Meta以20亿美元收购Manus，苹果2026年复兴，马斯克目标落空 | 2025年12月30日

欢迎各位收看《The Information》的TI TV。

Welcome everyone to the Information's TI TV.

我是阿卡什·巴什拉查。

My name is Akash Bashracha.

今天是12月30日，星期二。

It is Tuesday, December 30.

我们首先请AI记者报道昨晚Meta宣布的重大Manus收购案。

We are kicking off the show with our AI reporter talking about Meta's big manus acquisition announced last night.

接着，我们将采访Core Weave和OpenAI的投资人，了解他对Neo云的看法，以及他对OpenAI在2026年业务前景的思考。

We'll then speak with a Core Weave and OpenAI investor about how he's thinking about Neo Clouds and about OpenAI's business going into 2026.

我们还将开启预测系列的第一期内容。

We are also kicking off the first installment of our prediction series.

《The Information》正在发布一系列深度文章，探讨我们对2026年重大新闻的预测，今天聚焦于苹果公司如何重振其AI战略；最后，我们将回顾埃隆·马斯克今年为其企业设定的目标，并评估他是否真正实现了这些目标。

The Information is publishing a set of long form articles on what we think will be big stories in 2026, and today's focus will be on Apple revitalizing its AI strategy, and we will wrap with a look at the goals that Elon Musk set for his businesses this year and whether or not he actually reached them.

这将是一场精彩的节目，让我们马上开始。

It's going to be a fun show, so let's get right on into things.

Meta 正以一笔并购交易为今年画上句号。

Meta is closing out the year with some M and A.

该公司正在收购来自新加坡的 AI 代理公司 Manus，该公司由 Butterfly Effect 公司运营。

The company is buying Manus, the AI agent out of Singapore operated by a company called Butterfly Effect.

Meta 支付了 20 亿美元。在这里为我们讲解这笔交易的是我们的 AI 记者斯蒂芬妮·帕拉佐洛。

Meta is paying $2,000,000,000 Here to talk about that deal is our AI reporter, Stephanie Palazzolo.

斯蒂芬，欢迎再次回到节目。

Steph, welcome back to the show.

很高兴你来到这里。

It's great to have you here.

很高兴能来。

Great to be here.

我记得当时所有人都对 Manus 感到震惊。

So I remember everyone freaking out about Manus.

我想是在今年早些时候，它还是一个代理工具。

I think it was earlier in the year, it was the agent.

人们说，天哪，我什么都能做。

People said, Oh my God, I can do everything.

我可以让它为我做任何事情。

I can ask it to do whatever I want.

它到底擅长做什么呢？

What actually was it good for doing?

它后来怎么样了？

What happened to it?

因为我好像听说过它，但之后就没什么消息了，对吧？

Because I seem to have heard of it and then kind of stopped, right?

是的。

Yeah.

本质上，Manus 是一个可以接管用户浏览器来完成任务的代理，比如预订旅行、分析股票、撰写研究报告。

Essentially Manus is an agent that can take over a user's browser to do tasks like booking travel, analyzing stocks, making research reports.

所以这不过是今年和去年我们听到的众多计算机代理的又一个例子。

So kind of just another example of a computer using agent which we've heard a lot about this year and last year.

正如你提到的，它在今年早些时候确实变得非常流行。

So as you mentioned, it really gained popularity earlier this year.

但它绝对不是第一个或唯一一个走红的智能代理。

But it's definitely not the first or only agent out there to go viral.

所以我认为Manus的不同之处在于，许多用户认为它在执行需要多个步骤的复杂任务时表现得更好，而且不需要太多人为干预。

And so I think what made Manus different was that a lot of users thought it was a lot better at doing more complex tasks that might take multiple steps without meeting a lot of human prompting.

你认为Meta为什么要做这笔交易？

Why do you think Meta did this deal?

是的，我认为第一部分原因是，你知道，Meta在今年的AI领域表现并不理想。

Yeah, so I think the first part is, you know, obviously Meta hasn't been having its best year when it comes to AI.

所以我认为这部分原因在于，正如你最近几个月看到的，它目前正处于积极收购的状态。

So I think part of this is like, as you've seen in recent months, it is just really in acquisition mode right now.

它正在收购任何拥有优秀AI人才的初创公司，而Manus正是这样一个典型的例子。

It is picking up any startup that has really good AI talent, which Manus is definitely an example of that.

我认为第二点是，智能代理这个概念在今年，甚至去年，都一直非常重要。

I think second, the idea of agents has obviously been a very big one both this year and even last year.

因此，很多人认为智能代理是人工智能的圣杯。

And so a lot of people think that agents are kind of the holy grail of AI.

所以，能够拥有这些AI机器人，它们可以代表用户在我们的电脑或设备上执行操作。

So being able to have these like AI bots, I guess, that can take action on behalf of users on our computers or our devices.

因此，MetaMask表示，它希望继续运营和销售Manus的服务，同时也希望将这项技术整合到自己的业务中。

And so MetaMask said that it wants to continue operating and selling Manus' service, but it also wants to integrate that technology into its business as well.

你可以想象，比如，一个在WhatsApp上的小企业主使用Manus代理技术通过WhatsApp应用回复客户问题。

So you can imagine, for instance, a small business owner on WhatsApp using Manus agent technology to respond to customer questions through the WhatsApp app.

不过我觉得这很有趣，因为一方面，智能代理现在非常热门，但在我看来，这笔交易未必是最自然的。

I think it is interesting though, just because this, you know, I think on one hand agents are super hot, but I don't know if this deal was the most natural one in my mind.

为什么？

Why

社交媒体代理，对吧？

does social media agents, right?

