AI 需求超预期：Andrew Feldman 谈芯片、交易和 IPO 后的生活

AI 的需求已经超出了所有人的预期和预测。所以每个人都在追逐——追逐芯片、追逐内存、追逐数据中心。OpenAI 我们在 1 月份宣布了一项巨大的合作，是硅谷历史上最大的交易之一。未来几年我们将为他们提供超过 200 亿美元的硬件。我担心 Nvidia 经常有办法行使市场力量，有办法利用其资产负债表限制竞争。我们有机会让我们的孩子或下一代不仅不会死于癌症，甚至不认识任何死于癌症的人。

The demand for AI has outpaced everybody's expectation, everybody's forecast. And so, everybody's chasing. They're chasing chips, or they're chasing memory, or they're chasing data centers. OpenAI we announced in January a huge partnership, one of the biggest deals done in Silicon Valley history. We'll be doing more than 20 billion dollars of hardware for them over the next several years. I'm concerned that there are ways frequently for Nvidia to exercise market strength. There are ways for Nvidia to use their balance sheet to limit competition. We have a chance for our children or the next generation not only to not die from cancer, but to not know anybody who dies from it.

Andrew Feldman，欢迎来到 sorcery。

Andrew Feldman, welcome to sorcery.

非常感谢你的邀请。

Well, thank you so much for having me.

我们现在在巴黎的 Raise 大会现场。

We are out here in Paris at Raise.

还不错吧？

Pretty okay, huh?

确实不错。我还听说你是这里最受欢迎的人之一。

It's pretty okay. I also heard you're one of the most popular people here.

这在我人生中还是头一回。你这么说真好。

Well, that would be first in my life. So, it's nice of you to say.

不过一个大主题是——我们刚才在镜头外也聊到——硬件很酷，半导体很酷，现在一切都时髦起来了。

A big theme though is, and we were talking about this just off camera, is hardware's cool. Semiconductors are cool. Everything is in fashion now.

我们正时髦，供不应求，很难搞到。这是一个巨大的变化。我认为 AI 的需求已经超出了所有人的预期和预测。所以每个人都在追逐——追逐芯片、追逐内存、追逐数据中心。有机会发明新东西、新架构。这真是一个思想大爆发的时代，非常有趣。

We are in fashion, we're in demand, we're hard to get. It's a big change. I think what's happened is the demand for AI has sort of outpaced everybody's expectation, everybody's forecast. And so, everybody's chasing. They're chasing chips, or they're chasing memory, or they're chasing data centers. There's an opportunity to invent new things, different architectures. It's really a time of a Cambrian explosion of ideas. That's really fun.

你去年也参加了 Raise。我记得，因为我觉得你的展位是最大的，周围还围了栅栏。

So, you were at Raise last year. I remember this because I think you had the biggest booth here. There was a fence around it.

是的。

Yeah.

而且人很多。去年和今年最大的不同是什么？

And a lot of people. What is the biggest difference between last year and this year?

嗯，我觉得去年是他们的第一届。对吧？今年人更多，参与者更多。所以参会者更多，参展公司也更多。我觉得大家的参与规模更大了。昨晚在秘密地点——凡尔赛宫——的晚餐真是太棒了。你知道，我们大部分时间都待在数字世界里。而当我们谈论 AI 时，却去看那些 400 年前不识字的工匠建造的东西，对吧？坐在里面欣赏那种宏伟，真是非凡。这不是 AI，不是 EDA 设计或 CAD，而是那些用纸笔与不识字的石匠和其他工匠沟通的人。他们建造了难以置信的美丽建筑。那真的很有趣。

Well, I think last year was their first year. Right? And I think there are more people here. There are more participants. So, there are more attendees. There are more companies participating. I think people are participating in a bigger way. I think the dinner last night at the secret location called Versailles was sort of awesome. You know, we spend a lot of time in the digital world. And the people where we talk about AI to go look at something that illiterate masons made 400 years ago, right? It is and to sit in it and to enjoy the grandeur is extraordinary. Right? This wasn't AI. This wasn't EDA design or CAD or these are guys with pen and paper communicating with masons and other tradesmen who couldn't read. And they built something unbelievably beautiful. And that was really fun.

那么你今年会上台演讲。主要会讲什么？

And so you're going to be on stage this year. What are you mainly going to talk about?

和 OpenAI 的 Sachin 一起。你知道，我们在 1 月份宣布了一项巨大的合作，是硅谷历史上最大的交易之一。未来几年我们将为他们提供超过 200 亿美元的硬件。我们会聊一些关于部署硬件和快速 AI 的话题，以及快速推理在新兴的推理 AI 领域有多重要。

With Sachin from OpenAI. You know, we announced in January a huge partnership. One of the biggest deals done in Silicon Valley history. We'll be doing more than 20 billion dollars of hardware with them over the next several years. We'll talk a little bit about deploying hardware and fast AI and how important fast inference is in the emerging inference AI landscape.

推理是个热门话题。人人都爱推理。

Inference is a hot topic. Everyone loves inference.

