迪士尼投资10亿美元于OpenAI的Sora，中国AI人才大战，甲骨文100亿美元现金消耗 | 2025年12月11日

欢迎各位收看Information的TI电视。

Welcome everyone to the Information's TI TV.

我叫阿卡什·帕什里查。

My name is Akash Pashricha.

今天是12月11日，星期四。

It is Thursday, December 11.

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，我们带来一些突发新闻。

First up, we have some breaking news.

迪士尼将投资10亿美元于OpenAI，并允许其角色用于Sora视频中。

Disney is investing $1,000,000,000 into OpenAI and allowing its characters to be used in Sora videos.

我们稍后将详细讨论这项新协议。

We'll discuss the new deal in just a moment.

接着，我们将探讨独家报道：随着中国AI人才竞争加剧，腾讯挖角字节舞动研究人员的内幕。

We're then talking about some exclusive reporting, the information published about Tencent poaching bite dance researchers as the China AI talent race heats up.

接下来，我们将分析甲骨文的季度业绩，以及华尔街对这家云服务提供商昂贵的AI扩张计划的看法。

We'll then break down Oracle's quarterly results and how Wall Street is feeling about the cloud provider's costly AI expansion.

随后，我将在梅菲尔德公司与17次入选米达斯名单的投资人纳文·查达进行对话。

I am then sitting down with 17 time Midas List investor Naveen Chadda over at Mayfield.

最后，我们将以对TalkShop Live新任CEO的独家采访结束本期节目，这是一家直播电商平臺。

And finally, we will end the show with an exclusive interview with the new CEO of TalkShop Live, the live streaming e commerce platform.

这是一期内容丰富的节目，让我们马上进入正题。

It is a big show, so let's get right on into things.

今天早上，迪士尼和OpenAI宣布了一项新的合作。

Disney and OpenAI are announcing a new partnership this morning.

迪士尼将向OpenAI投资10亿美元，作为三年许可协议的一部分，允许其角色被用于OpenAI的Sora应用中的AI生成视频。

Disney will invest $1,000,000,000 into OpenAI, and as part of the three year licensing agreement, it will allow its characters to be used in AI generated videos on OpenAI's Shora app.

现在邀请我的同事、《信息》杂志的联合执行主编马丁·皮尔斯来讨论这笔交易。

Joining me now to discuss the deal is Martin Pierce, our co executive editor here at The Information.

马丁，早上好。

Martin, good morning.

这真是相当令人惊讶，你怎么看？

It is quite a What do you think?

我认为这对OpenAI来说是一笔绝佳的交易。

I think this is a great deal for OpenAI.

这是那些负责发布消息和谈判交易的人的胜利。

It's a triumph for the people who make announcements there and who negotiate deals.

我不太理解从迪士尼的角度来看这笔交易的意义。

I don't understand the deal from the Disney point of view.

他们能从中获得什么并不太清楚。

It's not quite as clear what they get out of it.

让我们稍微想一想。

Let's just think about this for a moment.

OpenAI的Sora视频生成应用已经推出一段时间了。

OpenAI has had this Sora video generating app out there for a while.

他们一直受到好莱坞的批评，因为该应用允许人们在未经许可的情况下使用名人形象制作视频。

They've had a lot of criticism from Hollywood that it allows people to make videos using famous people's images without permission.

例如，CAA发布了一份声明批评OpenAI，称他们确实应该达成某种协议。

CAA, for instance, put out a statement criticizing OpenAI and saying that they really should kind of reach some kind of deals.

因此，与迪士尼的这项协议首先使OpenAI免于遭受不与好莱坞合作的批评。

So this agreement with Disney firstly insulates OpenAI from criticism that it's not working with Hollywood.

这极大地提升了Sora应用的价值，因为现在人们可以要求生成带有卢克·天行者或钢铁侠形象的视频，而这正是协议的一部分。

It immeasurably increases the value of Sora, the app, because now people can go on, they can say, Make me a video with Luke Skywalker's image or make me a video with Iron Man, and that's part of the deal.

作为投资的一部分，OpenAI从迪士尼获得了十亿美元，而迪士尼也承诺使用OpenAI的产品。

OpenAI gets a billion dollars from Disney as part of the investment, and Disney has also committed to use OpenAI's products.

迪士尼从中得到了什么？

What does Disney get out of this?

这是一项授权协议。

Well, it's a licensing agreement.

因此，可以推测迪士尼获得了某些报酬，但公告中并未提及这一点。

So presumably, Disney is getting paid some money, but there's no mention of that in announcement.

迪士尼可以将部分视频用于其Disney+流媒体服务

Disney gets to use some of the videos on its Disney plus streaming service,

这

which

或许增加了一些价值。

maybe adds some value.

我真的不确定。

I really I'm not sure.

另一点是，这使迪士尼看起来走在AI前沿，从投资者的角度来看，或许有一定价值。

The other thing is it makes Disney look AI forward, which, you know, from the point of view of investors, maybe has some value.

但真正的风险是，这些角色会被过度使用，导致其价值随时间下降。

But really, the risk is that these characters will get overused and that the value will be diminished over time.

还存在一种风险，即人们可能利用米老鼠等著名角色制作糟糕的视频。

There's also the risk that people could create awful videos using famous characters like Mickey Mouse.

我的意思是，你可以想象人们用这些角色和AI能做出什么。

I mean, you can just imagine what people could do with some characters and AI.

这可能会严重损害迪士尼的品牌形象。

That could really hurt the Disney brand name.

