谷歌凭借TPU与Nvidia竞争，OpenAI通过预训练自动化华尔街 | 2025年11月25日

请大家到信息TI电视前。

Everyone to the Informations TI TV.

我叫阿尼塔·拉马斯瓦米。

My name is Anita Ramaswamy.

今天是11月25日，星期二。

It is Tuesday, November 25.

今天我们为大家准备了一场精彩的节目。

We have a great show lined up for you today.

首先，我们将带来关于谷歌为直接与英伟达在AI芯片领域竞争而采取的新举措的最新报道。

First up, we have new reporting on Google's fresh efforts to compete directly with NVIDIA on AI chips.

我们将讨论谷歌如何可能在此领域迎头赶上。

We'll discuss how it might be catching up there.

我们还将深入分析萨姆·阿尔特曼可能对谷歌Gemini三模型成功所存在的主要担忧。

And we'll break down one of the big concerns Sam Altman might have around Google's success with its Gemini three model.

此外，我们将深入分析甲骨文数据中心建设背后不断增长的债务融资，并展望2026年，我们邀请专家探讨IPO市场即将发生的变化。

Plus, a closer look at the rising debt funding behind Oracle's data center build outs and a look into 2026, we bring on an expert to discuss what's coming in the IPO market.

今天最后，我们将与ModelML的首席执行官对话，这是一家金融科技公司，在成立约一年后刚刚宣布了7500万美元的A轮融资。

And we'll end today on a conversation with the CEO of ModelML, a fintech company that just announced a $75,000,000 Series A round just about a year after launch.

今天的内容非常丰富，让我们马上开始。

It's a busy show, so let's get right into it.

《信息》发布了最新报道，显示谷歌在与英伟达直接竞争AI芯片业务方面取得了进展。

The information published new reporting that shows Google is making strides in its efforts to compete directly with NVIDIA in the AI chip business.

谷歌正通过与Meta等大公司洽谈，让它们在自己的数据中心使用谷歌的芯片来实现这一目标。

And it's doing so by talking to big companies like Meta about letting them use its chips in their data centers.

这篇报道发布后，英伟达的股价下跌了。

Shares of NVIDIA fell in the aftermath of the story.

今天请来深入探讨这一细节的是报道此新闻的记者之一，艾琳·吴。

Here to get into the details is one of the reporters on this story, Erin Wu.

嗨，非常感谢邀请我。

Hey, thanks so much for having me.

感谢你来到这里，艾琳。

Thanks for being here, Erin.

那么我想先请你给我们介绍一下背景情况。

So I want to start by having you give us a little bit of background.

谷歌目前在AI芯片方面的具体策略是什么？

What exactly is Google's current strategy when it comes to AI chips?

是的。

Yeah.

谷歌自主研发AI芯片已有十年时间。

So Google's been developing its own AI chips for a decade.

它们被称为TPU，张量处理单元。

They're called TPUs, tensor processing units.

这些芯片实际上是由谷歌制造，由谷歌为谷歌设计的。

And these are chips that are really made made by Google, designed by Google for Google.

因此他们一直在用这些芯片开发自己的人工智能模型，包括Gemini，正如你们讨论的，最近取得了很大成功。

And so they've been using them for development of their own AI models, including Gemini, which, as you discussed, has had a lot of success recently.

谷歌还有一个云业务。

And Google also has a cloud business.

因此，谷歌还通过云业务向客户提供TPU。

And so Google has also been providing TPUs for its customers through the cloud business.

他们已签署了许多知名交易，包括苹果、Anthropix和SAFE超级智能。

And they've signed a lot of high profile deals including Apple, Anthropix, SAFE Superintelligence.

但我们目前报道的是，谷歌也开始向客户推销将TPU部署在他们自己的数据中心。

But what we're reporting now is they're also starting to pitch customers on putting TPUs in their own data centers.

这意味着超越了我们目前所见的云模式，转向在本地部署芯片。

So going beyond the cloud model that we're seeing right now and into putting chips on premises.

所以，你能为我进一步解释一下这一点吗，埃琳？

So can you unpack that aspect a little bit more for me, Erin?

我的意思是，根据你的报道，你认为谷歌在芯片方面的AI战略发生了什么变化？

I mean, based on your reporting, what would you articulate as the change in Google's AI strategy when it comes to chips?

是的。

Yeah.

这意味着从将芯片放在谷歌自己的数据中心，转向将芯片放在NeoCloud或像Horoweave这样的客户的数据中心，甚至直接放在最终运营客户那里。

So it's going from having chips in Google's own data centers to having chips in either the data centers of NeoCloud to customers like or companies like Horoweave or directly to, like, the ultimate operating customer.

所以我们之前谈到过，谷歌正在与Meta讨论这样的交易。

So we talked about Google talking with Meta about a deal like this.

谷歌也在向那些可能需要更安全、更保密数据的大型金融机构推广这一方案。

Google's also pitching this to major financial institutions who maybe need to have data kept more safe and secure.

这些谈判，Erin，如果成功的话，它们如何可能缩小与NVIDIA之间的差距？毕竟目前NVIDIA在AI芯片市场遥遥领先。

And these talks, Erin, if they do succeed, how could they potentially close the gap with NVIDIA who is far and away the market leader at this point in AI chips?

没错。

Right.

目前，NVIDIA远远领先于谷歌。

So NVIDIA is way further ahead than Google is right now.

它们拥有许多优势，部分原因在于GPU是为方便客户使用而设计的，而TPU则是谷歌开发的芯片。

They have a lot of advantages because in part GPUs, they're made for customers to be able to use them versus TPUs, the chips that Google has developed.

从数据中心的设计，到电力和散热方式，一切都非常具体。

They're very specific from everything down to, like, the design of the data center to how power works and how cooling works.

因此，谷歌要想让客户安心使用这些芯片，无疑面临一场艰难的挑战。

So it's definitely an uphill battle for Google to climb in terms of getting customers comfortable using these chips.

但谷歌的看法是，这些数据中心存在着一个巨大的市场。

But the way Google is looking at it, it's like there's this massive market for all of these data centers.

