Meta’s 8,000 Job Cuts, SpaceX’s IPO by the Numbers, Google's Universal Shopping Cart

欢迎大家收看Informations TITV。

Welcome everyone to the Informations TITV.

我叫阿卡什·帕什里卡。

My name is Akash Pasricha.

今天是5月20日，星期三。

It is Wednesday, May 20.

今天盘后，华尔街都在关注英伟达的财报表现。

Wall Street is looking out for NVIDIA results after the bell today.

与此同时，我们也在关注SpaceX的关键一周，它正为一场重磅IPO做准备。

We're also tracking a critical week for SpaceX as it readies for its blockbuster IPO.

今天还有几条重要的媒体动态。

There's also some big media news out today.

詹姆斯·默多克收购了《纽约杂志》和福克斯播客。

James Murdoch is buying New York Magazine and Fox Podcasts.

而福克斯媒体旗下其他部分，包括The Verge和PopSugar等网站，将由现任福克斯媒体总裁瑞安·波利接手，组建一家新公司。

The other half of what we know as Fox, which includes websites like The Verge and PopSugar, will be will become part of a new company run by Ryan Polly, current president of Fox Media.

今天节目的第一件事，是谷歌正在举办年度开发者大会IO。

First up on the show today, Google is hosting its annual IO developer conference.

我们的谷歌记者王艾琳正在现场，稍后她会连线，跟我们分享她在现场看到的最新动态。

Our Google reporter Erin Woo is at the event and joins us in just a second to tell us about what she is learning on the ground.

我们还会用三张图表，深入分析SpaceX的IPO，帮大家理解这次事件的规模和背后的故事。

We'll also dig into the SpaceX IPO with three charts that can help us make sense of the scale and story of this moment.

《信息报》还独家报道了企业如何调整软件供应商合同，为AI领域的巨额投入腾出空间。

The information also published exclusive reporting about how companies are adjusting software vendor contracts to make room for heavy spending in AI.

我们也会聊聊Meta计划的裁员，最后用每周的财经专栏收尾，讲讲最近备受关注的金融科技公司Parker破产的内幕。

We'll also talk about Meta's planned layoffs, and we'll wrap the show with our weekly finance column and the inside story of the bankruptcy at Parker, a Fintech company that's gotten a lot of attention recently.

这期节目内容特别丰富，咱们马上开始。

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

今年的谷歌IO大会一开场就引爆了话题，推出了大量基于AI的新功能和工具。

Google has made a splash with its IO event this week, debuting a long list of new features and tools leveraging AI.

我们的谷歌记者王艾琳一直在现场跟进。

Our Google reporter Erin Woo has been on the ground.

我想请她来跟我们聊聊她听到的消息。

I want to bring her on to talk to us about what she is hearing.

Erin，欢迎来到节目。

Erin, welcome to the show.

很高兴你又回来了。

It's great to have you back.

昨天的主旨演讲上，谷歌发布了一长串新功能和新工具，真让人兴奋。

We had the keynote yesterday and a long list of new exciting announcements that Google announced.

你给我们梳理一下最重要的几项，你最关注哪些？

Walk us through the big ones and what you are paying attention to.

是的。

Yeah.

毕竟是谷歌，一整天下来有成百上千个新消息。

So this is Google, so there's a million different announcements.

有几个明显的趋势，比如智能代理就特别受关注。

Some major themes, obviously, agents were big.

搜索功能即将上线新的智能代理。

There's new agents coming in search.

这真的挺有意思的。

This is really interesting.

搜索现在能自动上网查找公寓房源之类的资讯，之前还举过球鞋发售的例子。

Search will be able to scan the web for information on things like apartment listings or like sneaker drop was the other example.

Gemini应用里还会新增一些个人助手功能，叫Gemini Spark。

Also, some personal assistant features coming to the Gemini app, Gemini Spark.

还有新的购物功能，比如一个通用购物车，可以把不同商家的商品都集中到一起，后台还会帮你做比价和研究。

And then there's some new shopping announcements as well, a universal cart that lets you take products from different retailers, like put them all in one place and it'll, like, do some research for you on the back end.

另外推出了一款新的视频生成模型，这点很特别，因为谷歌在这块的投入力度是其他公司没法比的。

And then a new video model, which is interesting because Google is really leaning into this one kind of like nobody else is.

咱们一个个详细聊聊吧。

So let's unpack each of these a little bit more.

视频这块，我觉得大家应该还能理解。

Video, I think people can kind of get their head around.

我的意思是，他们从一开始就在视频和图像生成方面做得非常出色。

Mean, you know, they've been they've been They've been doing some great things in video and image creation from the start.

搜索功能一直是个大问题。

Search has always been a big question.

我们之前聊过，他们怎么平衡搜索和Gemini之间的关系？

We've talked to a bit on how are they going to handle Search, Gemini, the balance between the two?

你对他们的做法有什么看法？他们真的在做了。

What was your reaction to I mean, they're doing it.

现在搜索基本上就是Gemini了吧？

The search is now really Gemini, right?

对。

Yeah.

我的意思是，昨天的发布让这个问题变得更突出了——搜索和Gemini到底有什么区别？

I mean, I think the announcements yesterday really re upped that question of like, what is the difference between search and Gemini?

为什么非要搞两个不一样的东西？

And like, why are there two different things?

说实话，我真的不清楚。

To be honest, like I do not know.

谷歌一贯的说法是，我们只是在满足客户当前的需求。

Answer that Google always gives is that, oh, like we're meeting like customers where we're at.

我个人搞不懂，为什么会有两个庞大的团队，各自开发几乎一模一样的产品，却把它们放在不同的地方。

I personally do not understand why there are two massive teams that are building two very similar products and then putting them in different places.

