OpenAI与英伟达紧密合作，Distyl AI估值达18亿美元，气候周议题转向 | 2025年9月23日

欢迎各位收看TITV资讯。我是阿卡什·佩斯里查。今天是9月23日星期二，我们从纽约市为您带来报道，这里正在举行气候周活动。今天我们将试图了解本周活动的最新进展。我们将邀请两位气候议题的核心人物，帮助我们理解当前时刻与往年有何不同。

Welcome everyone to the Informations TITV. My name is Akash Pesricha. It is Tuesday, September 23, and we are coming to you from New York City where climate week is afoot. We're gonna try to understand today how this week's events are shaping up. We're going to bring on two people who have been at the center of the climate story to help us figure out how this moment is different from years past.

我们还将连线人工智能与金融记者，讨论信贷领域最值得关注的数据中心领军企业。这些巨头正处在人工智能与信贷业务的交叉点。随后我们将采访一家Coastal Ventures投资的公司的CEO——我们率先报道该公司刚以18亿美元估值完成融资。但在进入这些话题前，我想先从本周最重磅的新闻开始。昨天英伟达宣布将向OpenAI投资高达1000亿美元，帮助这家ChatGPT制造商建设自有数据中心。

We're also bringing on our AI and finance reporter to talk about the data center leaders to watch in the land of lending. These are the biggest names sitting at the intersection of AI and credit. And we're then gonna bring on the CEO of co of a Coastal Ventures backed company that we are first to report just raised money at a valuation of $1,800,000,000. But before we get to any of that, I wanna start with one of the biggest stories this week. Yesterday, NVIDIA said it would be investing up to a $100,000,000,000 into OpenAI to help the chat GPT maker build its own data centers.

这立即在网络上引发大量 meme，调侃芯片公司、云服务商和AI企业之间资金流动的闭环现象。现在我要请出我们的英伟达专题小组——观众朋友们可能已经熟悉这几位常驻嘉宾。欢迎《信息报》云计算记者安妮莎·加德西、GPU托管公司Hydra Host首席执行官亚伦·金，以及Seaport Research Partners高级分析师杰伊·戈德堡。欢迎三位，最近怎么样？

It prompted a flurry of memes online about just how circular the flow of cash between chip companies and cloud companies and AI companies is all getting. I wanna bring on our NVIDIA panel that you might be getting familiar now to seeing on our show. Joining me now is Anissa Gardesi, our cloud and compute reporter at the information, Aaron Ginn, who is the CEO of a GPU hosting company called Hydra, Host, and Jay Goldberg, a senior analyst at Seaport Research Partners. Welcome to the three of you. How are doing?

挺好的。

Doing well.

嗯，还不错。

Doing Yeah, doing well.

好的，这个开场我接受了。那么安妮莎，我想先从你开始。昨天宣布了这项重大交易，请带我们梳理关键要点——你认为最需要关注什么？

All right, I'll take it. Okay. So Anisa, I want to start with you. We had this big deal announced yesterday. Walk us through the brass tacks of what we need to know, what your take was on that.

是的。昨天上午，萨姆·奥特曼与黄仁勋宣布了一项极其前所未有的协议，主要因其规模惊人——英伟达将向OpenAI投资高达1000亿美元，支持其建设10吉瓦的AI基础设施。我知道这些数字都很庞大，但关键点在于：英伟达实质是在资助OpenAI的数据中心建设。特别值得注意的是，这10吉瓦的数据中心容量全部是新增的。

Yes. Yeah. So yesterday morning, Sam Altman and Jensen Huang announced a very unprecedented agreement mainly because of the size in which NVIDIA would invest up to a $100,000,000,000 into OpenAI as OpenAI works on building out 10 gigawatts of AI infrastructure. So I know there's like lots of big numbers in that, but I think what people need to know is that NVIDIA is really helping finance OpenAI's data center build out. And one important takeaway is that this 10 gigawatts of data center capacity is all net new.

OpenAI宣布的其他事项仍在推进中。但现在又有一个新的10吉瓦项目，规模超过了当前主要云服务提供商。因此英伟达正与OpenAI紧密结盟，而获得AMD芯片并自研处理器的OpenAI也真正与英伟达站在了同一战线。此前人们担忧它可能不再仅依赖GPU。

So everything else that OpenAI has announced, you know, is still in the works. But now there's another new 10 gigawatt project, which is bigger than a major cloud provider is today. So NVIDIA is very closely aligning itself with OpenAI. And OpenAI, which got some chips from AMD and is working on its own chip, is really aligning itself with with NVIDIA. And I think there were concerns about it potentially splitting from only using GPUs.

这里有很多值得深挖的内容。若各位尚未阅读，今早Anisa的AI议程报告绝对值得一读——内容非常密集。我甚至需要做笔记才能消化所有信息。让我们逐条分析：你提到的关键点是OpenAI可能不再租用英伟达GPU。

Well, so there's a lot to dig in there. And I should say that if you haven't yet read it, Anisa's AI agenda this morning is definitely one to read because it was dense. There was a lot of information in it, I actually, I had to take notes on it all because there was a lot there. So let's go through it. So, you know, one key point you made is that OpenAI can move away from renting NVIDIA GPUs.

你提到的另一点是，这就像是他们自主宣布的'星际之门'计划。Anissa请简单说明下，然后Erin我想听听你的见解。

And, you know, the other thing you mentioned is that, you know, this is kind of like their own announcement of their own Stargate. Just tell us a little bit about that, Anissa, and then, Erin, I'm coming to you to get your thoughts on this.

好的。需要说明的是，OpenAI虽是GPU最大用户之一，但目前通过云服务商租用这些芯片——他们既不拥有芯片，也不拥有数据中心，所有设备都来自与微软、甲骨文等公司的租赁协议。但OpenAI希望拥有硬件，对这类公司而言直接采购GPU存在难度。

Sure. Just in case people don't remember, OpenAI is one of the biggest users of GPUs, but today it rents them from cloud providers. It doesn't own the chips, it doesn't have to buy them, and it doesn't own any of the data centers where the chips are located. These are all rental contracts with companies like Microsoft and Oracle. But OpenAI wants to to own the hardware, but it's it's hard for a company like OpenAI to to buy GPUs.

这些芯片极其昂贵，且需要良好信用才能获得优惠贷款条款。英伟达此次投资将真正助力OpenAI实现其目标——自主采购GPU。

They're they're very expensive and, you know, you sort of have to be credit worthy to get debt at good terms. And so this injection of investment from NVIDIA is really gonna help OpenAI actually be able to do what it wants to do, which is buy GPUs.

明白了。Aaron，你对此有何看法？

Right. Aaron, what do you think of all this?

我认为这仍面临许多与基础设施建设相同的难题——不仅是OpenAI，对黄仁勋和AI产业持续发展而言，真正的阻碍在于资金。摩根大通、伯恩斯坦等机构预测，到本年代末每年需万亿美元投入才能维持建设规模。而公有云厂商为保持市场份额已耗尽三倍现金流，显然必须转向私募市场和替代方案。

I think that it was still running to do a lot of the same problems with this, like, build out, which is that they're like, in build out, not just within particularly OpenAI, but the real impediment to Jensen and the continued growth of AI itself is just debt. It's like the the capital expenditures you actually need to do what is projected as an example. Everyone's like, sort of the end this decade, JP Morgan, Bernstein, etcetera, saying, like, it's gonna be a trillion dollars a year to, like, build, you know, annual TAM to continue to build out this stuff. And based on what's happening on public cloud side, you're looking at them more or less spending all of their cash times three just to maintain their minority market share. So clearly, this has to go to private markets, has to go to alternatives.

