本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
欢迎各位收看Information的TI TV。
Welcome everyone to the Information's TI TV.
我叫阿卡什·帕什拉查。
My name is Akash Pashracha.
今天是12月12日，星期五。
It is Friday, December 12.
今天我们为大家准备了一场精彩的节目。
We have got a great show lined up for you today.
首先，我们将深入探讨OpenAI的GPT 5.2，为此我们采访了曾率先接触该模型的哈维研究员。
First up, we are diving into OpenAI's GPT 5.2 with a researcher at Harvey who had early access to the model.
接着，我们会分析博通的财报，同时，Wealthfront今天也将迎来其IPO。
We'll then dive into Broadcom's earnings, and separately, it is IPO day for Wealthfront.
我几分钟后将采访其首席执行官。
I'm talking to the CEO in just a few minutes.
我还将与Rubrik的首席执行官对话，并带来另一期《编辑精选》。
I'm also speaking with the CEO of Rubrik, and we've got another edition of the Editor's Cut.
我们将与两位首席编辑讨论拉里·埃里森本周大胆押注的疯狂情况。
We're going to be talking with two of our top editors about how crazy a week this was for Larry Ellison making big bets.
最后，我们将带大家前往纽约AI峰会一探究竟。
Finally, we will end with a little trip to the New York AI Summit.
我与Nebius的联合创始人进行了交谈，讨论他快速增长的业务。
I sat down with the co founder of Nebius to discuss his fast growing business.
这是一期内容丰富的节目。
It is a big show.
那么，我们马上进入正题。
So let's get right on into things.
OpenAI发布了GPT-5的新版本。
OpenAI has released a new version of GPT-five.
它被称为GPT-5.2，考虑到该公司在不到两周前因谷歌竞争加剧而发出红色警报，如今任何OpenAI的发布都无疑引人注目。
It's called GPT-five point two, and any kind of release from OpenAI these days is certainly notable given the code red that the company called less than two weeks ago as competition from Google ramps up.
现在加入我为大家深入解析这一切并分享他对该模型看法的是哈维公司应用研究负责人尼科·格鲁彭。
Joining me now to break this all down and hear his thoughts on the model is Nico Gruppen, Head of Applied Research at Harvey.
尼科，欢迎来到节目。
Nico, welcome to the show.
很高兴你来参加。
It's great to have you here.
谢谢你们邀请我。
Thanks for having me.
那么，你觉得
So what did
你对它有什么看法？
you think of it?
5.2，给我们做个评价吧。
5.2, give us the review.
是的，我们确实提前获得了GPD 5.2的访问权限。
Yeah, so we did indeed have early access to GPD 5.2.
在评估任何新模型时，我们通常会问三个问题。
When it comes to evaluating any new model, we're really going to ask three questions.
这个模型固有的推理能力是什么？
What are the sort of innate reasoning capabilities of the model?
我们有自己的内部基准测试。
We have our own internal benchmark.
它叫做Big Law Bench。
It's called the Big Law Bench.
它衡量大语言模型完成实际法律任务的能力。
It measures the ability of LLMs to complete real world legal tasks.
第二个问题是，这个模型在产品中的表现如何？
The second question is, how does the model perform in the product?
我们会将模型集成到产品中，并在我们关键的产品功能上进行端到端的评估。
We're going to wire the model up in the product, and we're going to run end to end evals across our key product surfaces.
第三个问题是，律师们怎么看？
And then the third question is, what do the lawyers think?
这实际上是新模型评估过程中我最喜欢的部分，因为它是一种更非结构化的测试。
And this is actually my favorite part of the new model eval process, is it's more unstructured testing.
所以我们把模型交给我们的内部律师。
So we'll give the model to our internal lawyers.
他们真正关注的是法律工作等领域的显著改进。
What they're really probing for are step change improvements in things like legal work.
你把它给客户了吗？
You give it to the customers?
给我们的内部律师。
To our internal lawyers.
所以在Harvey，我们内部有一个法律研究团队。
So at Harvey, we have legal research team internally.
他们都是前大型律师事务所的律师。
They're all former big law attorneys.
明白了。
Got it.
好的。
Okay.
所以是三个维度。
So three vectors.
你有基准测试，还有律师。
You've got the benchmarking, you've got the lawyers.
那么人们是怎么看待它的呢？
So how are people thinking about it?
嗯。
Yeah.
这三个信号共同构成了我们对模型的整体看法，以及我们观察到的涌现行为。
So these three kind of signals coalesce into our general opinion of the model and sort of what emergent behaviors we've seen.
就整体表现而言，我们对这个模型的表现非常印象深刻。
In terms of overall performance, we're actually quite impressed with the model's performance.
在BigLawBench上，这是我们公开的基准测试。
So on BigLawBench, this is our public benchmark.
它是迄今为止我们见过的第二高分模型，因此它本质上是一个强大的法律推理工具。
It's the second highest scoring model we've seen to date, so it's inherently a strong legal reasoner.
但我认为，这个模型最让我们印象深刻的是它内在的防护机制。
But I think the thing that stood out to us most about this model in particular was the model's inherent sort of guardrails.
这一点在我们的非结构化测试阶段显现出来，我们注意到模型展现出一种内在的涌现行为，即具备更强的自我能力认知。
So this came out in our unstructured testing phase, we noticed an inherent sort of emergent behavior where the model displayed stronger capability awareness.
因此，它能够自问：我是否有足够的信息来回答这个问题？
So it's able to ask itself things like, Do I have enough information to answer this question?
我能做什么？
What can I do?
我不能做什么？
What can't I do?
我是否应该引入人类参与？
Should I bring a human into the loop?
这类行为非常明显且突出，实际上对我们的平台极为有益。
These sort of behaviors were pretty prominent and noticeable, and are actually extremely beneficial for our platform.
在基准测试中排名第一的是谁？
Who was number one on the benchmark?
所以这实际上是GPT-5.1。
So that is actually GPT-five point one.
对吧？
Right?
所以我们看到的是每个实验室的前沿模型，比如GPT-5.1、GPT-5.2、SONNET-4.0.5，当然还有Gemini 3 Pro，它们在我们的基准测试中的表现正趋同于同一个平均水平，这就是我们所说的前沿模型群体。
So what we see is the frontier models from each lab, so GPT-five point one, GPT-five point two, SONNET-four 0.5, and of course Gemini three Pro, they're converging to the same sort of mean or average level of performance on our benchmark, and that's what we consider sort of the frontier cohort.
我该如何理解这些模型公司在这方面的权衡？他们如何决定这只是一个零点几的提升，比如从5.1到5.2，还是说这已经是一个全新的模型？
How do I understand the calculus here for these model companies in terms of when they decide that this is a decimal point increase, say 5.1, a 5.2 increase, versus, Hey, this is a brand new model.
这里的差异究竟体现在哪里？
What do the deltas here look like?
他们又是如何判断，哦，这个进步足够大，值得推出一个全新的模型？
And when do they decide that, Oh, this is enough of a big step that we should launch a new model altogether?
是的，这是一个极其具有挑战性的问题。
Yeah, it's an incredibly challenging problem.
我认为模型提供商们是通过他们能构建的基准测试来间接推进的，对吧？
I think the model providers are working by proxy on the benchmarks that they can create, right?
而这里最大的挑战是，你能否构建出能够代表这些模型在现实世界中使用方式的基准测试？
And the big challenge there is, can you build benchmarks that are representative of the ways in which these models will be used in the real world?
对吧？
Right?
因此，模型提供商会在这些基准测试上不断优化，逐步提升性能。
And so the model providers will make progress, they'll hill climb against these benchmarks.
当他们看到显著的改进后，就会将模型提供给我们这样的用户，并征求我们的反馈。
When they've seen meaningful improvements, they'll give them to model consumers like ourselves, and ask us for feedback.
对吧？
Right?
因此，我们重点关注的是行为上的跃迁式变化和涌现行为，这是一个反馈循环。
And so a big thing we're looking for are sort of these step changes in behavior, emergent behaviors, and it's sort of a feedback cycle.
所以你总体上对这个模型印象深刻。
So you were overall impressed with the model.
我想知道，当你谈到反馈时，你给OpenAI提供了哪些反馈，说‘这个还需要改进’？
I wonder, when you talk about feedback, what feedback did you give to OpenAI in saying, Hey, this could use some work?
是的。
Yeah.
所以，我所说的最有力的反馈主要是关于模型的防护机制。
So, I mean, the strongest piece of feedback we gave was certainly around the model guardrails.
对吧？
Right?
这并不一定是负面反馈。
This isn't necessarily negative feedback.
我们实际上对它的表现印象深刻，但这是一种我们此前在GPD前辈模型中未曾见过的涌现行为。
We were actually impressed with this performance, but it was an emergent behavior that we hadn't seen from the GPD predecessors.
就改进空间而言，我认为我们发现的问题是，模型在风格和格式上有些僵化。
In terms of room for improvement, I think the things that we identified were the model is a bit rigid in terms of styling, formatting.
它的权重中内置了自己的一些偏好，因此通过提示来引导这些变化比较困难。
It has its own sort of preferences baked into its weights there, and it's harder to coax out these changes in prompting.
至于法律特定应用场景，我们确实发现模型在交易类法律工作上的表现明显优于诉讼类工作。
And then with legal specific use cases, you know, we did see the model perform notably better on transactional legal work than on litigation work.
所以这些就是我们提供的反馈信号。
So these are the kind of feedback signals we're giving.
我想再回到你之前说的，5.1 在你的基准测试中得分最高。
And I just want to go back to, again, where you said 5.1 was highest on your benchmark.
所以根据你对 5.1 和 5.2 与 Gemini 3 的比较，Gemini 3 当时引起了很大关注。
So based on your comparisons of five point one and five point two against Gemini three, Gemini three is the one that made a lot of noise.
你现在把 5.1 和 5.2 放在了 Gemini 3 前面。
You're putting five point one and five point two ahead of Gemini three right now.
我这样理解对吗？
Am I understanding that correctly?
对。
Yeah.
我们这里讨论的只是零点几分的差距。
And what we're talking about here are decimal points of difference.
对吧？
Right?
所以目前的排行榜是5.1、5.2、SONNA四十五和Gemini三Pro。
So the current leaderboard is 5.1, 5.2, SONNA four five, and Gemini three Pro.
再次强调，这些差异都属于微小的均值差异。
Again, these are within marginal mean kind of differences.