也许正因如此，我才不用谷歌。

But that maybe is why I do not write in Google.

但这两者正在越来越融合。

But they are very much coming together.

我们看到搜索功能越来越多地加入智能代理特性，能处理更复杂的回答了。

Like we're seeing more agentic features in search and like the ability to do much more like complex answers.

他们甚至在二十五年来首次扩展了搜索框，让它支持多媒体输入，也能处理更长的问题。

Like they're expanding the search box for the first time in twenty five years so that it's more multimedia and you can ask longer questions.

所以这两者之间的界限正在越来越模糊。

So there's a lot of convergence between these two things.

关于购物这块，我是说，咱们 broad 一点讲，大家反应怎么样？

On the shopping piece, what was the I mean, let's go broadly.

与会者对这些新功能的实际反馈如何？

What was the reaction on the ground to all of these features from attendees?

说实话，这事儿涉及的方面太多了。

I mean, again, there's a lot of different things.

我们还没聊到编程呢。

We haven't even talked about coding yet.

我接触的那些人，对新的搜索助手都很兴奋。

I think there's a the people that I was talking to, there's excitement about like the new search agents.

大家对通用购物车也很感兴趣。

There's excitement around the universal cart.

别忘了，谷歌在购物领域一直很成功。

Remember, like Google has had a lot of success in shopping.

就拿我个人来说，谷歌航班、谷歌酒店，还有那些能跨网站比价预订的功能，早就成了人们生活中不可或缺的一部分。

Like, I know like for me personally, like Google Flights, like Google Hotels, like a lot of this comparison booking stuff that like can pull across the Internet, it's stuff that they're really strong on and that really like becomes a part of people's lives.

我觉得这些智能助手的功能，对普通人日常生活来说，还是有点太复杂了。

I think some of the agent stuff, it feels like this is still a little bit more complicated than anyone actually needs in their personal life.

我跟那儿一个人聊过，他开玩笑说，每个谷歌高管的生活好像都得靠助手安排：周末有什么活动、去哪条徒步路线、晚上吃哪一家意大利菜，听起来挺棒的。

I was talking with someone there who was joking that like every Google exec, it seems like their life is like, and now I must look for events to do on the weekend and like what hikes I'm going to do and like the Italian food I'm going eat for dinner, which like great.

说实话，这种生活方式确实挺让人羡慕的。

That frankly seems like a great way to live.

但很多这类功能，普通人大概真不需要专门找个助手来帮自己规划周末行程。

But also like a lot of these things, it doesn't necessarily seem like a normal person needs, like, an agent, like, custom coding a schedule for their weekend.

其实这些事，自己动手安排一下也就完了。

Like, it seems like that's kind of something that, again, like, you can just kind of do.

话是这么说，但你可能不是每周都用得上，可一旦需要的时候，它就在那儿等着了。

Well, but, I mean, look, you might not need it every week weekend, but, you know, the moment that you do, that's what it's there for.

对吧？

Right?

我承认，也许是我才26岁，生活还没那么复杂，但我问过几个有孩子的朋友，他们也说：我们真不觉得需要这个。

I'm willing to accept that maybe I am just 26 years old and my life is not that complicated, but I did ask friends with kids and they were like, No, we don't feel like we need this either.

那也行。

So that's Okay.

有道理。

Fair.

那编程呢？

So what about coding?

对。

Yes.

对我来说，如果抛开普通网民的角度，从科技记者的视角来看，这确实是我在采访前最关心的问题。

So I mean, like for me, this was one of the big like if you take it less from like a person who uses the internet perspective and more from a tech reporter perspective, like This was really the big question that I had coming in.

就像我们在这档节目里说过无数次的，现在的AI领域，Anthropic在编程方面表现非常出色。

The AI framework right now, as we've talked about on the show a million times, is that Anthropic is really good at coding.

他们有个叫Methods的模型，虽然还没正式发布，但已经被炒得沸沸扬扬，Anthropic在编程这块确实一马当先。

They have this methods model that has been teased and very much hyped, not quite released, but Anthropic is really dominating on coding.

所以我一直很好奇，谷歌会不会突然推出一个重磅产品，迎头赶上，直接把其他人甩在身后。

So a big question was whether Google was going to come into this with some big splashy release towards like they catch up and they're going to blow everyone out of the water.

他们发布了一些与编程相关的更新。

And so they had some coding announcements.

他们演示了一个叫 Anti Gravity 的编程平台。

They had a demo of anti gravity, like their coding AFib platform.

他们还推出了一款较小的模型，Gemini 3.5 Flash，希望它能成为开发者更快速、更高效的工具选项。

And they also had not a big new model, but like a smaller model, Gemini 3.5 Flash, that they're hoping will be like kind of like a faster, like more like workforce option for developers.

虽然他们发布了几项编程相关的更新，但目前看来还没引起太大反响。

So they had some coding announcements, but I don't think anything that feels like it's made a big splash thus far.

我觉得他们主打的编程产品 Anti Gravity 在 X 上刚发布就收到了不少争议。

And I do think that anti gravity, their big flagship coding release is kind of getting mixed reviews on X immediately.

我想回到马克·扎克伯格的目标，就是专注打造个人超级智能。

Well, I just want to go back to sort of I'm thinking about Mark Zuckerberg's goal to prioritize personal super intelligence.

我也在想谷歌这次推出的那些面向消费者的產品。

I'm thinking about the consumer centric products that Google unveiled here.

现在的问题是，Meta 什么时候才能真正接近个人超级智能这个目标。

There is a question about when Meta will get truly close to that mission of personal superintelligence.

