cerebras on the podcast, Andrew Feldman, truly the perfect guest. So, Andrew, thank you so much for coming on the podcast on the week of your IPO.

非常感谢你们的邀请。非常荣幸。

Well, thank you so much for having me. What a pleasure.

当然。不如你直接开始吧？那个巨大的芯片，显然是真的，有餐盘那么大。从技术上讲，为什么这实际上是一种优越的架构形式，至少对于 AI 的某些方面来说？

Absolutely. Why don't you just start us off? The big giant chip, they're apparently real, they're as big as a dinner plate. What is the technical reason why this actually makes sense as a superior form of architecture for at least some aspect of AI?

我认为更大的芯片在更短的时间内处理更多信息。这会产生更快的结果。每个人都转向了更大的芯片。Nvidia 在五六年内从 400 平方毫米增加到 800 平方毫米，正是出于这个原因。在计算行业，晶圆级——制造这么大的芯片……

I think larger chips process more information in less time. And that produces faster results. Everybody had gone to bigger chips. Nvidia had moved from 400 square millimeters to 800 square millimeters over the course of five or six years for this exact reason. And in the compute industry, wafer scale, which is building a chip this big...

顺便说一下，对于正在收听的朋友，Andrew 现在举起了芯片，是的，老实说它看起来比餐盘还大。但那是一个大芯片。

By the way, for those who are just listening, Andrew is now holding up the chip, and yes, it looks actually bigger than a dinner plate, to be honest. But that is a big chip.

那是一个大芯片。很漂亮。它比任何其他芯片大 58 倍。它让我们能够使用一种不同类型的存储器。一开始有两种存储器：一种可以存储大量数据，但速度很慢；另一种每平方毫米存储量不大，但速度极快。历史上，所有图形处理单元都使用这种存储量大但速度慢的存储器。这就是它们推理如此缓慢的原因。所以，如果你现在使用 Claude 或任何非 ChatGPT 的产品，你经常会感觉到输入提示后要等待答案，对吧？那是因为存储器速度慢，它们必须将大量信息从存储器移动到计算单元。通过采用晶圆级，我们可以使用这种快速存储器。我们无法让这种存储器每平方毫米存储更多信息，但我们可以增加平方毫米。所以，通过制造这个大芯片，我们能够用这种快速存储器塞满它。这就是为什么我们比最快的 GPU 快 15 倍。在某些问题上，我们比图形处理单元快 50 倍、100 倍，甚至 1000 倍。

That's a big chip. Beautiful. It is 58 times larger than any other chip ever made. And what it did was allow us to use a different type of memory. At the beginning, there are two types of memory. There's memory that can store a lot, but it's really slow. And there's memory that can't store very much per square millimeter, but it's blisteringly fast. Historically, all graphics processing units used this memory that could store a lot, but was really slow. And that's the reason they do inference so slowly. So, if you're using Claude right now, or anything but ChatGPT, what you'll frequently feel is you'll enter your prompt, and you'll wait for an answer. Right? And that's because the memory is slow, and they have to move a ton of information from memory to compute. Now, by going to wafer scale, we could use this fast memory. We couldn't make that memory store more information per square millimeter, but we could add square millimeters. So, by building this big chip, we were able to stuff it to the gills with this fast memory. And that's why we're 15 times faster than the fastest GPU. On some problems we're 50, 100, even 1,000 times faster than graphics processing units.

等等，你能解释一下你们是如何做到的吗？因为我知道以前有人尝试过晶圆级，我似乎记得甚至在 1980 年代就有过早期尝试。你们是如何成功的？

Wait, can you explain how you actually managed to do this? Because I know there have been previous attempts to do wafer scale, and I seem to remember there was even an early attempt in the 1980s or something. How are you able to pull this off?

是的，那是一项雄心勃勃的事业，这是肯定的。

Yeah, it was an ambitious undertaking, that's for sure.

在我们行业 75 年的历史中，每一次之前的努力都失败了，包括 Gene Amdahl，他算是我们行业算力领域的 Mount Rushmore 级人物。他在 80 年代中期于一家名为 Trilogy 的公司遭遇了惨败。不仅如此，在我们成功之后，那些曾参观过我们实验室、试图模仿我们的人也失败了。所以，我们能够做到的是解决一系列真正根本性的问题，这些问题跨越了广泛的技术领域。它们涉及光刻技术，因此我们必须与台积电紧密合作，而他们最终成为了出色的合作伙伴。我们必须在材料和封装方面进行创新。这是如何将处理器、如何将一块硅片放在主板上，并为其提供电力和 I/O 的方式。我们必须在电力传输方面进行创新。对吧？当你制造一个巨大的芯片时，你需要输送的电力比一个邮票大小的芯片要多得多。我们必须发明冷却它的方法。我们必须编写在其上运行的新型软件。所有这些以前都从未做过，这是一个长达十年的过程。我们花了 5 年时间和大约 5 亿美元才交付了第一个芯片。从那以后，这是一段非凡的历程。去年 12 月，我们与 OpenAI 签署了超过 200 亿美元的协议，这是硅谷有史以来最大的合同之一。然后在 3 月，我们与 AWS 签署了协议，他们将在其数据中心部署我们的系统。所以，这真是一段非凡的历程，但花了很长时间。它需要非凡的工程能力。而且确实有很长一段时间，我们并不清楚能否成功。

Every previous effort in the 75-year history of our industry had failed, including Gene Amdahl, who's sort of on the Mount Rushmore of compute in our industry. He failed sort of spectacularly in the mid-80s at a company called Trilogy. Not only that, but after we succeeded, people who had visited us, who'd been in our labs, tried to copy us, and they also failed. And so, what we were able to do is solve a set of really fundamental problems, and those problems cut across a wide swath of technology. They cut across lithography, so we had to collaborate closely with TSMC, and they turned out to be a great partner. We had to make inventions in material and packaging. That's how you put a processor, how you put a piece of silicon on a motherboard, and deliver power and IO to it. We had to make inventions in power delivery. Right? When you build a giant chip, you're going to deliver way more power to it than if you do a chip the size of a postage stamp. We had to invent ways to cool it. We had to write new types of software that ran on it. All of these had never been done before, and it was a decade-long process. It took us 5 years and about 500 million dollars to deliver the first one. And it's been an extraordinary run since. In December, we signed a deal with OpenAI north of $20 billion, one of the largest contracts ever signed in Silicon Valley. And then in March, we signed a deal with AWS, where they would deploy our systems in their data centers, in their AWS data centers. And so, it's just been an extraordinary run, but it took a long time. It took extraordinary engineering. And there were certainly long periods of time when it wasn't clear we were going to make this work.

显然，你们已经达到了一个了不起的里程碑。你们实际上已经上市了等等。目前市场在 IPO 初期对你们公司的估值是 640 亿美元。为了让听众理解，这些芯片是仅用于推理，而不是训练吗？当我们想到 AI 时，我会想，好吧，有训练，训练模型，然后给出答案，那就是推理。这些芯片只用于推理吗？

Obviously, you've hit this remarkable milestone. You have, in fact, IPO'd and so forth. And right now, markets valuing your company at $64 billion early days of the IPO. Just for the listener to understand, the chips are they solely an inference as opposed to, you know, in training? When we think about AI, I think about, okay, there's training, training the model, and then answer giving, that's the inference. Are the chips for just for inference?