但问题在于，目前许多此类债务的结构方式使其高度集中于公司股权。这就好比说，当我申请抵押贷款时——就像阿卡什准备申请抵押贷款那样——现在他们支持的是你的会计科目表。但之后他们会将其出售给那些押注于抵押贷款市场或房产市场的其他人。当前债务信贷领域缺失的正是这一点：没有一个二级承接方来为初级贷款人创造流动性。

But the problem is that the currently, the way a lot of this debt is structured is that it's highly isolated to the equity of the company. And so that's like saying when I go get a mortgage so, like, you know, Akash was gonna get a mortgage. Right now, they're backing the chart of accounts of of you. But what happens afterwards is they sell that to somebody else who bets on the on the on the market of mortgages, on the markets of homes. That's what's currently missing in the debt credit space is that it it there's no sort of secondary off taker to which you can create liquidity for that primary lender.

没错。因此英伟达不得不扮演这个角色，本质上为项目提供流动性。我认为这才是当前现象的本质，而非人们嘲笑甲骨文时提及的那种循环逻辑。实际上，对于一家季度自由现金流高达1000亿美元的公司来说，这些根本不算什么——特别是考虑到数年累积的规模。真正应该关注的是：为何信贷市场陷入僵局？为何微软、亚马逊、NetVase、Fourier等公司都在尝试创设各类新型金融工具？

Right. And so then Nvidia has to be in this position to basically provide a liquidity for projects. And and I think that's more what's going on here than the kind of circular stuff that people make fun of about the Oracle, they make fun of. But in reality, it's like in the context of a company that's having, you know, a $100,000,000,000 of essentially free cash flow every quarter, like, this is this is kind of like nothing, like, terms of that, like, how the the and especially over the course of several years, really what you should be looking at is like, why are credit markets locked up? Why are Microsoft, Amazon, NetVase, Fourier, etcetera, trying to create these like different types of instruments?

是的。问题在于债务端和投资端的市场尚未跟上需求的发展。所以如果黄仁勋要实现销售，就必须解决这个资本难题。

Right. Is that that the world the world on the debt side, on the investment side, has not caught up to like where the demand is. And so if Jensen's going to make sales, he has to resolve that capital problem.

没错。杰，我想听听你的见解，但这次邀请你主要是因为你可能是最适合对这类循环逻辑提出质疑的人。跟我聊聊你对此事的反应吧。

Right. Jay, I want to get your thoughts here, but really, was excited to get you on because I thought you'd be the perfect person to cast a little bit of doubt maybe on the circularity of all this. Talk to me about what your reaction was here.

我的观点很简单：这看起来非常像供应商融资。英伟达这是在购买需求。亚伦提出了很好的观点——资本市场目前不愿提供替代性结构方案。部分原因在于这是个新事物导致的系统缺陷，融资渠道仍在适应阶段。

So it was just that. This feels very much like vendor financing. This is Nvidia buying demand. And I think Aaron makes a great point, is that the capital markets are reluctant here to come up with some alternative structures. And part of that is just a bug in that this is new and the financing sources are still getting up to speed.

我花了很多时间与私募信贷人士交流，试图理解这个市场、数据中心本质以及AI概念。系统确实存在这种缺陷和阻滞，但另一方面，一些相当成熟的融资渠道也开始提出质疑：我们需要多少GPU？整个体系如何运作？资金流向何处？目前仍存在大量待解问题。

I spent a lot of time with private credit people trying to understand what this market is, and what a data center is, and what AI is. So there is that bug and lockup in the system, but the other side of it too is there are some pretty sophisticated finance sources out there who are starting to push back a little bit and question how many GPUs we need, how's this all going to work, where's the money going, and there's a lot of questions still out there.

一个原因是

A reason for

你有一个

you to have a

筹集那笔资金的艰难时期。

hard time raising that money.

杰伊，听我说。我们今早在新闻编辑室讨论过这个问题。在这个OpenAI与NVIDIA深度绑定、NVIDIA又与Oracle紧密合作、Oracle反过来又与OpenAI深度交织的世界里，难道你不觉得这个故事中任何环节出错都会引发连锁反应吗？

Jay, hear me out here. We were talking about it in the newsroom this morning. Mean, in a world where OpenAI is so closely embedded with NVIDIA, who is so closely embedded with Oracle, who is so closely embedded with OpenAI. I mean, doesn't it just feel like anything that goes wrong in this story will really just cause a cascade of reactions?

我认为这正是我们一直以来所处的情境。回溯起点，OpenAI创立后，微软看到他们的进展便开始大力投入。其他企业见状也纷纷跟进，由此引发了这种正向循环——或者说恶性循环？所有人都在为不被落下而持续加注。这其中包含着对AGI（通用人工智能）的高度心理预期和信念，或者说某种未定义的终极目标。而这种心理驱动随时可能逆转。

I think that's very much where we've been all along, is if you go back to the start of this, OpenAI started, Microsoft saw what they were doing, they started spending heavily against that. Everyone else saw what Microsoft was spending, they felt that they had to do it too. And it sparked this positive cycle, this vicious cycle, I don't know, where everybody was spending to keep up with everybody else. And there's a very heavy degree of psychology and just belief in AGI, I guess, or something, some undefined goal out there. And at some point that can reverse.

虽然我不确定转折点何时到来——可能不会很快——但当前态势确实主要由心理因素驱动。

And I don't know when that's going to happen, probably not soon, but it is very psychologically driven.

阿妮萨，我想回到你今早的报道内容。我的两个疑问是：第一，这个数据中心合作协议（或者说投资意向书，随你怎么称呼）涉及1000亿美元的规模，是否会蚕食OpenAI与其他公司的现有数据中心合作？第二，这是否会影响——正如我们报道过的——OpenAI与谷歌在TPU芯片方面的合作？他们也在与博通合作。

Anisa, I want to come back to what you wrote this morning and some of the reporting you did. Two of the questions I've had are number one, could this data center deal or this LOI investment, intent to invest, whatever you want to call it, could this perspective, 100,000,000,000, could that cannibalize other data center deals that OpenAI has with other companies? That's the first question I have. And the second question is, could it also mean something for, as we've reported, OpenAI has also been doing some work with Google and its chips, its TPUs, right? It's been working with Broadcom.

有消息称他们正在自研芯片。请就这两个问题谈谈你掌握的信息，以及你是否认为当前事件会影响这些合作。

It's been reported to build its own chips. Talk about what you're hearing along those two questions and if you could see this story impacting those.

当然。目前OpenAI在地面上有许多数据中心项目正在筹备中。而与英伟达合作的这个项目可能达到10吉瓦的规模。我们必须意识到，市场上并没有那么多现成的吉瓦级电力供企业使用。这确实是个极具挑战性的基础设施问题。

Sure. So on the ground, OpenAI has many, many projects in the air when it comes to data centers. And this one with NVIDIA potentially is 10 gigawatts. I think we all have to remember that there aren't that many gigawatts just laying around for companies to use. This is like a really challenging infrastructure problem.