我几乎觉得，昨天的主旨演讲是我们迄今为止离目标最近的一次，撇开可穿戴设备不谈。

I almost feel like yesterday's keynote was the closest that we've sort of got putting wearables aside from them.

我知道可穿戴设备才是个人超级智能的真正方向。

I get that the wearables are supposed to be the true north of personal superintelligence.

但从消费产品来看，谷歌在用户使用搜索方面明显遥遥领先。

But insofar as consumer products, Google really seems to be far and away ahead of the pack in terms of people who are actually using search.

他们还有YouTube和视频业务。

They've got YouTube and video.

我的问题是，他们有没有编程工具，真的那么重要吗？

I mean, does it really matter that they don't have coding, I guess, is is my question to you.

是的。

Yeah.

我觉得你说得很有道理。

And so I think I like, I think that is a very fair point.

我觉得谷歌最大的优势在于，它无处不在——垄断地位、反垄断问题之类的。

I think that, like, again, the big advantage Google has is that Google is everywhere, illegal monopoly, antitrust, etcetera.

但谷歌无处不在。

But like Google is everywhere.

这些产品都是这样。

Like all of these products are

你一定是谷歌的记者吧。

things that You must be a Google reporter.

真厉害。

Wow.

包装上写得清清楚楚。

It's what it says in the carton.

这些产品都是人们每天都在用的东西。

All of these products are things that, like people use every day.

比如谷歌文档里的AI功能，你已经在用的文档生成工具，还有Gmail里的AI，搜索里的AI。

And so it's like AI in Google Docs, like the document builder you're already using, like AI in Gmail, like AI in search.

所以谷歌拥有巨大的用户覆盖面。

So like Google has this huge distribution surface.

说到编程，我觉得首先一点是。

Like what I would say about coding is one.

一方面，编程能赚很多钱。

So like there's one side of this, which is that coding makes a lot of money.

比如Anthropic，它是历史上收入增长最快的公司。

Like Anthropic is like the fastest growing revenue business in history.

这是因为所有这些企业都愿意为能写代码的AI支付高额费用，毕竟软件工程师的薪资太贵了。

And that's because like all of these enterprises are willing to pay a lot for AI that can code because like software engineers are very expensive.

另一方面，这还牵扯到通用人工智能和智能爆发的概念。

The other side of this is around the idea of AGI and AGI takeoff and artificial general intelligence.

所以有一种观点认为，要实现超级智能——不是个人层面的，而是真正意义上的技术奇点——就得打造能自我提升的AI。

And so there's a mode of thought, which is that the way to get to superintelligence, not like personal superintelligence, but like the real like singularity superintelligence is by building AI that can improve itself.

因此，Anthropic坚定地认为，实现这一目标的关键就在于编程。

And so Anthropic very firmly believes that like the way to get that is via coding.

编程。

Coding.

谷歌那边也有类似的情况。

And there's some of that at Google as well.

谢尔盖·布拉特顿特别重视这个编程方向，其中一部分目标就是为了实现通用人工智能。

Mean, Sergei Bratton has really focused on this coding push and part of that is trying to get to AGI.

谷歌的做法稍微有点不同，他们还特别关注构建世界模型，比如视频多模态技术就是这方面的工作。

Google does take a little bit different approach to this where a lot of it's also around trying to create a world model and this is where someone like the video multimodal stuff comes in.

但我觉得，从更大的角度看，编程的意义就在这里。

But I mean, that I think is the bigger picture of coding.

它既关乎利润，也关乎通用人工智能。

It's about money and it's also about AGI.

我昨天想到一个问题，想问问你有没有和一线的人聊过：当谷歌把AI嵌入到每一个产品里时，会不会影响它的利润率？

Now, one thought that I had yesterday that I wondered if you talked about with folks on the ground is the risk to Google's margins at all, as it embeds AI into literally everything.

我们都知道，部署这些技术的成本一点都不低。

We know that the cost of implementing all this is not negligible.

当然，我们也得承认，谷歌是少数几家完全负担得起这些投入的科技巨头之一。

Any discussion on the ground or any concern from folks that And we should say that, look, Google is one of the biggest tech companies that can afford a lot of this.

它手头有大量的自由现金流可以支配。

It has tons of free cash flow to go around.

但这个问题有人提过吗？

But did that come up?

我觉得开发者们并不会考虑这一点。

I don't think I mean, that's not something that developers think about.

参加大会的人也不会想这些，他们只会觉得：哇，太棒了。

That's not something that attendees of the conference, they're like, Oh, great.

现在我有这么多新的 AI 工具可以用。

Now there's all this new AI that I can use.

这显然也是财报电话会上讨论的热点。

I mean, this is something that's come up obviously on earnings calls.

这对公司来说是个大问题。

It's a big question for the company.

谷歌目前的说法，而且目前看来也站得住脚，第一点是他们在努力降低提供这些服务的成本。

And what Google has been saying, and I think thus far this seems to be very out, is A, that they're really working to reduce the cost of serving all of this.

谷歌早就大规模运营基础设施了。

So Google has operated infrastructure at scale for a really long time.

所以他们正在拼命降低这些服务的成本。

And so they're working very hard to make this stuff cheaper to serve.

另一面呢，今天正在举办的是谷歌营销大会，也就是他们的广告峰会。

And the other side of it, so like what's the other thing going on today is Google Marketing Live, which is their advertising conference.

但我们看到的是，他们也能赚更多钱，部分原因就是AI——广告投放量增加了，用户搜索也更频繁了。

But like what we're seeing is that they're also able to make a lot more money, like in part because of AI, where they're able to run more ads and like people are searching more.