所以，有几点。我认为你的描述完全正确。训练是我们制造 AI 的方式。推理是我们使用 AI 的方式。所以，发生的事情是，在 2025 年及 2025 年初，我们制造的模型已经足够智能，变得有用。于是出现了使用的爆炸式增长。我们通过推理来使用 AI。因此，推理需求如潮水般涌来。这种情况在 2026 年仍在继续，我们认为它还会持续很多很多年。这就是发生的事情。在 2015 年，当我们开始考虑创办公司时，我们就知道 AI 即将到来，并且会消耗大量算力。对吧？我们下了两个基本赌注。我们赌它需要专用的芯片。没错，图形处理需要专用芯片，于是有了 GPU。移动计算需要专用算力，于是有了 ARM 处理器。我们下了这个赌注，并且我们赌修改 GPU 架构是不对的。你需要从一张白纸开始。所以，我们从一个新的愿景开始。这个愿景既能做训练也能做推理，而且两者都快了几个数量级。但现在，我们看到的是推理需求的爆炸式增长，以至于我们业务中很大一部分是推理。尽管我们在训练上比 GPU 同样快那么多。

So, a couple things. I think you framed it exactly right. Training is how we make AI. And inference is how we use AI. And so, what happened was that in sort of 2025 and the first part of 2025, the models we made were smart enough to be useful. And there was an explosion of use. And we use AI by doing inference. So, there was a sort of tidal wave of demand on inference. And that has continued in 2026, and we think it will continue for years and years to come. And so, that's what had happened. In 2015, when we began thinking about the company, we knew that AI was on the horizon and that would eat a huge amount of compute. Right? And we made sort of two fundamental bets. We bet that it would need dedicated silicon. And right, graphics had needed dedicated silicon. That's how you got the graphics processing unit. Mobile compute had needed dedicated compute. That's where you got ARM processors. We made that bet and we made a bet that modifying the GPU architecture wouldn't be right. You needed to start with a clean sheet of paper. And so, what we started with was a new vision. And that vision could do training and it could do inference and it was orders of magnitude faster at both. But right now, what we're seeing is such an explosion in demand for inference that a lot of the business that's minute is inference. Even though we're just as fast at, you know, the same amount faster than GPUs on training.

这很有趣。也许我们稍后会更多地讨论理论上的训练市场。快速说一下推理，Ben Thompson，他写科技通讯，在一篇文章中区分了答案推理和智能体推理。答案推理就像是，格式化我的简历之类的，或者写一篇关于某主题的文章，或者回答一些问题。而智能体推理则是，这里有一个东西会四处活动，为你提供服务，而不是生成视觉答案。你区分这两者吗？在你看来，这是一个真正的分界吗？你的芯片能同时做这两件事吗？

That's interesting. Maybe we'll get more to the theoretical training market a little later. Just real quick on inference, Ben Thompson, who writes a newsletter about tech, he wrote a piece in which he distinguishes between answer inference and agentic inference. So, answer inference is like, you know, format my resume or whatever or write me an essay on X or Y or answer some questions. And then agentic inference is like, okay, here's this thing that's going to go around. Do you distinguish and do services for you, not producing visual answers? Do you distinguish between those two? Is that a real divide in your view? And can your chips do both?

我们的芯片可以同时做这两件事。我认为这是一个分界。我认为速度在两者中同样重要。我认为如果你在与 AI 互动，比如你在写代码，这是智能体式的，如果你在写代码或工作，没人愿意等待。我的意思是，我们可以反过来问：“慢速搜索的市场有多大？”零。拨号上网的市场有多大？零。为什么？因为没人愿意等待。对吧？所以，如果你在与 AI 互动，速度至关重要。但如果 AI 在做智能体工作，而你的竞争对手在 20 分钟内完成的工作量是你的三倍、五倍、十倍，你就会被淘汰。所以，那种认为速度在智能体流程中不重要的观点是大错特错的。速度在生产性工作的所有方面都很重要。你在更短时间内完成更多工作的能力是一种基本优势，会随着时间的推移而累积。对吧？如果你的竞争对手做一单位工作，你能做三单位。下一次他们做一单位，你能做六单位。对吧？这随着时间的推移会累积，你在任何工作领域都能击败他们。所以，速度，这算是我们的专长，在各个方面都很重要。

Our chips can do both. I think it is a divide. I think speed matters equally in both. I think if you are engaged with the AI, if you're writing code, which is agentic, if you're writing code or you're doing work, nobody wants to wait. I mean, we can just turn the question around and say, "Well, how big is the market for slow search?" Zero. How big is the market for dial-up internet? Zero. Why is that? Because nobody wants to wait. Right? So, if you're engaged with the AI, speed is of the essence. But if the AI is doing agentic work, and your competitor gets three times, five times, 10 times as much work done in 20 minutes than you do, you're going to get smoked. And so, this notion somehow that been proposed that speed isn't very important in agentic flows is dead wrong. That speed is important in all aspects of productive work. And that your ability to get more done in less time is a fundamental advantage that accrues over time. Right? If while your competitor is doing one unit of work, you can do three. And in the next time they do one unit of work, you do six. Right? This adds up over time, and you beat them in any line of work. And so, speed, which is sort of our specialty, is important across the board.

巨大的晶圆和速度实际上对 token 的经济性意味着什么？因为我的一种思考方式是，我脑子里有这样的画面：比如，我去买牙膏，我知道我偶尔需要牙膏，我去 CVS 这样的店，买一支牙膏，然后一周后再买一些。或者我可以去 Costco 买一大管牙膏带回家，可能更便宜。这有点像我对巨大晶圆的理解。也许这是个类比。但速度实际上对 token 的成本意味着什么？

What do giant wafers and speed in general actually mean for, I guess, the economics of tokens? Because one way I think about it, I have this sort of vision in my head, like, okay, if I'm out shopping for toothpaste, I know I need toothpaste every once in a while, and I go into like a CVS, a store, I get one thing of toothpaste, and then maybe a week later I get some more toothpaste. Or I could go to Costco and buy a giant thing of toothpaste and take it home, probably at a cheaper cost. And that's sort of how I think of the giant wafers. Maybe it's a analogy. But what does speed actually mean for the cost of tokens?

嗯，我认为有几点观察。到目前为止，人们选择对速度定价更高一些。例如，Anthropic 提供了一项高级服务，他们提供的 token 速度快两倍，但收费是六倍。而且他们卖光了，无法满足需求。现在，让你有个概念，我们比他们快两倍的速度还要快 15 倍。所以，人们看重速度，因为它能让他们做更多工作。他们看重自己的时间。当你能在更短时间内做更多工作时，你就提高了人们的生产力。这就是为什么人们选择对速度收取溢价。它们的制造成本并不更高。

Well, I think there are a couple observations. I think people have chosen so far to price speed a little higher. For example, Anthropic offered a premium service in which they offered tokens twice as fast and charged six times as much. And they sold it out. And they couldn't meet the demand. Now, just to give you an idea, we're 15 times faster than they're twice as fast. And so, people value speed because it allows them to do more work. And they value their time. And when you can do more work in less time, you are making people more productive. That's why people have chosen to price them at a premium. They don't cost more to make.

事实上，GPU 架构在生成极慢的 token 方面非常出色且高效。如果你不介意速度慢，GPU 上每个 token 的成本极低。但 GPU 有一个特点：当你试图加快速度时，每个 token 的成本和功耗都会增加。有点像开车时速度越快，每加仑汽油跑的里程就越少。所以，当你试图让速度足够快、足够有趣、足够让用户的注意力集中在这个产品上时，它们就变得极其昂贵且耗电。因此，问题不仅在于人们为 token 支付什么价格，或者他们选择如何定价，而在于它们实际的生产成本。GPU 能以极低成本生成极慢的 token，但在快速 token 上却贵得离谱。我们让快速 token 的成本远低于 GPU，并且只消耗一小部分电力。

In fact, the GPU architecture is extremely good and efficient at building very slow tokens. And if you don't mind slow, the cost per token on a GPU is extremely low. But the GPU has a characteristic: as you try to go faster, the cost and power used per token increase. Sort of like as you go faster in your car, your miles per gallon decrease. So what happens is as you try to get fast enough to be useful, interesting, and keep users' intelligence focused on this product, they become extremely expensive and power hungry. So the question is not just what people are paying for a token, or what people choose to price them at, but what they actually cost to make. GPUs make very slow tokens very cheaply, and they're unbelievably expensive at fast tokens. We make fast tokens vastly less expensive than GPUs and use a tiny fraction of the power.