谷歌试图进入这个市场，这对他们来说是一个巨大的机会。

And the idea that Google could try to compete with that market is a really huge opportunity for them.

因此，当谷歌云在考虑这件事时，他们就在问：潜在市场有多大？

And so when Google Cloud was thinking about doing this, they were asking like, Okay, what's the potential market?

有人提出了这样的数字：如果我们能拿到英伟达市场份额的10%，而英伟达的市场规模超过100亿美元呢？

And the numbers were thrown around like, Okay, well what if we could get to 10% of what NVIDIA is doing, which is like a $10 plus billion market?

所以，即使他们无法撼动英伟达的地位，这里依然存在巨大的机会。

And so there's a really big opportunity here for them even if they aren't able to dislodge NVIDIA.

我的意思是，这确实是一个巨大的机会。

I mean, it's definitely a big opportunity.

你认为谷歌在这一领域真正对英伟达构成实质性挑战的可能性有多大？

To what extent do you think it's realistic that Google really gives NVIDIA a run for their money in this space?

说实话，我觉得这很难。

Look, I think it's hard.

我提到过，NVIDIA 的 GPU 本质上已经为用户使用而设计。

I mentioned that NVIDIA that GPUs are already built essentially for customer use.

对于 TPU，谷歌需要解决的一大问题是：如何让人们使用它。

With TPUs, this is a lot of, like, what Google has to figure out is, like, how to get people to use it.

因此，我们报道的一部分内容是，他们不仅仅是向人们提供芯片。

And so part of what we reported, they're not just giving people the chips.

还有配套的软件，旨在让使用变得更加简便。

Like, there's some software that goes on top of it that which is designed to make it easier for people to use.

但目前，谷歌正处于真正的上升势头中，尤其是被广泛认为的势头——他们长期以来一直使用这些 TPU 来训练 Gemini。

But, I mean, right now, Google is in a place of real momentum and especially, like, perceived momentum where they've been using these TPUs to train Gemini for a long time.

但现在随着它的成功，我认为谷歌可能迎来一个切入市场、抢占份额的时机。

But now with the success of it, there's I think maybe a moment for Google to come in and take some market share.

而且他们一直在签署大量大型交易。

And they've been, again, signing all of these big deals.

比如，他们刚刚与 Anthropic 达成了一项采购一百万枚 TPU 的交易。

Like, there is just a deal with Anthropic, for example, for 1,000,000 TPUs.

对。

Right.

所以你注意到了一些这场竞争中的挑战。

So you noticed or you noted some of the challenges in this competition.

我经常听到的一个阻碍谷歌与英伟达竞争的问题是英伟达的CUDA软件。

And one roadblock that I've heard a lot about to Google competing with NVIDIA is NVIDIA's CUDA software.

这是一种AI开发者用来编程芯片的软件。

That's a software that AI developers use to program their chips.

据我所知，它在这些开发者中已经成为某种标准。

You know, from everything that I've heard, it's become somewhat of the standard amongst those developers.

谷歌究竟该如何在这个领域展开竞争？

How exactly will Google compete in that realm?

是的。

Yeah.

所以，在我们最近的报道中，谷歌正在向企业推广一个名为TPU应用前置的项目。

So what we've reported on in this most recent story, so Google's pitching companies this program called TPU app premises.

正如我提到的，这不仅仅是他们出售芯片那么简单。

And like I mentioned, this is not something where they're just selling the chips.

上面还附带了一些软件。

There's also some software on top.

这可能会被设计成一种按月租赁的模式，谷歌工程师仍然在管理其中某些方面。

Like, this is could potentially be structured as like a monthly rental fee where like Google engineers are still managing some aspect of this.

因此，谷歌完全明白，他们必须教会人们使用一种新的框架，这与GPU不同。

And so Google completely understands that this is something where they have to teach people a new framework and it's like different than GPUs.

但他们试图推销的方式是，这可能更具成本效益。

But the way that they're trying to pitch it is this idea that this could be more cost effective.

同样，这里还有Gemini的光环效应。

Again, there's this Gemini halo.

你报道中的大公司，还有

The big company in your story and in

埃琳，你的报道中的大公司是Meta。

your reporting, Erin, is Meta.

但你在报道中也提到，我注意到谷歌正在与大型金融机构讨论TPU的使用。

But you also said in the story, I noticed that Google is in talks with big financial institutions about the use of TPUs.

关于这一点，我们还知道些什么？

What more do we know about that?

我们了解得更多一些。

We know some more.

有些内容我们没能写进报道里。

There's some stuff that we can't couldn't put in the story.

我们确实知道他们正在与哪些机构洽谈。

We do have some examples of who they're talking to.

但这一方案被宣传为适合那些对数据安全要求极高、希望将系统部署在本地而非将信息发送到云端的公司。

But this is being pitched as something that's good for companies with that have very secure data needs and want need to have things on premise premises because they don't wanna send information to the cloud.

因此，这也是其中一部分原因。

And so that's part of it.

此外，还有一些有趣的细节，谷歌试图将TPU宣传为适合从事高频交易的公司的解决方案。

There's also some interesting details where Google is trying to pitch this as as something that's good for companies that are trying to do high frequency trading.

因此，他们显然正在积极进军金融机构市场。

And so they're definitely making a play at this financial institution market.

我个人非常期待看到你们没能在报道中包含的那些细节，我相信这些内容会在你们后续的报道中揭晓。

Well, I personally can't wait to see what details that you didn't get to put in the story that I'm sure will come out in your later reporting.

每次有你做客都真是愉快，埃琳。

Always a pleasure having you, Erin.

感谢你的到来。

Thanks for being here.

非常感谢你。

Thanks so much.

我们继续聊谷歌，Gemini 3 AI模型的发布，点燃了OpenAI及其首席执行官萨姆·阿尔特曼的紧迫感。

So sticking with Google here, the release of the Gemini three AI model has lit a fire under OpenAI and its CEO, Sam Altman.

正如我们上周报道的那样，他告诉员工要加紧努力，为所谓的艰难氛围和暂时的经济逆风做好准备。

He told his staff to buckle down and prepare for quote rough vibes and temporary economic headwinds, as we reported in the information last week.