另外，谷歌云业务也在快速增长。

And so I think like and there's also the growing Google Cloud business.

所以这在推高谷歌成本的同时，也大幅提升了他们的收入。

And so this is increasing Google's costs on the one hand, but it's also increasing the amount of revenue they are able to make.

很好。

Great.

埃琳，谢谢你今天来参加。

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

这位是 Erin Woo，我们《The Information》的谷歌特派记者。

That is Erin Woo, our Google Reporter here at The Information.

展望 SpaceX 的上市，我的同事 Cory Weinberg 和 Valita Pao 本周发布了一篇报道，里面有五组图表，能帮你理清这场历史性 IPO 的来龙去脉。

As we look ahead to the SpaceX IPO, my colleagues Cory Weinberg and Valita Pao have a story out this week with five charts that should help you make sense of this historic IPO.

我想请 Cory 上来，帮我们解读一下他整理的这些数据。

I want to bring on Cory to help us unpack some of the data that he pulled together.

Cory，欢迎再次做客节目。

Cory, welcome back to the show.

观众们对你的 SpaceX 评论和分析应该都很熟悉了，还有你做的原创报道，不过今天我们还是来仔细看看你整理的这些图表。

Viewers will be pretty familiar with your commentary and your analysis on SpaceX, also your original reporting, but I want to go through these charts that you put together.

一共有五张，但我觉得真正讲清楚故事的，其实是其中三张。

And there's five of them, but really three, I think, tell the story exceptionally well.

我想先从这张第一张图开始。

I want to start with this first chart here.

你对比了 SpaceX 和其他公司的 EBITDA 数据。

You looked at the basically, EBITDA figures for SpaceX compared to other companies.

我们在这张图上该关注些什么呢？

Tell us, what should we be paying attention to here on this chart?

是的。

Yeah.

我特别注意到的是，SpaceX在调整后的EBITDA利润和传统会计口径下的净利润之间，差距大得惊人。

What I found very striking was the very wide gap between how SpaceX will be describing its profit on an adjusted EBITA basis versus how it's more traditional accounting profit or loss in in net income.

很多观众应该都知道，公司向投资者展示调整后的利润指标是很常见的做法。

So as I think a lot of viewers will know, it's quite common for companies to sort of present adjusted profit metrics to investors.

这本身没什么奇怪的。

Nothing nothing weird about it.

但通常这些调整后的数字和实际利润没差这么多，这点让我觉得很突出。

But the they aren't usually this different than the actual profit, and that's what's stuck out to me.

这张图显示，即使在其他资本密集型行业，比如扣除折旧这类主要支出后，SpaceX的差距也要大得多。

The chart just shows, like, even in other capital intensive businesses where you would take out a key expense in depreciation, SpaceX's is gap is is much, much wider.

所以有几个问题。

So a couple of questions.

我知道调整后的EBITDA通常比净利润好看多了，图表就说明了这一点。

I know that adjusted EBITDA is often a much nicer figure than net income, as the chart shows here.

投资者看SpaceX这样的公司时，会自己做调整吗？

When investors look at a company like SpaceX, will they make their own adjustments?

但归根结底，这对SpaceX的故事真有影响吗？毕竟大家心里都清楚。

And ultimately, does that actually matter to the SpaceX story, given that everyone is sort of clear?

这可是埃隆·马斯克的公司。

This is an Elon Musk company.

你想怎么调整都行。

You can make whatever adjustments you want.

这并不会动摇核心逻辑。

It doesn't really take away from the core thesis.

是啊，这么做到底图个啥？

Yeah, what's the point of all this?

我干嘛要费这劲儿啊，阿卡什？

Why am I even doing this, Akash?

问这个问题方向不对。

That's not the right question to be asking.

不，听我说，我弄这个并不是想拿它当决定性证据，也不是说它利润没宣传的那么高。

No, look, I didn't put this together with sort of the idea of like, this is going to be the smoking gun, or this is going, this is like the Like, it's not as profitable as it says.

这只是为了更清楚地展现这家公司的实际运营状况，和其他类型的企业做个对比。

It's a way to visualize and understand, like, the actual business in contrast with other, you know, other types of businesses.

所以投资者会想搞明白，这家公司各个关键驱动因素到底是怎么运作的。

And so, you know, investors are going to be trying to understand, like, the different levers of the business.

比如星链的单用户盈利模型是怎样的？

Like, what's the unit economics of Starlink?

还有星舰到底什么时候才能真正起飞？

And, you know, what's the the actual timeline for Starship to get to get up and away?

但我还是觉得，投资者得真正吃透这些历史数据，搞清楚它们能告诉我们未来哪些信息。

But I still think, like, you know, investors are going to have to, like, truly digest, like, the historical figures and understand, you know, sort of like what it can tell us about the future.

明白了。

Okay.

那我们接着看第二张图，说到星链。

So let's move on to a second chart here, speaking of Starlink.

你收集了一些关于SpaceX发射任务的数据，区分了星链相关发射和非星链发射。

So you collected some data on the number of launches at SpaceX that were affiliated with Starlink, and then the non Starlink launches.

从这些数据里，你最注意到的是什么？

What stood out to you here about this data?

我觉得很多人没意识到的是，SpaceX作为主要的火箭发射服务商，正越来越多地把运力用在自己的互联网业务上，这个比例已经相当大了。

What I think people don't appreciate is broadly, is the fact of how large and how much of a growing percentage it is for SpaceX to use its dominant position as a rocket launch provider for its own Internet connectivity business.

大家都知道，SpaceX主要用猎鹰九号火箭把星链卫星送入轨道。

I think, you know, it's broadly known SpaceX uses Falcon nine, its its main rocket to bring its satellites into orbit that sort of help power Starlink.