假设我们承认这一切都是真的，每个人都想要最快的，每个人都觉得，你知道吗？这就是 Cerebras 的技术，一个大芯片，这才是真正的方向。展望明年、后年，你在推理市场的市场份额有多大程度上取决于你在台积电晶圆厂获得产能的能力？这在多大程度上是增长的瓶颈？

Let's say we stipulate that this is all true and everyone wants the fastest and everyone's like, you know what? This is the solution that the Cerebras technology, one big chip, this is really where it's at. How much of your market share for the inference market when you look out next year, the year after, etc., how much is your market share going to be dictated by your ability to get capacity at TSMC fabs? How much is that a gating mechanism for growth?

你知道，台积电是供应链中非常重要的一环。但我们有一些真正的优势。目前有三个领域限制了供应商构建 AI 算力。第一是 HBM 内存。就是我们之前描述的那种能存储大量数据但速度很慢的内存。它大约由三家公司生产：三星、海力士和美光。它面临着难以置信的供应压力，极难获得，交货周期很长，现在价格贵得离谱。我们不用它。第二个限制因素是台积电内部的一个工艺，叫做 CoWoS。这是英伟达和其他 GPU 使用的工艺。我们不用它。第三，在台积电，压力最大的工厂是他们的 3 纳米工厂。我们不用它。我们用的是 5 纳米。所以我们成功避开了一些最紧张的供应限制。当然，台积电仍然需要给我们有意义的产能分配，他们从一开始就是非凡的合作伙伴，而且是迄今为止地球上最伟大的制造公司。晶圆厂有点像现代的金字塔，令人难以置信，我强烈推荐你或你的听众，如果有机会去台北，一定要去看看。他们真的非常出色。

You know, TSMC is a huge part of the supply chain. But we have some real advantages. There are three areas right now that are limiting vendors in building AI compute. Number one is HBM memory. It's this memory we described earlier that can store a lot but is really slow. That's made by three companies approximately: Samsung, Hynix, and Micron. And it's under unbelievable supply pressure. It's extremely difficult to get. They have very long lead times. It's unbelievably expensive right now. We don't use it. The second part that's limiting is a process inside of TSMC called CoWoS. And this is the process that Nvidia and other GPUs use. We don't use it. The third thing is that at TSMC, the factory that is under most pressure is their 3 nanometer factory. We don't use it. We use 5 nanometer. So we have managed to avoid some of the most binding supply constraints. Now, TSMC still has to give us a meaningful allocation, and they've been an extraordinary partner from the get-go, and they are the greatest manufacturing company on Earth by far. A fab is sort of a modern pyramid. It's an unbelievable thing, and I highly recommend you or any of your listeners, if you get a chance to go to Taipei, go and see them. They are just extraordinary.

可以参观晶圆厂吗？

Can you do fab tours?

实际上可以。是的，可以参观晶圆厂。你可以去，他们有一个创新博物馆，非常了不起。他们有点像台湾的国家冠军。但我认为目前台积电已经给了我们所需数量的晶圆。现在的业务受限于数据中心。这就是最大的讽刺，对吧？你发明了在计算史上 75 年来无法制造、从未被发明过的技术。你编写了非凡的软件。你制造了比现有产品快得多的产品，而我们都被什么限制住了？建筑物。数据中心现在是整个行业每个人的限制。电力建筑，也就是房地产。现在这真是件神奇的事。而且这几乎是普遍现象。未来 15 到 18 个月内肯定不会改变。

You can, actually. Yeah. You can do fab tours. You can go, and they have a museum of innovation, and it is an extraordinary thing. They are the sort of the national champion of Taiwan. But I think today TSMC has given us as many wafers as we've needed. Business today is constrained by data centers. And that's the grand irony, right? You invent technology that has been unbuildable, never been invented for 75 years in the history of compute. You write software that is extraordinary. You build a product that is vastly faster than the incumbent, and what are we all constrained by? Buildings. Data centers right now are everybody's constraint in the entire industry. Power buildings, so real estate. It is an amazing thing right now. And that is true sort of across the board. And that will not change for the next 15 or 18 months, for sure.

既然我们在谈论物理限制，我想我应该问你，我们最近做了一期关于氦气的节目，关于氦气短缺，考虑到霍尔木兹海峡的情况。氦气的用途之一是半导体芯片的光刻。这对你有影响吗，或者这是你在关注的事情吗？

I mean, since we're talking physical constraints, I guess I should ask you, we did an episode about helium recently, the helium shortage, given the situation in the Strait of Hormuz. And one of the things that helium is used for is lithography on semiconductor chips. Has that affected you at all, or is that something that you're monitoring?

我们在关注，但能做的有限。而且有很多我们可以影响的事情需要担心。我们显然每天都与台积电沟通。我们每天都与整个供应链沟通。我们随时了解各种问题。但这对我们没有影响，我们把它归为我们的制造合作伙伴也会担心、但我们无法帮助的事情。

We monitor, but there's not a lot we can do. And there's plenty of stuff to worry about that we can affect. We obviously are in communication every day with TSMC. We're in communication with our entire supply chain every single day. And we stay abreast of the various issues. But it has had no impact on us and we put that in the bucket of things that our manufacturing partners worry about also and that we can't help.

除了制造这些芯片，实际上，我之前不知道，你们还有自己的云服务。是的。或者你们有自己的云服务。我对此有很多问题。但你们有自己的云服务，用户可以通过它访问各种开源模型等等。从视觉上看，它有点像 Open Router 的界面，大致相同的环境，只是全都是开源的。我好奇的是，也许你可以谈谈，在传统软件开源中，开源的一个好处是你不需要付费。所以它是免费的。但当我们谈论 AI 软件时，情况有点不同，因为即使它是免费的，你仍然需要支付芯片的折旧费和运行它们的电费。所以没有真正免费的 AI 开源软件。但我好奇的是，作为云服务商，你的经验中，开源模型在每单位智能上是否更便宜？如果我们有一种方法来表示智能的平准化成本——我不知道行业是否已经有了——开源模型在每个 IQ 点上是否更便宜，无论我们如何衡量智能？

You know, so in addition to manufacturing these chips, you actually, I didn't realize this, you have your own cloud. We do. And or you have your own cloud services. Which I have a bunch of questions about that. But you have your own cloud services through which a user can actually get access to various open source models and so forth. It looks a little bit sort of visually it looks a lot like the open router interface, roughly the same environment except it's all like the open source. What I'm curious about and maybe you can speak to this, you know, in traditional software open source, one nice thing about open source is you don't have to pay for it. So it's free. It's a little bit different when we're talking about there's no really such thing as like free AI software cuz even if it's like free, you still have to like pay for the depreciation of the chips and you have to pay for the electricity to run them. So there's no real such thing as like free open source AI software. But what I am curious about in your experience as a cloud vendor, are the open source models cheaper on a per unit of intelligence basis? If we had some way of saying levelized cost of intelligence, which I don't know if the industry has yet. Are open source models cheaper per IQ point, whatever we however we want to measure intelligence?

是的。便宜很多。

Yes. By a lot.

真的吗？

Really?