让我印象深刻的是，从新闻稿来看，他们提到了Manus在市场研究和编程功能方面的市场表现，这两个领域，除非我知道Meta内部确实有编程需求，否则这项技术可能对它们有帮助，但我觉得这更像是一次人才收购。

The thing that stood out to me was from the press release, I mean, they talked about Manus' traction in market research and coding functions, which are two areas that I unless I know that they must have internal coding that this might help with at Meta, but I actually feel like it's more of a talent play.

我想知道Manus和Butterfly Effect的这些领导者到底有多大程度被挖到了亚历山大·王的团队。

I wonder how much to what extent these leaders at Manus and at Butterfly Effect actually vicked away into Alexander Wang's group.

我想听听你的看法。

I wonder what you think.

是的。

Yeah.

这百分之百是一场人才争夺。

I mean, it's 100% a talent play.

我的意思是，Manus的这些人肯定会被纳入Meta的亚历山大·王团队。

I mean, these folks from Manus, I'm sure, are definitely going to make it into Alexander Wang's org at Meta.

是的，我同意。

And yeah, I agree.

乍一看确实有点奇怪，因为Meta的大部分业务——无论是Instagram还是Facebook——都围绕着人类用户浏览内容展开。

It does initially seem a bit strange just because so much of Meta's business, whether that's Instagram or Facebook, is centered around human users looking at content.

因此，一个可能从事研究或编写代码的智能代理产品，似乎并不完全契合这一点。

So that doesn't necessarily seem supernatural for an agent type product that might be doing research or you know, be writing code for instance.

但我觉得，至少如你所说，这对Meta来说是一次重大的人才引进。

But I think, you know, at the very least, to your point, this is a big talent get for Meta.

我想转到你今天发布的一份简报。

I want to turn to a newsletter that you published today.

你回顾了2025年AI领域最重要的新闻，我感觉在过去365天里，可能有366条重大新闻，因为每天似乎都有一波接一波的消息。

You recapped the biggest stories in AI in 2025, and I seem to think that for three sixty five days, there were probably three sixty six big stories because it seems like every day there was just, you know, a barrage of news.

我想聚焦于你文章中提到的三个关键转折点，实际上不止这三个。

I want to focus on three inflection points that you talked about in your story, and that there was more than that.

第一个发生在年初，那就是DeepSeek。

The first one came at the start of the year, which was deep seek.

我想知道你对DeepSeek引发的市场动荡有何看法？它对AI行业产生了哪些持久影响？你认为这将如何改变2026年AI的发展格局？

I wonder what you think the deep seek market mayhem, what lasting impact that had on the AI sector and how you think that sort of changes the AI story going into 2026.

是的。

Yeah.

我相信我们都能记得今年一月的DeepSeek时刻。

So I'm sure we can all remember the deep seek moment from January.

我认为最有趣的是，正如你最初提到的，当时人们对此充满了恐惧，担心天啊，中国AI会不会超越美国AI？

I think what was most interesting was, as you mentioned initially, there was so much fear around this moment of people wondering, Oh my God, is Chinese AI going to surpass American AI?

美国的AI公司是不是出了什么严重问题？

Is there something terribly wrong with American AI companies?

因为这个中国初创公司是怎么做到用远低于美国AI公司成本的方式，开发出性能如此出色的模型的？

Because how could this Chinese startup develop a model that performs so well for so much cheaper than what we've heard from American AI companies?

但我觉得，纵观全年，许多真相逐渐浮出水面，实际情况远比我们最初想象的要复杂得多。

But I think, you know, throughout the year, I think a lot of the story has come out and there's obviously a lot more nuance than I think what we'd originally thought.

例如，训练这些模型所耗费的资金很可能远超最初对DeepSeek模型的估算。

So for instance, it probably took a lot more money to train these models than what was initially thought about the deep sea models.

但确实产生了一些持久的影响。

But there definitely have been some lasting impacts.

首先，中国AI无疑在这个领域已成为强大的竞争对手。

Think, you know, first Chinese AI is definitely a very strong competitor in this space.

甚至到现在，我们仍能看到中国在开源AI领域占据主导地位。

And even now we see, you know, China is dominating the open source AI space.

因此，任何使用开源人工智能的开发者或公司，很可能都在使用中国的人工智能模型，这与我们今年年初的设想大不相同，因为以前很多人认为美国开发者绝不会考虑使用中国模型。

So any developer or company that's using open source AI is probably using a Chinese AI model, which I think is very different from what we thought at the beginning of this year because a lot of people, there's no way American developer would ever think of using a Chinese model.

但事实显然并非如此。

That just definitely hasn't been the case.

所以我认为中国确实在人工智能竞赛中占有一席之地，这是我们今年多次学到的教训，从DeepSeek时刻就开始了。

And so I think China's definitely in the AI race and that's something that we've learned many, many times this year, starting with the deep sea moment.

我觉得这并不一定让人们意识到开源的重要性，但确实让人开始更加关注它。

Well, I also feel like it not necessarily woke people up to open source, but it feels like people started paying attention more to it.

例如，我们曾邀请Pinterest的首席执行官做客节目，谈论他们对开源的青睐，而Meta显然也在积极投入开源。

For example, we had the CEO of Pinterest on the show talking about their affinity for open source, and Meta obviously is working in open source.

我感觉自从DeepSeek之后，人们开始更多地关注这一点。

I seem to feel like people started to focus more on that since DeepSeek.

我不知道我是不是理解错了。

I don't know if I'm reading it wrong.

是的。

Yeah.

开源AI长期以来一直是一个重点，但我认为在DeepSeek之后，这方面的投入明显加倍了。

Mean, source AI has been a focus for a very long time, but I think definitely was a doubling down post DeepSeek.

我认为在DeepSeek发布并广为人知之后，许多其他中国公司也推出了开源模型。

And I think after DeepSeek came out and became so well known, a lot of other Chinese companies also came out with open source models.

这对许多开发者来说是好事，因为现在你有了更多的选择。

And so that was good for a lot of developers because now you have so much more choice.