对我们来说更重要的不是平均分，而是我们在法律工作中最关注的子任务中，哪些地方存在明显的短板。
The thing that's more important for us is not the mean score, but where are the sharp edges for the subtasks of legal work that we're most interested in.
我想问你一个问题，关于这种争相开发最佳模型、寻找最强模型的趋势。
Let me ask you a question here about this trend that there's this race to develop the best model and find which model is the strongest.
同时也有讨论认为，人们并不总是需要最好的模型。
There's also a discussion saying, Hey, people might not need the best model all the time.
我们听到关于开源模型的讨论，说我们只需要根据自己的需求定制模型即可。
We hear this discussion coming up with open source models that, Hey, we just want to customize the model for what we need it for.
你在这条光谱上的立场是怎样的？是应该专注于打造最佳模型，还是应该专注于开发一个刚好满足公司需求的定制模型？
Where do you sort of land on this spectrum of, We should be focused on the best model, or We should be focused on making a custom model that is just good enough for what our company needs?
是的。
Yeah.
你看，整个AI生态系统正在以超快的速度发展。
Look, the framing here is entire AI ecosystem is moving at warp speed.
对吧？
Right?
我们昨天刚发布了GBD 5.2，Gemini 3 Pro，而GBD 5.1才一个月前，Sonnet 4.5更是两个月前的事了。
We had GBD 5.2 yesterday, Gemini three Pro, and and GBD 5.1 just a month ago, and and Sonnet 4.5 about two months before that.
所以事情真的、真的进展得非常快。
So so things are moving really, really fast.
在应用层面上，我们也当然熟悉这种紧迫感，以及推动紧迫感所需的东西。
At the application layer, we're, of course, also familiar with urgency and and sometimes what it takes to drive urgency.
我们优先考虑的事项是模型性能、生产就绪度和客户偏好。
The things that we are prioritizing above everything else are model performance, production readiness, and and client preferences.
因此，模型性能以我们刚刚提到的基准来衡量，生产就绪度也会产生影响，以及我们在质量上的权衡。
So model performance as measured by the benchmarks we just mentioned, things like production readiness do come into play and our trade offs with quality.
对吧？
Right?
因此，安全性、隐私性、区域可用性、延迟和可靠性等因素都会影响我们最终投入生产的模型。
So things like security and privacy, regional availability, latency reliability, all of these things do factor into the model we ultimately end up shipping to production.
很好。
Great.
尼科，感谢你参加本期节目。
Well, Niko, thanks for coming on the show.
很高兴听到你的见解，我们很快再和你交流。
It's great to have your review, and we will talk to you again very soon.
谢谢。
Thank you.
好的。
Okay.
博通昨晚公布了财报。
Broadcom reported earnings last night.
收入增长加速至28%。
Revenue growth accelerated to 28%.
自由现金流增长了36%。
Free cash flow jumped 36%.
但这还不足以激励投资者。
It wasn't enough to encourage investors.
昨天盘后交易中，股价最初有所上涨。
The stock initially moved up in after hours trading yesterday.
但在财报电话会议期间，这些涨幅被全部抹去，股价转为下跌，今天开盘时为负。
It then erased those gains during the earnings call and swung negative, and it opened in the red this morning.
为了全面分析，我想邀请Theory Ventures的普通合伙人托马斯·通努斯。
To break it all down, I want to bring on Tomas Tungus, General Partner at Theory Ventures.
托马斯，欢迎再次做客节目。
Tomas, welcome back to the show.
很高兴你能来。
It's great to have you here.
很高兴能参加
Great to be
在这儿，阿卡什。
here, Akash.
谢谢您邀请我参加。
Thanks for having me on.
您对这些结果有什么看法？
So what did you think of the results?
我认为，目前AI生态系统正处于悬崖边缘，任何被感知到的弱点——无论是真实还是虚构，或者任何预期的满足，都会导致显著的股价下跌。
I think the AI ecosystem is on a knife edge right now where any perceived weakness, real or not, or any satisfaction of expectations leads to pretty significant drawdowns.
我的意思是，看看博通相对于英伟达的整体增长，它的增速并不快，尽管它是一家比英伟达更侧重服务的公司。
I mean, you look at the overall growth of Broadcom relative to NVIDIA, it's not growing as fast, although it is much more of a services business than NVIDIA.
你有谷歌的TPU业务，特别是与Anthropic的合作，似乎增长得相当迅速。
You have the Google TPU business, particularly with Anthropic, that seems to be growing pretty quickly.
Meta是一个非常重要的客户，支出巨大，大幅增加了资本支出，代价是牺牲了其虚拟现实生态系统。
Meta is a very significant customer spending a lot, massively increasing their CapEx spend at the expense of their virtual reality ecosystem here.
因此，你拥有一个非常健康的企业，利润率高，经营活动现金流强劲，大约在25%到27%之间。
And so you have a very, very healthy business, great margins, great cash flow from operations, approximately between 25% to 27%.
你有大量的未完成订单。
You have a pretty significant backlog.
我认为唯一可以挑刺的地方是，一些剩余履约义务（RPO）可能比分析师预期的要延迟一些。
And I think the only thing that you could pick at the overall earnings statement was that some of that RPO might be a little bit of remaining performance obligations might be a little bit more delayed relative to where analysts' expectations had had had thought they would come in.
我的意思是，咱们直说吧。
I mean, look, let's call it what it is.
我觉得这些数字不错。
I thought the numbers were good.
懂吗？
Know?
而且，你的收入正在加速增长。
And I mean, you have accelerating revenue.
你有自由现金流。
You have free cash flow.
他们谈到了使用XPUs的大客户，比如苹果、SSI、Anthropic、谷歌，还有一个在电话中不愿透露名称的第五大客户，但显然他们与OpenAir在某种程度上有所合作，不管这个第五大客户是不是它。
They talked about the big customers they have using the XPUs, Apple, SSI, Anthropic, Google, a fifth customer that they didn't want to name on the call, but they obviously have been working with OpenAir in some capacity, whether or not it's that fifth big customer.
今天早上我关注的是博通相对于英伟达的估值。
The thing that I was looking at this morning was the valuation of Broadcom relative to NVIDIA.
根据收入预测，博通的股价实际上更高，我认为这可能就是人们想表达的意思。
And based on the revenue estimates, Broadcom is actually trading higher, which I think might be what people were getting at.
我觉得这是对的。
I think that's right.
我的意思是，市盈率非常高。
I mean, have very elevated multiples.
而且我们知道，从所有之前的硬件周期来看，无论是电信还是硬盘硬件，一旦出现任何失误，都会导致大幅下跌。
And I mean, we know from all the previous hardware cycles, whether it's telecom or also hardware hard disks, that any miss here, you have a very significant drawdown.
而且这里的资本支出非常剧烈。
And because of the CapEx here, so intense.
所以，如果出现任何微弱的迹象，或者市场 perceived 到未来可能的疲软——比如本周早些时候甲骨文的情况，其RPO规模巨大，但这些RPO的实现时间仍不明确。
So if there's any weakness whatsoever or any perceived sign of potential weakness in the future I mean, you saw it with Oracle earlier this week where the RPO was absolutely massive, but the timing of that RPO remains unclear.
人们最终真的会使用这些产品吗？
And will people actually end up using this?
我认为这是人们试图理解的主要因素。
I I think that's the major factor that people are trying to understand.
好吧。
Like, okay.
OpenAI 承诺投入特定金额。
OpenAI is committed to spending a particular amount.
谷歌也承诺投入特定金额。
Google is committed spending a particular amount.
资本支出正在增长。
CapEx is growing.
明年在数据中心建设方面的支出可能会接近美国 GDP 的 3% 或更多。
It'll probably be somewhere close to 3% or more of US GDP next year in terms of data center build out.
推理能力会到位吗？
Will the inference be there?
所有迹象都表明会的。
And all signs point to yes.
但如果这个计划有任何偏离的迹象，那么削减力度将会很大。
But if there's any indication that there's a deviation in that plan, then the drawdown will be significant.
所以我认为这就是为什么会有这种敏感性。
And so I think that's why you have this sensitivity.
我唯一能对市场做出的另一种解读是部分获利回吐。
The only other way I'd sort of interpret the market is a bit of profit taking.
今年股价已经大幅上涨，而年底之前通常会有一些获利回吐。
The stock has ripped this year, and there tends to be some profit taking before the end of the year.
所以这可能是另一个原因。
And so maybe that's another reason.
当然，这无疑会是一份很棒的圣诞礼物。
Well, it certainly would be a nice Christmas present for sure.
我想问你关于霍克·谭的事。
I want to ask you about Hawk Tan.
上周我们在《周末杂志》上刊登了一篇关于霍克的精彩专访，他是芯片界极具标志性的首席执行官。
We wrote this great profile about Hawk in our Weekend Magazine last week, and he's an iconic CEO in the world of chips.
我想知道硅谷对谭浩作为领导者的讨论是什么。
I wonder what the discussion is in Silicon Valley about Hawk Tan as a leader.
人们对他的领导风格以及他管理公司的方式有何看法？
What do people make of his leadership and the way that he's running the company?
是的。
Yeah.
我的意思是，我觉得他做得非常出色。
I mean, I think he's done a phenomenal job.
对吧？
Right?
你能看到那些核心的技术突破。
You see the the core technical advances.
我的意思是，博通在所谓的串行化和反串行化技术方面，是顶尖的技术之一。
I mean, Broadcom is some of the best technology to what's called a serialization and deserialization technology.
至于GPU上的速度，达到了每秒224吉比特。
And the speeds there on the GPUs, I mean, are 224 gigabit per second.
没人能接近他们的水平。
Nobody else can come close.
据说谷歌可能打算将这种芯片设计能力内部化，但考虑到博通相对于美满电子和新思科技的相对性能，这几乎不可能发生。
You know, there was a rumor that Google might be bringing some of this capacity in house for the chip design, but I think given the relative performance of Broadcom compared to Marvell and Synopsys, it's really unlikely.
所以我认为他确实做得非常出色，建立了一个相当卓越的业务。
So he's done I think he's done a really nice job building a pretty phenomenal business.
而且业务也很多元化，对吧？与英伟达不同，英伟达目前75%到80%的收入主要来自数据中心，而博通的AI业务年增长率约为75%，但仅占整体业务的30%左右。
It's also diversified, right, compared to NVIDIA, where 75% to 80% of the revenue now is primarily within data center, Broadcom has an AI business that's growing about 75% per year, but it represents something like 30% of the overall business.