是的。我认为在闭源世界里，你为那额外的一点点智能付出了很多，对吧？开源模型，没有开源模型能像闭源模型一样好。好吧，可以认为有 3%、4%、5% 的差异。大概在这个范围内。可能多一点，也可能少一点。但使用它们的成本呢？你现在就可以跳上去运行 Kimik 82。这是一个 1 万亿参数的模型。它是 Cerebras 上的开源模型。我们比其他方案快 10 到 15 倍。你支付的是我们的电力成本和一部分计算成本。你没有支付的是训练它的成本。这就是市场上正在进行的战斗。有 OpenAI 和他们的编码软件。有 Anthropic 和他们的编码软件。还有像 Cursor 和 Cognition 这样的公司在使用开源。我们为 OpenAI 和 Cognition 提供支持。

Yeah. I think in the closed source world you're paying a lot for that extra little bit of intelligence, right? The open source models, there are no open source models that are as good as the closed source models. Okay. Think of it as 3%, 4%, 5% different. So something in that range. And then, it could be a little more, could be a little less. But the cost to you using them, right? You can jump up right now and run Kimik 82. It's a 1 trillion parameter model. It's an open-source model on Cerebras. We're 10 or 15 times faster than others. And what you're paying for is the cost of our power and some cost of the compute that took to calculate it. What you're not paying for was the cost to train it. And that's a battle that is underway in the market. You have OpenAI with their coding software. You have Anthropic with their coding software. And you've got companies like Cursor and Cognition that are using open source. We power OpenAI and we power Cognition.

闭源和开源之间正在进行一场战斗。我认为这场战斗的赢家尚未确定。明确的是，闭源稍微好一点，好多少各不相同，而且更贵。

You have a battle underway between closed-source and open-source. And I think that the winners of that battle is yet to be determined. What is clear is that the closed-source is strictly better by a little bit. By how much varies, and it's more expensive.

是的，我想我们之前讨论过这个。但我听说很多美国大公司都在悄悄地从一些闭源模型转向开源模型，比如中国的 Kimi 和 Quan。抱歉追问这一点。但如果你必须下注，20 年后，主导的 AI 模型会是便宜的开源模型，还是更贵、略优的闭源模型？

Yeah, I think we've talked about this before. But, like I've heard of a lot of big companies in the US who have been very quietly shifting from some of the closed-source models to the open-source models, like the Chinese ones, like Kimi. Is that what it's called? Kimi and Quan. I'm sorry to press you on this point. But, if you had to make a bet, like in 20 years, is the dominant AI model going to be a cheap open-source thing or a more expensive, incrementally better closed-source model?

我不认为会只有一个赢家，对吧？SaaS 软件也不是只有一个，对吧？有一些大玩家，比如 Salesforce，还有其他一些巨头，还有很多专家。我想不出有多少市场最终只由一个玩家主导。看看半导体市场，x86 有 AMD 和 Intel 两大玩家，然后 ARM 及其芯片制造商占据了整个相邻市场，此外还有定制芯片。我认为 AI 也会是这样。OpenAI 会继续做出非凡的事情，会有竞争对手，也会有开源。我认为这些都不会消失。

I don't think there's going to be one, right? There's not one SaaS software, right? There are some big dogs, right? There's Salesforce, there's some other sort of giant players, and there are lots of other specialists. I can't think of many markets where we've sort of settled onto one player, right? If you look at the semiconductor market, you've got x86 where you've got two major players in AMD and Intel, and then you've got a whole adjacent market owned by ARM and the companies that build ARM parts, and then you've got custom silicon around that. I think that's the way you're going to have this. We're going to have, you know, OpenAI is going to continue to do extraordinary things. There will be competitors to them and there will be open source. I don't think any of those go away.

既然我们谈到软件，当讨论新芯片公司挑战 Nvidia 时，常听到一种观点：Nvidia 芯片固然好，但真正的护城河是 CUDA 这个软件栈。你怎么看？对于试图挑战像 Nvidia 这样庞大且深度嵌入软件系统的公司，这是一个现实的担忧吗？

Since we're on the topic of software, one of the things you often hear when talking about new chip entrants going up against Nvidia is this idea that, well, Nvidia chips are great and all, but the real moat of Nvidia's business is CUDA, that software stack that goes with it. What's your take on that? Is that a realistic concern for someone who's trying to go up against a company as big and as embedded in the software system as Nvidia currently is?

Nvidia 可能是本世纪上半叶最伟大的公司。Jensen 是我们这个时代最伟大的 CEO 之一，还有 Broadcom 的 Hock Tan 和 AMD 的 Lisa。这非常了不起。CUDA 在创造 AI 格局中确实非常重要，但现在它不重要了。

Nvidia is probably the greatest company in the first part of the century, right? Jensen's one of the great CEOs of our era along with Hock Tan at Broadcom and maybe Lisa at AMD. It's just extraordinary. And CUDA was really important in the creating of the AI landscape. But it's not important now.

嗯。

Mhm.

而且它在推理中完全没有作用。如果你想从今天在 GPU 上运行模型转向在我们这里运行，只需 10 次按键。只需迁移，指向我们的 API。

And it has no role whatsoever in inference. If you want to move from running a model on GPUs today to running it on us, we can move it in 10 keystrokes. Just move, point to our API.

嗯。

Mhm.

这是第一部分。第二部分是，一年前每个主要前沿实验室的模型都建立在 CUDA 基础上。而今天，三个中有两个不是。所以他们失去了 70% 的市场份额。三个领先的前沿模型是 Gemini、Claude 和 GPT。Gemini 由 Google 在 TPU 上构建、训练和服务，没有 CUDA。

So that's the first part. The second part is that a year ago every major frontier lab model had been built on a CUDA foundation. And today two of three haven't. So they lost 70% market share. Their three leading frontier models, Gemini, Claude, and GPT. Gemini, built by Google on TPUs, trained on TPUs, served on TPUs, no CUDA.

Anthropic 的模型在 Trainium 上训练，没有 CUDA，在 TPU、Trainium 和 GPU 上服务。而 OpenAI 的 GPT 在 GPU 上训练，使用 CUDA 环境。所以，今天三个领先模型中有两个不使用 CUDA。这是市场份额的大量流失。因此，我认为三到五年前 CUDA 在中心占据主导地位的情况已经显著缩小。在推理中完全不重要，在训练中的作用也在缩小。

Anthropic's models, trained on Trainium, no CUDA, served on TPUs, on Trainium, and on GPUs. And OpenAI's GPT, trained on GPUs in the CUDA environment. So, two of the three leading models today use no CUDA. That's a hemorrhaging of share. And so, I think what was true three or five years ago in which CUDA had a dominant position with central, has shrunk significantly. And not important at all at inference, and shrinking in its role in training.

既然我们在讨论推理的经济学等问题。我真的很想听听你的看法。就在过去几周，出现了一波试图将算力市场金融化的公告。比如，你可以购买一些容量，H100 基准等等。人们可能理论上想对冲。我不完全信服。在我看来，这似乎不是必须的，但另一方面，推理提供商可以与数据中心建立非常长期的双方关系，不需要这种现货对冲市场。你认为市场会发展到对允许推理提供商对冲价格风险的金融工具有显著需求吗？

Since we're talking about the economics of inference and all this stuff. I've actually I would love to get your take. One of the things that literally in the last couple of weeks, there's been this flurry of announcements of these attempts to financialize the market for compute. And so, it's like, oh, you're going to buy some capacity, the H100 benchmark, etc. And people want maybe theoretically hedging it. I'm not entirely convinced. It still seems to me like it's not a maybe, but on the other hand, like an inference provider can lock in a very long-term relationship bilaterally with a data center and so forth, and no need for like these spot hedging markets. Do you think the market is going to evolve in such a way that there will be significant demand for financial instruments that allow inference providers to hedge their price exposure?

我不知道。首先，我不是金融工程师。但我们可以稍微看看历史。

I don't know. I'm not a financial engineer, is the first thing. But we can look a little bit at history.

好的。

Okay.