是的，我们通过训练创造 AI，通过推理使用 AI。所以，随着模型和我们创造的 AI 变得有用，每个人都想使用它。推理就是我们使用它的机制。现在我们有聪明的 AI，人们想用它。当他们想用的时候，他们希望它快。这就是我们的用武之地。而且快不是快一点点，而是快 20 倍。所以每个人都在使用他们的 AI，尝试新东西，部署 GPT 或 Claude Code 或这些超级模型之一。这是一个构建、尝试新事物、新工作方式的大爆发。这很有趣。

Yeah, we make AI with training. Right? And we use AI with inference. And so, as the models and as the AI we made becomes useful everybody wants to use it. And inference is the mechanism through which we use it. And so, now we have smart AI. People want to use it. And when they want to use it, they want it to be fast. And that's sort of where we come in. And where the fast is not by a little bit, but by 20x. And so, everybody's using their AI, they're trying new things, they're deploying GPT or Claude Code or one of these super models. And it's an explosion of building, of trying new things, of a new way to work. That's pretty fun.

那么，距离你 IPO 已经快两个月了。有一条帖子我觉得特别酷。你发帖说你肩上扛着一块巨大的芯片。

So, we're nearly 2 months after your IPO. And there was one post that I thought was really cool. You posted that you had a very large chip on your shoulder.

是的。

Yes.

而且那是字面意义上的、物理上的。

And it was quite literal and physical.

我希望既然我们经营一家上市公司，我不会失去我喜欢的自嘲幽默，以及我社交帖子中的那种调调。所以我们把一块巨大的芯片放在我肩上，像这样。我们还做了一个背带。嗯，那很有趣。我认为人们假设作为 CEO 或任何规模的企业家，总是顺风顺水。但事实并非如此。有大量的辛苦工作，有你做出的牺牲，也有你家人做出的牺牲。他们见你的时间更少。而且不是一小段时间，不是像周末或连续两周努力工作，而是持续多年。所以分享一点这些，我觉得会受欢迎。

I hoped that now that we're running a public company, I wouldn't lose the self-deprecating humor that I enjoy, the tone in my social posts. So, we put a giant chip on my shoulder like this. And we built a harness for it. And yeah, that was a fun one. I think that people assume that as a CEO or an entrepreneur of any size, it's always peaches and cream. And it's just not the case. There's an enormous amount of hard work. There's sacrifice that you make and that your family makes. They see you less. And it's not for a little bit. It's not like a weekend or 2 weeks in a row you work hard. It's for years. And so, sharing that a little bit is something I thought would be well received.

那么，IPO 后对你来说最大的变化是什么？

So, what's been the biggest difference for you post-IPO?

想要东西的人数。各种形式的请求爆炸式增长。我们在分类处理上稍微好了一点。我的行政助理、幕僚长，以及每天收到 80 封不同人请求的邮件。那不是工作，那只是各种人想要某种东西——见你、占用你的时间、让你演讲、让你捐款给他们的事业、让你……哇。这有点出乎意料。

The number of people who want something. It has exploded of one form or another. And we got a little better at triaging those. Between my executive assistant, chief of staff, and just if you get 80 emails a day of different people asking you for something. That's not work. That's just this range of people who want something of one form or another to meet you, your time, for you to present, for you to donate to their cause, for you to whoa. That was a little unexpected.

你上一家公司帮助创造了上百个百万富翁。这家公司，我不知道，更多？

Your last company, you helped create a hundred millionaires. This company, you have I don't know, countless more?

大概一千个。

Maybe a thousand.

大概一千个？

Maybe a thousand?

大概一千个。

Maybe a thousand.

也许未来还有成千上万个？对吧？

Maybe future thousands? Right?

IPO 时大约有一千个。

At IPO, it was about a thousand.

包括投资者和员工吗？

Would that be investors and employees or?

还有员工，现任和前任。

And employees. Current and former.

好的。哇。那么作为 CEO，你从中学到了什么？你心里是怎么想的？

Okay. Wow. And so, what lesson do you learn from that as a CEO? Like what is in your mind?

我觉得有几件事。要做好这份工作，你必须热爱创造。我认为赚钱很棒，为你关心的人赚钱更是棒极了。当你有机会为那些押注于你、押注了他们职业生涯大部分的人带来回报时——对吧？你的投资者，为他们带来回报很棒。他们押注于你，但他们是多元化的。

I think a couple things. I think to do the job, you got to love to build. I think making money is really great and making money for people you care about is really really great. And when you get a chance to deliver for people who bet on you, who bet chunks of their career. Right? Your investors, it's great to deliver for them. They bet on you, but they're diversified.

他们押注在你和另外 20 家公司身上。当一个人押上五到七年的职业生涯——职业生涯总共也就 30 年，对吧？——那就是押上了他们职业生涯的六分之一。当你能够为他们带来回报，让他们实现自己想要的财务目标时，那种感觉非常棒，我每天都为此感到自豪。

They bet on you and 20 other companies. When someone bets five or seven years of their career and a career is 30 years, right? They're betting a sixth of their professional career. And when you get to deliver for them, and they get to achieve the financial goals that they wanted to, that's a great feeling and one I'm proud of every day.