所以在我看来，这看起来更像是有利于OpenAI的交易。

So to me, this seems like a deal that's really in the interest of OpenAI.

很难说这对迪士尼真的有什么帮助，除了让迪士尼看起来像是跟上了时代（抱歉，应该是让迪士尼看起来像是跟上了时代）。

It's not clear that it really helps Disney at all, apart from making it look Disney the OpenAI sorry, making it look like Disney is kind of up with the times.

让我问你一个问题。

And let me ask you one question.

从迪士尼的历史来看，迪士尼有没有投资过这种极具前瞻性的技术？还是说，从这个角度看，这似乎令人惊讶？

Just in the history of Disney, does Disney have a track record of making investments in really forward looking technology like this, or is this seemingly surprising from that sense?

我不觉得这有什么令人惊讶的。

I'm not sure it's that surprising.

我认为在迪士尼CEO鲍勃·艾格的领导下，他在技术方面一直非常前瞻。

I think under Bob Iger, the CEO of Disney, been very forward in terms of tech.

他多年前曾与苹果的史蒂夫·乔布斯会面，正是在那时，他成功谈判收购了皮克斯，并与苹果展开合作。

He met with Steve Jobs at Apple years ago, and that's when he was able to negotiate to buy Pixar and also to work with Apple.

我认为他非常清楚技术如何影响娱乐产业。

I I think he's very conscious of technology and how it affects entertainment.

但我真的不确定，在这种情况下，迪士尼从这笔交易中究竟获得了什么价值。

I'm just not sure in this case though, what the value that Disney really gets out of this deal.

对。

Right.

很好。

Great.

好了，马丁，感谢你前来做客。

Well, Martin, I want to thank you for coming on.

这是一个进展迅速的故事，正如你所说，有些情况下问题多于答案，但今天早上这确实让我们感到意外。

It is a quick moving story, and like you said, more questions than answers in some cases, but it was certainly one that took us by surprise this morning.

这位是马丁·佩尔，我们《信息》杂志的联合执行主编。

That is Martin Pierce, our co executive editor here at The Information.

好的。

Okay.

人才大战已经蔓延到了中国。

The talent wars have reached China.

《The Information》今天独家报道，腾讯从字节跳动挖走了研究人员，这一幕与美国正在上演的故事如出一辙。

The information reported exclusively today that Tencent has posed researchers from ByteDance in a familiar story to what is playing out here in The United States.

现在加入我们的是我们的亚洲记者吉罗·小川，他是这篇报道的合著者。

Joining me now is Jiro Oshawa, one of our Asia correspondents who co authored the story.

吉罗，欢迎来到TI TV。

Jiro, welcome to TI TV.

很高兴你来到这里。

It's great to have you here.

嘿，阿卡什。

Hey, Akash.

又见到你了。

To see you again.

所以给我们讲讲你们今天发布的关于亚洲人才大战的故事吧。

So tell us about this story that you published today on the talent wars shaking out in Asia.

当然。

Sure.

我们的报道聚焦于中国顶尖AI人才争夺战的加剧。

So our story is about how the battle for top AI talent is heating up in China.

我们关注腾讯如何积极从字节跳动挖走顶尖人才。

And we are looking at how Tencent is aggressively trying to poach the best talent from ByteDance.

字节跳动是一家拥有TikTok的公司。

ByteDance is a company that owns TikTok.

为了吸引这些研究人员，腾讯有时会提供双倍薪资，并且已经成功聘用了部分人。

And to attract those researchers, Tencent sometimes offering to double their salaries, and it has already hired some of them.

腾讯是中国最大的科技公司之一。

And Tencent is one of the biggest tech companies in China.

但在开发AI模型方面，它一直落后于人。

But when it comes to developing AI models, it has been kind of lagging behind.

我们的报道显示，腾讯终于开始认真打造更强大的研究团队。

And our story shows that Tencent's really finally getting serious about building a stronger research team.

它正努力吸引更多人才，以在AI模型领域展开更激烈的竞争。

And it is trying to bring more talent to compete more aggressively in AI models.

你现在提到，在某些情况下，一些薪资已经翻倍了。

Now, you mentioned that some of the salaries have been doubled in some cases.

这些薪资包现在相比我们在美国看到的AI研究人员获得的数亿美元薪酬，有多大差距？

How big are these salary packages now compared to the hundreds of millions of dollars that we see AI researchers getting paid here in The US?

是的，我认为总体而言，我们在国内听到的薪资水平，远不及今年早些时候一些报道称Meta提供的数千万美元，甚至一亿美元。

Yeah, I think overall, definitely the salary levels, what we hear in China, is nowhere near what some of the reports said about Meta earlier this year offering tens millions of dollars or sometimes $100,000,000 yes.

但在中国，这些数字确实要小得多。

But in China, definitely the numbers are smaller.

我的意思是，整体工资水平较低，生活成本也更低。

I mean, overall wage differences and lower cost of living.

而且历史上，中国公司支付的薪酬一直不如那些大型美国公司。

And historically, Chinese companies haven't paid as much as those big US companies.

但即便如此，在中国国内，人才竞争和薪酬争夺的激烈程度依然非同小可。

But still, it is a big deal that within China, the competition over talent and paying for talent is just as intense.

如今，腾讯正在提供更高的薪资，从其他公司挖角人才。

And Tencent's now offering higher salaries to bring people from bike towns.

因此，在中国，这种竞争确实越来越激烈。

And so within China, definitely it is getting more intense.