CoreWeave 的人在为其大规模部署融资方面极具创新性。他们是最早使用以 GPU 为后盾的债务工具的公司之一。这使他们能够真正脱颖而出，在新云领域获得先发优势。这是金融工程上的创新，极具创意。其他人纷纷效仿，现在有一个庞大而活跃的债务市场，用于资助数据中心的建设和装修。

The guys at CoreWeave were enormously innovative in how to fund some of their massive deployments. They were some of the first to use a debt instrument that had a backstop with the GPU. And this enabled them to really leap out and sort of have first mover advantage in the neo cloud space. And that was an innovation in financial engineering and extremely creative. Others followed and now there's a big and active debt market in funding the building and the fit out of data centers.

是的。

Yeah.

当市场如此庞大且活跃时，人们会想在两边下注。我认为随着时间的推移，这些赌注会正常化和规范化，你可以打包它们，让下注变得容易。当 CoreWeave 是最早以 GPU 为抵押贷款的公司之一时，这确实很有创新性。不仅 CoreWeave 因创造这种工具而获得赞誉，交易的另一方也因参与其中而获得赞誉，并成功进行了创新性下注。随着越来越多的人加入，这些可以规范化，更容易定价。一旦规范化并形成市场，历史上该市场的衍生品就容易产生。这就是我看到的演变方式。随着数据中心和算力市场的成熟，会有人们在两边下注，金融工具将被创造出来。至于这是否是个好主意，我目前没有意见。

When you have a market that is that big and that active, you have people who want to make bets on either side. And I think over time those bets normalize and regularize and you can wrap them up and you can make it easy to make the bet. When CoreWeave was one of the first to loan money against GPUs for CoreWeave, this was really innovative. And not only does CoreWeave get credit for the creating of the instrument, but so does the other side of the deal for doing it. And making a successful innovative bet. And as more and more people jumped in and these could be regularized, they could be more easily priced. And then once it's regularized and you have a market, then derivatives of that market are easy to make historically. And that's the way I see this unfolding. That as this market for data centers and compute matures, there'll be people making bets on either side and financial instruments will be created to do it. Whether it's a good idea or not, I have no opinion at this time.

既然我们提到了金融，我看了 IPO 文件中的一些实际数字。我知道你现在有 OpenAI 的交易，但你收入的很大一部分来自阿布扎比一家叫 G42 的公司。我认为他们既是你的最大客户，也是主要投资者。G42 用这些芯片实际做什么？

Since we brought up finance, I was looking through the IPO filing and looking at some of the actual numbers in there. And I know you have the OpenAI deal now, but a huge chunk of your revenue comes from this company called G42 in Abu Dhabi. And I think they're both your biggest customer and also a major investor. What does G42 actually do with all these chips?

当然。去年他们是我们业务中非常重要的一部分，占了很大比例。他们是少数投资者。他们是阿联酋的国家冠军，国家 AI 冠军。他们建立了一个云，用于整个阿联酋的生态系统。所以，它被那里的顶尖大学、顶尖公司使用，比如 ADNOC（他们的主要石油公司），以及 G42 的九家运营公司。

Sure. Last year they were a really important chunk of our business, a lot of it. They're a minority investor. They are the national champion, the national AI champion of the UAE. And they built a cloud that is used across the UAE's ecosystem. So, it's used by leading universities there, it's used by leading companies there, companies like ADNOC, their leading oil company. It's used by G42's nine operating companies.

迄今为止的部署都在美国。我们在圣克拉拉、明尼阿波利斯、德克萨斯州达拉斯，很快还有多伦多，拥有为 G42 运行设备的大型数据中心。他们正在进行训练和推理。在训练方面，他们开创了一些领先的英阿模型，还做了基因组学工作。他们提供模型服务，并作为云服务商运营，尤其服务于阿联酋生态系统，也服务于全球公司。

The deployments to date have been in the US. We have data centers that massive data centers that run equipment for G42 here in Santa Clara, but also in Minneapolis and Dallas, Texas, soon in Toronto. And so, they're doing training and they're doing inference. The training they're doing they have pioneered some of the leading English-Arabic models. They've done genomic work. They're doing serving of models and they're operating as a cloud, particularly for the UAE ecosystem, but also for global companies.

你认为随着时间的推移，企业用户（也许还有个人用户，但主要是企业用户）是否会希望推理服务由独立于模型制造商的另一家公司提供，以确保他们不会泄露信息，从而训练可能取代他们的公司？你看 Anthropic 每隔几天就宣布一些新东西。哦，我们有一个新的 markdown 文件，可以用于税务或其它什么，然后一堆公司就倒下了。使用 AI 的公司是否会越来越倾向于使用非模型制造商本身的数据中心和推理提供商？

Do you think that over time corporate users and perhaps individual users but corporate users will want inference served from a company that's separate from the model maker such that they can be certain that they are not revealing and thus training the company that might replace them? I mean look Anthropic every couple days announces some new thing. Oh, we have a new markdown file that could do this for taxes or that could do this for whatever and then a bunch of companies fall. Like are companies that you use AI increasingly going to want to want want to use data centers and inference providers that aren't the model themselves?

嗯，首先我认为有一类专业人士、一类工作最直接受到 AI 的威胁。

Well, first I think there is a type of professional a type of job that is most directly under threat from AI.

好的。

Okay.

而且他们几乎总是白领。他们需要你对某个知识体系有专长。对吧？这就是会计师。对吧？他们对裁决、先例、税案判例法等知识体系有专长。这正是 AI 目前擅长的。没错。所以律师、会计师这类专业人士，他们一直站在对 IRS 税务规则一无所知的普通人和税务规则之间。

And they're almost always white-collar. And they required you to have expertise over a body of knowledge. Right? That's what an accountant is. Right? They have you have expertise over a body of knowledge of rulings of previous examples of tax case law etc. That's exactly what AI is good at right now. Exactly. So lawyers, accountants, this sort of these professionals who have stood between sort of the ordinary person who doesn't know anything about IRS tax rules and the tax rules.

这正受到威胁。而且这是 OpenAI 和 Anthropic 这样的公司很容易就能攻克的。还有其他领域，比如药物设计、遗传学、基因组学，像葛兰素史克这样的公司拥有卓越且独特的数据集。我们的大客户之一梅奥诊所也是如此。葛兰素史克和我们其他制药客户也是如此。他们拥有独特的数据。

That is under threat. And that is something that it will be very easy for companies like OpenAI and Anthropic to chew through. There are other areas like say drug design genetics, genomics where companies like GlaxoSmithKline have remarkable and unique data sets. This is true for one of our large customers, Mayo Clinic. It's true for GlaxoSmithKline and other of our pharma customers. They have unique data.

是的。

Yeah.

他们将能够从这些数据中发现洞见并获得价值。他们当然不想与基础模型制造商共享这些数据，除非得到保证不会让通用模型变得更聪明。

And they will be able to find insight in that data and they will be able to get value from that data. And they will certainly not want to share that data with the foundation model makers unless they are guaranteed that it will not sort of make the general model smarter.

对。

Right.

这些公司花了二三十年，每年花费数百亿美元收集数据。对吧？患者护理记录或药物设计的测试结果。他们将挖掘这些工作中的洞见，并发现非凡的东西。这些数据受到更多保护，因为洞见就在数据中，而他们拥有数据。

And these are companies that have spent 20 or 30 years spending tens of billions of dollars a year gathering data. Right? Patient care records or test results for drug design. They're going to mine the insight in this work. And they're going to provide find extraordinary things. And those are much more protected because the insights in the data and they have the data.

你知道，你之前谈到台湾的晶圆厂，我现在后悔在台北时没去参观晶圆厂。但当时没想起来。下次吧，希望如此。根据《芯片法案》和其他一些产业政策，美国已经做出了各种努力，试图建立更多的芯片制造能力。在你看来，实际执行的主要障碍是什么？是的，A 是它正在发生吗？B 是为什么看起来这么难实现？

You know, you were talking about fabs in Taiwan earlier and I'm now regretting not going on a fab tour when I was in Taipei. But it just didn't cross my mind at that time. Next time, yeah, hopefully. And there've been various efforts under the CHIPS Act and some other industrial policies to try to build more chip making capacity in the US. In your view, what's the big, I guess, impediment to actually do it? Yeah, A is it happening and then B why does it seem so difficult to actually make happen?