因此，如果出现下滑，你仍然拥有优质的核心知识产权。
And so if there were to be a drawdown, you have this you have nice core intellectual property.
即使是超大规模云服务商也很难复制。
It's very difficult even for the hyperscalers to replicate.
你需要雇佣数千名工程师，投入大量资金，历时多年才能勉强接近，更何况还有一个多元化的业务布局。
You're talking about hiring thousands of people, pretty significant engineering investment over the course of multiple years to come anywhere close, and then a nicely diversified business.
所以，正如你所说，考虑到这样的业务表现，它的股价表现确实令人费解。
So to your point, it's kind of a head scratcher about the overall stock performance given this business.
昨晚在电话会议期间，我一直在和一些分析师发短信，如果你看一下股价走势，基本上在分析师问完第一个关于谷歌是否将产能内部化的问题后，
Well, I was texting with some analysts last night during the call, and if you look at the stock movement, basically after this first question that analysts asked about Google bringing the capacity in house, right?
投资者对这个回答并不满意，我说的是投资者。
That was the answer that they were not very impressed with, investors I'm talking about.
分析师们认为，他对他们在该领域优势的解释并没有他们期望的那么令人安心。
And analysts sort of said he didn't give as reassuring an answer as they would have hoped about their advantage there.
我想知道，你认为客户自研工具这一想法对博通的业务构成多大威胁？也就是说，他们能否自行设计自己的定制芯片。
I wonder how much of a threat you think that is to Broadcom's business, this idea of customer owned tooling, I guess, that they could design their own custom chips on their own in house.
这似乎是一项庞大的工程。
It seems like a big endeavor.
这是一项巨大的工程。
It's a huge endeavor.
这些系统极其复杂，涉及多个不同层面。
These are massively complex systems, and there are many different dimensions to it.
你知道，两三年前，就有传言说谷歌正在尝试内部构建这一能力。
You know, I think two or three years ago, there was the rumor that Google was trying to build this in house.
根据分析师的估计，谷歌在芯片设计上花费了大约30%到50%的预算。
Google pays something like 30 to 50% according to analyst estimates of the the bomb for chip design.
因此，你可以想象，能够做到这一点对他们来说将是一件非常重要的事，而且这在技术上极其困难。
And so you can imagine that this will be an important thing for them to be able to do at some point, and it's incredibly technically difficult.
在某些吞吐量方面，尤其是序列化和反序列化组件上，还没有人接近过，你可以想象GPU正在并行处理海量数据。
No one has come close on some of these throughputs on the particularly the serialization, deserialization component where you can imagine a GPU is processing a huge volume of data in parallel.
它必须将这些数据聚合起来，发送到下一个步骤。
It has to aggregate that to send it to the next step.
然后下一个步骤需要将其解聚合并重新并行化。
And then that next step has to take it and de aggregate it and parallelize it.
除了内存周边之外，这正是许多GPU系统的主要瓶颈。
That's the bottleneck for a lot of these GPUs systems aside from in and around memory.
因此，只要博通在这一领域拥有明显优势，我认为它们的业务依然非常出色。
And so as long as Broadcom has head and shoulders advantage there, then I think they still still have a phenomenal business.
但如果你是谷歌，试图在TPU的利润率上与博通竞争，而Meta等公司也在尝试，你会怎么做？
But if you're Google and you're trying to compete with them more margins on the TPUs and and Meta and others, will you be trying?
我认为你会尝试的。
I think you will be trying.
但我无法想象在未来两三年内会发生变化。
But I can't imagine this changes in the next two or three years.
很好。
Great.
好了，托马斯，和你交谈总是很愉快。
Well, Tomas, it's always a pleasure to talk with you.
谢谢你来参加节目。
Thank you for coming on the show.
我们非常感谢。
We really appreciate it.
我的荣幸也是我的。
My pleasure is mine.
谢谢你，阿卡什。
Thank you, Akash.
很快再聊。
Talk to you soon.
好的。
Okay.
财富前沿今天通过IPO上市，包括所有股票期权在内，公司估值为26亿美元。
Wealthfront is going public today in an IPO that values the company at $2,600,000,000 with all the stock options included.
该公司一直以机器人顾问和服务为基础开展业务。
The company has built its business on robo advisory and services.
它还提供现金管理产品，并正逐步进入借贷业务领域。
It also has cash management products and is moving into the lending business as well.
现在加入我们的是财富前沿的首席执行官大卫·福蒂纳托。
Joining me now is David Fortinato, CEO of Wealthfront.
大卫，欢迎来到节目。
David, welcome to the show.
很高兴你来到这里。
It's great to have you here.
非常感谢。
Thanks so much.
很高兴能来到这里。
Excited to be here.
恭喜整个Wealthfront团队迎来重要的一天。
Well, congrats on a big day for the Whole Wealthfront team.
我想聊聊IPO窗口期，这是我们节目经常讨论的话题。
I want to talk a little bit about the IPO window, which is something we've talked a lot about on our show.
我们经常谈到开启和关闭。
Look, we talk about an opening and closing.
你最终为什么决定现在是上市的最佳时机？
Why did you decide ultimately that this was the right moment to go public?
这个问题有很长的答案，也有一个简单的时机因素。
Well, so there's a there's a long answer to that and sort of a short timing thing.
简单的时机因素是，我们原本计划在十月上市，但显然受到了政府停摆的影响。
The short timing thing is we were planning to go public in October, and obviously, we're impacted by the government shutdown.
随着政府停摆结束，我们与顾问团队会面，与投资者沟通，得到的反馈是我们已经做好了上市的准备。
We got together with our advisers as the shutdown lifted, talked to investors, and the feedback that we got was we were in good shape to go.
如果回顾威富通公司历史的长期轨迹，我们多年来一直是一家盈利公司，持续增长，实现了26%的年收入增长率，并不断向成为一家成熟公司迈进。
If you think about the longer sort of trajectory of Wealthfront's history, we've been a profitable company for years, have been growing consistently, have done, you know, 26% year over year revenue growth, and have continued to make progress towards being mature company.
我们认为，这是我们的下一步发展。
We felt like this was the next step for us.
你们建立的业务基于机器人顾问，这种模式已经存在很长时间了。
Now, you guys have built a business on robo advisory, which has been around for a long time.
现在我们进入了人工智能的新时代，这似乎是这项技术的自然应用。
Now we're in this new era of AI, which seems like a natural application for the technology.
我想知道，在财富管理领域，是否存在人工智能并不适合的场景？你们在业务中哪些部分没有应用它？
I wonder if there are scenarios here where you think AI is not well suited for wealth management and what parts you are not applying it to in your business.
因此，我们没有将人工智能应用于我们的核心投资策略。
So we don't apply it to our core investment strategy.
如果你想想人工智能擅长的方面，比如自然语言的使用以及做出复杂判断的能力。
If you think about the things that AI is amazing at, you know, its use of natural language and being able to to apply complex judgment.
在最佳被动投资策略方面，这些策略长期以来都是由学者们提出的。
In the case of kind of best of breed passive investing, these are these are things that have been set out by academics for long periods of time.
我们应用这些经过时间检验的学术策略。
We apply those time tested academic strategies.
我们以极低的成本实施这些策略，并将节省的成本与客户共享，使他们能够以极低的成本获得最佳收益。
We make them very low cost to implement, and we share the savings with our clients so that they can get the best outcomes at extremely low cost.
那么你们用它来做什么呢？
And so what are you using
那么你们用它来做什么呢？
it for then?
我们将其应用于客户服务和工程团队的各个方面。
You know, we use it across our client support, engineering teams.
在财务规划方面，还有一些非常有趣的应用，我们将来会逐步开展。
There are very interesting applications in financial planning that we'll work on over time.
我们对这些可能性感到非常兴奋。
We're very excited about the possibilities.
但对于投资而言，我们希望专注于经过时间检验、学术验证的策略。
But for investing, we want to stay focused on time tested, academically proven strategies.
但我只想再回到这一点。
But I just want to go back to it.
关于被动投资，你提到学者们已经研究了这些内容数十年，那么你难道不认为在某些应用场景中，AI可以根据其收集的所有数据为你做出决策吗？
So with respect to passive investing, and you talked about the academics having studied this stuff for decades, So so you don't think there are any applications or any suitable sort of scenarios where AI should be able to make the decision for you based on all the data that it's collected?
被动投资的核心本质上是基于预期收益和风险的优化问题。
I mean, the core of passive investing is basically, looking at an optimization problem based on expected return and risk.
这是一个相对简单的数学方程，由哈里·马科维茨在20世纪60年代初提出。
It's a it's a relatively straightforward mathematical equation to solve, you know, defined by Harry Markowitz in the in the early nineteen sixties.
此后，黑-利特曼等人也对此进行了研究。
There have been there's been research that's gone into it with Black Litterman.
当然，你可以用AI工具表达一些观点，也有人尝试过这样做。
There are obviously views that you could express with AI tools, and some people have tried to do that.
但目前这仍处于非常早期的阶段。
It's still very early to do that.
投资的一个不同之处在于，它是一个竞争性市场。
One of the things that's I think different about investing, is that it's a competitive market.
你的每一个行为都会引发反应，你知道的。
There are pause you know, everything that you do, you get reactions to.
因此，如果你看看那些正在使用或开始使用人工智能的大交易公司，他们会非常谨慎，因为其他所有人都在与他们的每一步行动竞争，并实时更新策略。
And so if you look at the big trading firms that do or are starting to use AI, they are being very careful with it, because everyone else is competing with everything that they do and updating on a real time basis.
你不能只是实施一个AI策略，然后让它永远运行下去。
It's not something that you can just implement an AI strategy and then let it run forever.
让我问你另一个炫目的技术，我相信你至少在近期会避开它，那就是预测市场已经席卷全球。
Let me ask you about another flashy technology that I trust you're probably staying away from, at least in the near term, which is prediction markets have taken the world by storm.
请再听我一会儿。
Stick with me for a minute here.
作为一个外行，看着围绕预测市场兴起的这种快速变化的交易生态，你对此有什么看法？
Just as an outsider looking at this flashy sort of quick moving trading ecosystem that is popping up around prediction markets, what are your thoughts on
它？
it?
好吧，宾夕法尼亚大学已经发表了一系列令人惊叹的研究。
Well, so there's been a bunch of amazing research that's come out of UPenn.