但这些任务加起来，已经占了SpaceX全部发射次数的四分之三左右。

But it makes up a a about three quarters of the total launches that SpaceX runs.

所以当SpaceX的业务板块重新划分时，主要的三个部分会是：星链的互联网服务、火箭发射业务，以及AI业务——也就是xAI和推特。

And because of that, when you're gonna look at SpaceX's three business segments when the s one flips, it's going to be connectivity, which is Starlink space, which is the rocket launch business and AI, which is xAI and and Twitter.

你会发现，火箭发射业务的增长率只会是个位数。

And you'll see the space launch business growing single digits.

每年增长百分之七到八，你可能会想，为什么SpaceX会这样呢？

It's growing seven or 8% year over year, which you would think, why why is that happening for SpaceX?

它明明主导着整个市场。

Like, dominates the market.

因为真正为外部客户执行的发射任务，只占它总发射量的一小部分。

And it's because the actual launches it's doing for outside customers is a very small percentage of the overall launches that it's doing.

没错。

Right.

所以，如果你看这张图，你是不是觉得红色条代表的发射占比，会继续保持多数？

And so in other words, if you look at that chart, is your expectation that the red bar here in the chart, a proportion of total launches, the red bar should remain a majority.

而且未来可能还会进一步上升。

And if it might actually increase proportionally going forward.

我们的预期就是这样吗？

Is that our expectation?

是的。

Yeah.

我的意思是，这在航天行业里是个热门话题，尤其是当SpaceX开始更多谈论把计算能力部署到太空时，这种发射主导地位对整个行业意味着什么？

I mean, I think that's like a big hot topic in the space industry broadly, especially as SpaceX starts talking more about using its rockets and satellites to put compute into space, which is what is this kind of launch dominance going to mean for the rest of the industry?

其他公司是不是得排长队，才能把载荷和卫星送入太空？

Will they have to wait in a long queue to be able to deliver their payloads and their satellites into space?

任何想把计算设备送上太空的公司，都得通过SpaceX这一关吗？

Will any company will any other company that wants to put compute in space need to run through the the toll that is SpaceX?

还是说SpaceX会直接用它的火箭掌控整个链条，自己把所有事情都做了？

Or will SpaceX just be using its rockets to kind of own the entire stack and do it all themselves?

说实话，现在还说不清楚未来会怎样。

Like, there's gonna be we can't quite the story is not clear today.

但我认为，未来几年，围绕SpaceX的市场地位，以及它如何用自家火箭服务自身业务而非第三方客户，一定会引发巨大讨论。

Like like but I think in the years to come, there's going to be a huge conversation around SpaceX's market position and how it uses its own rocket launch capabilities for its own purposes versus third party customers.

明白了。

Okay.

你刚才提到了这些领域的增长。

So now you talked about the growth of all these segments.

我想提一下第三张图，它展示了 SpaceX 内部 xAI 部门的增长情况，也就是大家常说的 SpaceX 人工智能部门。

And so the third chart I want to bring up here looked at the growth of the xAI segment within SpaceX, otherwise, I guess known as the SpaceX AI segment here.

但这一块的收入增长相当缓慢。

But the revenue growth here is pretty meager.

我们预期未来会好转吗？你那边听到什么风声了吗？

Do we expect this to get better over time, or what what are you hearing?

他们刚把一批算力卖给了 Anthropic，这会是一笔不错的收入来源。

Well, they just sold a bunch of their compute to Anthropic, and that'll probably be a nice little revenue jewel.

他们还收购了 Cursor，预计今年晚些时候会把这家公司整合进 xAI 业务里。

They bought, you know, they have agreed to buy Cursor, and, you know, we'll expect they'll fold that into the xAI business later this year.

这会带来一波明显的提振。

So that'll be a nice jolt.

但我觉得，这正好解释了为什么他们最近急着做这么多大手笔的商业交易——花六十亿美元收购 Cursor，把大量算力租给 Anthropic。

But I think this helps explain why they're sort of seemingly scrambling to make all these pretty extreme, you know, sort of business deals, spending 60,000,000,000 on Cursor, renting out a bunch of their compute to Anthropic.

根本原因在于，他们的模型和核心 AI 团队相比行业领头羊 OpenAI 和 Anthropic，落后得太明显了。

It's simply because their models and their their kind of key AI division have been a a pretty significant laggard compared to the bells of the industry, which are are OpenAI and Anthropic.

当然，这些数据也有一些局限性。

And, obviously, there's some caveats with this data.

xAI 还包含了 Twitter，而 Twitter 本身已经有稳定的收入来源，这也会影响增长表现。

X xAI obviously also incorporates Twitter, which has, you know, sort of a an established revenue base, which can affect growth as well.

但不管怎么分析，都能看出去年 AI 业务的增长远不如 OpenAI 和 Anthropic 那么亮眼。

But I I think any way you slice it, you know, sort of it shows that last year, the AI business was was not at

根本没达到我们看到的 Anthropic 和 OpenAI 那种爆发式增长的水平。

all the kind of huge growth driver that we saw for for Anthropic and OpenAI.

大家是不是还期待更多类似 Anthropic 那样的交易？比如把算力租出去，让 xAI 进入云服务领域？

Do do people expect more deals like the the Anthropic deal, renting out compute, you know, xAI getting into the neo cloud business.

这短期内能带来收入增长，帮助他们充分利用现有的算力资源。

Again, this is a short term jump in revenue that could help them make use of all the capacity that they have.

在 SpaceX 这圈子里，有没有人讨论过，也许 xAI 最终就变成一家云服务商？

Is there a discussion at all amongst folks in the SpaceX orbit that maybe xAI just becomes a neo cloud company?