要知道，这类数据中心的供应本就稀缺，无论英伟达是否承诺提供芯片。由于OpenAI同时推进着多个项目，所有这些企业都将寻找数据中心来承载芯片——无论是CoreWeave租给OpenAI，还是Oracle为那4.5吉瓦项目向OpenAI出租，亦或是OpenAI与英伟达共建数据中心。我认为供应链上的建设公司最终会发现，虽然他们在与甲骨文、微软和OpenAI洽谈，但终端客户其实都是OpenAI。此外，关于OpenAI试图通过自研芯片、谷歌芯片或AMD产品来摆脱对英伟达依赖的问题...

And the, you know, there's there's not a lot of supply of of these types of data centers, whether or not NVIDIA promises you the chips or not. And so because OpenAI has all of these balls in the air, all of those companies are gonna be looking for data centers to to host the chips, whether it's CoreWeave renting to OpenAI, Oracle renting to OpenAI for that 4.5 gigawatts, or OpenAI building its own data centers, with NVIDIA. And so I do think that, you know, at the in the supply chain, the actual companies that build these data centers might realize that they're talking to Oracle and Microsoft and OpenAI, but the end customer is all OpenAI. So I do think there should be a little bit of that there. And then to your last question about what does this say about OpenAI's efforts to diversify away from NVIDIA, potentially use its own chip, use Google's chip, you know, potentially use AMD.

我不认为OpenAI会因此停下其他技术路线的探索。但从Sam Altman昨天的访谈可以看出，他们在这个巨型项目上基本押注了英伟达。现阶段OpenAI自研芯片超越对英伟达承诺的可能性尚不明确——毕竟昨天官宣的是与英伟达的合作，所以焦点自然集中在英伟达。

I don't think OpenAI is gonna stop in its tracks and, and not pursue those other efforts. But but one of my takeaways from from the interview that Sam Altman did yesterday was that, you know, they're kind of all in on NVIDIA for this this massive project. And so I think, you know, I I don't really know what the chances are that it OpenAI's own chip would surpass the commitment to NVIDIA at this point. But, you know, yesterday was an announcement with NVIDIA, so I think that's why it was all about NVIDIA as well.

是的。Jay，你认为这次合作会让他们暂缓自研芯片计划吗？哪怕只是稍微？

Yeah. Jay, do you think they do you think they sort of put any brakes on on their own chip efforts because of this at all?

我认为所有企业都在多线并行。没人愿意完全受制于英伟达，但定制芯片项目风险很高，分散投资才是明智之举。

I think everybody is pursuing every strategy that they can. Yeah. Nobody nobody wants to be totally dependent on NVIDIA, but these custom silicon projects are risky, and so it's good good to hedge your bets.

Aaron，你怎么看？

Aaron, what do you think?

我基本同意Jay的观点，这本质上是种制衡英伟达的策略。就这笔交易而言，资本规模其实很小。正如其他嘉宾提到的，关键问题是：Stargate究竟是什么？

Yeah. I think that they most of it, as Jay was saying, I think is the actual strategy, which is, like, creating leverage over NVIDIA. And because the number one concern in terms of this deal like, the capital is, like, again, very small. And I agree with what what the rest of the panel was saying around, like, okay. What is Stargate?

这就是现在的星际之门吗？因为另一个星际之门是什么？那边那些其他的门又发生了什么？对吧？而且没人完全明白，因为每个人出现、宣布一下，然后就没有后续了。

Is this Stargate now? Because what is the other Stargate? What happened to those other ones over there? Right? And nobody fully understands because everyone shows up, announces it, then there's no follow on.

而且我认为，在某种程度上，开放这方面，我可以做到，因为目前它与Jensen一起是决定成败的关键。所以他们可以继续这样做，而没人真正质疑。但就定制硅芯片或AMD而言，上周早些时候或晚些时候有个AMD的公告，我认为是为了创造选择性。当NVIDIA在技术曲线上遥遥领先，同时赚取如此巨额利润时，作为最大买家，你甚至有动力把这当作游戏资金，而这游戏资金可能是几十亿美元，它有助于你在这些谈判中。但正如你之前所说，很多这类事情只是以意向书的形式构建的。

And and I think that that's where, like, in some respects, opening, I can do it because it's it's currently the kingmaker with with Jensen. So they can just continue to, like, kinda do this and, like, nobody really questions. But, like, in terms of, I think, the like, what the custom silicon does or AMD with the there was an AMD announcement, I think, earlier or late last week, is to create op you know, optionality. And when when NVIDIA is so far ahead in terms of the technical curve and as well as it making as much money as it is, that they as being a largest buyer, you're incentivized just even as just play play money, and that play money can be a couple billion dollars, is it helps you in these negotiations. But but as the as you were saying earlier, like, you know, a lot of this stuff is just structured as a letter of intent.

所以谁知道它最终会变成什么？如果他们签署的结构有点像，你知道，就像一份可以用Microsoft Word编辑并签署的文件，我还能以此获得贷款，那么你可能应该假设后续步骤仍然悬而未决。但我认为真正的关键在于，市场上现在存在NVIDIA面临的限制，这些限制与需求无关。是资本，是电力。他们认为这些会减缓他们现在的增长。

And so who knows what it actually turns into? So so if the construct itself that what they signed was a little bit like like just, you know, Microsoft Worded a document that you can sign, and I can get a loan against, yeah, you should probably assume that the in terms of the the follow on steps that it's still kind of up in the air. But but I think I think what what the real takeaway is that there are constraints now in the market that NVIDIA is looking at that is not demand related. It's capital, it's power. It's like, that's what they think is gonna slow their growth now.

因此，他们正在寻找不同的方法来增强生态系统。而我必须

And so they're trying to find different ways of enhancing the ecosystem. And I have to

告诉你，这完美引出了我们节目的下一位嘉宾。我们有AI财经记者Myles Krupa来讨论债务问题，之后还有一个气候周的小组讨论。所以这个故事远不止是一份意向书那么简单，虽然，好吧，1000亿美元，我想这值得报道。

tell you, this perfectly tees up the next guest we have on the show. We've got Myles Krupa, our AI finance reporter coming on to talk about debt, and then we've got a Climate Week panel coming on. So this story is so much bigger than just an LOI, which fine, 100,000,000,000. I guess it's news worthy.

我们都想要那样。我也想要一份1000亿美元的意向书。

We would all want that. I'd love an LOI of $100,000,000,000 All

好的。那么，感谢你们三位再次来到节目。这是一周内的第二次了，我相信还会有更多新闻需要报道。这就是我们的NVIDIA全明星小组，有Anissa、Aaron和Jay。感谢大家的参与。

right. Well, look, I want to thank the three of you for coming on. This is the second time in a week, and I trust that there's going to be more news to cover. That is our all star NVIDIA panel with Anissa, with Aaron, and with Jay. I want to thank you all for being here.