我那位率先报道此事的同事，现在正聚焦于阿尔特曼可能担忧的一个具体问题：谷歌的模型究竟是如何变得如此强大的。

My colleague who broke that story is now zeroing in on one specific concern that Altman might have, which is around how Google's model got so good in the first place.

这是通过一种称为预训练的过程实现的。

That's through a process known as pre training.

现在有斯蒂芬妮·帕拉佐洛加入我们，来谈谈这个话题。

Stephanie Palazzolo joins me now to talk about it.

斯蒂芬，很高兴你回来。

Steph, good to have

你回来了。

you back.

谢谢。

Thanks.

很高兴能在这里。

Great being here.

那么，为了开始，你能先提醒一下

So just to get started, can you first remind

我们预训练到底是什么吗？

us what exactly pre training is?

完全正确。

Totally.

预训练是训练模型的第一步，在这个过程中，你会向模型展示来自互联网、书籍和大量其他来源的海量数据，从而帮助模型理解这些不同概念之间的关联。

So pre training is kind of the first step of training a model where you essentially show the model tons and tons of data from the internet, from books, from tons of sources to, you know, kind of teach the model connections between these different concepts.

这基本上赋予了模型对世界及其运行方式的总体理解。

So essentially kind of gives the model like a general understanding of the world and how the world works.

明白了。

Got it.

那么，这与AI模型中的后训练过程有何不同呢？

And how is that different from the process known as post training when it

来了。

comes to AI models?

是的。

Yeah.

后训练是训练过程中的一个后期阶段，在模型已经掌握了对世界的总体表征之后，后训练主要帮助它在特定领域进一步提升。

So post training is a later stage in the kind of training process where, you know, once the model has learned this kind of general representation of the world, post training basically helps it improve either in specific areas.

例如，你可能想教它更多关于数学、编程或法律的知识，或者帮助模型以人类更喜欢的方式做出回应。

So for instance, maybe you want to teach it extra knowledge about math or coding or law, or it can also help the model kind of respond in a way that humans like more.

所以你可能想教模型更礼貌、更友好，或者拥有某种特定的性格。

So maybe you want to teach the model to be more polite or friendly or to have a certain personality.

所以

So

为什么萨姆·阿尔特曼特别担心预训练，尤其是在ChatGPT的预训练进展方面？

why is pre training something that Sam Altman is specifically worried about, especially when it comes to CHAT GPT and its progress in pre training?

是的。

Yeah.

萨姆之所以如此担心预训练，是因为许多研究人员认为，预训练是让模型更好地泛化的一种更有效的方式。

So the reason why Sam is likely so worried about pre training is because a lot of researchers think that pre training is a much better way for models to generalize.

所谓泛化，我的意思是模型能够根据训练数据之外的信息生成内容或做出新的发现。

And by generalize, I basically mean models being able to, you know, create content or make new discoveries outside of the data that it was trained on.

因此，这之所以如此重要，是因为OpenAI和其他许多模型开发者正在构建这些模型，因为他们声称希望最终创造出能够治愈癌症、发现新科学，甚至进行AI研究以推动AI自身进步的AI。

So the reason why this is so important is that OpenAI and lots of other model makers are building these models because, you know, they're saying that they want to create AI that can cure cancer at some point or, you know, come up with new science discoveries or even do AI research to help improve AI.

显然，所有这些对模型来说都是训练数据中从未见过的全新概念和想法。

So obviously all of these are like brand new concepts and ideas that the models have never seen before in their training data.

这意味着它们必须超越最初训练时所学的内容进行泛化。

So this means that they have to kind of generalize beyond what they were originally trained on.

因此，由于预训练对此更有帮助，你可以想象预训练再次赋予了模型这种通用的全球知识，从而帮助模型学习超越训练数据的内容。

So because pre training is more helpful for that, you can kind of imagine that pre training again gives you this general world knowledge and so it helps the models kind of learn things beyond the training data.

这就是为什么它如此重要，因为它确实有助于这一点；而事后训练实际上有时可能会对教会模型掌握训练数据之外的知识产生负面影响。

So this is why it's so important because it does help with that versus post training can actually sometimes either, you know, it can actually have a negative effect on teaching models knowledge outside of the data that it was trained on.

例如，如果你想对模型进行事后训练，教它如何更好地进行数学证明。

And so for instance, if you, you know, want to post train a model on how we, how to be better, better at to be math proof.

你希望它在回应时更加逻辑和理性。

So you want it to be more logical and rational in its, in its responses.

我实际上也听说过它在其他领域的影响，比如幽默感。

I can actually heard it in other areas like being funny, for instance.

这相当惊人，因为在训练模型时，有时确实存在直接的权衡。

That's pretty wild that there's a direct trade off sometimes when it comes to training the model.

我的意思是，我只是好奇想了解更多关于谷歌是如何在预训练模型方面做得如此出色的细节。

I mean, I'm just curious to get a little bit more context on how exactly Google got so good at pre training the model.

是的。

Yeah.

我们并不清楚谷歌内部具体发生了什么，才让他们在预训练方面如此出色。

So it's not really clear to us exactly what happened within Google to get them to be so good at pretraining.

正如我的同事Aaron Wu所写，谷歌有一些特定的研究人员，比如在Character AI以27亿美元收购案中被引入的Noam Shazir，据消息人士称，他在提升预训练过程方面发挥了重要作用。

As my colleague, Aaron Wu has kind of written about, there are specific researchers at Google like Noam Shazir who was brought in during the $2,700,000,000 acquisition of Character AI, who sources have told us has been really important to the process of improving pre training.

但很多人只是笼统地说，这归根结底在于拥有大量优质的数据，以及在模型训练方面拥有非常完善的过程。

But a lot of people kind of just say generally, it basically boils down to having a lot of really great data and then very good processes when it comes to training model.

因此，研究人员需要进行大量、大量的实验。

So essentially the researchers need to run lots and lots of experiments.

关键在于很好地执行实验过程，测试各种不同的想法，然后继续推进那些

So it's kind of the process of running the experiments really well, testing a bunch of different ideas, then moving forward with the ones that

最有前景的。

are the most promising.