感觉每个月左右都会有一个新模型问世，超越所有其他模型。

It feels like every month or so there's a new model coming out that's beating all the others.

所以这对开源开发者来说很棒，因为他们现在有了更多可用的选择。

And so yeah, it's been great for open source developers because they have access to so many more options now.

我想关注的第二个转折点是谷歌在今年的崛起。

So the second inflection point that I want to focus in on is the rise of Google throughout the year.

提醒一下，用一两句话说说，你认为谷歌今年取得成功、Gemini获得如此多关注的原因是什么？

And remind us, I mean, in one or two sentences, what do you attribute the success for Google this year and all the traction that they've been getting with Gemini?

是的。

Yeah.

我的意思是，这归结为两个方面：资源和分发。

I mean, I think it boils down to two things, like resources and distribution.

在资源方面，谷歌显然拥有大量的计算能力，以及大量的人力来解决这些极其困难的问题。

On the resources side, obviously Google has a ton of compute, a ton of people to throw at these very difficult problems.

因此，这是它的一大优势。

And so that is one thing that has going for it.

我认为在另一方面，它通过数百亿人每天使用的众多产品实现了广泛的分发。

I think on the other end, it has distribution obviously through so many products that hundreds of millions, billions of people use every day.

所以，即使你不是 Gemini 聊天机器人的用户，你每天使用 Google 搜索时，也会看到并交互 AI 概述的回答，从而接触到 Gemini。

So for instance, like even if you're not a user of the Gemini chatbot, you probably see and interact with Gemini every single day by using Google search and seeing the AI overviews answer.

因此，无论怎样，你都在通过 Google 的模型与 Gemini 互动。

So no matter what, you know, you're interacting with Gemini with Google models.

我认为这也是它今年取得巨大进展的重要原因。

I think that has been a really big boon for it this year as well.

你认为 Gemini 会在 2026 年在使用量和活跃用户方面超越 ChatGPT 吗？

Do you think that Gemini overtakes ChatGPT in 2026 in terms of usage and active users?

这是个好问题。

That's a good question.

我认为这很大程度上取决于谷歌愿意在多大程度上用AI彻底改造其现有产品。

I think a lot of this boils down to just how much is Google willing to just completely revamp its existing products with AI.

所以，比如现在，当我进行谷歌搜索时，我确实能看到屏幕顶部的AI概览。

So I think right now, for instance, yes, I see AI overviews at the top of my screen when I make a Google search.

但很多体验仍然很相似。

But a lot of the experience is pretty similar.

我得费点劲才能进入AI模式，也就是谷歌提供的那种更对话式的AI搜索选项。

I would kind of need to go pretty out of my way to reach AI mode, which is like the more conversational kind of AI search option that Google provides.

所以，没错，我现在确实在与谷歌的AI模型互动，但这并不是主要体验。

So yes, right now I'm interacting with AI models from Google, but it's not really like the main experience.

我认为谷歌有点走极端了，直接说：我们要抛弃谷歌搜索的原有体验，完全替换成Gemini聊天机器人。

I think Google kind of just went crazy and was like, we're getting rid of the Google search experience and we're just going to completely change it out for the Gemini chatbot.

我认为他们确实有可能赶上甚至超越ChatGPT。

I think there's definitely a possibility that they could match or even beat ChatGBT.

但我认为，只要他们保持相对保守，将这些体验保持分离，因为他们不希望破坏消费者的使用体验，我觉得到2026年他们仍有可能落后。

But I think as long as they're kind of being a little bit more conservative, kind of keeping those experiences separate because they don't want to mess up the experience for consumers, I feel like there's a pretty good chance that they might still be behind in 2026.

尽管我认为空白差距肯定会大幅缩小。

Although I think the gap will definitely narrow a lot.

这感觉非常像一份商学院案例研究，即如果你只是像你说的那样换掉那个按钮，明天你完全有可能在使用量上获胜，对吧？

It feels very much like a business school case study, which is that you very much could win in terms of usage tomorrow if you just swapped out the button, as you say, right?

我的意思是，我们已经看到过一些截图，人们说：‘如果谷歌搜索按钮直接换成Gemini，或者AI模式成为全部体验，那当然，你明天就能超越ChatGPT，但商学院案例研究的问题是：你的核心产品会怎样？这显然是一个值得思考的权衡。'

I mean, we've seen these screenshots of people saying, Well, if the Google search button was just Gemini or if AI mode was everything, I mean, yeah, sure, you could pass ChatGPT tomorrow, but then the business school case study is what happens to your core product, obviously, which is certainly an interesting trade off to think about.

我想聊的第三个故事是这些循环融资交易。

The third story I want to talk about is these circular financing deals.

我们看到了不少这样的案例，比如OpenAI、甲骨文和英伟达的三角关系，还有英伟达、微软和Thropic的合作。

We saw a bunch of them, whether it was the OpenAI, Oracle, and NVIDIA Triangle, there's also NVIDIA, Microsoft, and then Thropic.

你在通讯中提到一个观点，我觉得挺有意思，那就是这些循环融资交易在2025年之前其实并不新鲜。

You made a point in your newsletter that I thought was kind of interesting, which is that these circular financing deals weren't actually new as of 2025.

事实上，我们已经看到这类交易持续了好几年了。

We've actually been seeing them for a few years now.

是的，没错。

Yeah, exactly.

你知道，今年由于这些交易规模巨大，这已经成为新闻中的一个热门话题。

You know, I think this has become a really big topic in the news this year just because of the sheer size of some of these deals.

我们看到，在某些情况下，这些交易高达数千亿美元。

We see them, you know, in some cases hundreds of billions of dollars.

但没错，这种现象已经持续了相当长一段时间，自2019年微软首次投资OpenAI以来就一直存在。

But yeah, this has been going on for quite a while, ever since, you know, 2019 when Microsoft made its first investment in OpenAI.