这是一种非常无私的立场。我的意思是，这真的很有趣，因为如今的大背景是，我们正处于一个 AI 的超超级周期，随便你怎么称呼。这些新兴公司，无论是做半导体、模型还是编程智能体，很多都飞速崛起，达到了十亿美元估值。有些公司会有流动性事件，比如 OpenAI 和 Anthropic 即将发生的那样。但其他公司则获得了账面增值，看到了那种财富创造效应。你如何在这种环境下保持健康的心态？

It's a very selfless position on that. I mean, it's really interesting because today, in the backdrop, we are in a hyper super cycle, whatever you want to call it, of AI. And so, these companies that are coming up, whether they're in semiconductors or they're in models or coding agents, there's a lot of companies that are rising up really fast and hitting that billion-dollar mark. Some are having liquidity events, like OpenAI and Anthropic to come. But others are getting that paper markup and they're seeing that kind of wealth creation event. How do you maintain a healthy mindset around that?

我觉得你带着这种心态就好。这不是一种改变，对吧？对于我喜欢共事的那种人来说，他们喜欢造东西，喜欢在回报微薄、潮流不在的时候造难的东西。你知道，当硬件不酷的时候，他们仍然在造硬件，因为他们就是喜欢造东西。现在硬件流行了，他们还是喜欢造硬件，他们对此保持平稳心态，他们的热情在于创造本身。我认为在硅谷——这是我比较了解的领域——追逐金钱并不是通往金钱的道路。通往幸福的道路是和你喜欢的同事一起，为正直的人做你喜欢的项目。如果你这样做，钱自然会来。但更重要的是，你会做你喜欢的事情，你会和那些你能从他们身上学到东西、也能教他们一些东西的人一起工作。如果你以这种方式来追求职业生涯，那么当情况很糟时，你能保持平稳；当情况很好时，你也能保持平稳，因为你享受你所做的事情。

I think you bring it. It's not a change, right? For the type of people I love working with, they like building stuff and they like building hard stuff when it paid a little, when it was out of fashion. You know, when hardware was uncool, they were still building hardware because that's what they like to build. And now that it's in fashion, they like to build hardware and they're sort of even-keeled about that, that their passion is the building. And I think in Silicon Valley, which is sort of what I know, that chasing money is not the path to money. The path to happiness is working on projects you like with colleagues that are interesting for people with integrity. And if you do that, the money will come. But even more importantly, you'll work on things you like. And you'll work with people you learn a little something from and you can teach a little something from. And if that's the way you set about pursuing your career, I think when things are really bad, you're on even keel. And when things are really good, you're on even keel because you're enjoying what you're doing.

本期节目由我最喜欢的 Brex 赞助。你花时间做什么，你就会成为什么，而我拒绝把时间花在那些本不该存在的工作上——比如报销单、追发票和手动结账。那些正在构建未来的公司，从 Vercel、OpenAI、Anthropic、Granola 到 Deepgram，都做出了同样的选择：他们都在用 Brex。Brex 是一个智能金融平台，将信用卡、费用管理和银行服务整合到一个技术栈中，内置智能金融功能。AI 智能体自动处理费用，在消费前执行政策，并在几分钟内完成结账。这就是 Sourcegraph 使用 Brex 的原因，这样我就可以把时间花在构建上，而不是琐事上。是时候用 Brex AF 了。了解更多请访问 brex.com/sourcegraph。网址是 b r e u r c e r y。再见。

This episode is brought to you by Brex, my favorite. You become what you spend on, and I refuse to spend my time on work that shouldn't exist. Expense reports, receipt chasing, and manual closes. The companies building what's next from Vercel, OpenAI, Anthropic, Granola, and Deepgram all made the same call. They all run on Brex. Brex is the intelligent finance platform that combines cards, expenses, and banking into a single stack with intelligent finance built in. AI agents that handle expenses automatically, enforce policy before spend happens, and close your books in minutes. That's why Sourcegraph runs on Brex, so I can spend time on building and not busy work. It's time to get Brex AF. Learn more at brex.com/sourcegraph. That's b r e u r c e r y. Bye.

那么，展望未来，我相信这也只是你旅程中的一个里程碑，因为你是一个创造者。在接下来几个月，当我们跨过今年时，你最期待什么？

So, looking forward, I'm sure this is also just a mark in your journey because you're a builder. So, what are you most looking forward to over the next couple months as we sit across this year?

你看，实现 IPO 并不是旅程的终点。它更像是一个平台期，是到达企业成年期。它是达到一个平台，这样你就可以攀登其他平台。我们的机会变得更大了。我们有了更多资源。我们得到了更好的认可。我们可以触达更多人。我们可以用更多的燃料来推进我们的愿景和雄心。所以，这就是我们每天兴奋的原因。你知道，制造更多的芯片，建设更多的数据中心，发明推动行业前进的技术。这就是驱动我们每天起床的动力。

Look, getting to an IPO is not the end of a journey. It's sort of a plateau. It's the arrival at corporate adulthood. It is achieving one plateau so that you can climb others. And our opportunity has gotten bigger. We have more resources. We are better recognized. We can reach more people. And we can sort of prosecute our vision and our ambitions with more fuel. And so, that's what we're excited about every day. You know, building more chips, building more data centers, inventing technology that moves the industry forward. That's what drives us and gets us out of bed every day.