所以，如果我们真的想看到这些模型性能的显著提升，就必须在预训练上实现基础性的改进。

So it does sound like if we're going to really see a step change in how well these models work, it has to stem from a foundational improvement in pre training.

我的意思是，我在想你如何看待这个故事的下一阶段，因为听起来OpenAI似乎有点落后了。

I mean, I'm wondering how you're thinking about the next stage of the story because it sounds like OpenAI is a little bit behind.

他们究竟需要做些什么来改进预训练，以确保在这一领域保持竞争力？

What exactly will they need to do to improve on pre training just to make sure that they're staying competitive in this area?

是的。

Yeah.

正如我们写的备忘录所示，萨姆·阿尔特曼似乎非常有信心，认为公司正在押注一些他认为会取得回报的举措，这些举措将帮助OpenAI在预训练和模型质量上追上谷歌。

So as we saw from the memo that we wrote, Sam Altman does seem pretty confident that the company is making bets that he thinks is going to pay off and is going to help OpenAI catch up to Google when comes to pre training and the kind of quality of their models.

所以我认为，不管OpenAI正在做什么，他们显然都在大力推动这些工作。

So I do think that, you know, whatever OpenAI is working on, they do seem to be pushing ahead pretty strongly on it.

我猜测，他们希望在未来几个月内就能看到成果。

And I imagine that they will wanna see results sometime in the next couple months or so.

我们之前已经写过，比如他们正在开发一个代号为‘shallotpete’的新模型，旨在解决过去在预训练中遇到的一些问题和缺陷。

We already wrote, for instance, that they are working on a new model which is code named shallotpete, which basically aims to kind of fix some of the pre training bugs and issues that they had experienced in the past.

所以，你知道，这可能需要几个月的时间，但我预计我们很快就会看到OpenAI的Shallot Pete模型。

So, you know, it might take a couple of months, but I imagine that we're gonna see that shallot peat model from OpenAI pretty soon.

然后从那时起，我们可以看出OpenAI在预训练方面追赶了多少。

Then I think from then going forward, we can see how much OpenAI has been able to catch up with pre training.

他们是否在

Have they put

后训练上也投入了大量努力？

in significant efforts into post training as well?

是的。

Yeah.

到目前为止，OpenAI一直非常专注于后训练。

So, so far OpenAI has been pretty focused on post training.

这涉及从医疗或金融等行业的公司获取大量非常具体的数据，以在这些特定领域大幅提升模型性能。

So this involves acquiring lots of really specific data from companies in healthcare or finance, for instance, to really improve the models along those specific areas.

此外，我们还写过，像OpenAI和Anthropic这样的公司花费了大量资金用于数据标注，以及一种名为强化学习微调的数据标注方式——这基本上是模型用来学习如何操作新应用的虚假应用副本。

And then we've also written about how companies like OpenAI and Anthropic have spent a lot of money on data labeling, but also on a specific type of data labeling called reinforcement learning gems, which are basically kind of like fake copies of apps that the models can use to learn how to navigate these new apps.

所以毫无疑问，OpenAI 目前也非常专注于后训练。

So definitely OpenAI, I think right now is definitely also very focused on post training.

明白了。

Got it.

谢谢，斯蒂芬。

Well, thanks, Steph.

我想这还有待观察，我们必须密切关注 OpenAI 将如何应对前方这些所谓的困难局面。

I guess it remains to be seen and we'll have to watch exactly how OpenAI navigates these quote unquote rough vibes ahead.

我知道你将站在第一线报道这一切。

And I know that you're going be on the front lines reporting on all of it.

所以希望我们能再次邀请你来参加节目。

So hopefully we'll get you back on the show.

感谢你的到来。

Thanks for being here.

华尔街越来越担心，为建设 Oracle 数据中心的公司提供融资的债务正在增加。

Wall Street is growing increasingly worried about rising debt funding for companies that are building data centers for Oracle.

我的同事迈尔斯·克鲁帕就这个故事写了一篇精彩的文章，今天他来到这里和我一起讨论。

My colleague, Myles Krupa, wrote a great piece on this story, and he joins me here today to talk about it.

迈尔斯，你好吗？

Miles, how are you?

很好。

Good.

谢谢，阿妮塔。

Thanks, Anita.

你呢？

How are you?

我很好。

I'm doing well.

我准备好谈谈债务和数据中心了。

I'm ready to talk about debt and data centers.

为了让我们观众更好地理解这个故事，目前围绕甲骨文数据中心项目的借款规模到底有多大？

So just to put the story in perspective for our viewers, how much borrowing exactly are we seeing around Oracle's data center projects?

是的。

Yeah.

今年，为甲骨文建造数据中心的开发商已通过项目融资筹集了至少650亿美元，这基本上帮助他们建设数据中心，并为甲骨文入驻并部署芯片做好准备。

So we've seen data center developers building facilities for Oracle raise at least $65,000,000,000 this year in project finance, which basically helps them build, the data centers and get them ready for when, Oracle's ready to come in and put their chips in there.

所以，迈尔斯，你能详细说说甲骨文与OpenAI签订的那些大合同吗？大家似乎都在谈论这些合同，它们构成了甲骨文债务的很大一部分。

So can you tell me a little bit more about the big contracts with OpenAI, Miles, that everyone seems to be talking about that represents a substantial chunk of this this Oracle debt?

是的。

Yeah.

真正推动这一切的是，甲骨文与OpenAI签订了一份为期五年、价值约3000亿美元的云服务合同。

The thing that's really driving this is, of course, that Oracle signed this five year, roughly $300,000,000,000 contract with OpenAI for cloud services.

因此，甲骨文四处收购所有正在建设的、功率超过一吉瓦的数据中心。

And, basically, as a result of that, we've seen Oracle go around and sort of snap up any, like, gigawatt plus data center that's being built.

而这一切现在都打上了OpenAI今年年初在白宫提出的“Stargate”品牌标识。

And this is all now being branded under the kind of Stargate branding that OpenAI laid out in the White House at the beginning of the year.