菲利普·泰洛克教授谈到了专家团队的重要性，以及他的‘良好判断项目’、预测市场及其运作方式。
Philip Tedlock as a professor talking about the importance of both teams of experts and his and his judgment project, the good judgment project, and prediction markets and how they work.
我认为这是一种非常有趣的技术。
I think it's a fascinating technology.
它对于学习事物非常有帮助。
It's amazing for learning about things.
它显然被用于投机。
It it obviously is used for speculation.
很难看到预测市场有长期投资的角度，因为它是一个零和游戏。
It's very difficult to see a a long term investing angle to prediction markets because it's a negative sum game.
运行交易所的费用最终会消耗掉预测市场的预期回报。
The fees to run the exchanges end up taking out, you know, expected return from prediction markets.
因此，如果你只是随机投资预测市场，长期来看你可能会亏钱。
And so, if you were just to invest randomly in prediction markets, you would expect to lose money over time.
那其他针对私营公司的代币化产品呢？比如购买某种OpenAI等公司的股份代理？
And what about these other tokenized offerings for for private companies and, you know, being able to buy some kind of a proxy for a share for OpenAI and stuff like that?
现在市面上有太多这样的东西了。
There's so much out there now.
你怎么看？
What do you make of it?
是的。
Yeah.
我的意思是，我认为在这个市场中取得成功非常困难。
I mean, I think it's a it's a it's a tough market to be successful in.
显然，这些公司的估值极高。
Obviously, companies are valued extremely highly.
这些公司存在很大的风险。
There's a lot of risk in those companies.
我认为，如果你把风险投资作为一类资产，与纳斯达克100指数或其他科技导向指数相比，一个分散投资于多种资产的投资组合，很可能为个人投资者带来更好的回报。
I think if you look at the venture returns as an asset class and compare that to the Nasdaq 100 or other tech focused indexes, really a diversified portfolio that invests in many things is likely to do better for individual investors.
你们平台上的用户平均年龄是多少？
What's the average age of the people on your platform?
威富通当前客户的中位年龄是35岁，但新客户群体的中位年龄大约在24岁左右。
The median age of a client on Wealthfront today is 35, but new client cohorts are coming in with a median age of around 24 years old.
你创办这家公司已经有一段时间了，我想知道你能否谈谈你观察到的年轻人投资观念发生了哪些变化？
And you started the company a while ago, I wonder if you can speak a little bit to how you've seen investing change among young people.
你提到了新加入的用户。
You talked about the new joiners.
你经常和二十出头的年轻人交流。
You talk to people in their early 20s all the time.
他们谈论投资的方式，和你刚创业时、或者公司刚成立时相比，有什么不同？
How are they talking about it differently compared to when you started company or when the company started, I should
我认为，每一代人都有一批投资者热衷于投机，试图利用他们认为自己能为市场带来的机会。
Well, I think what happens is there's a there's a group of investors in every generation that's interested in speculation and taking advantage of the opportunities that they think that they can bring to the market.
还有一部分人则更关注家庭、事业，以及长期积累财富。
And there's a group of folks that are more focused on their families, their careers, and building wealth over the long term.
我们吸引的是那些对市场中的投机机会不感兴趣的人。
We don't attract people that are interested in the speculative opportunities that exist in the market.
我们吸引的是那些关注家庭、专注事业，并通过每月储蓄、让资金增值来逐步积累财富的人。
We attract people that are focused on their families, focused on their careers, and building wealth gradually by saving every month and putting their money to work.
他们希望将这件事委托给他人，而我们希望成为他们委托这一责任的首选。
They wanna delegate that to someone and we want to be the folks they delegate that responsibility to.
很好。
Great.
大卫，恭喜你迎来这一重要时刻。
Well, David, congrats on the big day.
这是重要的一步。
It's a big step.
正如你所说，这背后是数月乃至数年的努力。
And like you said, it was months and years in the making.
我们就不打扰你回去参加庆祝活动了。
We'll let you get back to the festivities.
这是财富前沿公司的首席执行官大卫·福蒂纳托。
That is David Fortinato, CEO at Wealthfront.
谢谢。
Thank you.
好的。
Okay.
数据管理公司Rubrik的股价自去年上市以来已上涨了一倍以上。
Shares of data management company Rubrik have more than doubled since its IPO last year.
该公司最近公布了第三季度业绩，营收同比增长48%。
The company recently reported its third quarter results, 48% revenue growth from the prior year.
自由现金流是去年同期的四倍以上。
Free cash flow was more than 4x what it was this time last year.
现在邀请到与我讨论公司策略的是联合创始人兼首席执行官比普尔·辛哈。
Joining me now to discuss the company's playbook is Bipul Sinha, co founder and CEO.
比普尔，欢迎来到节目。
Bipul, welcome to the show.
很高兴你来到这里。
It's great to have you here.
谢谢，阿卡斯。
Thanks, Akas.
很兴奋能来到这里。
Excited to be here.
所以，我想和你聊聊你们最近看到的增长。
So, look, I want to talk to you a little bit about the growth you guys have been seeing.
48%的收入增长来自哪里？
Where is 48% revenue growth coming from?
我们为全球每个企业和政府提供网络韧性服务。
We are delivering cyber resilience to every business and government around the world.
而网络韧性是头号网络安全类别，因为每个人都花了大量资金用于预防和检测攻击，但攻击仍然在发生。
And cyber resilience is the number one cybersecurity category because everybody has spent so much money in prevention and detection of attack, but attacks are still happening.
你无法预防不可预防的事情。
You can't prevent the unpreventable.
我们的策略是为不可避免的攻击做好准备，确保即使遭受攻击，您的业务也能持续运行。
And our strategy is that be ready for attack that is inevitable and ensure that your businesses are up and running, even when they are confronted with an attack.
我们现在来谈谈攻击。
Now, we talk about attacks.
我想稍微谈一谈更广泛的AI领域及其对网络安全的意义。
I do want to talk a little bit about the broader landscape here of AI and what it means for cybersecurity.
但回到增长话题，这种增长主要来自现有客户的追加消费，还是新客户？
But sticking with the growth here, is the growth largely coming from existing customers spending more or is it new customers?
Rubrik 的业务有三大核心支柱。
Rubrik has three core pillars for our business.
我们的核心支柱是数据保护，这是已经具备规模并快速增长的业务。
Our core pillar is data protection, which is the business that is at a scale and we are growing it quickly.
下一个业务是身份韧性，这是我们大约三个季度前推出的新业务，仅在三个季度内就实现了2000万美元的年度经常性收入（ARR）。
Then the next business is identity resilience, which is a new business that we created about three quarters ago, and we hit 20,000,000 ARR in just three quarters.
我们在第三季度客户数量翻了一倍，这是一个增长非常迅速的业务。
We doubled the number of customers just in Q3, and it's a very fast ramping business.
还处于早期阶段，但非常令人兴奋。
Early days, but very exciting.
我们业务的第三部分是智能操作，这是我们刚刚推出的新产品。
And the third part of our business is our agentic operation, which is a new product that we just launched.
因此，我们在数据、身份和AI方面全方位提供韧性保障。
So we are delivering resilience everywhere across data, identity, and AI.
如果你看看这三者的结合，这就是推动需求和增长的原因。
And if you look at the combination of all three, this is what is driving the demand and what is driving the growth.
当你思考你所服务的这些客户时，我们在这个节目中经常谈论的一点是，公司为了提供这些创新产品，不得不进行大量的并购，因为在这个快速变化的环境中，他们似乎无法足够快地自主开发。
And if you think about these customers that you're serving, you know, one of the things that we've been talking a lot on this show is how much M and A companies need to undergo to be able to offer these new innovative products, given that they can't seem to build things fast enough in this fast moving landscape.
看吧，你们的股价至少在市盈率上高于其他软件公司。
Look, your stock is trading higher than other software companies, at least on a multiples basis.
你们有考虑进行并购吗？
Are you looking to do any M and A?
Rubrik在文化上是一家独特的公司。
Rubrik is a unique company in terms of our culture.
我们显然既有有机增长策略，也有非有机增长策略。
And we obviously have both organic and inorganic strategy.
但如果你想想身份韧性这个业务，这是我们完全从零开始有机创建并成功推出的，这表明即使在如此规模下，我们的工程和产品团队依然非常创新。
But if you think about identity resilience, a new business that we created completely organically from scratch and able to launch it, so that shows that our engineering, product even at this scale is very, very innovative.
然后我们正在此基础上叠加团队、产品和技术，以加速我们的路线图。
And then we are layering on top of it team, products, technology to accelerate our roadmap.
所以我们始终保持开放态度。
So we are always open.
我们正在关注优秀的团队和出色的技术。
We are looking at great teams, great technology.
我们刚刚收购了PrediBase，这是一个AI模型微调与服务平台。
We just acquired PrediBase, which is an AI model fine tuning and serving platform.
无论我们是自己开发还是收购，我们的目标都是服务好客户。
We want to serve our customers, whether we can build it or we can acquire.
你认为AI对网络安全是利大于弊还是弊大于利？
Do you think AI is net good or net bad for cybersecurity?
展开剩余字幕（还有 369 条）
因为在这档节目中，我们经常听到AI让攻击者以各种方式、形态更容易地对你发起攻击。
Because we've been hearing a lot on this show about how AI makes it a lot easier for attackers to come at you in different ways, shapes, and forms.
我确信AI也有保护你的方法，但看起来它确实让这场游戏变得可怕多了。
I'm sure it has ways to protect you as well, but it seems like it's making the game a lot scarier.
我在IT行业已经工作了大约三十年，长期从事技术工作。
I'm in the overall IT industry for about thirty years and doing technology for a long time.
你看，任何新技术的出现，都会带来正面和负面的影响。
Look, any technology that comes in, it has both positive and negative impacts.
如果看看AI，情况也是一样。
If you look at AI, it's no different.
AI为每个人带来了百倍的机会，但也创造了百倍的风险。
AI has given 100x more opportunity to everyone, but also it creates 100x more rest.
我们需要做的是，为组织提供一个治理AI的框架，使其在扩展AI时充满信心，同时确保能够部署智能代理工作，真正加速自身业务，增强市场竞争力。
And what we need to do is we need to give organization a framework to govern AI, to have confidence in scaling AI while ensuring that they can deploy agentic work to actually accelerate their own business and be competitively more viable in the marketplace.
因此，我们推出了Rubrik Agent Cloud，在管理风险的同时提供智能代理运营。
And that's why we launched Rubrik Agent Cloud that delivers agentic operations while managing risk.