我记得我听过你参与的一期 Harry Sublings 的节目，谈到数据中心及其需求，以及我们实际上并没有在努力满足未来的需求。我们实际上只是远远落后于当前的需求。那么，你能解释一下我们在数据中心建设方面处于什么位置吗？

I think I was listening to a Harry Sublings episode that you did and talking about data centers and the demand for them and how we're actually not trying to meet future demand. We're actually just really behind on where we are. So, could you explain where we are with the data center build out?

我们确实落后了。

We are behind.

在目前的需求上。

on where we are.

没错。事情是这样的：数据中心历来以房地产的速度发展，对吧？有人决定建一栋楼，两年后拿到许可，浇筑混凝土，大楼拔地而起。而两年前没人关心 AI。好吧，两年前我们才刚刚开始。AI 的爆发来得太快了。对算力的需求如此巨大。我们已经超出了对算力和内存的需求。这些东西放在哪里？它们放在这些建筑里。而我们建造的速度不够快。所以，人们在全球范围内追逐数据中心。这就是正在发生的事情。因此，这目前对每个人来说都是一个主要的限制。

Right. So, what happened is data centers had historically moved at sort of the speed of real estate. Right? Somebody would decide to build a building and 2 years later after permits and they poured concrete and the building would go up and 2 years ago nobody cared about AI. All right, 2 years ago we were at the beginning of this. And the AI explosion has happened so quickly. There's so much demand for compute. We've outstripped the demand for compute, for memory. And where do these things go? They go to these buildings. And we haven't made them fast enough. And so, people are chasing data centers around the world. And that's what's happening. And so, it is a major limitation for everybody right now.

我们刚刚请来了 Tony，我们讨论了数据中心的新架构。那么，你是如何与其他公司合作的，无论是通过你的建设，还是创建正确的技术栈和架构？

We just had Tony come on and we were talking about the new architecture of data centers. So, how are you working with other companies, whether it's through your build outs, creating the right stack and architecture?

当然。你知道，数据中心有趣的一点是，它们在 20 年里没有太大变化。里面的很多基础设施也没有变。我的意思是，我们用于备份的发电机大约 20 年没变过。我们用来备份冷却器和 CDU 的电池类型——突然之间，需求激增，人们开始寻求创新，他们使用燃料电池，还有人使用喷气发动机来为这些数据中心发电，涡轮机也在被重新思考。所以，数据中心领域出现了一股创新推动力。对我们来说，当我们制造这种超级大的芯片——比任何其他芯片大 58 倍——它被放在一个服务器里。它被放在一个金属外壳里，大小大约相当于一个宿舍用的小冰箱，我们把几个这样的东西放在一个标准机架里。所以，物理需求与部署竞争对手的产品并没有太大不同。Tony 真正在说的是，这些建筑在一代人的时间里没有得到改进，而现在它们成了制造 AI 的工厂的一部分。人们正试图优化那个工厂的每一个部分。

Sure. So, you know, the interesting thing about data centers is they hadn't changed a lot in 20 years. And a lot of the infrastructure in them hadn't changed. I mean, the generators that we use for backup have been unchanged for about 20 years. The type of batteries we use to back up the chillers and the CDUs and the — and suddenly there's this intense demand and people are looking to innovate and they're using fuel cells and guys are using jet engines like boom to generate power for these data centers and the turbines are being rethought of. And so, there's an innovative push in the data center. Now for us, while we build this super big chip, the chip that's like 58 times larger than any other chip, it goes in a server. It goes in a metal enclosure that's about the size of a fridge for a dorm room little fridge and we put a couple of those in a standard rack. And so, the physical requirements weren't very different from if you're going to deploy a competitor's product. What Tony's really talking about is these buildings were unimproved for a generation and now they're part of a factory that makes AI. And people are trying to optimize every part of that factory.

是啊。现在他们还要把它们放到太空里。

Yeah. And now they're also putting them in space.

是的。嗯，他们在讨论把它们放到太空里。

Yeah. Well, they're talking about putting them in space.

我喜欢——我觉得就像很多技术一样，在实现之前你会谈论很长时间。

I like — I think like a lot of technology you talk about it for a long time before it happens.

你们有计划吗？

Do you have plans?

嗯，你知道，我们非常适合太空，因为太空中最难的问题之一就是让所有这些小芯片相互通信。

Um, you know, we are really good for space because one of the hardest problems in space is getting all these little chips to talk to each other.

而且因为我们是大芯片，我们没有那个问题。我认为短期内不会有太空数据中心的风险。我觉得那至少是 5 年以后的事。

And because we're a big chip, we don't have that problem. I don't think we're in danger in the near term of having a data center in space. I think it's more than 5 years away.

5 年？

5 years?

5 年。

5 years.

好的。

Okay.

而在我们这一行，那是很长的时间，对吧？我是说，3 年前还没人用 AI。所以，我认为在真正拥有太空大型生产数据中心之前，我们在地球上还有很多工作要做。

And that's a long time in our world. Right? I mean, 3 years ago, nobody was using AI. So, I think we have a lot of work building data centers on Earth before we actually have big production data centers in space.