你知道，甲骨文在云服务领域的规模小于微软、谷歌和亚马逊。

So, you know, Oracle is a smaller cloud player than Microsoft, Google, and Amazon.

而且，历史上，它比这些公司更依赖外部的数据中心开发商。

And, historically, it's relied a lot more, than those companies on outside data center developers.

我们看到，即使在努力拓展与OpenAI的合同之际，它仍在继续这一策略。

And we're seeing it basically continue that playbook, even as it's trying to really grow into this OpenAI contract.

所以，我想知道，米尔斯，所有这些借贷和甲骨文通过直接和间接方式所承担的债务，究竟存在什么主要风险或担忧？

So I'm wondering, Miles, what exactly is the big fear or the big risk around all of this borrowing and all this debt that Oracle has been taking on through, you know, both directly and indirectly?

是的。

Yeah.

现在还为时过早，但活跃在项目融资市场的贷款方已经开始担心，甲骨文在市场中占据了巨大份额，是每月流入市场项目规模的重要组成部分。

It's, it's still early days, but, basically, lenders who are active in this project finance market are starting to get worried that basically Oracle is a huge component, a huge, portion of the market, and of what is sort of, like, coming to market, the kind of flow that they're seeing month to month.

这些贷款方虽然受到甲骨文与开发商签订的租赁合同中某些条款的保护，但仍对过度集中于任何一家公司——尤其是甲骨文——感到警惕，因为甲骨文极度依赖OpenAI，而众所周知，OpenAI预计要到2030年才能实现盈利。

And, you know, these lenders, even though they are sort of protected by some of the contractual terms, around the lease that Oracle signs with these developers, they're still leery about getting overexposed to any one company and in particular Oracle because it's so reliant on, OpenAI, which as we all know, isn't projecting, that it'll be profitable until 2030.

在这个特定市场中，除了甲骨文的债务外，还包括哪些其他内容？

Within this particular market, what else is included besides Oracle's debt?

我的意思是，还有哪些其他类型的项目正在获得融资？

I mean, what what other sorts of projects are being financed?

是的。

Yeah.

我的意思是，我们看到其他科技公司越来越多地使用第三方数据中心开发商，而这些开发商也会以类似的方式为Oracle融资来筹集债务。

I mean, we're seeing, other tech companies, use third party data center developers more and more, and then those developers will go out and raise debt in a similar fashion to the way they're doing it for Oracle.

Core Web 也是一个重要的参与者。

Core Web is also a pretty big player here.

他们的一家数据中心合作伙伴本月筹集了超过23亿美元的债务。

One of their data center partners recently raised more than $2,300,000,000 of debt this month.

因此，这是一个非常活跃的市场，2025年的活跃程度前所未有。

So it's a very active market, more so in 2025 than ever before.

我认为，到2026年，我们还会继续看到交易量持续增长。

And I think we're just gonna continue to see increased volumes going into 2026.

读过你的报道后，Miles，我觉得你花了很多时间思考系统性风险的问题。

Having read your reporting, Miles, I think you spend a lot of time sort of thinking about questions about systemic risk.

我想知道，你对这个市场更广泛的潜在风险有什么看法？

And I was wondering what your perspective was on sort of the broader risks to the market.

随着数据中心建设的快速推进，我们看到各大公司都在拼命加快建设进度，尽快将数据中心投入使用。

As we see this data center build out progress really quickly, we see companies trying to build data centers and scramble to get them out as quickly as possible.

是的。

Yeah.

我的意思是，我认为

I mean, I think

值得关注的是，那些在这些交易中承担风险的银行。

the thing to watch for is the banks that are taking on the risk in these deals.

这些交易规模变得如此庞大，银行能否将债务成功 syndicate 给众多其他银行和资产管理机构呢？

Are they able to get the debt syndicated to a wide array of banks and other asset managers just because these deals are getting so huge.

你知道，高达380亿美元。

You know, $38,000,000,000.

这么大的金额，仅靠几家银行是难以消化的。

That's a lot for just a few banks to swallow.

而在这种更广泛的银团贷款中，保险公司基金又扮演着怎样的角色呢？

And then in that sort of broader syndication, you know, what kind of role are insurance funds playing?

你知道，我们曾报道过，由阿波罗、KKR拥有的寿险基金正开始在数据中心融资中扮演越来越重要的角色。

You know, we've reported on how life insurance funds, owned by Apollo, KKR, how these are starting to play a bigger role in data center financing.

我认为，如果情况出现恶化，这绝对是需要关注的领域。

I think, that's definitely an area to watch if, things go sideways.

你知道，这些保险保单持有人会受到怎样的影响？

You know, how how are these, insurance policyholders affected?

所以，这些是我会关注的几个要点。

So those are a few sort of points I would I would follow.

如果你是甲骨文的投资者，有没有哪些具体的问题是你应该留意的？

And if you're an Oracle investor specifically, are there any any specific concerns that you should be watching for?

是的。

Yeah.

我的意思是，甲骨文极度依赖第三方开发商去外部融资，如果市场开始堵塞，甲骨文不得不自己承担更多为建设数据中心融资的负担，我们知道它的资产负债表已经非常紧张了。

I mean, because Oracle is so reliant on these third party developers to go out and raise debt, you know, if the market starts clogging up and Oracle has to bear more of the burden of raising the debt to build data centers, You know, we already know that its balance sheet is pretty stretched.

它的信用评级目前正被标普和穆迪列入观察名单。

It's on credit rating watch from S and P and Moody's.

它自身继续借款的能力是有限的。

It has a sort of limited ability to continue borrowing even more on its own.

因此，如果这个市场出现严重堵塞，可能会对甲骨文构成潜在问题。

And so if this market were to clog up in a really material way, that could potentially be problematic for Oracle.

听起来投资者需要了解很多风险。

Sounds like there's a lot of risks that investors need to be apprised of.

非常感谢你前来讨论这些问题。

Really appreciate you coming on to talk about them.

非常感谢你，迈尔斯。

Thanks so much, Miles.

谢谢你，阿妮塔。

Thanks, Anita.