好的，谢谢。当AI公司需要巨额资金时，股权并非总是解决之道。实际上，在建设数据中心方面，债务融资是常见的资金来源。今天，《The Information》发布了一篇关于为AI数据中心提供资金支持的贷款方报道，这些机构值得你高度关注。

Okay. Thank you. Well, when you need as much money as AI companies do, equity isn't always the answer. In fact, as it relates to building data centers, debt is a popular source of funds. And so today, The Information published a story about the lenders backing AI data centers that should absolutely be on your radar.

我想请我们的AI与金融记者迈尔斯·克鲁帕来聊聊这些机构的名字，以及为何它们在他的重点关注名单上名列前茅。迈尔斯，欢迎回到节目，很高兴你能来。

I want to bring on Miles Krupa, our AI and finance reporter, to talk about those names and why they are at the top of his watch list. Myles, welcome back to the show. It's great to have you.

谢谢，阿卡什。

Thanks, Akash.

好的。在进入名单讨论之前，希望你能先解释下，在当前这个报道背景下，债务融资与AI领域是如何产生交集的？

Okay. So before we get into the list, I'm hoping you can just explain to us how is it that debt and AI intersect right now in this story?

是的。这是一个巨大的交叉领域，我认为这是当前金融界最引人入胜的角落之一。为人工智能建设所需的数据中心规模正变得越来越大。过去常见的数据中心功率可能是50兆瓦或100兆瓦，而现在我们看到企业正在建设超过1吉瓦的设施。

Yeah. It's it's a huge intersection, and I think it's kind of the most interesting corner of finance right now. Basically, data centers that are needed for the AI build out are just getting bigger and bigger. It used to be common to have data centers with maybe 50 megawatts of power or 100 megawatts of power. And now we're seeing companies build a gigawatt plus of power.

随着这些数据中心规模持续扩大，建设成本也在不断攀升。因此，亚马逊、谷歌、微软等科技巨头正越来越多地将这部分业务外包给第三方开发商，比如Vantage和QTS这样的公司。我们看到这些公司及其他企业正通过债务市场筹集大量资金，用于建设这些超大规模的数据中心。目前我们看到的交易规模已达数百亿美元，贷款机构也在积极应对，不得不开出巨额支票，并越来越多地与其他贷款方合作。

And so as these data centers keep getting bigger, the cost to build them keep getting bigger. And so the big tech companies like Amazon, Google, Microsoft, they're increasingly outsourcing a lot of this to third party developers, companies like Vantage and QTS. And we're seeing those companies and others go to the debt markets to raise a lot of money for some of these really big data centers. So, you know, we're seeing deals in the 10s of billions of dollars. And we're seeing lenders step up and have to write really big checks and work with other lenders increasingly.

没错。在我们继续之前，先简单提一下

So yeah. And very quickly, before we get

说到这些名字，你看，我在本科学习商业时记得债务比股权更便宜。我记得这一点。这是古老的格言。这是人们转向债务市场的唯一原因吗？还是有其他原因使得债务，相对于其他融资来源，特别适合AI数据中心？

into the names, look, when I was studying business in undergrad, I remember debt is cheaper than equity. I remember this. This is the age old adage. Is that the only reason that people are going to the debt markets? Or is there another reason why debt, as opposed to other funding sources, is really perfect for AI data centers?

是的，没错。通过债务，你能够以项目融资的方式逐个为这些数据中心提供资金。这样，数据中心开发商或像Meta这样的科技公司就不必增发股权稀释股东权益，而是能够逐个为这些特定项目筹集资金。这提高了效率。

Yeah, that's right. With debt, you're able to kind of finance these data centers, one by one on a kind of project finance basis. So instead of the data center developers or a tech company like Meta having to raise more equity and dilute shareholders, they're able to sort of raise capital for these specific projects one by one. And that just makes it more efficient.

对。好的，那我们来看看名单。你名单上的第一个名字是摩根大通。为什么他们在你关注的贷款机构名单中名列前茅？

Right. Okay. So let's get to the list. The first name on your list was JPMorgan Chase. Why are they at the top of your list of the lenders to watch?

是的。考虑到他们作为拥有庞大资产负债表的大型贷款机构，摩根大通位居榜首或许并不令人意外。但他们在数据中心融资方面的行动确实非常迅速。他们与数据中心公司Crusoe达成了一项重大交易，你可能知道Crusoe是Stargate项目的一部分。

Yeah. That's right. And maybe they're not a very surprising name to have at the top of the list given that they're such a big lender with such a big balance sheet. But, you know, they've really moved with speed into data center financing. They did a big deal with this data center company Crusoe, which, you know, you may know Crusoe is part of the sort of Stargate project.

他们正在阿比林为甲骨文和OpenAI建造这个大型数据中心。摩根大通独自承接了全部94亿美元的债务交易，而不是与其他银行合作。现在他们将通过银团贷款的方式开始向其他银行出售这部分债务。但独自承担全部贷款是相当大胆的举动，这显示了他们的积极态度。

They're building this big data center in Abilene for Oracle and OpenAI. And so JPMorgan took down that entire 9,400,000,000 debt deal, instead of, you know, working with other banks on that. Now they're going to start selling that to other banks through a process called syndication. But it was a pretty bold move to kind of take that entire loan themselves. So they're, you know, active in that way.

他们还参与了其他几笔非常大规模的交易。

And they're also involved in a bunch of other deals that are really big.

好的。跟我谈谈名单上的第二名——三菱和住友。

Right. Talk to me about the number two list here. Mitsubishi and Sumitomo.

没错。这两家日本银行在基础设施债务领域活跃已久。主要原因是日本的利率极低，使得这些银行的融资成本远低于部分美国同行，因此它们长期深耕于基础设施投资领域。

Right. So these are two Japanese banks that have been really active in infrastructure debt for a while. And a big reason for that is that interest rates in Japan are so low. So these banks have a much lower cost of funding than some of their American counterparts. And so they've been really active in infrastructure for a while.

过去十年间，它们也将数据中心纳入投资范围。其中住友银行早在2016年就率先为数据中心企业提供建设贷款，并推动信用评级机构开始对这些贷款进行评级，从而扩大了可参与的贷款机构范围，发挥了关键作用。

And they've been getting into data centers as part of that for, you know, the past decade. So Sumitomo in particular was one of the first movers in doing these kind of construction loans for data center companies going back to 2016. They've been pushing credit rating agencies to start giving ratings to some of these loans, which widens the pool of lenders that can participate. So they've been playing a really key role.

这其实引出了我想在讨论下个名单前请教的问题：这些贷款机构如何竞争？当然利率是一个方面，但这是纯粹的价格战吗？还是像风投那样，有些机构会通过提供运营支持等软性服务来竞争？

And that kind of hints on sort of the next question I wanted to ask you before we get to the next name on the list. You know, I've wondered about how these lenders sort of compete with each other. One way is, of course, just the rate that they give for these loans. Is it merely a price thing? Or just like in venture, the way some firms can say, you know, we will support you with sort of more qualitative things about running your business?

贷款机构在争夺数据中心项目时有哪些竞争手段？

Are there ways that lenders compete against each other for these data center deals?

确实存在激烈竞争。银行与私募信贷基金之间的较量已持续多年——过去十年私募信贷的爆发式增长就是明证。

Yeah. Definitely. I mean, you're seeing a lot of competition between the banks and private credit funds. Now that's not a new phenomenon. You know, the growth of private credit has been a huge thing over the past decade.