那笔投资至少部分是以云服务积分的形式进行的。

That was at least partially made up of cloud credits.

但我觉得，过去一年里，这类交易的规模确实达到了前所未有的水平。

But yeah, I think this has definitely reached new levels of just the sheer size of these deals this past year.

这确实是新出现的现象。

And that's something that is definitely new.

你认为明年我们会看到更多这样的交易吗？

And do you think we see more next year?

当然。

For sure.

是的。

Yeah.

我的意思是，我确信这将继续延续到新的一年。

I mean, I think this will definitely continue into the new year.

我认为现在这些交易的不确定性增加了，人们对这些交易的疑问也更多了。

I do think that there is more uncertainty around these deals now and more questions around these deals.

我的同事阿尼萨写了一篇非常棒的文章，谈到我们将在2026年看到更多数据中心的延误。

My colleague Anissa wrote a really great story about how we're going to see more data center delays in 2026.

我认为，如果我们继续看到这些延误，甚至出现某方退出某项交易的情况，这种循环融资的故事可能会崩溃。

I think if we continue seeing these delays or maybe even some cases where a party might back out of a certain deal, that could bring this kind of circular funding deal story crashing down.

但我认为至少在今年上半年，这种情况肯定会持续下去。

But I think at least for the first half of the year, will definitely keep on going.

很好。

Great.

好了，斯蒂芬，感谢你来参加节目。

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

这位是斯蒂芬妮·帕拉佐洛，我们《The Information》的AI记者。

That is Stephanie Palazzolo, our AI reporter here at The Information.

好的。

Okay.

继续我们的2025年回顾，Neo Clouds是今年备受关注的一类公司。

Continuing with our 2025 recap, Neo Clouds were one class of companies that got a lot of attention this year.

Suro Capital将其基金的很大一部分投资于CoreWeave，正是这类公司之一。

Suro Capital holds a big chunk of its fund in CoreWeave, one of those exact companies.

我想邀请Suro的合伙人威利·李，来谈谈他对这个领域未来走向的看法。

I want to bring on Willie Lee, a principal at Suro, to talk more about where he sees that space going.

威利，欢迎来到节目。

Willie, welcome to the show.

很高兴你能来。

It's great to have you here.

嘿，阿卡什，很高兴来到这里。

Hey Akash, it's great to be here.

所以我非常期待聊聊苏罗对新云公司的投资，但先快速提一下，我们昨天看到了Meta收购Manus的交易。

So I'm excited to talk all about Suros investments in Neoclass, but very quickly, I mean, we saw this deal yesterday, Meta acquiring Manus.

你对这笔交易怎么看？

What did you make of that deal?

我认为这实际上与你提到的新兴云公司和计算支出密切相关。

So I think it actually ties very well to what you're talking about with the neo clouds and compute spend here.

所以，如果你看看Manus在宣布其年经常性收入达到1亿美元时的公告，就能看到该公司在令牌使用和虚拟计算机使用方面的某些数据。

So if you look at what Mannus' announcement was when they passed a 100,000,000 in ARR, you can see some components of the token usage as well as the virtual computer usage that the company was using.

因此，他们宣布自己使用了超过8000万个虚拟计算机实例，并且到目前为止，其生命周期内总共使用了超过147万亿个令牌。

So they announced that they did over 80,000,000 of virtual computers instances, and then they also did over a 147,000,000,000,000 tokens over their lifetime so far.

我认为这反映了当前人工智能领域中计算使用量的激增，尤其是初创公司也展现出万亿级别的令牌使用量。

And I think that speaks to a lot of the compute usage that you're seeing directly within the AI world right now is this explosion of token usage, even within startups, showing trillion token usage.

我认为这也说明了我们在苏罗看到的需求——这些公司迫切需要计算资源来推动人工智能的运行。

And I think it speaks to the demand we're seeing here at Sero for a lot of these companies that are just hungry for compute for to make AI work.

所以你的意思是，这反映了Manus的计算使用情况。

So you're saying that, I mean, this speaks to the compute usage that Manus has.

我只是想知道你对Meta与之的契合度有什么看法，你觉得Meta会如何使用智能体？

I just wondered if you had any thoughts on fit with Meta, how you could see Meta using agents.

我的意思是，我们都在这里做预测。

Mean, we're all making predictions here.

你们那边有没有什么头脑风暴？

Any brainstorms over on

你的想法呢？

your end?

听我说，当你第一眼看到这个头条时，是不是有点困惑？

Look, I think when you first see the headline, it's a bit of a head scratcher, right?

因为你知道，他们的使用场景并不符合你预期中Meta的核心业务，但没错。

Because, you know, their use case is not really within the core competency of what you would expect Meta to do, but Right.

这主要是社交媒体方面的东西。

It's something that's sort of social media largely.

对。

Right.

但我觉得有几点值得串联起来，科赫。

But so there's a couple of points that I think is worth stringing together, Koch.

我认为是他们在这一年中的一些收购，比如语音AI，以及他们一直在做的企业业务。

And I think it's a couple of the acquisitions they made during the year, whether that be Voice AI, an enterprise that they've been doing.

此外，他们显然也是开源模型领域的领导者之一。

And then they've also obviously been one of the leaders within open source models.

当你思考开源模型及其使用场景时，其中一个最重要的应用场景就是企业本地部署。

And I think when you think about open source models and the use case for it, one of the greatest use cases for it is on prem deployments for enterprises.

我认为你看到的是，Meta正开始逐步涉足扩展其企业工具领域。

And I think what you're seeing is Meta starting to tiptoe within that area of expanding their enterprise tooling.

对。

Right.