随着年末临近，我们正在审视IPO市场以及2026年的发展前景。

As we head into year end, we're taking a look at the IPO market and what's ahead in 2026.

现在加入我们的是芬威克律师事务所资本市场与上市公司业务的合伙人兼联席主席罗恩·本瑟。

Joining me now is Ron Benser, partner and co chair of Capital Markets and Public Companies at Fenwick.

罗恩，很高兴你能来参加，因为政府停摆结束后，你一定在忙着帮助公司准备IPO，对吧？

Ron, glad that you're here to join us because now that the government shutdown has ended, you must be extra busy helping companies prep for IPOs, right?

谢谢你邀请我，阿妮塔。

Thanks for having me, Anita.

是的，这真得很有趣。

Yeah, it's been really interesting.

今年秋季，IPO市场原本势头强劲。

So we obviously had a ton of momentum heading into the fall this year in terms of the IPO markets.

但过去一个半月左右，由于政府停摆，市场明显放缓了。

That was obviously slowed down over the last month and a half or so with the government shut down.

现在政府工作人员已经返岗。

The government's back in the office now.

他们明确表示，还有大量待处理的申报文件积压着。

They are very clear that they have a long backlog of filings to still go through.

因此，我们看到许多公司都出现了显著的延迟。

So we are seeing significant delays for a number of companies.

好消息是，政府正在优先处理那些在停摆前已接近上市的公司。

The good news is that the government is prioritizing companies that were very close to going public before the shutdown.

因此，我们看到他们正在努力处理这些公司，帮助它们尽快启动路演。

And so we are seeing them move to clear those companies and get them out on the road.

我们希望他们能在未来几周内清理完积压的文件，让市场在明年一月恢复正常运作。

And we're hoping that they are able to clear the backlog here over the next few weeks so that the markets are up and running in January.

罗恩，这个积压具体是什么情况？

Ron, what exactly does that backlog look like?

我的意思是，我们说的是多少份申请？

I mean, how many filings are we talking?

这个待处理名单里都有哪些类型的公司？

What sorts of companies are in that pipeline?

是的，他们曾公开表示，在停摆期间提交了大约900份注册声明。

Yeah, mean, they've been on record to say that they had about 900 registration statements that were filed during the shutdown.

因此，他们确实有一大堆积压文件需要处理。

So they do have a significant backlog to clear through.

其中很多显然是科技公司，我们正在处理多个这样的IPO项目，但你知道，从传统IPO流程中常见的注册声明，到现有上市公司必须提交并由监管机构审查的注册文件，各种类型都有。

A number of those obviously are technology companies and we're working on a number of those IPOs but you know, it's everything from you know, the registration statements that you would see for a traditional IPO process to registration statements that existing public companies have had to file and that they are required to review as part of their mandate.

你现在能说出任何正在处理的公司名称吗？

Can you name any of the ones that you're working on at this point?

很遗憾，不能透露，因为其中一些已经公开提交并处于静默期，但我们在芬威克今年的IPO业务表现非常出色。

Unfortunately, can't because a few of them are publicly filed and are in the quiet period but we've had a fantastic year at Fenwick in terms of IPO work.

今年我们参与了CoreWeave和Figma的IPO，这两家都是今年最大的风险投资支持的IPO项目。

This year we worked on the CoreWeave IPO and on Figma which were two of the largest venture backed IPOs this year.

明白了。

Got it.

我想展望一下2026年。

So I want to look ahead to 2026.

每年这个时候，大家都喜欢谈论他们的IPO预测。

You know, everyone talks about their IPO predictions around this time of year.

你认为2026年会是科技公司IPO的大年吗？

Do you think 2026 is going to be a big year for tech IPOs?

如果是的话，为什么？

And if so, why?

是的，我的意思是，我们确实看到许多公司主动联系我们，开始考虑上市准备。

Yeah, I mean, you know, I think we're certainly seeing a number of companies approach us and start to think about IPO readiness.

我认为有很多公司积压着，它们显然在2020年、2021年就筹集了大量资金。

I think there's a lot of backlog of companies that had obviously raised significant capital, even kind of going back to 2020, 2021.

在过去几年里，它们一直在调整自己的商业模式。

And they've spent the last few years kind of retooling their business models.

再加上我们目前看到的AI领域的发展，以及AI被融入它们的商业模式中。

And then obviously with what we're seeing in terms of developments in AI and incorporating AI into their business models.

我认为公司们对自己讲好一个精彩的AI故事越来越有信心。

I think companies are increasingly confident around their ability to tell a really nice AI story.

我认为这预示着2026年上市市场前景良好。

And I think that bodes well for the IPO market heading into 2020, into 2026.

我还觉得，不幸的是，由于政府停摆，许多原本因我们之前提到的积压问题而被推迟的公司，将在明年年初集中上市。

I also think, unfortunately, again, with the government shutdown, you will have you will see a number of companies that were just kind of delayed as a result of the, you know, this backlog that we were we were talking about that will go out early next year.

但这至少应该为我们进入2026年及全年奠定一些势头。

But that should, you know, at least kind of set us up with some momentum heading into 2026 and throughout the year.

罗恩，从你观点的反面来看，我们长期以来也观察到一个趋势：许多公司，尤其是科技公司，选择长期保持私有状态。

Sort of on the flip side to your point, Ron, I mean, we've also been seeing the trend for a while now that many companies, especially in tech are choosing to stay private for longer.

上周《信息报》独家报道，Databricks 尽管已在私募市场融资约150亿美元，目前正洽谈新一轮融资。

We had a scoop at the information last week that Databricks, for example, which has already raised about 15,000,000,000 in the private markets, is in talks to raise another round.

我想知道，你认为是什么真正推动了这一趋势。

And I wanted to know what you thought was actually driving this trend.

是的，我们代表Databricks，我不能就他们具体的情况发表评论，但是

Yeah, we represent Databricks, I can't speak to them specifically, but

是的，你知道，是

Yeah, you know, is

这确实是一个我们看到的趋势：公司希望更长时间地保持私有状态。

it a is a trend that we've certainly seen companies are wanting to stay private longer.