对。首先，困难是因为这是一个难题。它们很困难。建一个晶圆厂要花 300 到 400 亿美元，需要 5 到 6 年时间。所以，这笔钱和这段时间跨越了多届政府。对吧？这就是美国政治的问题所在。很难制定出跨政府、跨时间持久的政策。这是第一点。第二点，这些建筑极其复杂。我们有一堆杂乱无章、奇怪的本地和区域建筑规范，晶圆厂制造商必须逐一应对。第三，我们正在努力。台积电已为其亚利桑那州的晶圆厂投入了数百亿美元，并承诺再投入数千亿。三星已为其德克萨斯州的晶圆厂投入了数百亿美元，并承诺再投入数千亿。但这需要很长时间，我们必须持续投入，不仅建设晶圆厂，还要建设周边生态系统，不是三五年，而是 20 年或 25 年。因为你不仅想要一个晶圆厂，而是想要一整个晶圆厂的发展轨迹。你希望它们不仅在今天的前沿工作，还要在明天、明年和 10 年后的前沿工作。这些在美国已被证明极具挑战性。我认为我们需要它。它们是战略资产。我认为我们需要找到与拥有专业知识的人合作的方式，并找到制定持久政策的方法，以建立一个充满活力的晶圆厂及相关生态系统。

Right. The first thing is difficult because it's a difficult problem to They're hard. They cost 30 or 40 billion dollars and take 5 or 6 years to build. So, that amount of money and that amount of time cuts across administrations. Right? And that's a problem with the politics in the US. Is it's hard to make policy that's durable across administrations and across time. It's the first thing. The second thing is these are remarkably complicated buildings. And we have a sort of a hodgepodge, a sort of strange lattice work of local, regional building codes that that a fab maker has to negotiate. Third is we're trying. TSMC has dedicated tens of billions of dollars to their fabs in Arizona and have committed hundreds of billions more. Samsung has dedicated tens of billions of dollars and committed hundreds of billions more to their fabs in Texas. But they take a long time and we have to remain committed to building not just the fab, but the surrounding ecosystem, not just for three or five years, but for 20 years or 25 years. Because you want not just one fab, but you want a whole trajectory of fabs. You want them working at today's cutting edge, but tomorrow's and next year's and in 10 years' cutting-edges, well. And those are things that have proven really challenging in the US. And I think we need it. They're strategic assets. And I think we need to find ways to collaborate with those that have the expertise and to find ways to build policy that is durable over a length of time that can build a vibrant ecosystem in the fab and the associated elements.

那么，关于半导体，另一个重要的政治经济主题是，它们实际上是一种具有战略重要性的技术，因此美国应该对其海外使用施加一些限制。所以我们看到了出口管制、出口限制等措施。你是一家实际的芯片公司，所以我很好奇，在运营层面，你对这些出口管制的实际体验是什么？它们占用了你多少时间？此外，考虑到你最大的客户之一是阿布扎比的一家国际公司，这些出口管制的走向对你未来的业务有多重要？

So, the other big political economy theme, I guess, when it comes to semiconductors is this idea that they are in fact a strategically important technology and so the US should place some limitations on their use abroad. And so we've seen things like export controls, export restrictions. You're an actual chip company, and so I'm very curious at an operating level what your experience of these kind of export controls has actually been. Like how much time does that take up for you? And then also, given that one of your biggest customers is an international firm in Abu Dhabi, like how important is the trajectory of those export controls to your future business?

我认为，你知道，三四年前我会说完全不重要。今天我认为它们非常重要。在上一届政府中，我认识了商务部以及负责许可的工业与安全局（BIS）的领导。我认为这是一项极其困难的工作，我们看到非常勤奋、聪明的人在从事一项非常非常困难的工作。我认识了本届政府中的人，发现同样如此。他们每个人的收入都只是他们在私营部门能赚到的零头，他们这样做是因为他们相信这是一项重要的使命。问题在于，对于如何正确执行存在不同看法。对于如何正确实现目标——即不把最宝贵的技术交给工业敌人——也存在不同看法。我认为我们可以同意，在当今环境下，中国是一个工业敌人。善良、善意的人可能对正确的策略是否是限制他们获取技术存在分歧。

I think, you know, three or four years ago I would have said not important at all. I think today they're really important. In the last administration, I got to know the leadership in the Department of Commerce, and in the BIS division of Commerce, which oversees the licensing. I think this is an extraordinarily difficult job, and we saw really hard-working, smart people doing a job that is very very difficult. I got to know the people in this administration, and I found the same. Every single one of them is earning a tiny fraction of what they could earn in the private sector, and is doing this because they believe that this is an important mission. The problem is that there are differing views about the right way to do this. And there are differing views on the right way to achieve the goal, which is to not give your most precious technology to your industrial enemy. And I think we can agree that today, in today's environment, China is an industrial enemy. Good, well-meaning people can disagree on whether the right strategy is to limit them from gaining access.

其他人，比如英伟达的人认为，正确的策略是让他们访问，并让他们继续使用美国制造、美国设计的产品。我持相反观点。我理解双方都有合理的论据。我认为限制我们最宝贵技术的扩散是有道理的。

Others argue, as those at Nvidia have argued, that the right strategy is to give them access and to keep them working on US-made, US designed product. I come down on the other side of that argument. I understand there are good arguments in both directions. I think limiting the diffusion of our most precious technologies makes sense.

嗯。

Mhm.

而且我认为我们必须深思熟虑，并认识到这意味着一些市场将对我们关闭。我对此可以接受。

And I think we have to do it thoughtfully and we have to recognize that means some markets will be foreclosed to us. And I'm okay with that.

快速问一下当前业务的事。你提到了与 AWS 的交易。那是怎么运作的？现在客户，比如 AWS 的客户，能付钱让他们在你们的芯片上专门运行推理吗？

Just quickly on the current business stuff. You mentioned the deal with AWS. How does that work? Can customers right now, like can customers of AWS pay them to have inference served specifically on one of your chips?

还没有，但很快了。

Not yet, but soon.

好的。

Okay.

它将在 Bedrock 中提供，那是他们的 AI 即服务产品。他们将能够下拉菜单，获得超快推理，这通过一种称为解耦解决方案的组合来交付，部分推理工作使用 Tenium，其他部分使用 Cerebrus 技术和我们的 CS-3 系统。

They will be served in Bedrock, which is their AI as a service offering. And they will be able to go down the click-down menu and get super fast inference, which will be delivered via a combination of what's called a disaggregated solution, using some Tenium for some of the inference work and using the Cerebrus technology and our systems called the CS-3 for other parts of the work.

那么，有人下拉选择它，大概会为这种超快推理支付溢价。

And presumably someone who scrolls down and selects that, they would pay some premium for that ultra-fast inference.

我认为他们会支付溢价。我们拭目以待。这完全由亚马逊定价。他们的产品。

I think they will pay a premium. We will see. This is entirely as Amazon wishes to price it. Their product.

所以你们这周上市了。现在是 2026 年 5 月。这不是你们第一次尝试或考虑上市。早在 2024 年就有关于你们想上市的新闻。去年也有新闻，特别是因为与 G42 的关系，涉及 Cerebrus 和一些国家安全担忧，可能影响了 IPO。但去年 9 月，你们在 G 轮融资中，其中一个投资者是 1789 Capital，这家公司当然与小唐纳德·特朗普有关联，这有很多含义。然后 IPO 就发生了。我是个怀疑论者，所以我想知道，小唐纳德·特朗普的投资是否让你们更容易从国家安全角度获得 IPO 的绿灯。

So you IPO'd this week. It's May 2026. This is not the first time that you've tried to or looked towards going to the IPO market. There were headlines going back to 2024 about wanting to try for the IPO market. And then there were headlines last year, especially because of the relationship with G42, about Cerebrus and some of the national security concerns and maybe that was an issue with the IPO. And then but also last September you got in one of your looks like G round, one of the participants in the G round investor was 1789 Capital, which is of course the firm associated with Donald Trump Jr., which is a lot of things. And then the IPO happened. I'm a cynic, so I wonder if the participation if Donald Trump Jr.'s investment in your company made it easier to get the green light from these national security concerns to do an IPO.