随着交易规模不断扩大且日益定制化，银行和私募信贷基金开始在条款和交易结构设计上展开全方位竞争，目前所有可能性都在谈判桌上。

But what you're seeing is with these deals just getting bigger and bigger and more bespoke, you're seeing banks and the private credit funds compete on terms, different ways to structure the deals. And so basically, everything is on the table right now.

明白了。我们再来看看其他几个案例。黑石集团大家都很熟悉，请谈谈他们承销的Coraweave项目。

Right. Let's get to a couple other names here. Blackstone is a name that we know very well. Tell us about the the Coraweave deal that they underwrote.

是的。Coraweave这笔交易与常规数据中心交易有所不同。在这个案例中，他们为云服务公司CoreWeave的芯片采购提供融资。这是一笔由黑石集团主导的75亿美元交易，他们投入了超过40亿美元。这堪称芯片融资领域的标志性交易。而黑石也参与了许多更传统的数据中心交易。

Yeah. So the Coraweave deal is a little bit different than your normal data center deal. In that case, they were financing the purchase of chips for CoreWeave, which is a cloud company. And that was a $7,500,000,000 deal that Blackstone led, they put in more than $4,000,000,000 And so this was kind of the marquee deal in this chip financing world. And you've seen Blackstone also do a lot of more conventional data center deals.

他们拥有两家大型数据中心公司，同时也为这些企业提供贷款。

They own two large data center companies. So they've been lending to those as well.

在我们结束前，我想快速请教的是，你在报道过程中产生了哪些疑问？随着这些数据中心的建设，我们未来会在信息中看到更多关于债务的内容。基于你的调研，有哪些报道方向或需要持续关注的问题？

So, you know, one of the things I just want to ask you very quickly before we let you go is what questions this raised for you as you were doing your reporting. I take it we're going to read a lot more about debt in the information as these data centers get constructed. Were there reporting questions or things that you want to watch for going forward as a result of the research that you did here?

确实。我了解到最基础的一点是：这个市场实际上相当不透明。许多交易都是私募信贷，或是银行发放并转售给其他银团的贷款。关于交易结构、条款、利率和参与方的公开信息非常有限，这让我意识到应该对此进行更深入的报道。

Yeah. Well, I think really one of the most basic things that I learned is that this market is actually kind of opaque. I mean, a lot of these are private deals, private credit, or loans that banks are making and syndicating to other banks. And so there's actually really not a ton of information out there about the way these are structured, the terms, the rates, the participants. And so it just made me think I should be doing more reporting on this.

世界需要更清楚地了解这些关键交易是如何达成的。

The world needs to know a bit more about how these deals are getting done because they're so central Right.

关乎

To

人工智能基础设施的建设。

how AI is being built out.

太好了。迈尔斯，这个故事确实与众不同，应该说，目前关于债务的报道还不多，所以这个故事特别吸引我的注意。非常感谢你能来和我们分享这些。这位是TheInformation负责AI与金融领域的记者迈尔斯·克鲁帕。好的。

Great. Well, Miles, it was a story unlike a few other well, I should say, the information hasn't yet written too much about debt, and so it was a story that really caught my eye. Thank you so much for coming on and talking to us about it. That is Miles Krupa, our AI and finance reporter at TheInformation. Okay.

现在正值纽约气候周，对清洁能源和绿色科技领域来说是个重要时刻。政策层面有很多动向，这为相关科技初创公司创造了有趣的发展环境。抱歉打断一下。我想请两位深谙此道的专家来帮我们解析——弗兰克·麦萨诺是政策咨询公司Bracewell的创始合伙人，拉吉·卡普尔则是气候主题风投基金Climactic的联合创始人兼管理合伙人。

Well, it is Climate Week here in New York City, and what a time it is for clean energy and the green tech sector. There is a lot of movement on the policy front, and that has created an interesting environment for tech startups in this space. I'm sorry about that. I want to bring on two people who know this space very well, and they're going to help us make sense of it. Frank Maisano is a founding partner at the policy consulting company called Bracewell, And Raj Kapoor is the co founder and managing partner at Climactic, a climate focused venture capital fund.

欢迎二位到来，很高兴能邀请到你们。

Welcome to the two of you. It's great to have both of you here.

很荣幸参与。

Great to be here.

谢谢，非常荣幸。

Thanks. Pleasure to be here.

弗兰克，我想先请你谈谈。你参加过多次气候周活动，今年与往年相比有什么不同？

Frank, I want to start with you. You're here at Climate Week. You've been to a number of these Climate Weeks before. How is this year different than past years in your experience?

很有意思，我认为有两个关键因素：一是企业参与度显著提高，部分原因可能是大家对即将在巴西举办的COP30存在不确定性。就像几年前在迪拜时，由于交通便利，企业代表就大规模出席了。

Well, it's very interesting. I think there are two factors at play here. One, industry is way more engaged. And I think partly that's because I think there's some uncertainty around what will happen at of who will go to Brazil for COP thirty. Like for a couple years ago when we were in Dubai, industry showed up in force because it was easy, right?

他们已习惯去迪拜。那里有许多酒店之类的设施。我认为由于活动地点及出行方式的不确定性，谁将前往巴西存在一些疑问。因此更多人选择投资这里。第二点不仅是特朗普政府及其政治立场，全球政治局势也从先前激进的气候立场有所退缩。

They're used to going to Dubai. There are lots of hotels, things like that. I think there's some uncertainty about who will go to Brazil because of where the events are and some of the uncertainty around how you can get around. And so I think more people have invested in here. I think the second thing is not just in the Trump administration and the politics of the Trump administration, but also the global political scene has really kind of pulled back from some of the aggressive climate positions that it had previously.

你会看到能源现实正在取代那些对严格限制的幻想期望。

And you're seeing some energy reality kind of take the place of those wishful hopes of meeting strict limitations.

Raj，你如何看待今年气候周与风险投资、初创企业以及科技行业关注点的关联？

Raj, how are you thinking about Climate Week this year as it relates to venture capital and startups and how the tech sector is looking at it?

是的，可以说我们的同行们确实在舔舐伤口，但我想我们已经过了那个阶段。我们确信的是未来永远充满不确定性，而令人欣慰的是大家已接受这点。现在的关键不是居高临下地谈论绿色与气候，而是弄清客户需求并真正倾听——归根结底他们要的是效率。

Yeah, so I would say that our brethren has definitely been licking their wounds, but I think we're beyond that stage now. And the certainty we have is that we're always going to be in a state of uncertainty. And the comfort there is with that. And the key is now to not try to talk down to about green and climate, it's to figure out what the customer wants and to really listen. And the bottom line is they want efficiency.

他们需要供应链的韧性。所以你会看到初创企业不仅追求减碳，更在探索如何提升效率。而当前我们拥有两大绝佳技术来实现这点：人工智能和机器人技术。这两大领域正掀起热潮，探讨如何产生实际影响。

They want resilience for their supply chains. So you're seeing startups that are going after not just reducing carbon, but how do we drive efficiency? And we have the two of the greatest technologies right now to do that, AI and robotics. And there's a big push around both those things and how we can make an impact.

Frank，今年关于数据中心的讨论有多少？

Frank, how much discussion is there this year around the data centers?