而且我们有时也会看到中国公司从美国引进人才。

And sometimes we do see also Chinese companies bring people from The US.

在这些情况下，显然他们必须提供与美国公司相当的薪资。

And in those cases, obviously they have to offer more comparable salaries to what US companies offer.

在这则故事中，我们提到腾讯今年早些时候聘请了一位来自OpenAI的研究员。

And in this story, we mentioned how Tencent earlier this year hired a researcher from OpenAI.

显然，在这些情况下，他们确实必须——

And obviously, in those cases, they do have to offer- They

必须

have to

支付高额薪酬。

pay up.

为保持竞争力而采取行动。

Movement to make it competitive.

对。

Right.

我很好奇，中国大型科技公司之间的人才竞争，与DeepSeek这样的初创公司相比如何？

I'm curious, how does the talent war between the big tech companies in China, how does that compare to the talent at startups like DeepSeek?

在美国，研究人员面临的这个抉择也是一个话题。

This is a story here in The US as well around the decision researchers face.

我要自己单干吗？

Do I go it on my own?

我要加入一家可能成长为下一个OpenAI的公司吗？

Do I join a company that can grow into the next OpenAI possibly?

DeepSeek那里一定有一些非常强大的研究人员。

DeepSeek must have some pretty strong researchers over there.

是的。

Yes.

DeepSeek一个非常独特的地方是，它主要依赖国内的AI人才。

And one thing that's quite unique about DeepSeek is that it did rely a lot on domestic AI talent.

因此，许多在中国接受教育、毕业于中国大学的中国研究人员，都是中国大学的博士。

So a lot of Chinese researchers who were educated in China from Chinese universities, PhDs from Chinese universities.

在中国，清华大学等高校培养了大量人才，这些博士研究实验室的研究人员经常被聘为DeepSeek或其他初创公司的实习生。

And in China, there is a big pool of talent coming out of places like Tsinghua University, where PhD research labs, those researchers often are hired as interns for DeepSeek or other startups.

因此，有许多这样的学生实习生、博士实习生参与了DeepSeek在R1等模型上的开发。

And so there were a lot of those student interns, PhD interns, that participated in DeepSeek's development of models like R1.

所以这是一个庞大的人才库。

So that's a big talent pool.

对于大型科技巨头而言，你也确实能看到一些在美国接受教育的人，因为有许多中国研究人员在美国顶尖高校获得博士学位。

In the case of big tech giants, you also do see people who have studied in The US because there are a lot of Chinese researchers who come out of who get their PhDs in The US at the top US schools.

其中一些人留在美国，为硅谷公司工作，但另一些人则回到中国，为中国的公司效力。

Then some of them stay in The US to work for Silicon Valley companies, but others also go back to China and work for Chinese companies.

像拼多多或阿里巴巴这样的公司，也拥有部分这类研究人员。

Places like Pike Dance or Alibaba, they also have some of those researchers as well.

对。

Right.

你之前提到过腾讯最近在人工智能领域的进展如何。

You talked a little bit about how Tencent's progress is shaping up these days in AI.

是超出预期了，还是低于预期？

Has it been above expectations, below expectations?

你还提到他们正努力从落后中追赶上来。

You talk a little bit about how they are trying to come back up from behind.

为什么像这样一家大型科技公司的进展，没有你最初预期的那么强劲？

Why hasn't their progress been as strong as you might've thought originally for such a big tech company?

我认为有多方面的原因，但腾讯绝对是中国人最富资源的科技公司之一。

So I think there are multiple factors, but Tencent is definitely one of the most resourceful tech companies in China.

它曾是中国移动互联网的主要推动力，尤其是通过微信。

And it was one of the biggest drivers of mobile internet in China with WeChat.

但人们通常认为腾讯的优势在于能够打造最优秀的应用，并让它们广受欢迎、吸引大量流量，比如微信和其他应用，对吧？

But its strengths has people see Tencent's strengths as the ability to build the best apps and make them really popular, drive a lot of traffic, like WeChat and other apps, right?

因此，当竞争转向全新的AI领域时，他们面临的是另一种挑战。

So when the battle became really the new AI battle, they face a different kind of challenge.

开发最佳模型的能力与他们擅长的产品和应用开发截然不同。

The ability to develop the best model is very different from what they're good at with developing products, the applications.

历史上，腾讯并不被视为研究重镇。

So historically, Tencent wasn't seen as like the research powerhouse.

但我们现在看到，他们正在努力迎头赶上。

But now what we are seeing is that they are really trying to step up.

我的意思是，他们一直像其他公司一样拥有AI研究团队，但他们现在意识到自己需要做得更多。

And I mean, they have always had an AI research team like other companies have, but they really seen that they need to do more.

他们需要重组团队，引进外部人才，比如来自OpenAI的研究员。

They need to shake up the team bring some outside talent, like the researcher from OpenAI.

这些就是我们正在看到的变化。

So those are the changes that we are seeing.

在我们结束之前，我再问你一个问题。

Let me ask you one more question before we let you go.

这一切都涉及现有的研究人员，而他们在美國和中國都十分稀缺。

This is all dealing with the pool of existing researchers, and they are certainly in scarcity in The US, in China.

我想知道你能否谈谈中国的培训项目或教育体系，以及它们在中国或更广泛的亚洲地区发展得如何。

I wonder if you can speak to at all any of the training programs or any of the education, how that is shaping up in China or in Asia more broadly.