我很想聊聊芯片设计是如何变化的。协同设计已成为为未来 3 年打造这类平台的重要部分，因为你是在为未来 3 年构建。那么，你如何看待这一点，又是如何进行芯片设计的？

I'd love to talk about how chip design has changed. Co-design has become an important part of how do you create that kind of platform for the next 3 years because you're building for the next 3 years. So, how do you think about that and how do you go about chip design?

嗯，从历史上看，你制造芯片，然后在上面运行软件。令人惊讶的是，两者之间并没有紧密的交互，因为芯片和软件之间有一个叫操作系统的层。所以，英特尔和 AMD 制造芯片，人们为操作系统或芯片编写软件。AI 变得如此庞大，速度如此重要，以至于他们开始一起考虑设计。我们可以在软件中做哪些改变来有利于硬件，或者在设计硬件时，我们可以做哪些改变来让软件更容易运行？所以，它们是同时设计的。就像任何事情一样，当你一起设计时，优势是巨大的。这正是目前正在成形的事情。我们与 OpenAI 的关系优势之一是，我们能够确切地看到前沿方向，并有机会将其融入我们的设计。谷歌的优势之一是，他们的 TPU 可以与构建 Gemini 的团队（DeepMind 团队）合作设计。所以他们可以来回反馈各自的选择。这是一件非常强大的事情，令人惊讶的是，在我们这个领域它还相对较新。

Well, historically, you made chips and you ran software on them. And there wasn't, surprisingly, a close interaction because there was a layer called an operating system that lived between the chip and the software. So, Intel and AMD made chips and people wrote software for the operating system or for the chip. AI has gotten so large and speed is so important that they're thinking about the design together. What changes could we make in software that would advantage the hardware, or as we're designing the hardware, what changes could we make that would make the software easier to run? So, they're being designed at the same time. And like anything, when you design things together, the advantages are enormous. This is something that's really taking shape right now. One of the advantages of our relationship with OpenAI is we get to see exactly where the frontier is going and we get a chance to roll that into our designs. One of the advantages Google has is their TPU can be designed in collaboration with the team building Gemini, the team of DeepMind. So they can inform their choices back and forth. That's an enormously powerful thing that is surprisingly relatively new in our space.

你认为这个过程中最大的误解和挑战是什么？

What do you think the biggest misconception with that process is and the challenges are?

我认为误解在于认为这很容易，只需要聚在一个房间里就行。这是一个非常困难的问题。软件人员是一种思维方式，硬件人员是另一种略有不同的思维方式。你为了让软件更容易编写所做的任何事情，都会让硬件更难做。这些都是非常艰难的权衡。把他们聚在一起意味着这种妥协——在这里更难，以便在那里更容易。这意味着某人的进度会受到影响，某人需要增加资源。这些讨论极其困难。

I think the misconception is that it's easy and all you need to do is get in a room. It's a very hard problem. The software guys think one way. The hardware guys think a slightly different way. Anything you do to make it easier to write the software makes it harder to do the hardware. These are really hard trade-offs. Bringing them together means these sorts of compromises where it will be harder here to make it easier here. That means somebody's schedule is going to be impacted. Somebody's got to add resources. Those discussions are enormously difficult.

我很好奇，因为我是外部视角，而你有内部视角。现在到处都有很多大交易。我不知道标题下面到底是什么。所以当 SpaceX 出来说他们现在与谷歌和 Reflection 都有数十亿美元的交易时，这实际上意味着什么？他们在卖什么？

I'm really curious because I come from an outside perspective. You have an inside perspective. There's a lot of big deals that are being thrown around left and right. I don't know what's actually under the headline. So when SpaceX comes out and they say they now have multi-billion dollar deals with Google and also with reflection. Like what does that actually mean? What are they selling them?

很难说。不，内部也很难说。

Tough to tell. No, it's tough to tell inside as well.

真的吗？

Really?

是的，我认为有些宣布的交易并没有实质内容。有些交易以后可能会有实质内容。我认为 X 有可用的算力。你得问为什么他们有可用算力——他们的 Grok 模型用得不多。所以，他们有这些闲置的 GPU，这是个坏主意。于是，他们卖了一整批或者租了一整批给 Anthropic。然后他们抬头一看，说：‘哇，这主意不错。我们有这么多 GPU，我们的模型不成功，但哇，我们可以通过进入一个受限市场来开展一项伟大的业务。很难搞到大量 GPU 并出租它们。’然后他们环顾四周说：‘嗯，我们还能做什么？还能租给谁？’事情就是这样开始的。至于这些交易的具体细节，我不是特别熟悉。

Yeah, I think there have been announced some sorts of deals that didn't have teeth. Deals that could have teeth later. I think X had available capacity. And you have to ask why they had available—their Grok model wasn't used very much. So, they had these GPUs that were sitting around, and that's a bad idea. So, they sold a whole block of them or leased a whole block of them to Anthropic. And they looked up and said, 'Wow, it's a pretty good idea. We had all these GPUs. Our model wasn't a success, but wow, we can have a great business by stepping into what is a constrained market. Very hard to get lots of GPUs and lease these.' And then they looked around and said, 'Well, what else can we do? Who else can we lease to?' And that's how it started. Now, the specifics of those deals I'm not super familiar with.