这确实是个全球性热门话题，对吧？华盛顿在讨论，各国首都和各州也在讨论，这里同样热议不断。

Well, that is a big topic everywhere, right? That is a big topic in Washington. That is a big topic in capitals around the globe, and in states. That's a big topic here.

人们都在说什么？我是说，就是啊。直截了当

And what are people saying? I mean, just Yeah. Flat out

真的，这是个重大关切，因为我们并不清楚未来需要多少电力。但我们确实知道，我们需要几乎能找到的每一个电子，对吧？现在，在弗吉尼亚和东北部等地，我们看到连海上风电这样的项目也遭到政府某种程度的抵制。那些地方的人都在问，我们该怎么办？如何满足数据中心和消费者需求这种爆炸式增长？

Really, it's a big concern because we don't really know how much power we're going to need. We do know that we're going to need almost every electron that we can find, right? Right now, we're seeing that in places like Virginia and in the Northeast where, you know, even a thing like offshore wind has where the administration has kind of pushed back on it some. You've seen those places say, hey, what are we going to do? How are going to meet that data center and that consumer demand that is exploding?

所以如果我们想拥有大量先进技术，想拥有数据中心，并继续让客户做他们想做的事、满足这些需求，我们就必须找到获取电力的途径。因此这成为讨论焦点，以及如何以最高效、最清洁的方式实现。

So if we're going to have a lot of advanced technology, if we're going to have data centers, and we're going to continue to let the customer do what the customer wants to do and meet those needs, we're going to have to find places to get electrons in. So that becomes a discussion point here, and how you can do it in the most efficient and cleanest way.

Raj，你接触的投资组合公司和初创企业，他们如何看待在满足数据中心需求的同时不损害环境这个问题？

Raj, how are your portfolio companies and startups that you're talking to, how are they thinking about sort of helping us meet the demands for data centers, but do it in a way that doesn't hurt the environment?

我认为这里有两大关键因素。一是我们将需要比现有更多的能源，无论政府如何评价能源类型，目前对中国尤其有效的是太阳能和电池。据我所知，天然气涡轮机有90个月的积压订单，所以问题不在于政治，而是一两年内我们如何必须让更多能源上线。我们已准备好用众多初创企业来助力正在爆发的太阳能和电池产业，而且说实话这个趋势尤其在公用事业规模上并未放缓。

So I would say that there's two factors that are really big here. One is that we're going to need a lot more energy than we have, and whatever an administration wants to say about energy sources, right now, the here and now that's working well, especially for China, is solar and batteries. There's a ninety month backlog, from what I hear, on natural gas turbines, and so the reckoning is not going be politics. It's going to be in the next year or two how we just absolutely have to bring more energy online. So we're ready with lots of startups that can help the solar and battery explosion that we're seeing, and that's not frankly slowing down, especially at the utility scale.

而在数据中心本身，从冷却系统到芯片的创新层出不穷。我们正在显著降低每个令牌的能耗。而且，单就这个领域可能就有200家初创企业在攻关。

And then at data centers themselves, there's so much innovation that's going on from cooling to chips. We're bringing down the energy consumption per token significantly. And again, there's probably 200 startups going after that just to look.

那么你们如何看待政策倒退？比如我们没谈到的风电，现政府正在撤销对风电和太阳能的激励措施。你的公司如何看待这点？

But then how are you thinking about sort of the policy rollbacks? We didn't talk about wind, for example, but the current administration, the way that they're rolling back the incentives for wind and solar. How are your companies thinking about that?

因此我认为，在关键需求场景下——比如我们讨论数据中心能源问题时——政策调整不会产生太大影响，因为我们需要快速增加能源供应，而能采取的措施有限。从边缘角度来看确实延缓了进程，但数据中心的核心需求将决定我们能快速部署什么。好消息是这些新增能源都是可再生能源。

So I think in critical conditions where there's a need, like when we talk about data center energy, the policy's not going to make much of a difference because we need to bring on energy fast, and there's only a certain amount that we can do that. So it certainly has slowed things down when you think about the edges, but the core need that we're going to have around data centers is going to drive what can we deploy quickly, and it's just great news that it's renewables.

如果我理解正确的话，你是说尽管政策已经改变，但对于这些大规模数据中心项目而言，在能源供应方式上不会造成实质影响？我这样理解对吗？

So if I hear you correctly, you're saying that, hey, the policy has changed, but for these large scale data center projects, that's not going to make a difference insofar as how we supply the energy, even though the policy is different now? Am I hearing you

对吧？

right?

我认为当前这些政策变化的影响将在多年后显现。眼下关键是一两年内我们能采取哪些有效措施。另外，数据中心由超大规模运营商掌控——尽管媒体报道给人的感觉不同，但84%的企业仍在坚持或加强气候承诺，他们正积极寻求包括核能在内的可再生能源解决方案。

I say would that the here and now, these policy changes have impact years from now. The here and now is what do we do the next year or two that's going to make a difference? And also, the data centers are owned by the hyperscalers. The hyperscalers, 84% of all companies are still maintaining or increasing their climate commitment, even though it doesn't feel that way when you look at the media, and they are looking aggressively at renewables, including things like nuclear as well to solve the problem.

坦白说，我认为地方审批是更严重的问题。正如Raj所说，相比政策的长期影响，地方阻力才是更大障碍。无论是化石能源还是可再生能源项目，我们都遭遇过各种地方性抵制。

Frankly, I think a larger problem too is local permitting, right? I think that's a bigger problem than what these policy changes will make, as Raj says, the long term. We're still seeing some pushback. Know, of course, I've worked on lots of projects from fossil projects to renewable projects. They all see some sort of local pushback at times.

数据中心同样难以幸免这类抵制。因此在我看来，短期内更大风险在于地方政治——当遇到邻避主义者反对时，地方监管者可能阻碍或帮助社区叫停这些项目。关键在于我们必须让这些电力接入电网，必须采用包括可再生能源在内的全方位解决方案。

And data centers are no, no, are subject to the same type of attacks, right? And so to me, that's the larger risk in the short term is those local politics and local supervisors getting in the way of communities, or helping communities stop these projects when you have, you know, some NIMBYs that don't want them there. So really the larger point is, you know, we do have to have these electrons coming on the grid. They have to be all of the above. They have to be renewables.

我们还需要配套电池储能、海上风电，甚至保留部分化石能源。这才是满足未来需求的解决之道。

Then we have to have battery storage. We have to have offshore wind. We have to have fossil. You know? And so that's where we're going to see how we meet those demands.

但是罗杰，巩固我也是

But Roger, firming I is also a

想再回到你这边稍作讨论，但在我们继续之前。正如弗兰克所说，这是我们报道过的故事，关于孟菲斯市和xAI在田纳西州建设两座大型数据中心的情况。他提到的重点正是我们报道的核心——当地城镇才是面临这些数据中心问题的群体。与此同时，我们还看到英伟达宣布将向OpenAI投资1000亿美元建设数据中心。这就引出了疑问：真的有足够的空间和能源来实现这一切吗？

want to come back to you for one second, but just before we go. I mean, this was a story, as Frank said, this was a story that we wrote in the information about the city in Memphis and xAI building these two massive data centers in the state of Tennessee. And the point that he hit on is exactly what we wrote about, which is that the local towns are the ones that have the issues with these data centers. I mean, the same time, we have this deal, you know, NVIDIA saying, Hey, we're going to invest $100,000,000,000 into OpenAI to build these data centers. And there's a question around, Hey, is there even enough space and enough energy to do that?