那里是否在积极努力培养更多的机器学习研究人员和工程师，以增加这种尖端人才的供给？

Is there a concerted effort there to train more machine learning researchers and engineers to create more supply of this cutting edge talent?

毫无疑问，中国现在在该领域培养了大量博士。

Definitely China does produce a lot of PhDs in that field now.

就像我提到的，例如清华大学位于北京。

And like I mentioned, Tsinghua University, for example, is in Beijing.

它是该领域的一个重要机构。

It's a big institution for that.

而且也有一些初创公司是由该大学的研究人员创办的。

And there are some startups also that came out of that university's researchers.

例如，北京有一家名为Chubu的初创公司，我们之前曾报道过。

And there's a startup, for example, in Beijing called Chubu, which we wrote about before.

因此，它们是由该校的一些顶尖研究人员创立的。

So they were set up by some of the top researchers from that university.

他们还会从那所大学招聘一些学生实习生。

And then they also hire some student interns from that university.

所以虽然不一定有正式的培训项目，但招聘博士实习生的现象非常普遍。

So there is a sort of They're not necessarily training programs, but there is a lot of hiring of PhD interns.

这不仅仅是那家初创公司，许多其他大型科技公司也这样做。

And it's not just that startup, but a lot of other big tech companies also do that.

这些做法在某种程度上已经成为了培训项目。

And those are sort of becoming the training programs in a way.

在许多情况下，这些实习生都被赋予了很大的责任。

Those interns are being given a lot of responsibility in a lot of cases.

他们参与了DeepSeek和Great公司重要模型的开发。

And they did participate in the development of very important models at DeepSeek and Great.

其他

Other

好了，Jiro，感谢你参加我们的节目。

Well, Jiro, I want to thank you for coming on the show.

这是一个很棒的故事，我认为我们会更加密切关注。

It's a great story and one that I think we'll be watching more closely.

感谢你做客节目。

Thank you for coming on.

这是吉罗·小川，我们《信息》杂志的亚洲记者。

That is Jiro Oswawa, one of our Asia reporters here at The Information.

好的。

Okay.

甲骨文公司公布了最新的季度财报。

Oracle reported its latest quarterly results.

尽管云收入增长了68%，但该公司本季度仍亏损了约100亿美元。

The company burned around $10,000,000,000 in the quarter, even as cloud revenues jumped 68%.

现在邀请到我身边讨论财报的是RBC资本市场的股权研究主管里希·贾卢里亚。

Joining me now to discuss the results is Rishi Jaluria, Managing Director of Equity Research at RBC Capital Markets.

里希，早上好。

Rishi, good morning.

很高兴你能来。

It's great to have you here.

非常感谢。

Thanks so much.

很高兴来到这里，阿卡什。

Great to be here, Akash.

所以甲骨文的股价今天波动很大。

So Oracle shares are moving quite a bit today.

从这些结果中，什么最让你印象深刻？

What stood out to you from the results?

是的，我认为让我印象深刻的是，有些方面可能没有像投资者希望的那样得到充分讨论，对吧？

Yeah, look, I think what stood out to me is what wasn't maybe addressed as much as investors would have liked to see, right?

因为数据还不错，基本符合预期。

Because numbers are fine, numbers are pretty in line.

我们也可以争论一下OCI是否略有不及预期之类的。

We can debate if there was a slight miss on OCI and the like.

但归根结底，投资者真正关心的是：你们已经获得了来自OpenAI及其他公司的巨大承诺。

But really what matters at the end of the day is investors want to know, hey, look, you have this giant commitment from primarily OpenAI as well as others.

那么，你们打算如何支付这笔费用呢？

Now, how are you going pay for this?

尽管他们显然谈了很多关于维持投资级评级、筹集资本并按时完成所有事项的信心，但仍然没有回答关键问题：你们是否已经筹集了这么多债务？

And, you know, even though they obviously talked a lot about their confidence in maintaining their investment grade rating, in raising capital and being able to meet everything on time, there's still just no answered questions of, Okay, have you raised this much debt?

你们是否设立了这些特殊目的实体（SPVs）？

Have you done these SPVs?

或者你们是否与外部供应商达成了这些战略合作？

Or have you made these strategic deals with outside vendors?

因此，在当前这种极其易变的市场环境下，任何导致更多不确定性、或至少未能澄清或增强确定性的事情，都会被视作负面。我认为，这是一次很少见的事件，因为对这个问题根本没有给出明确的答案。

And so anything, especially in this environment, in this incredibly fickle market, anything that results in just more uncertainty, or at least doesn't address or add certainty, is to And be I think this is a kind of seldom used type event just because there were no concrete answers to that question.

他们甚至没有明确说明最终需要筹集多少资金。

They didn't even really even say how much they needed to raise at the end.

他们只是笼统地说，分析师们推测这个数字是1000亿美元。

They just sort of said, Analysts are pointing to this $100,000,000,000 figure.

这不对。

That's not right.

有点模糊。

Like, was a little vague.

是的，实际上这很有趣，因为我们都算过账，即使认可他们30%到40%的AI基础设施毛利润率，最终会怎样呢？

Yeah, no, it's actually interesting because, you know, a lot of us have done the math and said, all right, even if you give them credit for that 30 to 40% AI infrastructure gross margin, where does that end up?

我们很多人都算出来，大概在1000亿美元左右，我的数字比那还高一点，因为我对毛利润率的估计可能偏低。

A lot of us have come somewhere, you know, around the $100,000,000,000 My number was a little bit higher than that in terms of the capital they'd have to raise because I'm probably on the lower end of gross margins.