你是否担心正在发生的循环交易？

Are you concerned at all about the circular deals that are going on?

我担心这些常常是英伟达行使市场力量的方式。它们是英伟达利用其资产负债表限制竞争的方式。如果他们投资一家新云服务商，那家新云服务商就不太可能使用非英伟达芯片。如果他们投资一家模型构建商，就会有不使用别人芯片的压力。这才是更让我担心的。

I'm concerned that they are ways frequently for Nvidia to exercise market strength. They're ways for Nvidia to use their balance sheet to limit competition. If they invest in a neo cloud, that neo cloud is less likely to use non-Nvidia chips. If they invest in a model builder, there's pressure not to use other people's chips. That's what I'm more worried about.

是的。这在代币世界里也在发生，所有那些免费代币都被提供给这些初创公司。

Yeah. This is happening in the token world with all the free tokens that are being offered to these startups.

完全正确。我认为这些是毒贩子。‘来，试一点，就一点点。’我认为初创公司应该做的是接受它，但永远不要依赖。从 AMD 拿一些，也来找我们，看看我们能不能也给你一些，避免依赖。依赖从来不会有好结果。

That's exactly right. I think these are drug pushers. 'Here, try a little bit, just a little bit.' I think what the startup should do is take it and then never be dependent. Take some from AMD and come to us and see if we can get you some as well and avoid dependence. That never ends well.

如果你正在构建 AI 的下一个前沿，你需要了解 MongoDB——这个开发者喜爱的数据库平台，专为你所处的时代而构建。MongoDB 实时存储、搜索和推理你的数据。通过 Voyage AI 的向量搜索和嵌入，全部在同一个系统中。无需单独的管道，无需拼接 10 种不同的工具。这就是为什么 75% 的财富 100 强和领先的 AI 原生初创公司都在 MongoDB 上运行。从你的第一个用户到数十亿向量，构建并扩展。访问 mongodb.com/ai 了解更多。那就是 mongodb.com/ai。再见。

If you're building what's next in AI, you need to know MongoDB, the database platform developers love and built for the age you're running. MongoDB stores, searches, and reasons over your data in real time. With vector search and embeddings from Voyage AI, all in the same system. No separate pipelines, no stitching together 10 different tools. It's why 75% of the Fortune 100 and leading AI native startups run on MongoDB. Build and scale from your first user to billions of vectors. Go to mongodb.com/ai to learn more. That's mongodb.com/ai to learn more. Bye.

AssemblyAI 是一个语音 AI 基础设施层，数百万开发者在其上构建。他们构建了业界最佳的语音转文字、语音智能体和语音理解模型，为 Grammarly、Haystack、Ashby 和 ClickUp 等公司提供关键基础设施。他们的语音转文字模型在准确性和质量上领先业界，而他们的语音理解模型帮助你超越转录，通过从语音数据中挖掘洞察、识别说话者和突出关键信息。你可以今天就在 assemblyai.com/sourcery 开始使用，并获得 50 美元的免费额度来构建语音 AI 产品。那就是 assemblyai.com/sourcery。

AssemblyAI is a voice AI infrastructure layer millions of developers build on. They build the industry's best speech-to-text, voice agent, and speech understanding models that serve as critical infrastructure for companies like Grammarly, Haystack, Ashby, and ClickUp. Their speech-to-text models lead the industry in accuracy and quality, and their speech understanding models help you go beyond transcription by uncovering insights, identifying speakers, and highlighting key information from voice data. You can get started today at assemblyai.com/sourcery and get $50 of free credits to start building voice AI products. That's assemblyai.com/sourcery.

我的意思是，这实际上是一个很好的过渡，引向 Karp 不久前在 CNBC 上说的内容。他基本上是在谈论主权 AI，以及每家公司都应该拥有全栈。

I mean, this is actually a good transition into something that Karp said not too long ago on CNBC. He was pretty much talking about sovereign AI and how every company should own the full stack.

那么，无论是他们的产品数据还是模型，你都不想卖掉，因为模型会重新生成，然后出现一个新的 Figma。对于主权 AI 以及这种转向拥有完整技术栈的趋势，你怎么看？

And so whether it's their data for their products and everything in between and models, too, because you don't want to sell that and then the models will recreate it and then come out a new Figma. How do you feel about sovereign AI in this shift over to owning the full stack?

我认为，不依赖的概念——你的意思是不应该依赖 Nvidia，不应该依赖单一的模型制造商——很少能行得通。作为一个国家或大公司，你希望有选择权。我不确定你是否需要拥有整个技术栈，但你在每一层都需要有选择。另一种思考 Alex 所说的话的方式是：想想你的优势在哪里。如果你有独特的数据，确保不要把它送出去。确保你在技术栈的不同部分有多个选择，但你要从你的数据中获得回报，而不是让它帮助改进别人的模型。

Well, I think that the notion of not being dependent — you mean you shouldn't be dependent on Nvidia. You shouldn't be dependent on one model maker. I think that rarely works out well. What you'd like as a nation or a big company is to have choices. I don't know if you need to own the entire stack, but you want choices at each layer. A different way to think about what Alex said is to consider where your advantage is. If you have unique data, be sure you don't give that away. Be sure you have multiple choices in different parts of the stack, but you get credit for your data, and it doesn't help make somebody else's model better.