我认为Manus就是一个很好的例子，展示了他们如何将强大的超大规模计算基础与他们在企业开源领域所追求的目标结合起来。

And I think that Manus is a great example of how they can combine their massive hyperscaler type compute backdrop with what they're trying to actually accomplish within the enterprise open source world there.

所以我认为，这正是他们试图开始构建的东西。

So I think I think is really what they're trying to start to form.

但没错，阿卡什，你请说。

But yeah, go ahead Akash.

不，我觉得这是一个很好的观点。

No, I think that's a great point.

实际上，我之前并没有太关注Meta在AI领域的企业级布局。

I actually, I didn't think too much about the enterprise build out that Meta had been embarking on in AI.

所以我觉得这是一个很好的观点。

And so I think that's a good point.

我想转向你在Suro对CoreWeave的投资。

I want to turn to your investment in CoreWeave at Suro.

所以，你知道，这家公司将其基金的很大一部分投在了CoreWeave上，我想聊聊Neo Clouds的事，但先快速确认一下，你是通过SPV投资了CoreWeave，对吧？

So, you know, the company holds a significant chunk of its fund in CoreWeave, and I want to get to the business of Neo Clouds, but very quickly, I mean, so you're an investor in CoreWeave through an SPV, is that right?

没错。

Correct.

是的。

Yeah.

我们确实和Magnetar合作。

We're with Magnetar, exactly.

能给我简单介绍一下SPV这个业务吗？

Walk me through a little bit of this business of SPVs.

我最近好像经常听到这个术语。

I seem to be hearing the term a lot more.

为什么投资者使用SPV会有帮助？

Why is it helpful for investors to use SPVs?

它们通常用来做什么？

What are they typically used for?

这仅仅是一种获取投资机会的方式吗？

Is this just a way to get access to investments?

我的意思是，帮我理解一下。

I mean, help me understand it.

所以我认为，你在成长阶段投资领域看到的是对某些轮次的资金分配变得更加严格。

So I think what you're seeing within the growth stage investment sphere is a tightening of allocations to certain rounds.

因此，你会看到一些公司——资金分配收紧只是其中一部分，但更重要的是，这些公司非常专注于执行，而投资者必须加大投入、提供资金，并希望与风投分担一些工作，以筹集这些资金。

And so you'll see some companies which, and maybe a tightening of allocations is one part, but really it's that the companies are very focused on execution and investors are having to step up and having to bring the funds and they want to share some of that work with VCs to bring those funds in.

因此，你看到了SPV的创建和普及，尤其是在这些轮次规模变大时，用来填补这类融资需求。

And so you've seen this creation of SPVs and proliferation, especially as these rounds have gotten larger to fill in these types of rounds.

所以，是的。

And so yeah.

请说。

Go ahead.

你认为在未来一两年内，SPV会变得更普遍吗？

Do do you think that SPVs become more common in the next year or two?

我认为它会变得越来越普遍。

So I think it will become increasingly common.

我认为非常明显的是，特别是如果你在纽约，关注过有关SPV的新闻，就会知道显然在预售IPO前SPV股权方面出现了停滞。

What has been very apparent, I think, especially as you've seen if you're in New York and you've seen some of the news around SPVs is there's obviously been an arrest within, you know, selling pre IPO SPV stakes.

因此，我认为这展示了作为散户投资者尝试这样做的一些风险。

And so I think it shows some of the dangers of trying to do this as a retail investor.

Searo 这家公司一直在努力为散户投资者提供进入私募市场的机会。

Searo as a as a company has really been trying to provide retail act retail investors access to the private markets.

我认为，通过展示 SPV 世界中的复杂性，这进一步印证了 Soro 所做工作的价值——即提供经过筛选的私募市场准入，由我们来处理这些复杂 SPV 和公司的尽职调查工作。

And I think showing the complications of the SPVs in the worlds there, this gives more credence to what Soro's doing, which is trying to provide that curated private market access where we are handling really the diligence of some of these complicated SPVs and the companies.

我认为，要求散户投资者完成所有这些工作是极不现实的，因此你需要真正理解这个市场的人，他们了解 SPV 的双层、单层结构以及随之而来的复杂性，因为这是一个日益复杂、且 SPV 正变得越来越普遍的领域。

And I think that's a really hard expectation for a retail investor to be able to do all that type of work, which is why you need to have people who really understand this type of market, who understand SPVs, double layer, single layer, and the complications that come with that, because it is an increasingly, I'd say, complex space and they are becoming increasingly common.

那我们来谈谈 CoreWeave。

So let's talk about CoreWeave.

我们在节目中一直在讨论 Neo Cloud，这些 AI 专用云提供商的看涨理由是，正如你所说，市场需求巨大，市场增长如此迅速，即使存在竞争，也仍有空间让 Neo Cloud 进入，或许不一定抢走市场份额，但至少能获得足够的业务实现可持续发展。

We've been talking about Neo Clouds on the show, and the bull case for these AI specialized cloud providers is that, just like you said, I mean, there's so much demand and the market is growing so quickly that, hey, sure there's competition, but there is room for Neo Clouds to come in and maybe not steal share necessarily, but at least generate enough business to be sustainable.

这就是看涨的理由。

That's the bull case.

那你认为这些 Neo Cloud 的看跌理由是什么？

What do you think is the bear case for these Neo Clouds?

我认为，看空的理由之一是GPU领域的效率大幅提升，你会看到部署这些技术的能力变得商品化，大型云服务商可以自行满足全部需求。

Look, I think for the bear cases that you see a huge increase of efficiency within the GPU space, and I think that you'll see a commod a commoditization of the ability to deploy these things where, you know, these hyperscalers can build entire demand on their own.

但我不认为我们正在看到这种情况。

Now I don't think that's what we're seeing.

我认为我们看到的是这些系统的复杂性进一步增加，因为技术发展得太快了，如何高效地部署这些设备成了难题。

I think what we're seeing is even further complication of these setups because the technology is moving so fast and how do you rack these efficiently.