但他们现在把这个数字放出来了，对吧？

But now they put that number out there, right?

他们说，这个数字会更少，甚至远少于这个数。

And they said it'll be less, if not substantially less than that.

但现在我们都固守在1000亿美元这个数字上了。

But now we're all anchored to that $100,000,000,000 number.

对吧？

Right?

而且我认为这里面有很多假设。

And and I think there's a lot of assumptions in there.

而且，你知道，如果你想想他们可能需要采取的那些低效手段，我觉得这很清楚。

And, you know, I mean, if you think about the low level of levers they may have to take on that, I think it's pretty clear.

他们必须考虑除了直接债务之外的其他选择，比如灵活融资、特殊目的载体，或者甚至再次提到外部投资。

They're gonna have to look at other options besides just straight debt, whether that's flex financing, SPVs, or, you know, even, again, outside investment.

我曾提出过主权财富基金作为潜在资本来源的想法，特别是考虑到那里正在进行的所有活动，但我们就是不知道答案。

I've thrown out the idea of sovereign wealth funds as a potential source of capital, especially, you know, given all the activity that's going on out there, But we just don't know the answer.

现在，他们谈到了一些降低开支的方式，或者至少在初期保护利润率不致过低的方法。

Now, they talked about some ways in which they are moderating expenses, or at least ways in which they're protecting against margins falling too low at the outset here.

他们提到的一种方式是，客户可以自带芯片，对吧？

One of the ways they talked about it was they said, Hey, customers can bring their own chips, right?

我们可能不需要完全承担成本。

We may not have to front the cost entirely.

那这会不会影响他们向客户收费的能力，以及最终自身的利润率？

Wouldn't that then impact what they can charge customers and ultimately their own margins?

我认为这完全正确。

I think that's exactly right.

而且，我之前提到过，OCI 能拿到这么多交易，原因之一是他们恰好在正确的时间身处正确的位置，并拥有正确的战略合作伙伴关系，对吧？

And you know, one point I've made is with OCI, the reason they're able to get all these deals is almost a, you know, they're in the right place at the right time and have the right strategic partnerships, right?

他们建成了所有的 OCI 容量，在所有人都急需的时候，他们恰好有富余的产能。

They built out all this OCI capacity, they had excess capacity at a time where everyone needs it.

而且，他们还与英伟达建立了非常关键的关系，当所有人都在争相获取 GPU 容量时，他们成为了首选的供应商或合作伙伴。

And then they also have this very key relationship with NVIDIA, right, where they're, you know, a kind of vendor or partner of choice for GPUs when everyone is kind of clamoring for GPU capacity.

现在，正如你所说，如果你自带 GPU 甚至 TPU，这会不会导致他们的定价能力有所下降？

Now, exactly to your point, if you're bringing your own GPUs or even TPUs, right, does that create a setup where, you know, their ability to charge comes down a little bit?

而且，这又会对他们在分析师日上公布的长期 OCI 收入和长期每股收益目标产生什么影响呢？那个目标虽然感觉像是一年前的事，但实际上非常近期。

And, you know, again, what does that do to that long term OCI revenue and long term EPS target that they laid out at the Analyst Day, which, you know, was very recent, even if it feels like it was more than a year ago.

现在，他们还谈到了在某些情况下从供应商那里租赁芯片的能力。

Now, the other thing they talked about was their ability to lease chips from vendors in some cases.

你觉得这种降低投资和支出的方式，其利弊分别是什么？

What do you see as the trade offs with that way of moderating the investment and expenses that they have to put in?

是的，我认为这是一个明智的决定，去思考是否有更好的融资方式，比如租赁之类的。

Yeah, look, I think it's a smart decision to kind of figure out like, are there better ways of financing this of like, you know, using lease and the like.

我认为这仅仅是表明，我们如何才能最快地满足需求。

And I think this is just a way of saying how do we just meet that demand as quickly as possible?

因为我一直对这些合同有一个疑问：它们是否包含具体的里程碑、时间表、阶梯式条款和规定？如果甲骨文在规定时间内未能达到一定的容量，这些收入会不会无法实现？

Because one question I've always had with a lot of these contracts is are there specific milestones, timelines, ratchets and provisions on them, that if Oracle doesn't meet or get certain amount of capacity up in a certain time period, you know, does that revenue maybe not materialize, right?

所以我认为，他们正在利用手中所有可用的工具，尽可能快地上线容量，即使这意味着采用租赁这种方式。

So I think they're using every tool they have in their arsenal to be able to get capacity online as quickly as possible, even if it means doing these sort of, you know, leasing.

从长期经济角度看，这种方式可能不如自己购买划算，但至少能缩短上线时间。

The long term economics probably aren't as good as the other way, but at least it shortens that timeframe to getting things online.

但当你谈到长期经济性时，你觉得租赁安排会带来哪些权衡？

But when you talk about the long term economics, what are the trade offs that you could see coming from a leasing arrangement?

是的，说到底，如果他们拥有自己的芯片，并且按照六年折旧周期来处理，那又是另一个复杂的问题了。

Yeah, look, mean, the end of the day, right, and if they own their own chips, and you know, they're they're doing it on this six year depreciation schedule, and that's a whole other can of worms.

没错。

Yeah.

你知道，也许我们可以谈谈这个

You know, maybe maybe to talk about it at

在某个时候。

some point.