展望未来，从宏观角度看，AI 的普及以及我们能够创造、构建和实现的一切，你对随之而来的外部性感到兴奋的是什么？

As we look forward, more on the macro lens, the proliferation of AI and everything we're able to create and build and do, what are you excited about on the externalities that come with all of this?

我认为我们有机会让我们的孩子或下一代不仅不会死于癌症，甚至不认识任何死于癌症的人。我认为这是一个在 25 年内可以实现的目标。那该多好啊？当我们思考 AI 能做什么时，写出更好的代码很酷，而且市场巨大。但它能为人类带来的更大好处是消除成年人的头号杀手。当我思考我们做这一切是为了什么时，就是这样的结果。胰腺癌最近取得了巨大突破。现在突破的机会前所未有，而 AI 是攻克重大人类杀手的非凡工具。如果你消除了癌症和车祸，你就消除了每年大量的死亡。自动驾驶——人类是糟糕的司机，非常糟糕。不仅仅是在 16 岁或 18 岁，或者当我们不注意时，或者我们 80 多岁的父母。我们就是不注意，我们看手机、跟妻子说话、担心工作。这甚至还没算上喝酒或做其他明显坏事的情况。我们就是开不好车。机器现在可以比我们开得更好。15 到 40 岁人群的头号杀手是车祸。通过自动驾驶消除它，再消除像癌症这样的主要杀手，你会说，‘哇，这 20 年的技术发展真不错。’

I think we have a chance for our children or the next generation not only to not die from cancer but to not know anybody who died from cancer. I think that is a real achievable goal in 25 years. Wouldn't that be something? When we think about what AI can do, writing better code is cool and there's a huge market for that. But what it can do to better humanity is rid us of the number one killer of adults. When I think of what we're all doing this for, it's an outcome like that. Pancreatic cancer had a huge breakthrough recently. The opportunity for breakthroughs right now has never been better, and AI is an extraordinary tool in pursuit of knocking down major human killers. If you took out cancer and automobile accidents, you're taking out huge numbers of deaths a year. Self-driving — humans are terrible drivers. Horrible. It's not just when we're 16 or 18 or when we're not paying attention or our parents in their 80s. We don't pay attention. We're on the phone, talking to our wife, worried about work. That's even before people drink or do things that are obviously bad. We're just not good drivers. Machines can drive better than we can today. The number one killer of people aged 15 to 40 is car accidents. Take that out through self-driving, take out a major killer like cancer, and you go, 'Whoa, that's a pretty good 20-year run of technology.'

你是肽类爱好者吗？

Are you a peptide fan?

我是肽类爱好者。

I'm a peptide fan.

是吗？

You are?

是的。

Yes.

我认为不是特指某一种肽，而是我们有机会增进对身体生物学的了解，如何实现性能和长寿——我们才刚刚开始。我们将取得重大进展。无论是通过肽还是其他东西，下一个 GLP-1 抑制剂或其他什么，我们都会取得巨大进步。

I think not in the specifics of there is a peptide, but rather that our opportunity to advance our knowledge about the biology of our bodies, how to achieve performance and longevity — we're just beginning. We're going to make some big strides. Whether it's with peptides or something else, the next GLP-1 inhibitor or whatever, we're going to make giant strides.

是的。看到所有新药发现和新药发现公司真的很酷。Brian Armstrong 刚成立了一家名为 New Limit 的公司，就是其中之一。我认为他们的目标是根除所有疾病。

Yeah. It's been really cool to see all the new drug discoveries and new drug discovery companies. Brian Armstrong just came out with this company called New Limit, that's one of them. I think their goal is to eradicate all diseases.

这不对。我的意思是，即使这有点狂妄，但在十年前连这种狂妄都是不可想象的。我们现在处于一个领域，哇，这有点疯狂的大，但并非荒谬。这有多酷？

That's not right. I mean even if that's a little hubris, it wasn't even hubris that was thinkable a decade ago. We're now in a realm where wow, that's sort of crazy big, but not insane. How cool is that?

是的。很多人喜欢谈论 AI 的末日论，但我认为我们正在进入一个讨论实际成果的新阶段。所以正如你所说，无论是自动驾驶还是药物发现，这都非常好。

Yeah. I mean a lot of people like to talk about the doomerism of AI, but I think we're entering a new mix of talking about the actual outcomes with it. So to your point, whether it's with AVs or drug discovery, it's really good.