我想请教你的问题其实和之前类似：我们是否有足够的空间？是否有足够的能源？考虑到环境问题，你认为这些计划真的能以预期规模实现吗？

I mean, you know, I guess the question for you, it's kind of the same before, is do we have enough space? Do we have enough energy? Do you think these are even going to happen on the scale that they happen given the environmental concerns?

正如詹斯·王最近所说，我们已从算力匮乏转向能源匮乏。我们观察到数据中心技术正在发展——物理空间因设备密度增加和冷却技术改进而变得可用，但输入能源才是当前面临的挑战。为此我们必须做两件事：提升能效（这也有益环保），其次是引入兼具可靠性、可再生性、间歇性和储能特性的新型能源。要实现你提到的目标，综合采用这些方案是基本途径。这些公告并未解决所有问题。

So, Jens Wang recently said that we've moved from being compute starved to energy starved. We're looking at is we're looking at data center technologies, that there's actually physical space that's available because things are becoming more dense, the cooling is becoming better, but what we're seeing is that the input energy is the challenge that's there. So we have to do two things: drive efficiency, which also helps the environment, and secondly is bring in new forms of energy that are both reliable and renewable and intermittent and storage. All of the above is basically the answer to get there, if we have a chance of getting there, to your point. These announcements have not figured everything out.

现在是

Now it's

该我们

our turn to

解决问题的时候了。

figure things out.

是的。我认为我们还有很多话题可以探讨。首先感谢两位嘉宾的到来。正如我所说，这是气候周，但我们应该越来越多地讨论这个话题。期待未来能再次邀请两位。

Yeah. Well, I think there's a lot more to talk about. I want to thank you both for coming on. Like I said, it's climate week, but you know, we should be talking about this more and more. So I look forward to having you both on.

这位是Climactic VC的罗德和Bracewell的弗兰克。非常感谢。好的。在我们最后一个环节中，人工智能为咨询公司带来了巨大商机，它们正帮助全球企业探索如何运用这项新兴技术。这也使得专注于AI的小型咨询公司即便行业资历尚浅，也能实现快速增长——虽然我今天似乎总念错台词。

That is Rod from Climactic VC and Frank from Bracewell. Thanks so much. Okay. Well, for our final segment, AI has meant big business for consulting firms that are trying to help other enterprises around the world figure out how to use this new emerging technology. It has also meant that smaller AI focused consulting firms that have seen rapid growth, even if they are relatively new to the industry, can't seem to read my lines today.

重点在于，各位，AI咨询正在蓬勃发展。我们本周报道了其中一家公司Distill AI，它以18亿美元的估值获得了巨额融资。现在有请CEO阿尔琼·普拉卡什谈谈他将如何运用这笔新资金推动业务发展。阿尔琼，欢迎来到TITV，很高兴你能来。

The point is, folks, AI consulting is booming, and we wrote a story this week about one such company, Distill AI, that has raised a ton of money at a $1,800,000,000 valuation. I wanna bring on CEO Arjun Prakash to talk about where he plans to take his business with the new funding. Arjun, welcome to TITV. It's great to have you.

非常感谢邀请。消息是昨天曝光的，我们自己的新闻稿今天正在发布。能在这里与您畅谈Distill的发展方向，我实在感到非常兴奋。

Okay. Thanks a lot for having us. And the information broke the news yesterday. Our own release is going out today and as we speak, and I'm I'm really excited to be here chatting with you about Distill and where we're headed.

是的，我也很期待这次对话。你看，你们业务的核心是这个新职位——前置部署工程师。对吧？

Yeah. I I I'm excited to talk to you. So look. At the center of your business is this new title, the forward deployed engineer. Okay?

现在这个词已经被人们随意使用，但我不想假设大家都了解这个概念。能否详细解释下这个AI领域炙手可热的新职位究竟是什么？

And we you know, it's become something that people throw around now, but I don't want to assume that people know what the forward deployed engineer is. So tell us what exactly is this new job that AI has become obsessed with?

请允许我先做个背景说明。这项技术确实极具价值，而且当前存在一个相当独特的局面：企业对这项技术的需求已经迫在眉睫，他们不愿等待技术完全成熟，而是需要立即借此获得竞争优势。

Well, let me take a step back here. I think it's fair to say that there is really valuable technology here. And in what I think is a fairly unique situation and circumstance, the demand from enterprise for this technology is here and now. It isn't waiting for the technology to be perfected. It's they need it right now to gain a competitive edge.

我们在此真正所做的是与客户紧密合作，以当下实现成果，这主要通过三方面结合达成。首先是人才与团队——我们派遣工程师、研究员和项目经理直接进驻客户方，与他们携手共担成果责任，既解决技术难题也攻克项目管理挑战。

And so what we really do here is we partner with clients very closely to manifest outcomes today, and we do it through a combination of three things. Number one is talent and people. So we forward deploy engineers, researchers, program managers to be at our clients to own the outcome hand in hand with them, both to solve the technology problems and the program management problems. Number two is we deploy our products. Right?

其次是部署我们的产品。拥有经过实战检验、能可靠服务数亿人需求的规模化产品，能让你从一开始就获得加速优势。第三则是研究——在行业应用这项技术的同时，我们也在持续探索其边界。因此，垂直整合强大的研究团队、优质产品与服务来实现协同运作至关重要。

So it is helpful to have products that work at scale reliably that have been battle tested to meet the needs of hundreds of millions of people. That gives you acceleration straight out of the gate. And number three is research. We are figuring out this technology as we deploy it as an industry. So it's really important to vertically integrate a very strong research team, good products, and the services to actually make this work and make it all come together.

正因如此，我们的客户已开始见证成效。Distill坚信这种模式的关键在于其直接性——采用AI绝非像从Windows XP升级到Vista那么简单，而是对企业运作方式的重构。AI时代的赢家不会是那些仅将其视为年度工具更新的企业。

And our clients are seeing results because of it. And the reason why we here at Distill believe this is very important is it's really straightforward. And adopting AI isn't as simple as upgrading from Windows XP to Windows Vista. It's it's a fundamental rearchitecture of how your company works. And the winners of this AI age aren't gonna be companies that just think of it as a tool year rollout.

真正的赢家将是那些围绕AI重构自身架构、人才体系和商业模式的公司。我们希望通过垂直整合的方式成为他们的合作伙伴，共同实现这些转型成果。

It's gonna be companies that think about rearchitecting them themselves, the people, and their business models around this. And so we wanna be partners with them in a vertically integrated way to make these outcomes possible.

明白。长话短说——为何不直接称他们为工程顾问？这个'前沿部署'究竟指什么？

Right. Very quickly, and I don't want to dwell on this, but why don't they just call them engineering consultants? What is the forward part of this, very quickly?

前沿部署的概念很简单。传统硅谷模式是让最优秀的工程师坐在帕洛阿尔托，听产品经理指挥开发。我们反对这种模式——坚信顶尖工程师和研究员应该活跃在客户的走廊里，深入客户...