但到了那个时候，你能从中获取更多价值。

But I you know, at that point, like you you get you extract more value out of it.

对吧？

Right?

而如果你从别人那里租赁，长期来看你支付的租金会比自己购买更多。

Whereas if you're leasing from others, you're actually paying more in rent over time than you would if you if you owned it.

所以，在租赁模式下，你的长期利润率和贡献利润率会低于自有的模式。

So your long term margins, contribution margins on that will be lower under a lease world versus an own world.

但话说回来，如果这能加快部署容量、满足需求，那也未必是坏事。

But, know, again, if that means that it just speeds up the ability to get capacity online and meet these demands, that's not necessarily a bad thing.

我想再问你一些关于更长远未来的问题。

I want to ask you a little bit looking further into the future.

假设他们能够构建起这套基础设施，并最终实现30%到40%的毛利率，正如他们所谈论的那样。

So let's say they're able to build out this infrastructure and eventually get to positive gross margins, the 30%, 40% range that they've talked about.

我想问问你，Oracle相对于其他云厂商，以及所有这些新兴云服务商的差异化战略是什么。

I want to ask you about what Oracle's differentiation strategy is against other cloud vendors, and also broadly, all these neo clouds even as well that are coming up.

我的意思是，虽然我不说Oracle是新的，但它在试图开拓的这个业务领域确实是新的。

I mean, how do these newer I'm not calling Oracle new, but it's new in the business that it's trying to build out.

Oracle如何与三大云服务商竞争并实现差异化？

How does Oracle look to differentiate itself against the big three?

这些新兴云服务商又该如何实现自身的差异化？

How do the Neo Clouds look to differentiate themselves?

这是否只是基于一个假设，即市场足够大，足以容纳所有参与者？

Is this just a bet that the market is so big that there will be enough demand for everyone?

还是说，最终只会有一部分玩家胜出？

Or is this not going to be a situation where there are winners at the end of the day?

是的，很高兴你问到这个问题。

Yeah, look, I'm glad you're asking that question.

我认为这是一个非常关键的问题，这也触及了核心：在这个环境下，为什么我更看好微软而不是甲骨文？

I think that's a very critical one, and that gets to the crux of, you know, in this environment, why do I favor Microsoft over Oracle?

对吧？

Right?

但甲骨文以及一些新兴云厂商可能会辩称，从AI架构的角度来看，他们更适合AI工作负载。

But the argument that that Oracle and and potentially some of the neo clouds would make is they're more suited from an AI architecture perspective for AI workloads.

对吧？

Right?

他们从GPU入手构建。

They build GPU first.

想想扩展方式——是纵向扩展还是横向扩展，以及这如何适用于当前情况。

You think about scale up versus scale out and how that lends itself here.

我认为他们所押注的是，当工作负载从训练转向推理和推理时，这种架构具有可互换性，能够随之适应。

And I think the bet that they're making is as we get workloads that move from training to inferencing and reasoning, this is fungible architecture that will be able to adapt as a result of that.

而甲骨文所押注的是，他们拥有整个产品套件，对吧？

And the bet that Oracle is making is they have this entire suite of products, right?

他们显然拥有AI数据库，现在还有这个数据湖，所有数据都已经存储在Oracle数据库中，并且拥有SaaS应用程序。

They obviously have the AI database, they have now this data lake, that they have all this data that's already sitting in Oracle database, they have the SaaS applications.

他们所押注的是，由于拥有这套AI基础设施，拥有GPU，并且声称其架构与AWS、Azure和谷歌不同，因此他们能从整体产品组合中获益，对吧？

The bet they're making is that because they have this AI infrastructure, because they have, you know, the GPUs and the argument they make is it's built in a way that's differentiated from AWS, Azure and Google, that they're gonna be able to benefit from a whole portfolio perspective, right?

因为，你知道，他们公布的AI基础设施30%到40%的毛利率，我对他们短期内能否实现这一目标持一定怀疑态度。

Because, you know, even that 30 to 40% gross margin number that they gave on AI infrastructure, and, know, I would I would be a little skeptical on their ability to hit that at least in the near term.

但长远来看，你希望这个业务的毛利率远高于这个水平。

But, you know, you want this business to do way higher than that in gross margins over time.

实现这一点的方式是，AI工作负载开始催生其他工作负载，而OCI的其他部分也开始带动数据库和SaaS的采用。

And the way that happens is if the workloads in AI start to beget workloads and the rest of OCI starts to beget adoption of database, starts to beget adoption of SaaS.

对吧？

Right?

没错。

Right.

这就是Oracle所押注的重大战略。

That's the big bet that Oracle is making.

待定，看情况如何发展。

TBD, how that plays out.

很好。

Great.

嗯，我认为我们就到这里吧。

Well, I think that's a good place to leave it.

里希，非常感谢你前来做客。

Rishi, thank you so much for coming on.

我们非常感谢。

We appreciate it.

这是来自RBC资本市场的董事总经理里希·贾卢里亚，正在TITB节目上。

That is Rishi Jaluria, managing director at RBC Capital Markets here on TITB.

好的。

Okay.

我们的下一位嘉宾是17次入选米达斯榜单的投资人。

Our next guest is a 17 time Midas List investor.

他有18家投资的公司成功上市，27家被收购。

He has seen 18 of his companies IPO and 27 get acquired.

他的投资覆盖了从人工智能到加密货币，再到网约车和电子商务等各个领域。

He's invested across everything from AI to crypto to ride hailing to e commerce.