事实上，我们说：释放智能代理，而非风险。
In fact, we say, Unleash agents, not risk.
但这足够吗？
But will it be enough?
我的意思是，你是否预计由于人工智能，我们会遭遇更多大规模攻击？
I mean, do you anticipate we're going to have more of these large scale attacks because of AI?
攻击是不可避免的，恶意行为者会比良性行为者更多地利用人工智能，因为他们无需应对合规性和治理问题。
Attacks are inevitable, and bad actors will use AI in some ways more than good actors because they don't have to deal with compliance and governance.
好人必须做到人工智能加加，这意味着他们必须更积极地基于人工智能进行构建、利用人工智能，并确保其业务不仅通过预防和检测，还要通过确保韧性来抵御这些攻击。
The good guys have to do AI plus plus which means that they have to be more aggressive in building on top of AI, utilizing AI, and making sure that their businesses are protected against these attacks by not just doing prevention and detection, but ensuring resilience.
让我问你一个关于Rubrik之外的简单问题。
Let me ask you a quick question outside of Rubrik for just a minute.
你提到你已经在IT行业工作了几十年。
You mentioned you've been in the IT sector for decades now.
你职业生涯始于甲骨文，而这家公司自你加入以来经历了多次转型。
You started your career at Oracle, and it's a company that has gone through many transformations since you were at the company.
你如何看待他们如今在云计算上所下的巨大赌注？
What do you make of the scale of the bet that they're making these days in the cloud?
听好了，你不能跟拉里·埃里森对着干。
Look, you don't bet against Larry Ellison.
我为他工作了将近十年。
I worked for him for almost a decade.
他是那个能追踪市场、发现机遇并迅速行动的人。
He is the person who tracks the market, spots the opportunity, and switches quickly.
这正是科技行业和创始人生态系统的标志。
And that is the hallmark of tech industry and founder ecosystem.
如果你看看他进入云计算领域、特别是针对AI的速度，以及他如何迅速提升需求，我真心被他的成就所鼓舞。
And if you look at how fast he has gone into the cloud business, specifically for AI, and how fast he's ramping up demand, I'm honestly inspired by what he has done.
这正是美国科技行业和整个硅谷经济所需具备的特质，以抓住机遇并快速前进。
And that is what is needed for American technology sector and for overall Silicon Valley economy to support the opportunity and move fast.
但接着你就会看到这些数据。
But then you see these stats.
我的意思是，阿波罗公司的托尔斯滕·斯拉赫上周发布了一张图表。
I mean, Torsten Slach over at Apollo, he put out this chart, I think it was last week at some point.
我的意思是，他目前展示的数据表明，大型企业对人工智能的采用实际上已经开始趋于平稳。
I mean, the data that he showed at this point is that for large enterprises, AI adoption is actually starting to plateau.
归根结底，正是需求最终给了那些为债务提供担保的人信心。
And at the end of the day, it's the demand that ends up giving confidence to people who are underwriting the debt here.
你对此怎么看？
What do you think of that?
大型企业对人工智能持谨慎态度的原因不是技术本身，而是关于治理，以及担心有人在朝鲜通过代理漏洞操控他们的业务运营。
Look, the reason that large enterprises are a little hesitant on AI is not the technology, it's about the governance and confidence that somebody sitting in North Korea is not running their business operations through agentic compromise.
因此，这种情况会很快发生转变。
And so, you'll see this flip very quickly.
随着Rubrik及其他公司提供越来越多的技术，信心将会提升。
As Rubrik and others are providing more and more technology, the confidence will go up.
我个人认为，我们正处在推理能力大规模增长的临界点，这真正属于企业业务，也是消费级人工智能技术的一部分。
I personally believe that we are at the cusp of massive ramp up in the inference, which is truly the enterprise business and part of the consumer AI technology.
我对未来感到非常兴奋。
I'm very excited about what's ahead.
很好。
Great.
比普尔，非常感谢你参加我们的节目。
Well, Bipul, thank you so much for coming on the show.
这是一次很棒的对话，期待很快再和你交谈。
Was a great conversation and look forward to talking to you again very soon.
谢谢，阿卡什。
Thanks, Akash.
好的。
Okay.
现在是我们每周的编辑精选环节，这是我们周五的特别板块，我们会邀请编辑们来分享过去几天里主导他们编辑会议的讨论内容。
It is time for our weekly Editor's Cut, our Friday special segment where we bring on the information's editors to give us a glimpse into the discussions that are dominating their editorial meetings over the past few days.
本周，科技和媒体新闻头条都被埃里森一家主导，尤其是拉里·埃里森领导甲骨文，并支持他的儿子大卫·埃里森推动派拉蒙收购华纳兄弟探索公司的计划，我们想聚焦于拉里所下的巨大赌注。
This week, the tech and media news headlines were dominated by one family, the Ellisons, and specifically with Larry Ellison leading Oracle and also backing his son, David Ellison, and Paramount's ambitions to buy Warner Brothers Discovery, we want to focus on the scale of the bets that Larry is making.
现在加入我们讨论这一话题的是我们的联合执行主编马丁·皮尔斯和特稿编辑尼克·温菲尔德。
Joining me now to discuss that is our co executive editor, Martin Pierce, our features editor, Nick Wingfield.
欢迎你们两位。
Welcome to you both.
很高兴你们能来。
It's great to have you here.
嘿，阿卡什。
Hey, Akash.
谢谢。
Thanks
感谢邀请我。
for having me.
好的。
Okay.
这是马丁和尼克的节目。
It's the Martin and Nick show.
马丁，我先来问你。
Martin, I'm going to come to you first.
你如何看待拉里·埃里森在甲骨文公司乃至媒体领域所下的赌注规模？
How are you thinking about the scale of the bets that Larry Ellison is making, not just at Oracle, but also in media?
我应该先说，尼克在科技方面比我聪明多了。
Well, I should start by saying that Nick is much smarter than I am about tech.
在编辑会议上，我们通常也不会争执太多。
And in the editorial meeting, we don't tend to argue too much.
所以让我先打个底，把这一点说清楚。
So let me just lay the groundwork and make that clear.
尼克和我实际上已经合作了二十五年，所以我们彼此非常了解。
Nick and I have actually worked together for twenty five years, so we actually know each other pretty well.
我认为这些赌注的规模很大，但对我来说，他更大的赌注显然是在甲骨文公司。
I think the scale of the bets are great, but it's clear to me that the bet the much bigger bet is the one he's making at Oracle.
我的意思是，他正在投入一笔数额不明的资金来扩展其人工智能业务。
I mean, is investing an undefined amount of money in expanding its AI business.
它非常依赖于OpenAI，而没有人真正知道这家公司究竟能表现得如何，以及它是否能在四五年后负担得起向甲骨文购买云计算服务的承诺。
It's very reliant on open AI and no one really has any idea how well that company will actually do and whether or not it will be able to afford the commitments it has made to buy cloud computing from Oracle in four or five years' time.
另一方面，埃里森在娱乐业上押了重注，但以他的个人财富来看，这一赌注的规模并不算大。
On the other hand, Ellison is making a big bet on the entertainment industry, but the scale of it in terms of his personal wealth is not that great.
他向派拉蒙投入了约60亿美元，并在华纳交易中至少承诺了120亿美元。
He put about 6,000,000,000 into Paramount and he's committed at least 12 into the Warner deal.
他需要为410亿美元的全部股权提供担保，但他实际上并不需要全部出资。
He's on the hook to back the entire equity amount of 41, but he really won't have to put up all of that.
所以，即使派拉蒙价值归零，即使股权价值归零，他也不会损失太多。
So even if Paramount goes to zero, even if the equity goes to zero, he won't lose that much.
而他的大部分财富都集中在甲骨文，如果由于AI押注导致股价大幅下跌，他将损失更多。
Whereas most of his wealth is in Oracle, and if that stock falls a lot because of the AI bet, then he would lose a lot more.
好的。
Okay.
尼克，你是否同意甲骨文的押注是更冒险的？
Nick, do you agree that the Oracle bet is the riskier bet?
这似乎是一个无懈可击的论点。
Seems like a foolproof argument.
是的。
Yeah.
我同意马丁的观点，但为了使这场讨论更有趣一些，我会试着反过来说。
I agree with Martin, but I'm going to try and argue the opposite in the interest of making this discussion a little more stimulating.
首先，我要带大家回到记忆中，那是我和马丁二十五年前相遇的那天。
First, I'm going to take a trip down memory lane, which when Martin and I met twenty five years ago.
我们相遇的那天，一家如今已被遗忘的公司AOL收购了时代华纳——也就是华纳兄弟的前身，或者至少是华纳兄弟探索公司中的华纳部分。
We met on the day that a now long forgotten company called AOL acquired Time Warner, the predecessor to Warner Brothers, or at least the Warner portion of the Warner Brothers discovery company.
这成了媒体并购史上，甚至可能是企业并购史上最惨重的失败之一。
And that turned into one of the biggest fiascos in certainly media deal history, if not corporate mergers and acquisitions.
这是一笔糟糕的交易。
It was a terrible deal.
事实上，大多数媒体交易都是糟糕的。
And the fact remains that most media deals are bad.
但为什么那个交易那么糟糕呢？
Why was that one bad though?
哦，原因太多了。
Oh, there were so many reasons.
你这是在考验我的记忆力。
You're going to stretch my memory.
但基本上，AOL的股票因为许多可疑的交易而被高估了。
But basically AOL had an inflated currency, its stock, because of a lot of questionable deals that they struck.
马丁，你还记得当时其他具体原因吗？
Martin, do you remember the other particulars around why-
其他原因在于，AOL的业务模式是向人们销售互联网接入服务，而这笔交易发生时，宽带技术刚刚兴起，这彻底摧毁了AOL的业务。
Well, other reasons were that AOL was built on a selling internet access to people and the deal happened right as broadband was emerging and that completely destroyed the AOL business.
另一个问题是，这两家公司之间存在巨大的文化冲突，导致它们根本无法协同工作。
And the other issue is that there was such a cultural clash between these two companies that it prevented them from actually working together.
好的。
Okay.
所以，尼克，现在说完你的观点。
So Nick, finish your point now.
你刚才说什么来着？
So you were saying?
对。
Right.