The forward part is really simple. You know, traditionally Silicon Valley has had its smartest engineers sitting in Palo Alto building products that a product manager tells them what to build. And we don't believe in that. We believe that the smartest engineers and researchers should be at our clients' halls. Going to the client

现场。懂了。

side. Okay.

没错。这是我们唯一能推进此事的方式。明白了吗？就是这样。

Exactly. And that is the only way we Taking your do this forward Got it. Exactly.

好的。那么前向部署工程师，或者工程顾问，随你怎么称呼。这在AI领域变得非常流行。不过，我想问你的是，当你与所有这些希望实施AI的企业合作时，我们记录了他们面临的诸多挑战，尤其是人为因素方面。你知道，如何教会人们使用它？

All right. So the forward deployed engineer, the engineering consultants, call it what you want. This has become very popular in the land of AI. Know, one of the things I wanted to ask you though is as you're working with all of these enterprises that are looking to implement AI, we've written about several challenges these enterprises have at the information, not the least of which is the human side of this. You know, how do you teach people how to use it?

另一方面是领导层表示，他们甚至不知从何开始。你认为当前企业在战术层面实施这项技术时面临的首要挑战是什么？

Another side is leadership saying, I don't even know where to start. You know, what do you see as the number one challenge right now for enterprises tactically in implementing this technology?

嗯，好问题。这正是我们日常新闻的背景。对吧？95%的AI项目都失败了。

Yeah, great question. It's the backdrop to the news that we all read about. Right? 95% of AI projects are failing.

是啊。

Yeah.

为什么这些项目会失败？根据我们的经验，我们成功地将100%的倡议投入生产。我认为有几个关键点。首先，必须承认企业内部领域专家的重要性。让AI发挥作用的关键在于为模型提供正确的上下文，使其能准确可靠地执行任务。

Why is it that these projects are failing? And I can tell you that based on our own track record, we're being successful in getting a 100% of our initiatives into production. I I think it's a few key things. Number one is it's important to acknowledge the role of the subject matter expert or the domain expert at the enterprise. The key to making AI work is being able to provide the right context to these models so they can carry out these tasks in a in a accurate and reliable way.

因此，我们非常重视构建能吸引领域专家的产品和参与模式。在我们合作过程中，这些专家的角色会从执行日常任务转变为即时管理者，负责策划让AI处理这些日常任务的接触点。这正在引发企业内部的根本性变革管理。

And so what we really place a lot of emphasis on is building products and engagement models that engage the subject matter experts. And what that looks like over the lifetime of our engagement is the role of these experts goes from people who are doing mundane tasks to instant managers who are curating contacts that ask the AI doing these mundane tasks. And there's a fundamental change management that's going on in these enterprises.

让我来转述一下你的意思。你是说，基本上需要在客户端有专人负责AI项目，不仅要对其盈亏负责，还要能运营整个流程。这在当前是个很大的制约因素。

So let me translate what you're saying. You're saying that basically having somebody on the client side, being able to own the AI initiative, being accountable for it, not just from a P and L perspective, but also being able to run the entire operation. That's a great limiting step right now.

阿加什，说得好。完全正确。你需要把这看作一个必须由客户端主导的跨职能项目，并且需要引入领域专家——这些专家通常就是客户方的员工。

Agash, great point. That's exactly right. You need to think of this as a cross functional project that has to live on the client side. And you need to bring in the domain experts. They're usually employees off the client.

他们是最了解自身供应链的人，最清楚临床运营细节的人。必须让他们以一流方式参与项目成为贡献者。因为如果不让他们全程参与，AI系统就无法明确自己该做什么。

They are people who understand their supply chain best. They're people who understand their clinical operations best. And they need to be brought into these projects in a really first class way to be contributors. Because if you don't get them along for their journey, the AI is not going to be able to know what it is that it's supposed to do.

明白。那么请允许我快速请教最后一个问题——你认为企业首席AI官这个职位是被高估还是低估了？

Right. Okay. So let me ask you this, and we've to let you go. So very quickly, just want to get your take on this. The role of a chief AI officer at an enterprise, do you think it's an overrated or underrated position?

我认为这是极其重要的职位，需要高度的战略规划意识。

I think it's an extremely important role, and it requires a degree of intentionality.

但有人可能会反驳——既然需要协调组织各部门而不仅是盯着AI，为什么非要设立这个职位？团队协作不是比单设一个职位更合理吗？

But you don't And I'm just going to pose an argument here. Some people could say, if you have a chief AI officer, right? I mean, you need somebody who's drawing on all the different parts of the organization, not just looking at AI. Some people could say, well, you know, you need a you need a team, not just a person. Why are you naming a person?

我来谈谈我的看法。首先我认同你的观点。美国企业的现状是存在大量信息孤岛，这是工业化时代遗留问题——当时决策集中化而执行分散化。

Well, let me let me tell you how I think about it. And and I think you're right. Let me start by saying that. The American enterprise is one where we've just grown a lot of silos. I think it's a legacy artifact of industrialization where we had centralized decision making and decentralized execution.

因此，装配线的不同环节之间互不了解彼此的工作内容。这有其优势，对吧？它使我们能够扩大生产规模，但也形成了这些信息孤岛。我认为，首席人工智能官的职责应该非常明确。

So you had different parts of the assembly line that didn't know what each other were doing. And that had its benefits. Right? It allowed us to scale up production, but it's created these silos. And what I think that a chief AI officer's role should be is very straightforward.

就是将业务、技术、治理、法律、安全这些孤岛整合起来，通过团队协作机制，有目的地实现使命。这一使命必须与企业试图达成的目标同步，同时遵循IT部门旨在保护的治理框架。这正是我认为优秀的人工智能官应做到的。

It's in bringing these silos together, the business, the technology, the governance, the legal, the security, and a team of teams construct and add intentionality to achieve a mission. And that mission needs to be in lockstep with what the business is trying to achieve within a governance paradigm that IT is meant to protect. And that's what I think a great AI officer does.

我们非常荣幸能与

And we're very fortunate to partner with

客户们杰出的首席人工智能官合作。

great chief AI officers of our clients.

没错。太棒了。阿琼，非常感谢你参加本期节目。这位是Distill AI的首席执行官阿琼·普拉卡什，该公司本周以18亿美元的估值完成了融资，这一消息由《The Information》独家首发报道。好的。

Right. Great. Well, Arjun, thank you so much for coming on the show. That is Arjun Prakash, the CEO of Distill AI, and they have raised money at a $1,800,000,000 valuation this week, exclusively reported first in the information. Okay.

今天的节目就到这里。提醒大家，我们每周一至周五太平洋时间上午10点（东部时间下午1点）在此直播。我要感谢本节目首席赞助商亚马逊云服务，也感谢各位的收看。我们真心感激每一位观众。我已经开始期待明天的节目了。

Well, that does it for today's show. A reminder that we are live on this stream Monday through Friday at 10AM Pacific, 1PM Eastern. I wanna thank Amazon Web Services who is our presenting sponsor for this production, and I wanna thank you for tuning in. We really do appreciate your viewership. I am already excited for our next show tomorrow.

那么下次再见，各位拜拜。

And so until then, bye bye for now.