现在加入我们的是梅菲尔德公司的管理合伙人纳温·钱达。

Joining me now is Naveen Chadda, Managing Partner at Mayfield.

纳温，早上好。

Naveen, good morning.

很高兴你来到我们的节目。

It's great to have you here on the show.

当然。

Absolutely.

能来到这里我感到非常荣幸和愉快。

It's a pleasure and delight to be here.

所以，我想先谈谈今天早上引发关注的新闻，那就是甲骨文股价下跌。

So, I want to start by talking about what's making news this morning, which is Oracle shares shares are down.

我们刚刚邀请了里希·贾卢里亚，他谈到了甲骨文相对于其他超大规模云服务商的价值主张，以及这些新兴云公司如何打算与传统云提供商竞争。

We just had Rishi Jaluria on talking a little bit about Oracle's value proposition against other hyperscalers, and broadly how these neo cloud companies are going to look to compete against the traditional cloud providers.

他提出的观点，我认为他们也会这么说，就是：我们专注于AI工作负载。

The argument that he made, which I think is what they'll say too, is that, look, we are specialized in AI workloads.

在甲骨文的情况下，他们表示，这有点像一个飞轮效应，对吧？

In Oracle's case, they say, well, it's a bit of a flywheel, right?

我们拥有云服务、应用程序和数据库。

We've got the cloud, we've got the applications, we've got the database.

我想问你的是，你认为这是一个可行的价值主张吗？

My question for you is, do you think that is a viable value proposition?

你认为这些新兴云公司能否在长期与这些超大规模云服务商的竞争中生存下来？

Do you see these neo cloud companies even surviving in the long run against

这些超大规模云服务商？

these hyperscalers?

我认为这将是一个双城记。

So I think it's going to be a tale of two cities.

首先，要在云服务领域赢得AI市场，你需要有强大的资产负债表。

First, in order to win in the cloud for AI, you need to have strong balance sheets.

一些超大规模云服务商在其核心业务中赚取了巨额利润，因此能够资助这种增长。

And some of the hyperscalers make so much money in their core businesses that they can fund this growth.

新兴云服务商必须持续融资，甲骨文也是如此，因为其经营活动产生的现金流并不高。

The neo clouds have to continuously keep raising money and so does Oracle because the cash flow from operations is not that high.

只要这种天文数字般的需求持续存在，而我们又处于供给受限的市场，所有人都将受益。

So as long as there is this astronomers demand, extraordinary demand, and we are a supply constrained market, everybody is going to rise.

一旦需求开始放缓，问题就会出现。

Once the demand starts slowing down, it's going to be a problem.

因此，在短期内，只要这些公司能持续融资，它们就会安然无恙。

So in the short run, as long as these companies can keep raising money, they'll be fine.

但一旦需求放缓或资本枯竭，就会成为一个巨大的、巨大的问题。

But once the demand slows or that capital dries up, it's going to be a huge, huge issue.

现在，考虑到融资可能枯竭的方式，正如你所提到的，需求是一个因素。

Now, thinking about the ways that the financing could dry up, as you mentioned, demand is one factor there.

一旦这种情况开始令人担忧，债务市场可能就不会再那么慷慨了。

And as soon as that starts to get worrisome, perhaps the debt markets won't be as generous.

如果我们考虑一下你认为可能导致债务市场枯竭的其他因素，还有哪些其他因素可能影响人们是否愿意为这些企业提供融资？

If we sort of think about the other levers that you think could dry up debt markets, what are some of the other levers that you think could have an impact on whether or not people are willing to finance these businesses?

我认为，除了债务市场之外，股票市场的表现也很重要。

I think it's also going to be, besides the debt market, how's the stock market doing?

以及‘七巨头’或AI公司关键驱动因素的市值能否继续上涨。

And can the market cap of the magnificent seven or the key drivers of the AI companies keep rising.

如今，资本供应几乎是无限的，不仅在债务市场，也在私募市场。

And today, there is infinite supply of capital, not only in the debt markets, but also in the private markets.

但所有这些资金都在关注公开市场中的增长，而这些公司所赚取的利润正在重新投入到这些领域中。

But it's all looking at the growth that is happening in public markets and that money from the profits that is being made is recycling in to some of these plates.

因此，问题不仅仅是需求放缓，还会因为这些上市股票因增长放缓而价值下跌，引发多重连锁效应。

So it's not only going to be the slowdown in demand, it's also going to be multiple effects that will be created as these public stocks start coming down in value because the growth slows.

这两方面都会产生影响。

Both those things are going to affect.

我想说的第三点是，你的差异化在哪里？

And then the third thing I would say is, where is your differentiation?

在这方面，我们已经看到平台型公司开始将硬件与软件捆绑在一起。

And this is where we have seen the platform plays start bundling hardware with software.

所以，如果你只是基于英伟达和第三方模型构建，你的差异化在哪里？

So essentially, if you're just going to be built on Nvidia, and third party models, where is your differentiation?

随着市场从训练转向推理，垂直整合将变得至关重要。

And as the market moves from training to inference, vertical integration is going to matter a lot.

而微软、谷歌、亚马逊和Meta等公司将在这一点上拥有巨大的优势。

And that's where companies like Microsoft, Google, Amazon, and Meta are going to have a huge, huge advantage.

所以，简而言之，我从你这里听到的是，巨头似乎会继续变得更强。

So, in short, what I'm hearing from you is it feels like the big will continue to get bigger.