收购 Cursor 让这种可能性变小了。

Buying Cursor of makes that less likely.

我觉得他们确实是开发者AI和企业级应用领域的重要玩家，这个市场体量巨大，xAI迟早得去抢这块蛋糕。

I think you know, the they they are a real player in the kind of, you know, sort of developer AI and broadly enter enterprise application world, which is a huge market that xAI is going to need to go after.

而且他们需要海量的算力。

And so and they need a ton of compute.

所以，你是说我就干脆变成个Neo Cloud算了？

So, yeah, like, is that to say I just can become a Neo Cloud?

不可能。

Like, no.

这反而会拉低他们的估值。

It won't help their valuation.

这类业务的估值倍数通常都比较低。

Those businesses are usually valued on a lower multiple.

所以，要是Cursor这笔交易真成了，你估计就看不到太多这种转向了。

And so, yeah, I think if this cursor deal gets done, you're gonna not see a ton of that

我再问你一个问题，关于xAI的。

Let me let me ask you one more question on the xAI point here.

我想知道，当你跟人聊SpaceX上市的事时，xAI这个话题会占多少分量？毕竟太空业务才是支撑整个估值的核心。

I wonder when you're talking to people about the SpaceX IPO, how much does xAI actually come up in the conversation given that you know, the space part of the business is really the thing that is anchoring a lot of this?

其实，上市故事的重心完全在算力和太空中的数据中心，还有地面上的数据中心。

I mean, so much of the IPO story is about compute and data centers in space and data centers on the ground.

至于Croc正在开发的模型，或者xAI业务带来的实际收入，大家可能根本没太关注。

And so the actual model that Croc is developing and the xAI business, like the actual revenue you see coming through there, you know, people are are maybe not too focused on that.

大家关心的，是能不能拿到算力，以及怎么把算力送到太空去。

It's all about the actual access to compute and the actual access to get compute into space.

但这些现在都还很理论化，完全是着眼于未来。

But that's still very you know, that's very theoretical at this point, or it it's very it's very forward looking.

所以，你要想花一百到两千亿美元买下SpaceX，就得相信这套未来叙事。

And so that's what that is the story you need to believe to buy SpaceX at $1 to $2,000,000,000,000

明白了。

Great.

Cory，非常感谢你今天来分享。

Cory, I want to thank you for coming on.

这位是科里·温伯格，我们《信息》杂志财经部的副总监。

That is Cory Weinberg, our Deputy Bureau Chief of Finance here at The Information.

随着人工智能成本持续攀升，我的同事劳拉·布拉顿最新调查发现，企业软件买家已经开始调整合同，为持续的人工智能投入留出空间。

As the costs of AI show no signs of getting any less severe, my colleague Laura Bratton has new reporting that buyers of enterprise software are starting to adjust those contracts to make room for these ongoing AI investments.

我想请她来跟我们分享一下她的发现。

I want to bring her on to share with us what she knows.

劳拉，欢迎再次做客我们的节目。

Laura, welcome back to the show.

很高兴你能来。

It's great to have you here.

你能跟我们说说，你这篇文章里提到的这些情况具体是指什么吗？

Tell us a little bit about what you mean here in your story.

你提到一些公司正在尝试缩短企业软件的合同期限。

You talk about companies trying to make enterprise software contracts shorter.

能详细解释一下吗？

Unpack that for us a little bit.

是的。

Yeah.

所以企业软件的客户希望合同更有弹性。

So customers of enterprise software want more flexibility in their contracts.

他们希望既能继续使用传统软件，又能在将来替换为AI工具，或者改用其他软件系统时，合同周期更短一些。

They want to be able to use traditional software, but for shorter terms in the case that they want to eventually rip out those traditional software systems for AI tools, or they want to switch to use different software systems.

我觉得这一点很重要，因为人们谈论SaaS的末日时，总是非黑即白。

And I think this is important because we hear the SaaS pocalypse talked about in really black and white terms.

要么说软件已经完蛋了，要么说AI会让软件大幅升级，成为最大的赢家。

Either software is dead or software is going to be improved a lot by AI and is going be a big winner in all of this.

但在我看来，我跟客户的交谈，还有我收集到的那些案例，最清楚地反映了软件公司眼下面临的压力。

And I think to me, my conversations with customers and the anecdotes I gathered were really the clearest signal of the near term pressures that software companies are facing.

那我们来聊聊这些压力吧。

So let's talk about those pressures.

我是说，你具体指的是Anthropic调整定价模式这类情况吗？还是其他什么压力？

I mean, are we talking specifically here about Anthropic changing its pricing model, or what pressures are you talking about?

是的。

Yeah.

所以这其实是一个间接的关联——随着这些AI工具的兴起，比如Anthropic的Claude Code、Claude Enterprise、Claude CoWork这些产品，尤其是Claude Enterprise这套企业级工具，现在在企业里最受欢迎。

So it's really an indirect link where the rise of all these AI tools, and I think we've seen that Anthropic Claude Code, Claude Enterprise, Claude CoWork, like the Claude Enterprise sort of accounts and suite of tools, are the most popular among enterprises.

随着AI提供商和软件公司普遍转向按使用量收费，企业客户在选择软件系统、分配预算以及评估软件回报率方面变得越来越挑剔。

As AI providers and software providers in general begin to charge customers based on usage, we're seeing that enterprise software buyers are becoming a lot more picky about which software systems they use and how they're going to spend their budgets, what return they're seeing from the software that they use.

你能不能帮我理清楚这一点？

So help me understand this.