因此，这笔交易的失败最终将成为涉及双方公司的许多人的讣告中的第一行或第二行。
So the collapse of that deal ended up being the first line or will be the first or second line in the obituaries for a lot of the people who are involved on both sides of the company.
所以，基本上，如果这笔交易达成，存在风险，它可能会搞砸，成为埃里森家族的耻辱。
So basically, If this deal happens, there is risk that it ends up souring and being an embarrassment for the Ellison family.
但同时也存在一个问题：这笔交易能否完成。
But there's also a question of whether the deal can get done.
它仍需通过监管审批。
It still has to go through regulatory approval.
我认为，埃里森家族认为，他们有明确的途径获得监管机构的批准，因为他们与特朗普政府有着，怎么说呢，友好的关系。
The Ellisons, I think, have argued that they have a clear path to getting it approved by regulators because of their, shall we say, friendly relations with the Trump administration.
但这并不是板上钉钉的事。
But it's not a slam dunk.
如果没有华纳兄弟的发现，你可以说派拉蒙注定会失败。
And without Warner Brothers' discovery, you could argue that Paramount is destined to be a failure.
而如果这种情况发生，即使没有像甲骨文倒闭那样严重损害财富，也会让拉里·埃里森蒙受巨大耻辱。
And that could end up, if not damaging the wealth seriously or as seriously as Oracle's collapse would, it would be a huge embarrassment for Larry Ellison.
那么，马丁，这里是否可以提出一个观点：派拉蒙这笔媒体交易——也就是说，这个行业正在经历转型。
Well, and Martin, is there an argument to be made here that the Paramount deal, the media deal, I mean, this is an industry that is going through transformation.
很大程度上，它正在衰退，对吧？
Much of it is in decline, right?
我的意思是，你把钱投进了一个缺乏全球趋势支持的领域。
I mean, you're putting your money into this area that does not have the rest of the world sort of on its tailwind.
而人工智能这个领域，从资本角度看是有风险的。
Whereas the AI play, it's risky from a capital perspective.
但人们谈论的正是这个，它在某些情况下将推动未来十年的整个经济。
But I mean, this is the thing that people are talking about is to carry much of the economy in some cases for the next ten years.
你不觉得对于拉里·埃里森来说，进入一个正在萎缩的行业风险更大吗？
Don't you think that it's riskier to go into the shrinking industry for Larry Ellison?
毫无疑问，几十年来，娱乐行业吸引了许多根本不了解这个行业的人，他们被参加奥斯卡颁奖典礼、感觉自己很酷的机会所吸引。
There's no question that the entertainment industry for decades has attracted people who don't know anything about it, are attracted by the chance to go to the Academy Awards and to just feel like they're really cool.
而这个行业本身其实是个糟糕的生意，现在尤其糟糕。
And the business itself is really a bad business, and it's particularly bad right now.
是的，无论埃里森家族是否成功完成这次收购，他们在派拉蒙上真正赚到一大笔钱的可能性并不大。
And yes, whether or not the Ellisons manage to succeed in this takeover bid, their chances of actually making any significant amount of money on Paramount are not very good.
换个角度看，我认为他现在已经83岁了，埃里森是在孤注一掷。
Look at it from the other point of view, which is that at 83, I think he is, Ellison is betting the ranch really.
他正在将自己四十五年前创立的甲骨文公司的未来，押注在这项未经验证的技术上。
He's really betting the future of Oracle, the company he founded, what is it, forty five years ago, on this unproven technology.
他正试图在云服务领域与亚马逊、微软和谷歌竞争。
He's trying to compete with Amazon, Microsoft, and Google on the cloud front.
这是一场巨大的豪赌。
It's an enormous bet.
我认为，即使假设派拉蒙这笔交易彻底失败，他可能损失的钱也远比在甲骨文公司多。
And I think the amount of money he stands to lose is much greater in Oracle than in Paramount, even if you assume the Paramount thing is a complete disaster.
我应该说，我刚谷歌了一下。
And I should say, just Googled it.
他是81岁，马丁，所以还没到83岁。
He's 81, Martin, so he's not quite 83.
抱歉，我把他的年龄说高了。
Sorry, I overstated his age.
这真是太糟糕了。
That's really terrible.
尼克，归根结底，这只是一个父子故事吗？
Nick, does this just come down to a father son story at the end of the day?
我的意思是，这里的商业逻辑就是这个吗？
I mean, is that the business case here?
你知道，我是个电影迷。
You know, I'm a movie guy.
我真的很、真的很喜欢电影。
I really, really love movies.
今年我最喜欢的电影是华纳兄弟的。
I of my favorite movies this year were Warner Brother.
让我们说清楚。
Let's be clear.
尼克对娱乐行业一无所知。
Nick doesn't know anything about entertainment industry.
当他自称是电影迷时，只是喜欢看电影。
When he says he's a movie guy, he likes to watch movies.
一开始我就说尼克比我想象的要聪明。
Just started off by saying Nick is smarter than I thought.
你开场时就是这么说的。
That was what you opened the segment with.
当然了。
Well, of course.
这正是我的意思。
That was my point.
我不是说我对这个行业比你更了解，但我热爱电影，也许，仅仅是也许，拉里·埃里森也觉得他更想掌控这个领域。
I'm not saying I know the industry better than you do, but I love movies, and maybe, just maybe, Larry Ellison feels like he'd rather own the scene.
这
This
这就是人们进入这个行业的原因。
is why people go into the industry.
他们热爱电影。
They love movies.
这就是这个行业的问题。
This is the problem with the industry.
所有这些蠢货进入这个行业，只是因为他们喜欢看电影。
All these morons go into the industry because they like to watch the movies.
他们根本不理解。
They don't understand.
这是一门糟糕的生意。
It's a terrible business.
没人会愿意投资。
No one's going be investing.
拉里·埃里森的钱多到花不完。
Larry Ellison has more money than he could ever spend.
你知道，生活中不止有金钱。
You know, there's more to life than money.
你知道，也许他只是想保护这些美丽的好莱坞资产。
You know, maybe he wants to be able to protect these beautiful Hollywood assets.
好吧，现在让我们把自己当作记者。
Okay, let's think of ourselves as reporters now.
当你思考报道问题时，马丁，你认为记者们应该去探究这些交易背后的哪些重大问题？
As you think about the reporting questions, Martin, what are the big questions that you think that journalists should be out there trying to figure out with these deals in play?
我想我会试着弄清楚埃里森个人投入了多少钱，以及这些钱究竟来自哪里？
I guess I would be trying to figure out how much money is Ellison personally putting in and where is that money actually coming from?
我的意思是，甲骨文在文件中表示，艾里森已经以其持有的约三分之一甲骨文股份作为抵押借款。
I mean, Oracle has said in its filings that Ellison has borrowed against about a third of his Oracle stake.
我们不知道甲骨文股价下跌对他的这一安排造成了什么影响。
We don't know what the impact of the Oracle price falling has had on that particular arrangement he's got.
但我很想知道，他是否通过甲骨文的贷款来为对派拉蒙的投资融资。
But I would be interested to find out, is he financing his investments in Paramount via his Oracle the loans.
所以这是一个问题。
So that's one question.
另一个问题是，甲骨文在人工智能领域需要筹集多少资金？
And then the other question is how much money does Oracle have to raise on the AI front?
这是最关键的问题。
That's the single biggest issue.
当本周被问及此事时，他们根本没有给出任何具体的答复。
And when they were asked about that this week, they really did not give any kind of specific answer.
尼克，对你来说，有哪些问题浮现在脑海中？
And Nick, what are the questions that come to mind for you?
甲骨文在云业务上的许多成功都依赖于一个大客户——OpenAI。
A lot of Oracle's success in cloud is based on a big customer, OpenAI.
因此，我认为需要了解OpenAI是否能兑现其对甲骨文的云服务承诺，是否有足够的资金，以及其产品是否有增长空间。
So I think understanding whether OpenAI can follow through on its cloud commitments to Oracle, whether it has the money to do it, whether the growth is there for its product.
这些因素在很大程度上确实影响着埃里森的财富。
A lot of that actually certainly influences Ellison's wealth.
将所有这些因素综合起来，看看它们如何影响这笔交易，是理解这个故事的一种方式。
So putting all those pieces together and seeing how they impact this deal is one way to approach the story.
让我先回到娱乐行业，稍微谈一下。
Let me just go back to the entertainment industry for one second here.
看，这些交易中可能有一个我们会详细讨论。
Look, one of these deals maybe we'll go through.
要么是Netflix，要么是派拉蒙。
It's either going to be Netflix or Paramount.
但我们还需要一段时间才能弄清楚。
It'll still be a while before we figure it out.
但在媒体公司不断并购、分拆的背景下，你认为这两者中哪一个最终会更好？
But in the story of media companies constantly taking over each other, merging, breaking apart, which of the two do you think may end up better?
在这两笔交易中，你认为哪一笔更有可能最终出售部分资产给其他公司、分拆，或者最终失败，马丁？
Which of these two deals has a higher likelihood of eventually just selling a piece off to another company or splitting apart or eventually going south, Martin, do you think?
我一直认为，派拉蒙这笔交易是更合理的选择。
I mean, I've argued all along that I think the Paramount deal, it's a more logical bet.
我认为奈飞是一家非常优秀、强大的公司。
Netflix, I think, is a very good, strong company.
它并不需要这笔交易。
It does not need this.
好的。
Okay.
尼克？
Nick?
根据我在行业媒体上看到的信息，如果你问好莱坞的人，奈飞这笔交易在就业方面会更好。
Well, if you ask people in Hollywood, from what I've read in the trades, the Netflix deal is going to be better in terms of jobs.
所以这取决于你从哪个角度来论证。
So it depends on what perspective you're arguing from.
派拉蒙在这笔交易中所宣称的协同效应意味着更高程度的冗员和削减。
The synergies that Paramount is projecting from this deal imply a much higher level of redundancies and cutbacks.
而奈飞的情况则没那么严重。
With Netflix, it's less.
当然，可能会有一些工作岗位流失。
There certainly probably will be some job loss.
但我认为，奈飞的交易对影院体验可能更不利，因为他们的重心在流媒体上。
But I also think a Netflix deal is probably worse for the theatrical experience just because their bias is in streaming.
而我仍然喜欢去电影院看电影。
And I still like to go to the movies.
很好。
Great.
好的。
Okay.
那么，这个周末你打算看哪部电影，尼克？
Well, what movie are you seeing this weekend, Nick?
我没什么计划。
I don't have any plans.