你知道，新一届SEC管理层的保罗·阿特金斯一直积极倡导让IPO再次变得辉煌。

You know, one of the really interesting things with the new administration over at the SEC, Paul Atkins has been very vocal around making IPOs great again.

其中一部分是简化适用于上市公司的多项规则和法规，使公开市场最终更具吸引力。

And part of that is streamlining a number of the rules and regulations that apply to public companies to make the public markets ultimately more attractive.

我认为我们不会看到完全逆转的情况，即公司在生命周期早期就急于上市。

You know, I don't think that we will see a complete reversal and move back towards IPOing early in a company's life cycle.

我认为公司选择长期保持私有状态有多种原因，但作为代表那些致力于突破性技术和商业模式的创新科技公司的顾问，能够避开公众视线运作，至少尽可能长时间地这样做，往往是一种优势。

I think companies are staying private longer for a number of different reasons but you know, as someone who represents very innovative technology companies that are kind of working on groundbreaking technologies and business models, oftentimes it's a benefit to be able to do it outside of the the public limelight, at least for as long as you possibly can.

因此，我们看到公司正在利用在私募和公开市场筹集资金的能力，而且能筹集到大量资金。

And so we are seeing companies, you know, take advantage of the ability to raise capital and significant amounts of capital in the public private markets.

然后最终推迟它们进入公开市场的时间。

And and and then kind of delay ultimately their their entrance into the public markets.

我很想知道，你认为本届政府是否出台了任何具体的行政或政策变化，真正推动了IPO管道的放松？

I'd love to know if there are any specific administrative or policy shifts that you think have taken place in this administration that really will encourage the IPO pipeline to be loosened up?

是的，我认为我们正看到他们正在讨论的几项措施。

Yeah, I mean, I think we're seeing a number of things that they're talking about.

例如，新兴成长公司身份，这是2012年《创业企业促进法案》生效时引入的一项规定。

So for example, emerging growth company status, which is something that came into effect back in 2012 with the Jobs Act.

事实上，我认为这为鼓励企业上市提供了强大的推动力。

It was actually, I think a significant tailwind in terms of encouraging companies to go public.

它简化了企业必须在申报文件中包含的大量披露要求，最终我认为这促使企业决定上市。

It streamlined a lot of the disclosures that they were required to include in their filings, which ultimately I think encouraged companies to decide to list.

他们目前正在探讨如何延长新兴成长型公司身份的益处，使企业能够更长时间地保持简化披露制度。

They are talking about now ways that they can extend the benefits of emerging growth company status so that companies can stay in a scaled disclosure regime for longer periods of time.

所以我认为，这是他们正在做的一件事；此外，他们还在考虑明年各种不同的提案，比如简化申报状态、股东提案，这最近在证交会引发了广泛讨论，同时也在探索帮助企业降低上市相关成本的各种方式，这些都属于本届政府为激励企业上市而正在考虑的议题。

So I think that is a, you know, one of the things that they're doing, they're also kind of looking at a variety of different proposals next year in terms of streaming filing statuses, shareholder proposals has been a big topic of of discussion recently at the SEC and and figuring out kind of ways for companies to be able to reduce some of the costs that that are associated with being being public are all things that are sort of on the table for this administration in terms of how they're trying to kind of spur companies to to go public.

展望2026年，你认为哪些类型的科技公司会率先上市？

As you're looking ahead to 2026, what kinds of tech companies specifically do you think are going to be going first?

是的，我认为我们普遍看到的是，要想成为成功的上市科技公司，你必须讲好一个出色的AI故事。

Yeah, I think, you know, one thing that we're seeing across the board is you have to tell a really good AI story in order to be a successful public technology company.

我们已经看到一些基础设施公司开始上市，也就是这个领域的核心企业。

You know, we have already started to see some of the infrastructure companies go public, so the core waves of the world.

显然，接下来会很有趣的是大型语言模型（LLM）最终会如何发展，以及它们是近期上市还是长期等待。

It's gonna be interesting to see obviously the LLMs and what what, you know, they're gonna ultimately do and whether they are near term IPOs or or longer term.

但我觉得，真正有趣的是，在2020年、2021年及之前，最主流的IPO公司其实是软件公司，而如今，许多这类公司仍在努力构建自己的AI故事。

But, you know, I think really the interesting thing to me is you know, in twenty twenty, twenty twenty one and before that, the bread and butter IPOs really like the software companies that we were taking public and you know, I think a number of those companies have still tried to figure out their AI story.

Figma就是一个很好的例子，它不仅开始将AI技术融入其平台，还能讲清楚为什么AI会成为推动其业务发展的助力，而非阻力——毕竟他们多年来在平台上展示了诸多创新。

I think Figma's a great example of a company that was able to both start to kind of incorporate AI technology into their into their platform and then be able to tell a story of why, AI is gonna be a tailwind for the their business and not a not a headwind given all of the innovations that they've been able to show to to their platform over time.

我认为，这为其他希望上市的企业软件公司提供了一个很好的范例：既能讲好AI故事，又能清晰说明AI如何影响其商业模式和成本结构。

And I think that is a really good model ultimately for other enterprise software companies that that want to go public being able to tell, tell an AI story and then kind of explain how that, that AI story ultimately impacts the business model and the cost structure.

我认为，除了AI之外，另一个可能的差异化因素，至少是我注意到的，是加密货币在当前IPO市场中扮演了积极角色。

I think in addition to AI, maybe one other differentiator, at least that I've noticed is crypto being a positive in this IPO market.

我们看到Circle成功上市，据传Kraken、BitGo等几家加密公司也正在筹备上市。

I mean, we saw Circle go public and then we see in the pipeline, reportedly Kraken, BitGo, a couple of other crypto firms are looking at it.

但如果你回过头看，还有一些近期上市的科技公司，比如Navon和StubHub，它们的股价表现似乎并不如IPO时那么理想。

But then if you look back, there are some other tech companies that recently took the leap like Navon, StubHub that maybe aren't doing so well relative to their IPO prices.

我想知道，你能否谈谈，是什么原因导致了这些不同类型科技公司之间的差异？

And just wondering if you could give me your thoughts on what explains the difference between these different types of tech companies.