如果他们希望和企业软件公司签的合同周期更短，这怎么能在短期内减轻开支呢？

If they want their contracts with enterprise software companies to be shorter, how does that help with the expenses in the short term?

在我印象里，我明白不签五年长期合同，而是改签两三年的合同，这层意思。

Because in my mind, I sort of understand how I'm not committing to a contract that is maybe five years in length, I'm committing to something that is maybe two or three years.

这部分我懂。

So I see that part.

但那怎么改善我短期的现金流呢？

How does it help me with my cash flow in the short term then?

是的。

Yeah.

我的意思是，缩短合同本身并不一定意味着短期内支出会减少，但确实让你在发现没达到预期回报时，能更早地停止付费。

I mean, it doesn't necessarily the part about the shortened contracts doesn't necessarily make it so that you're spending less in the short term, but it does make it easier to halt that spending at a nearer point in the future if you decide that you're not getting the return that you want.

所以这其实对软件公司的收入可预测性构成了威胁，而不是眼下收入的总规模。

So really it's a threat to the predictability of software firms revenues rather than the, you know, gross value of those revenues at this at this moment.

我还听顾问们提到，这些软件合同中的净留存率正面临压力。

And then another thing that I heard from consultants is that net retention is under pressure within these software contracts.

也就是说，续约的客户在增加支出方面，远不如以前那么积极了。

And what that really means is that customers that are renewing their contracts aren't growing their spending with their software vendors as much as they used to.

我觉得这清楚地表明，客户现在变得更挑剔了，不再像过去那样轻易扩大合同范围。

And so I think that that's another really clear signal about how customers are being pickier and not necessarily expanding the scope of their contracts as much as they were previously.

合同变短了，意味着他们一两年后就可能离开用了多年的软件供应商，而以前他们可是签三到五年合同的。

The contracts are shorter, which means that they could, you know, potentially leave their software vendors that they've had for a long time in one or two years, whereas they used to sign like three to five year contracts.

那软件供应商会满足客户，提供他们想要的短期合同吗？

And are the software vendors giving them the shorter contracts that they want?

他们有没有调整条款，或者涨价来弥补这一点？

Are they changing the terms at all or increasing the price to make up for it?

这得看情况。

It really depends.

我们看到很多企业级软件公司都在转向按客户使用AI的量来收费，比如ServiceNow就说过，这可能会让客户的支出大幅增加。

We've seen a broad shift among enterprise software players to shift to charge customers based on how much AI they use, which, know, for example, ServiceNow said that that could increase the spending by customers a lot or significantly.

具体数字我记不太清了，好像是未来几年会增长百分之十五到二十左右。

I can't remember the exact figure, I believe it was around 15 or 20% in the coming years.

所以我觉得，随着厂商调整定价模式，他们能让客户为AI工具支付比以前更多的钱。

And so I think that as vendors shift their pricing models, they're going to be able to see customers pay more than they were previously for their AI tool.

我觉得这可以看作是对行业变化的一种应对，也是他们想抢先一步的做法。

So I think that that can be seen as a response to this industry shift and them trying to get ahead of that.

但据我听说，有些厂商确实同意了这些更短的合同期限，这是从顾问和客户那里听来的零星信息。

But from what I've heard, some vendors are agreeing to these shorter contracts just by anecdotally what I'm hearing from consultants and customers.

另外，你之前也提到过，这种模式让企业客户更容易逐步减少合同支出。

Now, other and you hinted to this earlier, this idea that it makes it easier for these enterprise software buyers to maybe taper off their contracts over time.

你之前提到，这些企业软件公司的净留存率正面临压力。

You talked about net retention being something that was coming under pressure at these enterprise software companies.

能再给我们详细解释一下吗？

Can you unpack that a little bit more for us?

我的意思是，真正的风险在于，合同缩短后，客户不再续约。

I mean, this is basically the idea that this this is the big risk is is them not renewing after the shortened contract.

对吧？

Right?

是的。

Yeah.

净留存率下降，并不等于客户不续约了。

So net retention doesn't necessarily mean that customers aren't renewing their contracts.

而是说，续约的时候，他们不再增加对这家供应商的支出。

It just means that they're not increasing their spending with the provider when they do renew those contracts.

我跟帕森顿的一位高管聊过，那是安永旗下负责战略和并购的部门。

So I talked to an executive at Parthenon, which is the strategy and M and A arm of Ernst and Young.

她告诉我，帕森顿旗下数百家客户的净留存率在过去六到九个月里下降了大约15个百分点，我觉得这清楚地表明，软件公司正面临压力。

And she told me that net retention that she's seeing among Parthenon's hundreds clients is down about 15 percentage points, in the last six to nine months, which I think is a clear signal that software- That's

这幅度不小。

a lot.

确实挺大的，对吧？

That's a- Yeah, seems like a lot, right?

不过净留存率还是在100%左右。

Mean, it's still around 100%.

客户确实还在增加支出，但相比以前，增长幅度小多了。

So, you know, customers are growing their spending to some degree, but it's less than they did previously.

我觉得这正是软件公司当前所承受压力的早期信号。

And I think that that's an early signal of the pressure that software companies are seeing.

软件行业里，有没有哪些细分领域特别受影响？

Are there specific segments of the software sector that in particular are under pressure here?

我在想，比如客户关系管理软件、身份认证的安全软件，还有数据存储这些领域。

I'm just thinking about, there's CRM software, there's, I don't know, cybersecurity identity authentication software, you know, there's data a storage.

那哪个领域风险最大？

Like, which one is most at risk here?

我经常听到有人提到企业资源计划系统。

One that I kept hearing come up was enterprise resource planning.

就是

Which

你说的是哪种？

is like, like what?

比如领英这样的平台。

So For for a LinkedIn.