我该看什么我
What should I I
你想看《杰伊·凯利》。
want to watch Jay Kelly.
那部电影在Netflix上。
That's the movie It's on on Netflix.
我没说我想去电影院。
Well, I didn't say I wanted to go to a theater.
我说的是我想看这部电影。
I said I wanted to watch the movie.
好吧。
All right.
非常感谢各位的到来。
Well, thanks a lot for coming on, guys.
我们下周五再和你们聊。
We'll talk to you next Friday.
这位是我们的联合执行主编马丁·皮尔斯，以及我们《The Information》的专题编辑尼克·温菲尔德。
That is Martin Pierce, our co executive editor, and Nick Wingfield, our features editor here at The Information.
最近，AI基础设施的建设已成为最受关注的话题之一，数据中心如雨后春笋般涌现。
Okay, the AI infrastructure build out has become one of the most closely watched stories lately with data centers popping up left, right and center.
为了全面报道这一领域，《The Information》将与Nebius合作推出一份名为《AI基础设施》的每周简报，深入探讨AI数据中心和计算领域最重要的发展动态。
And to cover it all, the information is teaming up with Nebius on a new weekly newsletter called AI Infrastructure, where we'll go deeper on the most important developments in AI data centers and computing.
这份简报将于12月15日星期一正式上线。
That newsletter launches on Monday, December 15.
Nebius当然是一家在Neo Cloud领域快速成长的公司，我们此前已对其进行了详尽报道。
Nebius, of course, is a fast growing company in the business of Neo Clouds, which we have covered in great detail.
本周在纽约市举行的AI峰会上，我有机会与Nebius的联合创始人罗曼·切尔南就他在业务中观察到的各种动态进行了交流。
I had a chance to speak with Nebius' co founder, Roman Chernan, about all the dynamics that he is seeing in his business at the AI Summit in New York City this week.
这是那段对话。
Here is that conversation.
罗曼·切尔南，欢迎来到TI TV。
Roman Chernan, welcome to TI TV.
很高兴你来到这里。
It's great to have you here.
嗯。
Yeah.
谢谢你们邀请我。
Thank you for having me.
我很期待和你聊聊你们正在打造的业务。
Well, I'm excited to talk to you all about the business that you guys are building here.
首先，云服务商是一个人们非常熟悉的企业。
Look, I wanna start with cloud vendors are a business that people know very well.
它们已经存在很多年了。
They've been around for many years.
这个领域有一些大玩家。
There are some big players in the space.
你们属于新兴云服务商这一类，
You belong to this category of neo cloud players,
说实话，这个名字挺快的。
which is fast name, to be honest.
但什么
But What
你更愿意称它为什么？
would you rather what would you rather call it?
AI专用云。
AI specialized cloud.
AI专用云。
AI specialized cloud.
好的。
Okay.
但你知道，这让我想到一个问题，我想问问你，你们如何定义自己的独特性？
But, you know, it kinda gets at a point that I wanted to ask you about is how do you look to differentiate yourself?
嗯。
Yeah.
这就是为什么我更倾向于称之为AI专用云，因为它确实能传达出我们所做的工作。
That's why I'd I my mostly like AI specialized cloud because actually it gives the sense of what we do.
对。
Right.
而且我认为，总的来说，我们所做的就是专业化。
And I think, like, in general, what we do is specialization.
所以我们专注于AI工作负载，全力以赴。
So we focus laser focus on AI workloads.
我们会尽最大努力在这些特定场景中挖掘出每一丝性能和可靠性。
We do the best to extract every single point of a kind of performance and reliability in this particular scenarios.
这就是我们的竞争方式。
And that's how we compete.
所以帮我理解一下这一点。
So help me understand this though.
你知道，当人们说那三大云服务商时，他们其实也都推出了自己的AI专用云服务。
You know, when people say, well, the big three cloud providers, I mean, they all have their own AI specialized cloud offerings as well.
你如何看待与它们竞争？
How do you think about competing against that?
我是说，是靠软件吗？
I mean, is it on software?
是靠客户服务吗？
Is it on customer service?
是的，实际上是在整个堆栈的各个层面，但客户服务是个棘手的部分。
Yeah, it's across all the stack actually, but customer service is a tricky part.
我们的首席执行官说，当你只有100个客户时，对客户好是很容易的。
Our CEO says, it's easy to be good to your customers when you have 100 of them.
但当你有上万客户时，还能保持冷静就很难了。
Stay cool when you're, like, 10,100.
所以这并不是你一开始获胜的方式。
So it's not like it's the way how you win at the beginning.
对。
Right.
确实如此，但这并不是那种模式。
It's true, but it's not like the the mode.
对吧？
Right?
这种模式可能更多来自于构建整个堆栈。
When the mode probably comes more from building the stack.
而我们所做的，是做全栈。
And what we do, we do the full stack.
我们涵盖了从物理基础设施到平台层（而非应用层）的所有层级。
We we all the layers from physical infrastructure all the way up to the not application layer, but, like, platform layer.
对。
Right.
我们坐在这里，面前是Token Factory，这是我们最新推出的服务层
We sit here in front of Token Factory, which is our, like, the newest layer of the offering
对。
Right.
推理平台。
Inference platform.
因此，我们专注于全栈优化，目标明确。
So and we we optimize across the full stack on one purpose.
AI工作负载应该是最出色的。
AI workloads should be the best.
我认为，我不想对其他厂商妄加评论，但当你追求通用时，就无法做到完全专业化。
And I think that I don't wanna say for other players, but you you cannot be fully specialized when you are universal.
所以你们是在深度深耕。
So you guys are going you're going deep in a way.
完全正确。
It's totally right.
这源于许多细节，但如果你考虑平台层、编排层，甚至硬件层，专注于一两种工作负载就能取得优势。
It's it's come from many details, but if you think about like the platform layer, if you think about orchestration layer, if you think about even hardware layer, you can win by focusing on one workload or two workloads.
你们在Token Factory产品和软件方面关注的一件事是，你们是开源AI模型的坚定支持者。
Well, one of the things that you are focusing on with your Token Factory offering and with the software is you're a big proponent of open source AI models.
我们在这个节目中经常讨论的一个话题是：未来会是开源的吗？
And it's a conversation we've been having a lot on the show is, is the future open source?
我们知道目前领先的模型都是闭源的。
We know the leading models right now are closed source.
你似乎是开源的拥护者。
You seem to be a proponent of open source.
你认为开源模型最终会成为最强大的模型，我不
You think that open source models will inevitably be the most powerful models in the I don't
我认为这个问题并不成立。
think this is the question.
我认为未来的发展格局会比单一模型统治一切更加复杂。
I think that the landscape will be more complex than one model wins all.
所以我把它想象成一个三角形。
So I think about it as a maybe triangle.
你拥有最强大、最智能的模型，而今天它们都是闭源的。
So you have the most most powerful smart models, and today they are closed source.
你也有那些不需要最智能模型，但希望成本更低或延迟更低的模型。
You have the models where you need, you maybe not need the smartest models, but you want cheaper or lower latency.
这非常适合开源，因为你可以针对特定用例进行大量优化，以降低成本和延迟。
And this is very good fits to the open source because you can do a lot of optimizations, like for particular use case to reduce the cost and latency.
然后你面临如何应用你的数据的问题。
And then you have the question of applying your data.
如果你有一个非常专业的工作流程，你可以选择不是最强大、不是最智能的模型，而是足够好的模型，并用你的数据进行微调来提升它。
And then like, if you have a very specialized workflow, you can take maybe not the power the most powerful model, like not the smartest model, but good enough, and improve it with your data, like fine tune.
在你的特定用例中，它会比通用的智能模型更聪明。
And then in your particular use case, it will be smarter than the general smart model.
所以可能不是比其他模型更好，但对你来说更好。
So maybe not better than other models, but better for you.
是的。
Yeah.
更适合你的特定使用场景。
Better for your particular use case.
对，没错。
So Right.
开源用例主要由降低延迟和成本或应用数据的需求驱动。
Like, and then the use cases for open source, mostly driven by the need to reduce the latency and the cost or apply the data.
如果你需要通用智能，它可能仍然来自SODA模型。
And then if you need the general intelligence, it probably will still come from the SODA models.
但是
But if
如果你需要进行优化或应用数据，OpenSearch将找到它的位置。
you need to go down to optimization or apply data, OpenSearch will find its place.
你提到了成本，我想谈谈Nebius的成本，因为看啊，这是一个资本密集型的操作。
You mentioned cost, I I wanna talk about Nebius' costs for for a moment here because, look, it's a capital expensive operation.
我们明白。
We know that.
过去，Nebius不得不进入债务市场为运营融资，而公司正在快速增长。
In the past, Nebius has had to go to the debt markets to fund its operations, and the company's growing fast.
我的感觉是，你们在不久的将来还得回到债务市场。
My sense is you'll have to go back to the debt markets at some point in the near future.
在实际操作中，这个对话进展如何？
How is that conversation looking for you on the ground?
债券发行方和贷款方最关心哪些问题？
What are the big questions that bond issuers and people issuing these loans are asking you?
是的。
Yeah.
我想重要的是指出，到目前为止我们一直依赖可转换债务，而现在我们正在考虑资产支持融资，这对我们的业务模式非常自然。
So I think that important to say that until now we were in the market of convertible debt and we looking now in asset backed financing that is very natural for the business that we are running.
我们也在探索未来其他企业债务融资的机会。
And we also look for other opportunities of a corporate debt moving forward.
但我们DNA中非常重要的一部分是，我们在条款上非常谨慎。
But the important kind of part of our DNA is we wanna be very cautious to the terms.
对。
Right.
所以，我们所有筹集资金的财务操作、所有达成的财务交易，以及实际上所有签约的客户，我们都始终专注于健康的利润率，因为这既是规模驱动的业务，也是成本结构驱动的业务。
So all the financial all the fundraising we did, all the financial deals we did, and actually all the customers we signed, we always are very focused on the healthy margins because this is the business of the scale, but this is also the business of the cost structure.
所以我们非常善于抓住机会。
So we are very opportunistic.
我们会关注所有可能使用的金融工具，但我们始终优先选择健康的交易。
We look in like all the types of the tools we can use from the financial side, but we always prioritize like the healthy deals.
你现在谈到了成本结构。
Now you talked about cost structure.
我很好奇，在当今云服务商拥有如此强大定价权的情况下，你未来会考虑提价吗？
I'm curious, would you ever think about raising prices in the future given how much pricing power cloud providers have in this day and age?