我希望这么容易。不，完全没有影响。我们在 2025 年 3 月解决了所有 CFIUS 问题。

I wish it were that easy. No, it had no role at all. We resolved all CFIUS issues in March of 2025.

好的。

Okay.

那是在我们从 1789 拿钱之前。而且，我不会去问。那不是我的风格，也不是我们的行事方式。我们拿钱是因为他们是一家有思想的风险投资公司。

That was before we took money from 1789. Moreover, I wouldn't ask. That's not who I am, and that's not the way we roll. So, we took money because they're a thoughtful venture firm.

好的。

Okay.

我们不相信只有一种政治观点。有很多政治观点。它们都有一些优点，也有一些缺点。所以我们有右倾政治倾向的投资者，也有左倾的。这家公司有一些右倾投资者，我们只看他们帮助建设卓越公司的能力。我们问过，而且我们不会——我们从未要求，也永远不会要求政治渠道或任何类似的东西。

And we don't believe that there's only one point of political view. There are lots of political views. They all have some merit. They'll have some weaknesses, and so we have right-leaning political some investors. We have left-leaning. The fact that this firm had some right-leaning investors, we were looking only at their ability to help us build an extraordinary company. And we have asked, and we will not, we have never asked, nor will we ever ask for political access or anything of the kind.

一天之内成为亿万富翁是什么感觉？我假设这永远不会发生在我身上，所以不妨问问你。

What's it like to become a billionaire in a single day? This is something I assume will never happen to me, so I might as well ask you.

不，老实说，对我来说这没什么大不了的。我之前有一些财富，之后也有一些。我认为这是一种非常艰难的赚钱方式，对吧，作为科技 CEO。你必须热爱工作，热爱员工，每天思考如何让你的团队致富。比我个人财富变化重要得多的是，我们创造了超过 800 个百万富翁。

No, I think the honest truth is it was a big nothing for me. I had some wealth before and have some wealth after. I think this is a very difficult way to make money, right, being a tech CEO. I think what you have to do is you have to love the work. You have to love the people, and you have to think every day about how to make your team rich. And far more important than some change in my wealth was we made more than 800 millionaires.

不错。

Nice.

这是我每分每秒都感到自豪的事。在我上一家公司，我们创造了 100 个百万富翁。而在这家公司，通过 IPO，我们创造了超过 800 个。这让你每天醒来都自我感觉良好。

And that's something I'm proud of every minute of every day. And at my last company we made 100 millionaires. And at this company through our IPO we made more than 800. And that's something that you wake up feeling good about yourself every single day.

那本来是我的最后一个问题，但你的回答提醒了我。你知道，我说你一天之内成了亿万富翁，但这显然是多年工作的结果。如果想想技术硬件，大多数人会联想到很长的交付周期和巨大的研发预算。现在你们是上市公司了，你如何平衡季度财务业绩压力与仍然需要投资资本支出、新的芯片设计方式和现有芯片改进的需求？

That was going to be my last question, but actually you just reminded me in that answer. You know, the idea that getting here I said you became a billionaire in a day, but obviously this was the outcome of years and years and years of work. And if we think about technological hardware, one of the things most people associate with it is really long lead times and really big research and development budgets. Now that you're a public company, how do you sort of balance that quarter-to-quarter financial performance pressure with the idea that you still need to be investing in capex, in new ways of designing chips, new improvements to the existing ones?

首先，我们认为基于晶圆级引擎的创新机会，最好的工作还在前面。第一，我们看到未来几年有非凡创新的机会，能够实现与制造地球上最大芯片同样大甚至更大的飞跃。当你热爱构建硬件时，需要时间本身就是一部分。对吧？我们所做的不能在一天、一个月或一年内完成。这就是你接受的。每个行业都是如此。你接受好的和挑战性的。如果你是一个想要立即投入、快速迭代、快速失败、编写代码、投放市场看是否成功的人，祝你好运。那很好，但不适合我。你知道，在我们的业务中，我们量两次才剪一次，你必须把它融入灵魂，并且喜欢它。你必须喜欢我们的错误代价高昂的事实。你必须喜欢你把生命注入一块硅片，让它做到别人从未让硅片做到的事。如果这适合你，那么这种需要时间和金钱的过程，你也会喜欢。所以我认为，如果一周就能完成，我反而不会那么喜欢。我认为我喜欢共事的人也有同感。他们喜欢做工程师，不是因为这是赚钱的途径，而是因为他们喜欢构建东西。他们喜欢构建困难的东西。我正是因为这个原因喜欢和他们一起工作。

Well, first we think the opportunity for innovation based on our wafer scale engine, the best work is still ahead of us. Number one, we see an opportunity for extraordinary innovation in the years ahead to make leaps every bit as big and often bigger than what we made by building the largest chip on Earth. When you love building hardware, the fact that it takes time is part of the deal. Right? That what we do can't be done in a week or a month or a year. And that's what you sign up for. And that's true in every profession. You sign up for the good and the challenging. And you have to sort of make peace with that if you're a person that wants to dive in and begin iterating right away and fail quickly and code up something and look at it and throw it out in the market and see if it wins, godspeed. That's great and that's not for me. You know, in our business we measure twice before we cut once and you have to put that in your soul and you have to like it. You have to like that mistakes in our business are really expensive. And you have to like the fact that you breathe life into a chunk of silicon. And you get it to do things that nobody else has ever been able to make a chunk of silicon do. And if that's for you, then this process that takes time and money, you love that too. And so I think I would love it less if you could do it in a week. And I think the people that I love to work with, they feel the same way. And they like being engineers not because it's a path to money, they like being engineers because they like building things. And they like building hard things. And I like working with them for exactly that reason.

是的，你提到了把生命注入一块硅片。我父亲是物理学家，他总是喜欢指出碳和硅在元素周期表上相邻。

Yeah, you mentioned breathing life into a chunk of silicon. My dad, who's a physicist, always likes to point out how carbon and silicon are right next to each other on the periodic table.

它们就像是离生命最近的两个元素，而且它们实际上紧挨在一起。也许这其中有些深意。

And they're sort of like here are the two things that we have closest to life and they're literally touching each other. Maybe there's something deep in that.

我觉得你父亲说得很有见地。

I think that's a really thoughtful thing your father said.

谢谢。

Thank you.

我觉得这真的很酷。不过之前没人跟我提过这一点。我们盯着元素周期表看了很久，但我觉得，要制造人工生命，我们需要硅。

And I think that's really cool. And nobody pointed that out to me though. So we've stared at periodic tables for a long time, but I think to the extent we can make artificial life, we need silicon.

是啊，它们正好相邻。

Yeah, and they're right next to each other.

没错。碳是所有其他生命的基础，而人工生命——至少是智能部分——将建立在硅之上。

Right. Carbon is the heart of all other life and artificial life will be founded, at least the intelligent part will be founded on silicon.

硅的正下方是锗。也许下一个……我不知道。好吧，我们接下来留意一下锗。Andrew，非常感谢你来做客 Odd Lots。这场对话非常精彩，正好是我们感兴趣的点。非常感谢你抽出时间。

Right below silicon is germanium. Maybe the next I don't know. Well, let's keep an eye on germanium next. Andrew, thank you so much for coming on Odd Lots. Fascinating conversation. Right in the sweet spot of what we're interested. Really appreciate you taking your time.