对。

Yeah.

这类系统存储着关键的业务数据和应用程序，ERP软件主要由SAP、甲骨文、微软等公司提供。

That's it it stores critical business data and application ERP applications are offered by like SAP, Oracle, Microsoft, a bunch of companies.

我听到的主要是这个领域。

And that's one that I heard come up.

但我认为这还得走着瞧。

But I think it really remains to be seen.

我觉得首席信息官和首席技术官们，也说不准三年后到底该用哪些软件系统。

And I don't think that chief information officers, CTOs really know which software systems that they're going to want to use in three years.

他们也不知道哪家软件公司能在这场所谓的AI竞赛中胜出，哪些AI工具会更出色。

And they don't know which software vendors are going to win this, you know, quote unquote AI race and which AI tools are going to be better.

这恰恰说明，我们谁都不确定。

And that's kind of the point is that we don't know.

所以他们才更看重灵活性，更倾向签短期合同。

And that's why they want more flexibility and they want shorter contracts.

有些情况甚至还包括退出条款，这让我觉得特别有意思。

And in some cases, opt out provisions, which I found really interesting.

也就是说，软件供应商，如果你没能按约定日期提供相应的AI功能，我就有权提前终止合同。

And that is basically they're saying, software vendor, if you do not provide promised AI features by x date, I can exit this contract early.

这又是另一个被提到的点。

And that's another thing that came up.

所以不仅仅是合同周期变短了。

So it's not just shorter contracts.

还有一些更有利于客户的条款。

It's also some other sort of more favorable terms.

我觉得这清楚地表明，客户在与软件供应商谈判时，比过去有了更大的话语权。

And I think really this is just a broad signal that customers have more power in negotiations with their software vendors than they did in the past.

很好。

Great.

好了，劳拉，感谢你今天来做客。

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

这位是劳拉·布拉顿，我们TIT平台《应用人工智能通讯》的作者。

That is Laura Bratton, author of our Applied AI Newsletter here on TIT.

明白。

Okay.

Meta已经开始通知数千名员工，他们将被裁员。

Meta has begun notifying thousands of employees that they are being laid off.

公司今天将裁撤8000个职位，约占公司总员工数的十分之一。

The company is set to cut 8,000 jobs today, which amounts to roughly one tenth of the company's overall headcount.

公司同时也在重组组织架构，将把数千名员工调配到新的岗位上。

The company is also restructuring its organization and will move thousands of other workers to new roles.

为了深入分析，我们邀请到美银美林的董事总经理兼互联网股权研究分析师劳埃德·沃尔姆斯利。

For more analysis, I want to bring on Lloyd Walmsley, Managing Director of and Internet Equity Research Analyst at Mizuho.

劳埃德，欢迎来到节目。

Lloyd, welcome, show.

很高兴你能来。

It's great to have you here.

能来我也很高兴。

It's great to be here.

谢谢邀请我。

Thanks for having me.

据我们所知，Meta今天的裁员已经启动了，虽然我们早就预料到会有这一天。

So as best we know, these layoffs are underway today at Meta, and I know we've been expecting these for a long time.

依你看，这家公司裁员这个决定合理吗？

Tell me, in your opinion, are these layoffs warranted at the company?

我觉得，现在这些公司开发软件的方式正在发生变化。

Look, I think we're seeing a shift in how, you know, these companies develop software.

有了人工智能，他们用更少的人就能做更多事。

And, you know, with AI, they can do a lot more with fewer people.

所以这肯定是原因之一。

So I think that's part of it.

另外，我觉得马克在公司整体管理上确实做得很好，纪律严明。

And then, look, I think Mark has done a good job of establishing very tight discipline at the company in general.

随着他们在云基础设施和AI产品开发上不断增加资本支出和运营支出，他们也在严格控制利润率和营业利润。

And as they spend more in CapEx and for that matter, OpEx on cloud expense to build out AI product, AI infrastructure, they are holding the line on, you know, margins and operating income.

考虑到他们在这么多领域投入这么大，他选择持续收紧开支，这很明智。

And so I think it's wise of him to try to be constantly tightening the belt given how much they're investing in so many areas.

所以现在大家讨论的这些裁员——比如Block这类公司的情况——其实核心问题是：哪些是之前招人太多导致的，哪些才是真正因为AI带来的效率提升？

So this whole discussion about how much of these layoffs now I'm talking the layoffs broadly that we've seen at companies like Block, etcetera, is a question, which ones are the result of overhiring and which ones are the result of actual AI efficiencies?

你对这个比例怎么看？

Where do you stand on the split here?

这八千人的裁员，到底是AI带来的真正效率提升，还是之前招太多人了？

Are all these 8,000 AI efficiency true cuts or is it the result of over hiring?

我觉得两者都有，很明显。

Yeah, I think it's a little bit of both clearly.

Meta这家公司，从疫情结束后就一直很克制。

The company, Meta in particular, has been very disciplined ever since coming out of the pandemic.

说实话，他们在广告业务上遇到的挑战，不仅来自疫情后的复苏，还有苹果IDFA政策调整的影响。

Honestly, some of the challenges they saw in the ad business both coming out of the pandemic but also post the, you know, the IDFA changes from Apple.

当收入增长放缓时，他们就彻底重视起了利润和成本控制。

You know, they, you know, really when the revenue growth slowed down, they got religion on margin and cost.

所以他们在裁员和控制增长方面一直很严格。

So they've been pretty tight in terms of reducing headcount and being disciplined in growth.

我觉得，比起我们跟踪的很多公司，他们在人员调整上确实走在了前面。

So I think they, more than a lot of companies we follow, they've been ahead of the pack in right sizing.