我的意思是，这是一种稀缺资源。
I mean, it's a scarce resource.
是的。
Yeah.
如今市场的定价主要由供需关系决定。
Like the pricing today on the market is mostly defined by supply demand situation.
我应该说，今年年初我们是非常激进的定价者。
So, and I should say that, like, I would say we were very aggressive price player at the beginning of this year.
对。
Right.
因为对我们来说，抢占市场份额非常重要。
Because it was important for us to gain market share.
我们当时可以说是极为激进的，而且我们有能力这么做，因为我们的成本结构很优。
We were I think we were one of the very aggressive and we could afford it because our cost structure was good.
现在，正如我们在财报电话会上提到的，我们的产能已经基本售罄。
Now, today, and we told about it on the earning calls, and we are pretty sold out.
当你供不应求时，实际上就可以选择客户。
And when you are sold out, you actually can choose the customer.
而且，我们优先考虑客户时，有更多因素比价格更重要。
And again, like, we have much more factors to prioritize the customers rather than the price.
但现在，很多与我们合作的客户来找我们，并不是因为最低价。
But now a lot of customers we work with coming to us not for the lowest price.
为什么？
Why?
因为他们看重我们平台的优势，或者欣赏我们能多快帮他们实现价值，或者我们现在有他们急需的产能，他们愿意为此支付溢价。
Because they appreciate like the advantage of the platform or they appreciate how fast we can move them to the value or we have capacity that they need right now and they're ready to pay the margin for that.
为什么要提价？
Why raise prices?
不，我们并不是
No, we that's
我就是这个意思。
what I'm saying.
并不是我们在提价，而是每个交易都有其特定的条款。
It's not like we raise prices, but every deal has its specific terms.
当你在市场中占据有利位置，拥有合适的产能和平台时，你就可以选择那些愿意为你的优势付费的客户。
And when you're well positioned on the market, you have the right capacity, you have the right platform, you can choose the customers that's ready to pray for your advantages.
对。
Right.
我的意思是，你的投资者也会喜欢这一点，因为毕竟最终目标是实现盈利，当然。
I mean, your investors will like that too because, I mean, look, the end goal here is to get to profitability, of course.
是的。
Yeah.
当然。
Of course.
但我想说的是，整个市场都过于关注价格，而总拥有成本（TCO）这一更复杂的指标其实更重要。
But what I wanted to say also that I think the market in general is very much obsessed about like the pricing when the TCO, which is more important thing, like the total cost of ownership is kind of more complex metric.
我们换种说法。
Let's put it this way.
它比单纯的GPU价格要复杂得多。
It will more complex than just GPU price.
您拥有可靠性、合同中的其他条款，以及能够帮助您挖掘更多价值的平台。
You have reliability, you have the terms, the rest of the terms in the contract, you have the platform that enables you to extract more value.
在代币工厂中，我们提供托管服务。
You in the again, in the token factory, we provide the managed service.
因此，我们为客户提供性能优化，而客户并不那么关心GPU的成本是多少。
So we optimize the performance for the customer, and it's not so important for the customer how much GPU cost.
重要的是每个代币的成本。
It's important how much token cost.
我认为，市场将进入一个更成熟的阶段，届时人们会更关注投资回报率和总体拥有成本，而不是GPU每小时的成本。
So I think that the market will move in a more educated stage when the we'll think more ROI and total cost of ownership rather than the, you know, GPU hour cost.
您提到了GPU，显然您是英伟达的重要合作伙伴。
You you mentioned GPU, and you obviously are a big partner with NVIDIA.
我想知道您对其他芯片公司，比如AMD和谷歌的TPU，有什么看法？
And I wonder what you think about other chip companies, about AMD, about Google's TPU.
有计划扩展您的产品线吗？
Any plans to diversify the offering?
我认为市场有竞争是件好事。
I think it's great that the market is has competition.
我认为竞争总体上非常好，我们在各个层面都受益于竞争。
I think competition in general is very good, and we benefit from competition on all the layers.
到目前为止，我们非常专注于英伟达。
So far, we are very much focused on NVIDIA.
这是我们的关键合作伙伴和主要供应商。
This is our key partner, key vendor.
实际上，我们看到的大部分需求仍然来自英伟达。
And in reality, most of demand we see is coming still for NVIDIA.
因此，我们密切关注硅片竞争层面的每一件事。
So we are very close looking to everything that happens on the, like, silicon competition layer.
但再说一次，今天我们并没有看到强烈的原因或动力让我们从英伟达的产品转向其他选择。
But again, today, we don't see a strong reason, strong push for us to move from NVIDIA offering to other things.
所以，你们并没有和谷歌就使用TPU进行过任何对话，例如。
And so no conversations with Google about using TPUs, for example.
对话不代表商业行为。
Conversations, it's not business.
不，我们会和每个人交流。
No, like we talk with everyone.
对。
Right.
在工程方面，我们希望始终站在前沿。
And on engineering side of the things, we wanna be very much like in the frontier.
所以我们测试各种产品，与不同公司交流，但关键问题是需求和经济效益来自哪里。
So we test stuff, we talk with the different companies, but then the question like where demand comes from and where economics come from.
对。
Right.
让我问你一下，现在快到2026年底了，对于Nebius来说，情况如何？
Let me ask you this, we're coming to the end of the year now, 2026 for Nebius.
对你来说，未来会是什么样子？
What does it look like for you?
你们目前专注于哪些方面？
Where are you focusing?
首先，这是规模之年。
First of all, it's the year of scale.
我们实际上公布了明年非常雄心勃勃的目标，涵盖我们的业务规模和收入规模。
We actually announced quite ambitious targets for the next year, both on our footprint scale and on the revenue scale.
因此，我们预测到明年年底，IRR将达到70亿到90亿，这将是从我们目前水平出发的显著增长。
So we are forecasting 7 to 9,000,000,000 IRR by the end of the next year, which will be impressive growth from where we stand now.
数据中心集群和电力设施等方面也是如此。
And the same on the data center fleet and power fleet and so on.
但另一个我们非常关注的重点是产品组合的开发。
But another important thing that we very much focus on developing the product offering.
推动我们的原因是，我们看到客户画像正在发生变化。
And what drives us is we see that the profiles of customers are changing.
所以你们自己也在变得大得多。
So if You guys are getting much bigger too.
你们和Meta以及这家公司的合作
You have these deals with Meta and This
是其中一部分
is one
一部分。
part.
这只是其中一部分，但另一部分是，在这场会议上，一年前，绝大多数需求——我认为99%的需求——都来自构建模型的人，无论是大型模型、小型模型还是专用模型，都没关系，他们构建模型然后运行模型。
This is one part, but another part, and what's happening here in this conference is one year ago, most of demand, I think 99% of demand, actually came from people who built models, large models, small models, specialized, doesn't matter, but people who built model and then run models.
我们现在看到的是，很多客户来自产品端，那些建立垂直AI公司或在不同业务场景中应用AI的企业。
What we see now, we see a lot of customers coming from the product side, people who build vertical AI companies or enterprises that apply AI in different works in in different use cases.
他们需要不同的产品。
And they need different product.
没错。
Right.
他们不一定需要显存大的GPU。
They don't necessarily need RAM GPUs.
他们需要一个托管平台来运行推理。
They need managed platform to run inference.
他们需要一个托管平台来运行训练后处理。
They need managed platform to run post training.
因此，我们不断改进我们的技术栈，这不仅因为我们乐于开发这些内容，更因为我们明白，下一波需求将来自不同类型客户，他们需要不同类型的平台。
So and we develop our stack further, not only because we have a lot of fun developing stuff, but because we understand that the next wave of demand will come from different type of the customers that need different type of the platform.
对。
Right.
你可以想象像微软这样的公司，他们需要裸金属服务器和大型裸金属集群，因为他们自带完整的软件栈。
And you can think about it like Microsoft of the world, they need bare metal, large bare metal clusters because they have all the software stack that they bring there.
然后是下一层次的客户，比如规模较小的模型构建者。
Then the next layer of customers like smaller builders of models.
对。
Right.
他们需要多租户云服务，因为他们会训练模型，然后在大规模上运行，并掌控技术，但他们需要的是云形态的便捷服务。
They need multi tenant cloud because they they go, they train models, then run them on on scale, and they control kind of the technology, but they need cool service as a cloud.
对吧？
Right?
这就像传统的基础设施。
And it's like classical infrastructure.
每个人都需要一些不同的东西。
Everyone needs something that's something different.
但结构是一种服务。
But structure is a service.
对。
Right.
但当你想到垂直领域的AI公司，比如Coursers或Lovels这样的公司，他们是为了推理而来，或者来微调模型然后部署。
But then when you think about vertical AI companies, like Coursers of the world or Lovels of the world, they come for inference or they come to tune the models and then deploy them.
他们不想深入到堆栈底层去租用集群。
And they don't want to go down to the down to the stack to rent the clusters.
他们需要一个托管平台来运行推理。
They want managed platform to run inference.
而这正是我们需要应对的下一波理念。
And that's what that's the next wave of conception that we need to address.
对。
Right.
而2027年可能会成为企业真正需要它的年份。
And then 2027 will be probably the the year of the enterprises that it will need
那时它们才会最终上线。
That's when they'll finally come online.
对。
Right.
嗯。
Yeah.
这一直是我们在密切关注的故事。
Well, it's a story that we've been watching very closely.
罗马，感谢你抽出时间与我们交流。
Roman, I wanna thank you for making time for us.
这位是Nebius的联合创始人Roman Chernan，正在TI TV上做客。
That is Roman Chernan, the co founder at Nebius here on TI TV.
谢谢。
Thank you.
今天的节目就到这里。
That does it for today's show.
提醒一下，我们每周一至周五上午10点（太平洋时间）、下午1点（东部时间）直播。
A reminder, we are on this stream Monday through Friday at 10AM Pacific, 1PM Eastern.
我要感谢本次节目的冠名赞助商亚马逊云服务，也要感谢各位的收看。
I want to thank Amazon Web Services who is our presenting sponsor for this production, and I want to thank you for tuning in.
我们非常感谢大家的支持。
We really do appreciate your viewership.
我已经迫不及待期待周一的下一期节目了。
I'm already excited for our next show on Monday.
祝大家周末愉快，去看场电影吧。
Have a great weekend and go to the movies.
再见，暂时告别。
Bye bye for now.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。