你可以看到Manus和其他一些公司正在努力优化计算效率，这简直是金矿。

And you can see in Manus and some of these other companies trying to optimize compute efficiency is sort of the golden goose.

这是他们的利润率。

It's it's their margin.

这关系到他们产品的表现。

It's how their performance on their product goes.

因此，我认为我们仍处于优化机架部署、计算能力及其他相关方面的初期阶段，而这些新兴云服务商已经证明，大型云服务商在这些方面确实遇到了一些困难。

And so I think we are still at the very beginning stages of being able to optimize racking and compute and all these aspects that the neo clouds have really shown that, you know, hyperscalers have struggled with this a little bit.

因此，他们找到了一个巨大的细分市场，专门提供这种类型的计算与优化服务。

And so they found a huge niche providing that type of compute and optimization.

让我问你一个有点挑衅性的问题，关于

Let me ask you a bit of a provocative question here about

关于

about

CoreWeave。

CoreWeave.

你认为这家公司在未来十八个月内被收购的可能性有多大？

Do Do you think there's a chance that the company gets bought in the next eighteen months?

我的看法是，为什么云服务商不直接收购像CoreWeave这样的公司，以帮助整合这些资源呢？

I mean, from my view, I seem to think that why wouldn't a cloud provider just snap up a business like CoreWeave to help consolidate these things?

当然，这是完全有可能的。

Sure, it's certainly possible.

我认为，可能对这种多元化云服务感兴趣并试图收购的买家只有寥寥数家。

I think what you would see is that there's only a handful of suitors that might be interested in trying to buy sort of a diversified cloud approach.

你看，谷歌或微软实际上都专注于自己想支持的特定模式。

You see, you know, Google or you see Microsoft really focusing in on a specific model that they wanna support.

我认为，亚马逊正在更加广泛地扩展，试图在支持和优化哪些模型方面变得更加中立，以适应这类基础设施。

And I think, you know, you have Amazon, which is broadening a little bit more on, you know, trying to be a little bit more agnostic on who which models they're supporting and optimizing for that type of infrastructure.

因此，我认为可能会有潜在买家对这种多元化的计算基础设施感兴趣。

And so, I think you would maybe have a suitor that might be interested in that type of diversified compute infrastructure.

其他公司可能没那么感兴趣，但毫无疑问，一家公司可能会对CoreWeave在大规模计算方面的优化能力感兴趣。

Others might not be as interested, but it's certainly possible that, you know, a company might be interested on, you know, how optimized CoreWeave has been at a scaled approach to compute.

我认为，真正能够实现大规模LLM级别计算能力的公司并不多。

And I think there's not that many companies that have really got that sort of scale down, the ability to provide compute at sort of massive LLM scale.

现在，我最后问你一个问题。

Now last question for you.

你们也是OpenAI的投资人。

You are also investors in OpenAI.

你认为OpenAI在2026年面临的最大风险是什么？

What do you think is the biggest risk for OpenAI going into 2026?

我认为，OpenAI在计算或其消费者领域处于非常有利的位置。

Look, I I think that OpenAI is in a really great position in terms of computer or their kind of consumer segment.

他们显然是消费者体验LLM的默认领导者。

They are clearly the the the de facto leader for consumers to, you know, experience what it is to use an LLM.

我认为，他们正在努力解决的一个问题，而Anthropic已经证明在这方面取得了很大进展，那就是企业市场。

Now I think one of the things that they're trying to figure out and which Anthropic has shown that, you know, they have a lot of traction there is, you know, on the enterprise side.

我认为OpenAI正在继续专注于消费市场。

And I think OpenAI is, you know, continuing to lean into the consumer side.

我认为，到2026年，他们将开始专注于成为企业市场中的强劲竞争者，并希望这能继续展现出强劲的势头。

I think that, you know, in 2026, one of the focuses that they'll try to start to do and hopefully that will show continued traction in is, you know, being a a strong competitor within the enterprise.

这是一个巨大的市场，而我们已经看到了编码领域采用的冰山一角。

And that's such a huge segment, and we've seen, you know, the tip of the iceberg with coding adoption.

我认为，你还会看到其他领域成为竞争战场，LLM需要在企业市场的不同方面展开激烈竞争。

I think you'll continue to see other areas in which it'll be a battlefield and it'll be very competitive for the LLMs to try to, you know, compete on different aspects of the enterprise.

我认为，如果OpenAI能够竞争并证明自己在企业市场也能获胜，这将表明这类公司的估值潜力无限。

And I think OpenAI trying to compete and showing that they can win on the enterprise side too will show that, you know, the sky's the limit for the valuation for that type of company.

很好。

Great.

好吧，威利，感谢你前来做客。

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

这位是威利·李，Suro Capital的合伙人，出现在TITP节目中。

That is Willie Lee, a principal at Suro Capital here on TITP.

好的。

Okay.

今天我们正式启动2026年预测系列，由《信息》杂志推出，这意味着在未来几天，我们的记者将基于深入的调研发布长篇预测。

We are kicking off our twenty twenty six prediction series today at the information, which means over the next few days, our reporters will be publishing long form predictions based on their extensive reporting.

今天的预测是：苹果公司将扭转其人工智能领域的低迷态势。

Today's prediction is that Apple will reverse its AI slump.

我想邀请我们的苹果记者亚伦·蒂利登场，他撰写了这篇报道，为我们解读他的思考。

I want to bring on Aaron Tilly, our Apple reporter who published that piece to walk us through his thinking.

亚伦，欢迎再次做客我们的节目。

Aaron, welcome back to the show.

很高兴你来到这里。

It's great to have you here.

是的，谢谢邀请我。

Yeah, thanks for having me.

所以，根据亚伦·蒂利的说法，2026年将是苹果与人工智能的一年。

So 2026 is going to be the year of Apple and AI, according to Aaron Tilly.