我认为末日论者的问题在于他们只看账本的一面。要清醒地看待这个问题，你必须看两面。我们会消耗大量能源。AI 确实有一些真正的风险。但在账本的另一面，它有一些可以真正改变的事情。有些事情它可以在教育方面做到。两千年来，我们一直知道教育孩子的正确方法，但我们从未实践过。从未。我们知道教育亚历山大大帝的正确方法是请一位导师——最聪明的导师亚里士多德——并且因材施教，考虑他们的学习方式。我们从未那样做。我们把孩子扔进教室，按某种平均水平教学。每个孩子没有根据不同的学习方式得到不同的教学。现在我们可以做到了。我们可以为你找到适合你的导师。更重要的是，导师可以在后台运行，说：‘看，3% 的学生犯这种错误，教他们克服这个弱点的最佳方法是这种方法。’这有多酷？我们搞砸了两千年。现在我们可以把它带给每个孩子。把它放在账本积极的一面。清醒地看待消极和积极两面，看看我们是否在为社会做正确的事。

I think the problem with the doomers is they're only looking at one side of the ledger. To look at this with clear eyes, you've got to look at both sides. We're going to use a lot of power. AI has some real risks. On the other side of the ledger, here are some things it can do really differently. Here are some things it can do that take education. We've known for 2,000 years the right way to educate children, and we never do it. Ever. We knew the right way to educate Alexander the Great was to have a tutor — the smartest tutor, Aristotle — and you teach each child differently, thinking about their way of learning. We never do that. We throw them in a classroom, teach to some middle level. Each child does not get different teaching for their different way of learning. Now we can do that. We can get you a tutor that's right for you. What's more, the tutor could be running in the background saying, 'Look, 3% of students make this type of error, and the best way to teach them to overcome this weakness is with this approach.' How cool is that? We've been screwing this up for 2,000 years. Now we can bring it to every child. Put it on the positive side of the ledger. Look with clear eyes at both the negative and the positives and see if we're doing right by society.

那么，我们的赞助商之一是 Brex，我提到这个是因为我们关于绩效的问题都是关于更聪明地花钱和更快地行动。但就个人财务方面的绩效而言，我问这个问题是从个人角度出发。我确实认为个人的绩效取决于你身边是谁，或者谁激励你、指导你。你在创办公司方面成绩斐然，取得了巨大成功。所以我很好奇对你来说那些人是谁。

So, one of our sponsors is Brex and I bring this up because our questions around performance are all about spending smarter and moving faster. But performance in terms of your finances, I ask this question in terms of your personal side. I do believe that performance for individuals is kind of who you surround yourself with or who you're inspired by or mentored by. You've had a great run at building companies and have had great success. So, I'm really curious who those people are for you.

有几位导师。有一位风险投资人叫 Pierre Lamond，他现在 90 多岁了。他在 84 岁高龄时投资了我们的 Cerebras。

There were a couple mentors. There was a venture capitalist named Pierre Lamond. He's now in his 90s. He invested in us at Cerebras when he was at the ripe age of 84.

天哪。

Oh my gosh.

太棒了。非常感谢你，Andrew。很高兴能有这次交流。我知道我们这次聊了很多内容，覆盖了很多话题。说实话，我没想到我们会聊到肽类。

It's amazing. Well, thank you so much, Andrew. It's a pleasure to have time. I know we covered a lot of ground with this conversation. We covered a lot of ground. So many topics. I didn't think we were going to get to peptides, I'm going to be honest.

我就是忍不住想问。

I just had to ask.

感谢你邀请我上节目，我真的很感激。

Well, thank you for having me on your show. I really appreciate it.

谢谢。非常感谢整个 Raise 团队举办了一场精彩的活动，也感谢 Brex、MongoDB 和 Assembly AI 让这次行程和系列节目成为可能。如果你喜欢这次对话，你一定会喜欢 Raise 系列的其他内容，嘉宾包括 Blackrock 的 Tony Kim、Cognition 的 Scott Wu、Cerebras 的 Andrew Feldman、SambaNova 的 Rodrigo Liang、Momentum 的 Michael Hurlston、MongoDB 的 CJ Desai 等等。还有我们在秘密地点录制的“热辣观点”，你可以在 X、YouTube 和 Instagram 上找到。订阅 Sorcery 的 YouTube 频道，获取更多与 AI 塑造者对话的内容，并加入免费新闻通讯。你也可以在 sourcery.vc 上付费订阅，每周获取关于 AI、机器人、企业软件、消费半导体等领域的洞察。我说 AI 了吗？又是 AI，还有接下来的一切，比如融资公告和科技界的大事。谢谢。再见。

Thank you. Huge thank you to the entire Raise team for an incredible event and thank you to Brex, MongoDB, and Assembly AI for making this trip and series possible. If you enjoyed this conversation, you're going to love the rest of the Raise series with Tony Kim from Blackrock, Scott Wu from Cognition, Andrew Feldman from Cerebras, Rodrigo Liang from SambaNova, Michael Hurlston from Momentum, CJ Desai from MongoDB, and many, many more. Like our hot takes that we did at a secret location that you can find on X, YouTube, and Instagram. Subscribe to Sorcery on YouTube for more conversations with the people shaping AI and join the free newsletter. You can also do paid at sourcery.vc for weekly insights on AI robotics enterprise software consumer semiconductors. Did I say AI? AI again and everything that's coming next like funding announcements and all big things in tech. Thank you. Bye.