当然，不同业务模式差异很大，加密公司和传统企业软件公司的样子截然不同。

Yeah, obviously kind of different business, the different business, well, crypto company looks a lot different from, you know, traditional enterprise software company.

我想说，就表现而言，这可能是政府停摆带来的另一个影响。

I will say in terms of performance, and this was another, I think, impact of the government shutdown.

我的意思是，进入秋季时，股市似乎有很强的上升势头。

I mean heading into the fall, I think there was a ton of momentum in the stock market.

如果你现在把IPO价格与许多IPO当前的交易价格相比，我们会看到回调，这显然会影响2026年的股市和IPO市场。

If you're now looking at IPO prices relative to where a lot of the IPOs are currently trading, we have seen pullback and that will obviously impact the stock market and the IPO market into 2026.

很难确切判断这是否只是年末现象——投资者在获得不错收益后，倾向于在年末采取一些避险策略，然后展望2026年。

Hard to really kind of tell if that's sort of a end of the year phenomenon with investors essentially taking a little bit of risk off a risk risk off kind of approach towards towards year end after having some nice gains and then and then kind of looking forward to to 2026.

看来2026年IPO市场会让咱们俩都忙得不可开交。

Well, it certainly sounds like the IPO market is going to be keeping both of us pretty busy in 2026.

但愿如此。

I hope so.

好的。

All right.

非常感谢你今天来参加节目，罗恩。

Well, thanks so much for coming on the show today, Ron.

很高兴有你。

Great to have you.

谢谢。

Thank you.

很高兴认识你。

Very nice to meet you.

我们正在密切关注公司如何利用人工智能自动化白领工作。

We're keeping a close eye on how companies are using AI to automate white collar work.

投资银行以繁重的体力劳动和长时间工作而闻名。

Investment banking is notorious for its grunt work and its long hours.

因此，为了讨论金融服务业如何普遍使用人工智能，我们邀请了ModelML的创始人兼首席执行官查斯·英格兰纳，这家初创公司刚刚筹集了7500万美元，用于自动化其中一些任务。

So to talk about how financial services companies across the board are using AI, we're bringing on Chas Englander, founder and CEO of ModelML, a startup that just raised $75,000,000 to automate some of those tasks.

查斯，欢迎来到我们的节目。

Chas, great to have you on the show.

谢谢你们邀请我。

Thanks for having me.

谢谢。

Appreciate it.

让我们先来考虑一位投资银行家的典型一天。

So let's start by considering the typical day in the life of an investment banker.

投资银行家在日常工作中会做哪些任务，这些任务哪些容易被AI自动化，哪些难以自动化？

What kinds of tasks would an investment banker be doing on a day to day basis that are easy and hard to automate using AI?

我们对这个问题的看法是，从底层来看，我们构建的是三种主要的基于代理的系统。

Well, the way we think about it really is, you know, under the hood and what we build is kind of three main agent based systems.

其中一种负责数据收集。

There's one that does the data gathering.

另一种负责我们所说的材料生成，还有一种负责验证。

There's one that does what we call the material creation and one that does the verification.

对我们而言，尤其是材料生成这一环节，我们认为完全可以实现自动化。

So for us, particularly the element of material creation, we think is entirely automatable.

因此，我们设想的用户是那些在工作中材料生成占比最高的那些人。

So the way that we think about our users are the ones that are doing the most material creation by way of percentage of their work.

现在，大部分工作都已经可以完全自动化了。

Most of that work now is fully automatable.

当我提到材料创建时，我指的是PowerPoint、Word、Excel等。

And when I say material creation, I mean PowerPoint, Word, Excel, etcetera.

明白了。

Got it.

你们刚刚融资了7500万美元，这可是笔巨额融资，尤其是在金融科技领域的A轮融资中。

So you guys just raised $75,000,000 I mean, that's a really big fundraising haul, especially for a series A in FinTech.

你们计划具体如何使用这笔资金？

What exactly are you planning to use the capital for?

我们目前的主要瓶颈是客户入驻。

Our main bottleneck right now is onboarding.

主要在旧金山、纽约、伦敦和新加坡，我们根本没有足够的能力来完成客户入驻。

So across predominantly across San Francisco, New York, London, and Singapore, we just do not have the capacity to onboard customers.

所以我们首先要解决的是，需要为入驻团队招聘更多人，接下来是工程和产品团队，最后才是销售团队。

So the initial thing that we need to solve for is we need to hire more folks in our onboarding teams, then it'll be engineering and product, and then eventually it will be sales.

目前，我们只有一名销售人员。

Currently, we only have one salesperson.

明白了。

Got it.

关于你们的收入增长和营收，你们的增长速度有多快？

In terms of your revenue growth and your top line, I mean, how fast are you growing?

你能让我了解一下目前的业务进展如何吗？

Can you give me a sense of what the traction has been like?

我们在大约三个月内将业务规模扩大了十倍。

We've 10x ed the business in about three months.

是指销售方面吗？

In terms of sales?

是的，销售方面。

Sales, yep.

明白了。

Got it.

我想知道，查兹，你们有没有发现任何客户因为使用你们的产品而削减了初级岗位的人手？

And I'm wondering, Chaz, you know, are you seeing any of your customers actually say that they're cutting headcount at the junior level, for example, as a result of using your product?

目前还没有。

Currently, no.

事实上，我们看到一些初级员工，比如分析师和助理，他们的职责发生了轻微变化。

In fact, what we are seeing is some of the more junior folks, analysts and associates, their roles slightly changing.

实际上，这些人正在转向更偏向AI工作流类型的角色。

So actually people that are being moved across to very much AI workflow type roles.

我应该说，我们是一款工作流自动化产品。

I should say we're workflow automation product.

对吧？

Right?

我们不是像ChatGPT或Anthropic那样的通用研究型工具。

So we're not a generalist research type tool like a ChatGPT or an Anthropic.

你使用我们的产品，大约需要45分钟的培训，然后就可以构建完全自动化的流程。

You you come into our product, you need about 45 of training, and then you can build fully automated workflows.