我非常期待讨论这个话题。

I'm really excited to talk about this.

你为什么认为这会发生呢？

Why do you think that this is going to happen?

所以，我的想法是，苹果会以他们缓慢而稳健的方式显得很聪明。

So, yeah, the way I'm thinking about it is that Apple so Apple's going to look smart, I think, with their sort of slow and steady approach.

这就是我的看法。

That's the way I think about it.

我认为这主要会因为市场很可能会对科技行业其他公司大规模建设数据中心的做法产生逆转。

And the way I think it's going to be primarily it's going to happen because the markets will likely continue to turn on sort of the massive data center build outs we've seen with the rest of the tech industry.

而苹果是唯一一家没有卷入这场混乱的公司。

And Apple is really the only one that stayed out of that mess.

所以我认为他们的做法是，他们没有盲目加入这场竞争，反而显得很聪明。

So I think their approach is going to be they're going to look smart how they've taken that, you know, not really jumped into the fray there.

而且，他们将推出比以往更好的AI产品，这就足够了。

And you know, they're going to deliver something in AI that's going to be better than what they had before, and that's going to be enough.

我的意思是，他们不需要成为AI领域的领导者，但这种稳扎稳打的方式看起来才是正确的。

Like, just I don't think they need to be an AI leader, but just their slow and steady approach is going to look like the right one.

好的。

Okay.

所以，这一切都取决于大家翘首以待的Siri更新。

So a lot of this rests on this Siri update that everyone is waiting on.

我想知道，你的消息来源关于这个Siri更新进展如何？

I wonder what you've heard from your sources in terms of progress on this Siri update.

你有没有听到一些我们可以期待的更新内容？

Are there things that you're hearing we can expect?

谈谈这方面的情况吧。

Talk a little bit about that.

是的。

Yeah.

我的意思是，他们目前正在测试内部和外部的AI模型，以增强这个改进版的Siri。

Well, I mean, they're currently testing out an internal and external AI models to power this improved Siri.

我认为他们会依赖外部模型，来大幅提升Siri的诸多能力。

I think they're going to rely on external models to get this to improve a lot of its capabilities.

我认为这是正确的做法，因为苹果在试图自主研发时屡屡失误，一直难以掌握AI所需的大量计算资源和海量数据集。

And I think that's going to be the right approach because Apple has fumbled time and time again with trying to develop something in house and it really struggling to master a lot of what they need in AI, which is massive compute and massive data sets.

而这两点正是苹果所缺乏的。

And those are two things Apple doesn't have.

因此，我认为依赖外部模型是实现更好Siri、更好AI体验的必然选择。

So I think relying more on external models was just going to be an inevitability to delivering a better Siri, a better AI experience.

所以他们会大力依靠这一点。

And so they're going to lean on that.

我认为这将解决他们的一些问题。

And I think that will fix, you know, some of their problems.

现在，我们在这档节目中和你讨论过一件事，那就是无论苹果做什么，它仍然主要依靠硬件，尤其是iPhone来赚钱。

Now, one of the things that we've talked about on this show with you is the fact that whatever Apple does, it still makes most of its money on its hardware, obviously, and its iPhones.

所以我想问你一个问题：假设Siri更新发布了，效果非常好。

And so one of the questions I have for you is, okay, so let's say the Siri update comes out, it's great.

假设苹果看起来很聪明，就像你说的，他们采取了稳扎稳打的方式。

Let's say Apple looks smart, like you said, they've taken a slow and steady approach.

他们依赖这些外部合作伙伴。

They rely on these external partners.

我的意思是，这在未来一两年内能为苹果带来多少实际收入呢？

I mean, much money does this really generate for Apple, you know, in the next year or two?

因为，我不太清楚，即使Siri变得更好，人们是否会因此购买更多iPhone，对吧？

Because, I mean, it's not clear to me that people are going be buying more iPhones even if Siri is working better, right?

是的，我认为这不会带来iPhone销量的大幅增长。

Yeah, I don't think it's going to result in some massive wave of growth in the iPhone.

但我认为他们会制定一种策略，通过第三方AI来赚钱。

I do think they'll develop a strategy where they can make money off of third party AI.

我的意思是，苹果的生态系统策略在于应用商店以及向全球用户和设备销售访问权限和计费服务，这才是他们能够获得服务收入的地方。

I mean, it's the ecosystem play for Apple where the app store and selling access to their billing users and their billings of devices out in the world that they're going to find some sort of service revenue that they can make money from.

你有什么想法，他们该如何实现这一点呢？

You have any ideas on how they could do that at all?

就是与不同的AI进行集成。

Just just integration with with different AIs.

AI，你可以选择你的AI。

AI, you can choose your AI.

你可以选择你的潜在选项。

You can choose your potentially.

据我了解，有些讨论涉及让你自己选择驱动你苹果设备的AI助手。

This is I understand some of discussion potentially choosing your own AI agent powering your your your Apple.

所以可能并不一定是Siri，也不是通过提供这种访问权限来赚钱。

So maybe not necessarily Siri and having, you know, having in making money by offering that access.

所以，就像应用商店那样，只是通过AI来实现。

So just, you know, just somewhat similar to the App Store just through AI.

我们显然知道，服务业务是公司增长最快的类别，因此这肯定会对此有所帮助。

And we obviously know that services is, I think, the fastest growing category of the business, and so that certainly would help it out there.

最后一个问题。

Last question for you.

你在预测中没有提到这一点，但苹果公司最近发生了高管变动，有人离职，有人新加入。

You didn't talk about this in your predictions, but we've had these executive shuffles at Apple, people stepping away, new people coming in.

你预计2026年还会发生更多这样的变动吗？

Do you foresee more of that happening in 2026?

还是你觉得他们已经基本完成调整了？

Or do you think that they've kind of done their shuffling?