没错。

Absolutely.

这主要体现在基础设施层面。

And it's on the infrastructure side.

好的。

Okay.

而大部分价值将聚集在这些人身上。

And a lot of the value is going to accrete to those people.

但当你向上游移动时，故事可能会有所不同。

But the story might be different as you move up the stack.

这正是这个市场将要发生的事情。

And that's just what's going to happen in this market.

因此，我认为这很好地引出了你所称的‘认知即服务’这一领域，这也是你的重点。

And so, think this is a nice segue then into an area that you have called Cognition as a Service, which is your focus.

而且，看啊，CAS这个缩写并不是一个缩略词。

And look, it's not an acronym, the CAS acronym.

我之前并没有特别关注或听说过太多关于它的内容。

I haven't really focused or heard too much about it.

什么是认知即服务？

What is Cognition as a Service?

是的。

Yeah.

所以，我们的观点是，首先，人工智能不仅仅是一种技术。

So, our belief is, first and foremost, AI is just not a technology.

它将作为智能即服务提供。

It's going to be available as intelligence as a service.

我们称之为认知即服务。

And we are calling it, it's cognition as a service.

类似于云时代中基础设施即服务（IaaS）和软件即服务（SaaS）在业务层发生的变化，智能也将按需提供。

And similar to what happened in the cloud era with IAS, infrastructure as a service, on the business layer with software as a service, intelligence is going to be available on demand.

借助英伟达和这些模型，这将提供给每个人。

And using Nvidia and the models, that's what is going to get provided to everybody.

但在上层将出现更多机会。

But then there are going to be opportunities up the stack.

因此，一旦建立了认知即服务的平台，就需要数据。

So essentially, once you have cognition as a service platforms which get built, you need data.

你需要中间件和工具技术来构建AI应用。

You need middleware and tooling technologies to build AI apps.

随着时间推移，价值将逐渐聚集到原生AI应用中，还包括我们所说的AI同伴——这些基于代理技术的数字伴侣，能帮助人类达到超人水平。

And over time, the value will start accreting into AI native applications, but also what we call AI teammates, which are digital companions built on agentic technologies, which help humans reach superhuman levels.

这个市场并不是从价值6000亿美元的企业软件市场中分走份额。

And that market, you're not taking share from enterprise software, which is a $600,000,000,000 industry.

你瞄准的是每年3万亿美元的知识工作者支出。

You're going after the knowledge worker spend, which is 30,000,000,000,000 a year.

如果其中10%到20%转向基于代理技术的AI同伴，那将是一个3万亿至6万亿美元的市场，是企业软件市场的五到十倍。

And if 10 to 20% of that moves to this AI teammates built on agentic technologies, that's a 3 to $6,000,000,000,000 market, five to 10x of the enterprise software market.

没错。

Right.

这就是数字世界将要发生的变化。

So that's what is going to happen in the digital world.

但在物理世界，市场还要更大。

But in the physical world, the markets are even bigger.

而当认知能力介入时，你将拥有以机器和机器人形式存在的数字伙伴，它们将补充并增强人类的能力。

And that's where if cognition enters, you're going to have now digital companions in the form of machines and robots, which will complement and augment humans.

因此，这个市场正朝着这个方向发展。

So, that's where this market is heading.

人工智能不仅仅是一种替代技术。

AI is just not going to be a replacement technology.

它将成为我的数字伙伴或物理伙伴，让我能够完成以前不可能做到的事情。

It's going to be my digital teammate or physical teammate, which will allow me to do things that weren't just possible before.

因此，我们对智能即服务持非常非常乐观的态度。

So we are very, very bullish on intelligence being available as a service.

它被封装为我们所说的认知即服务。

It gets packaged into what we are calling cognition as a service.

它使我们进入一个AI与人类协同合作的智能时代。

And it allows us to enter an era of collaborative intelligence where AI and humans work together.

对。

Right.

让人类的能力提升100倍。

Make humans 100x.

让我问你一个问题。

Let me ask you this.

那么，在这个新的认知即服务生态系统中，商业模式是否类似于软件公司曾经采用的按使用量计费模式？

Is the business model then in this new Cognition as a Service ecosystem, is the business model similar to where software companies were going in terms of consumption based billing?

你是否看到新的商业模式正在出现？

Do you see new business models coming up?

你是怎么看待这个问题的？

How do you think about that?

当然。

Absolutely.

我认为，当你转向认知即服务时，你的定价和商业模式将转向以成果为基础。

So, I think when you move to cognition as a service, your pricing and business model is going to move into outcome based.

这一点我们在模型公司身上已经看到了。

And that's what we have seen with the model companies.

这不会是按座位收费的许可模式。

This is not going to be a per seat license.

这将是：你将我的生产力提升X，我就给你Y。

This is going to be, you improve my productivity by X, I give you Y.

你将我的收入提升A，我就给你B。

You improve my revenues by A, I give you B.

你帮助我将运营成本降低C，我就与你分享这部分收益。

You help me improve my operational costs by C, I'll give you a share of that.

因此，我们将完全转向以成果为基础的定价模式。

So we're going to move completely to essentially an outcome based pricing model.

而且它将是按使用量计费的，而不是按座位收费。

And it will be consumption based, it's not going to be per seat.

因此，公司不仅要在这个时代在技术层面创新，才能取得成功。

So the companies have to just not innovate in order to do well in this era on the technology front.

它们还必须在商业模式层面进行创新。

They have to also innovate on the business model front.