嘿，谢谢你们邀请我，我真的很感激。期待很快再见到你们。

Hey, thank you guys for having me and I really appreciate it. Look forward to seeing you again soon.

这真的很有趣。我对这个话题超级感兴趣，而且我觉得，尤其是推理的经济性以及推理能力和速度的市场，还处于起步阶段。你懂我的意思吗？

That was really fun. I'm super interested in this topic and it does feel to me like the economics of inference in particular and the market for inference capacity speed like it's still day one. You know what I'm saying?

是啊。我就是喜欢看那个巨大的晶圆。

Yeah. I just like looking at the giant wafer.

这真的有点像洋葱，不是吗？就像公司用一块巨大的芯片来解决推理问题。

It really does seem like an onion thing, doesn't it? It's like company solves inference with a giant chip.

世界上最大的芯片。

The biggest chip in the world.

这很有意思。我们当然和 Semianalysis 的 Ray Wang 做过那期节目，讨论了内存作为尖端芯片重要组成部分的作用。有趣的是，他们并没有遇到这个瓶颈，而且至少按照他的描述，他们并不追求最小的纳米级芯片，所以这可能也在产能上给了他们一些喘息空间。

It's interesting. We did that episode of course with Ray Wang from Semianalysis and talking about the role like memory as being this really important part of the sort of cutting edge chips and it's interesting to think it's like, okay, well, here is a bottleneck that they don't have and the idea that at least as he described it, they're not fighting to get the smallest nanometer chips and so maybe that gives them a little bit of breathing room on capacity there, too.

是啊。我的意思是，我确实想象得到大芯片有一些缺点，就像 Andrew 列举的优点一样。我还在想另一件事，我知道他论证了速度非常重要的原因，但我也能想象一个世界，也许速度没那么重要，你知道吗？我觉得在某个点上，速度的增量因素与产生额外速度的增量成本相比，会开始变得不那么重要。

Yeah. I mean, I do imagine there are some downsides to having giant chips, just as there are upsides that Andrew laid out. The other thing I was wondering, I know he made the case for the reason speed is very important, but like I can also imagine a world where maybe it's not that important, you know? Like I think at some point the incremental speed factor just starts to become less important when weighed against the incremental cost of generating that extra speed.

我认为这真的取决于你用它来做什么，对吧？比如，如果你说，“你知道吗？我很好奇为什么翼龙其实不是恐龙。你能给我解释一下吗？”那么我就不在乎那点时间。那零点几秒并不重要。

I think it really depends on what you're using it for, right? So, it's like if you're like, 'You know what? I'm really curious why pterodactyls aren't actually dinosaurs. Can you explain it to me?' Then it's like I don't care about that. Like that fraction of a second is not that important.

聊天机器人要花几分钟才能告诉你你错了，Joe。

Minutes for the chatbot to tell you you're wrong, Joe.

你其实没那么在乎。但如果你在做某种智能体式编程之类的事情，那肯定就累积起来了。而且我要说，随着你用得更多，就像其他所有事情一样，期望值会不断攀升。有些任务你 30 秒就能完成，而几年前可能需要 30 分钟，但你在那 30 秒里就会不耐烦，希望它 10 秒完成，这就是那种不断压缩秒数的竞争。我认为这种竞争永远存在。

You just don't really care that much. But if you're doing some sort of agentic coding thing or whatever, then yeah, definitely adds up. And I will say as you use it more, it's just like everything else, the treadmill of expectations. Here's some task that you can do in 30 seconds, which maybe several years ago would have taken you 30 minutes, and you get impatient in that 30 seconds, and you want it in 10 seconds, and that's just like that competition to shave down seconds. I think it's always going to be there.

我的意思是，没有人会对此感到满足。最终总会让人觉得像是在等待。

No one ever gets satisfied with this is my point. It always eventually becomes like it feels like waiting.

但对我来说，这似乎是 AI 估值争论的关键，即我们愿意为一个可能比开源模型稍好的闭源模型支付多少溢价？我们愿意为比另一种算力稍快的算力支付多少溢价？对我来说，这是一个未解的问题。Andrew 对闭源与开源的态度相当坦诚，但我认为在速度问题上，我们也会找到答案。

But to me this feels like the crux of the AI valuation argument, which is like how much of a premium are we going to place on a model that maybe a closed source model that is maybe slightly better than an open source model. How much premium are we going to place on compute that is slightly faster than this other type of compute? Like that to me it's an unanswered question. And Andrew is pretty upfront about closed versus open source, but I think on the speed question too, like we're going to find out.

我们会找到答案的。而且，我认为会发生的一件事——已经有很多关于“token 冲击”的报道，比如公司在 token 上花了多少钱。我的猜测是，在某个时候，会有更多关于“为什么我们要用这个超高端模型，而本可以用别的？”的讨论。就像有很多情况是直接把问题扔给 AI，账单飙升等等。在某个时候，人们会开始问：什么真正需要快速响应？什么真正需要最顶级的闭源模型来服务？公司可能会更擅长根据需求分配不同形式的推理资源。

We're going to find out, and you know, I think one of the things that is going to happen, and there've been all these stories about sort of like token shock, like how much companies are spending on tokens. My guess is one of the things that will happen at some point is there's going to be a lot more discussion about why are we using this ultra premium model when we could have done this? Like there is a lot of just like throw it at the AI, rack up those bills, etc. And at some point there's going to be this like okay, what really needs to be served fast? What really needs to be served on the most premium closed source models? And companies are probably going to get a lot more skilled at allocating from different forms of inference depending on the need.

是的，我完全同意。到那时，我们很可能会看到市场的一些动态在估值方面开始变化。嗯，我们就到这里吧？

Yeah, I think that's exactly it. And at that point like we could well see some of the dynamics in the market start to change in terms of valuation. Um, shall we leave it there?

就到这里吧。

Let's leave it there.

这是 Odd Lots 播客的又一期节目。我是 Tracy Alloway。你可以在 Twitter 上关注我 @TracyAlloway。

This has been another episode of the Odd Lots podcast. I'm Tracy Alloway. You can follow me at Tracy Alloway.

我是 Joe Weisenthal。你可以在 Twitter 上关注我 @TheStalwart。关注我们的制作人 Carmen Rodriguez @CarmenArmandas，Caleb @Dashbot，Caleb Brooks @CalebBrooks，以及 Kevin Lozano @KevinLloydLozano。更多 Odd Lots 内容，请访问 bloomberg.com/oddlots 获取每日新闻通讯和所有节目。你也可以在我们的 Discord 服务器 discord.gg/oddlots 上 24/7 讨论所有这些话题。

And I'm Joe Weisenthal. You can follow me at The Stalwart. Follow our producers Carmen Rodriguez at Carmen Armandas, Caleb at Dashbot, Caleb Brooks at Caleb Brooks, and Kevin Lozano at Kevin Lloyd Lozano. And for more Odd Lots content, go to bloomberg.com/oddlots for the daily newsletter and all of our episodes. And you can chat about all these topics 24/7 in our Discord, discord.gg/oddlots.

如果你喜欢 Odd Lots，如果你喜欢我们谈论巨大晶圆的内容，那么请在你最喜欢的播客平台上给我们留下好评。记住，如果你是 Bloomberg 订阅用户，你可以完全无广告地收听我们所有的节目。你只需要在 Apple Podcasts 上找到 Bloomberg 频道，然后按照那里的说明操作即可。感谢收听。

And if you enjoy Odd Lots, if you like it when we talk about giant wafers, then please leave us a positive review on your favorite podcast platform. And remember, if you are a Bloomberg subscriber, you can listen to all of our episodes absolutely ad-free. All you need to do is find the Bloomberg channel on Apple Podcasts and follow the instructions there. Thanks for listening.