我认为我们会看到更多变动。

I think we're going to see more.

意思是，今年二月，苹果将举行年度股东大会。

Mean, so in February, Apple will be holding its annual shareholder meeting.

我认为董事会将会有重大调整。

I think there's going to be some big board changes.

主席是亚特·莱文森。

It's Chairman Art Levinson.

他现在75岁了，可能不得不退任了。

He's going to have to step down now that he's 75 probably.

将会有一位新的董事长。

And there's going to be a new chairman.

我认为苹果公司需要就其首席执行官的继任计划提供某种明确性。

I think there's going to have to be some sort of clarity Apple offers in terms of its succession plans with its CEO.

我不认为这件事会像有些人认为的那样立即发生。

I don't think it's happening imminently like some do.

我认为这可能还有一两年的时间，但苹果公司将会有一些董事会变动。

I think it's still potentially a year or two out, but there's going to be some board changes in Apple.

向股东和员工提供明确的信息是明智的，因为目前围绕这家公司和蒂姆·库克作为首席执行官的地位，存在着大量疑问。

Think would be smart to provide clarity there to its shareholders and to its employees, because there's just a lot of questions swirling around this company and where Tim Cook stands as its CEO.

很好。

Great.

好了，亚伦，感谢你前来做客。

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

这位是亚伦·蒂利，我们在《信息》杂志的苹果记者。

That is Aaron Tilly, our Apple reporter here at The Information.

好的。

Okay.

埃隆·马斯克今年也是硅谷许多重大新闻的核心人物。

Elon Musk was also a central part of many of the biggest stories in Silicon Valley this year.

在很多方面，特斯拉的汽车业务退居次要地位，让位于埃隆追求的其他宏大目标。

In many ways, Tesla's automotives business took a bit of a backseat to other grand ambitions that Elon has pursued.

我的同事西奥·韦特本周发表了一篇专栏文章，探讨了埃隆今年为他的企业设定的更广泛目标及其达成情况。

My colleague Theo Waite has a column out this week looking at the broader goals that Elon has set for his businesses this year and how they shaped up.

我想请西奥来聊聊这些。

I want to bring on Theo to talk all about that.

西奥，欢迎再次回到节目。

Theo, welcome back to the show.

很高兴你来这里。

It's great to have you here.

能来真好。

Good to be here.

所以，快速问一下，你有没有看到XAI在圣诞节期间在孟菲斯的圣诞灯饰？

So very quickly, did you see the XAI Christmas lights in Memphis over the holidays?

我还没去过那里，但我觉得灯饰应该还在亮着。

I haven't made it over there yet, but I think they're still running.

好的。

Okay.

我们接下来会讲这个。

We'll be after this.

我鼓励大家去看看我们在圣诞前夜录制的那期节目，以了解更多背景信息。

Well, I encourage everyone to check out our episode that we taped on Christmas Eve for more context on that.

让我们谈谈特斯拉的汽车业务。

Let's talk about Tesla's automotive business.

他们将在本周五公布第四季度的交付数据。

So they've got the fourth quarter delivery figures coming out this Friday.

我们该期待什么？

What should we expect?

它们可能远低于第三季度，因为第三季度末电动汽车的联邦税收抵免政策到期了。

They're probably gonna be significantly lower than than the third quarter, which is because there was a federal tax credit for EVs that expired at the end of the third quarter.

所以很多原本打算在感恩节或圣诞节前后购买特斯拉的人提前下单了。

So a lot of people that were thinking about, you know, maybe buying a Tesla around Thanksgiving or Christmas bought it early.

当把第四季度的数据加到全年其他数据中时，几乎可以肯定2025年的销量将比2024年更差，而2024年本身也已经比2023年更差，情况并不乐观。

And then when you add that fourth quarter to the rest of the figures for the year, it's almost certainly gonna show that 2025 was a worse year than 2024 for sales, and 2024 was already a worse year than 2023, so not great.

好的。

Okay.

现在，我想谈谈埃隆更宏大的抱负，但你能解释一下，为什么特斯拉要公布这些交付数据吗？

Now, I want to talk a little bit about Elon's broader ambitions, but can you just explain to me why does Tesla report these delivery figures at all?

既然他们每季度都有汽车收入数据，为什么投资者还需要这些交付数据呢？

Why is it helpful for investors to have this when they have the automotive revenues every quarter anyway?

投资者喜欢这个，因为它更及时。

Investors like it because it's more up to date.

比如，他们在每个季度结束后的第一个交易日就公布交付数据，这样你就能立即获得更更新的信息。

Like, they they put out the delivery figures the first trading day after a quarter is over, so you can get, you know, a lot more up to date info immediately.

而财报通常要等到季度结束一个月左右才发布。

Whereas, you know, earnings happen, like, a month or so after the quarter's over.

这并不是特斯拉独有的做法，福特也做类似的事情。

It's not unique to Tesla, you know, Ford does something similar.

明白了。

Got it.

明白了。

Got it.

嗯，我的意思是，我不确定。

Well, I mean, I don't know.

特斯拉似乎经常打破常规。

It seems like Tesla breaks with habit a lot of times.

所以特斯拉唯一的一点是

So One of the only things that Tesla

是的。

does yeah.

与其他行业一致

In line with the rest

汽车行业的做法。

of the automotive sector.

没错。

Exactly.

所以现在，如果我们 broadly 看看埃隆这一年的表现，你曾在我们的周日洞察简报中写过一篇关于埃隆为2025年设定的雄心壮志，以及他实际实现这些期望和目标的程度。

So now if we look at Elon's year broadly, you wrote this piece on in our Sunday Insights briefing about Elon's ambitions that he set for 2025 and the extent to which he actually lived up to those hopes and goals.

谈谈这方面吧。

Talk a little bit about that.

是的。

Yeah.