Microsoft’s 18% Revenue Growth YoY, Meta’s AI Investments Weighing On Q2 Profits

欢迎大家收听The Information的TITV节目。

Welcome everyone to the Information's TITV.

我是Akash Pasricha。

My name is Akash Pasricha.

今天是7月30日，星期四。

It is Thursday, July 30.

在开始今天的节目之前，我想先提一下我们新闻编辑部刚刚发布的一篇关于台积电的报道。

Before we get to today's show, I want to flag a new story our newsroom just published on TSMC.

The Information的记者Liu报道说，台积电正在开发一种新的先进芯片封装技术，这项技术将能与英特尔目前的技术相抗衡，这也再次表明芯片制造已成为AI热潮的战场。

The information's channel Liu reported that the company is developing a new advanced chip packaging technology that would rival what Intel has right now and is yet another sign of the chip of chip making becoming the battleground of the AI boom.

这篇报道现在已经在我们的网站上了。

That is posted right now on our website.

建议大家去看看。

I encourage you to check it out.

今天节目里，我们会深入分析Meta和微软的季度财报。

Today on the show, we are unpacking both Meta and Microsoft's quarterly results.

我们的联合执行主编马丁·皮尔斯昨晚在专栏里称这是两家公司的故事。

Our co executive editor, Martin Pierce, called it a tale of two companies in his column last night.

股市显然也是这么想的。

The stock market certainly thinks so.

微软股价上涨了。

Microsoft is up.

Meta股价下跌了。

Meta is down.

我们将和两位分析师聊聊他们对这些业绩的看法，以及两家公司在AI建设方面进展如何。

We're going to chat with two analysts about their thoughts on the results and where both companies AI build outs are at.

接着，我们会和HashiCorp前CEO戴夫·麦詹尼特进行对话。

We're then going to have a conversation with former CEO of HashiCorp, Dave McJannet.

我会问他，代理式AI的转型和他职业生涯所依赖的云转型有哪些相似之处，又有哪些不同。

I'm going to ask him about how the agentic AI transition is similar, but also distinct from the cloud transition that he built his career on.

今天的节目会很精彩，我们直接开始吧。

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

Meta昨晚发布了第二季度的财报。

Meta reported second quarter results last night.

营收增速放缓至28%，低于上季度的33%。

Revenue growth decelerated to 28%, down from 33% last quarter.

自由现金流还不到去年同期的一成。

Free cash flow was less than a tenth of what it was a year ago.

我想请花旗的董事总经理兼资深互联网分析师Ron Josey来聊聊他的看法。

I want to bring on Ron Josey, Managing Director, Senior Internet Analyst at Citi for his thoughts.

Ron，欢迎来到节目。

Ron, welcome to the show.

很高兴你能来。

It's great to have you here.

谢谢，Akash。

Thanks, Akash.

很高兴来到这里。

Good to be here.

那么，Meta的财报中，你最关注的重点是什么？

So what was the headline from Meta's results for you?

你看，重点是广告业务依然表现非常出色，对吧？

Look, the headline is the ad business continues to do incredibly well, right?

在如此高的营收基数上还能实现28%的增长，这说明他们确实看到了强劲的需求。

28% growth on a revenue number that's incredibly high tells you they're seeing the demand.

我觉得股价的反应是由多种原因造成的。

I think you're seeing the reaction of the stock for a variety of reasons.

一个是第三季度的指引基本符合预期。

One is 3Q guidance came in relatively in line with expectations.

换句话说，数字并没有大幅上调。

In other words, numbers didn't really move up that much.

还有一些人可能原本期待更多。

You also had some people maybe expecting a little bit more.

毫无疑问，我们的观点是，在线广告行业实际上相当健康。

There's no doubt our view here is that the online ad industry is actually quite healthy.

你在Meta的美国营收增长中看到了这一点，大约在30%以上。

You saw that in US based revenue growth for Meta around 30 plus percent.

但也许在欧洲，从定向和个性化角度来看，未来会有一些小问题，这导致他们的第三季度指引下调了。

But maybe there is a little bit of, issues from a targeting and personalization perspective in Europe going forward that that brought their 3Q guidance lower.

但真正的争论焦点是，我们现在已经进入Meta超级智能实验室一年了。

But really the big debate is, okay, we are now a year into Meta Superintelligence Labs.

我们在资本支出和计算基础设施上花费了数千亿美元。

We are spending hundreds of billions of dollars on CapEx and compute infrastructure.

坦白说，我认为这是第一个季度，我们真正听到CEO扎克伯格谈论新产品以及产品推出的策略。

And frankly, I thought this was the first quarter where we actually heard CEO Zuckerberg talk about newer products and a strategy of products coming out.

然而，你知道，我的很多客户都在问，我们什么时候才能开始看到成果？

However, you know, a lot on a lot of my clients are asking, when are we gonna start seeing the results?

而我认为我们已经开始看到一些成果了。

And I would argue we're starting to see them already.

但是，你知道，越多越好。

But, you know, more is more.

我觉得我们现在正处于Meta这个新产品——可以称之为“文艺复兴”——的早期阶段，接下来几个月到几个季度，我们应该会看到更多新东西陆续推出。

And I think we're in the early stages of this new product, call it Renaissance at Meta, where we should start seeing just more coming out here in the next couple months and quarters.

你说已经看到成果了。

You say see them already.

为什么这么说？

Why do you say that?

对。

Yeah.

你看啊。

I mean, look.

其实想想挺有意思的，Instagram的全球互动量在双位数增长。

It's actually pretty fascinating to to to think about, engagement on Instagram growing double digit globally.

Instagram已经存在很久了。

Instagram has been around forever.

Reels也推出好几年了，但我们的用户互动时长还在以双位数增长。

Reels has been around for a number of years, and we're still growing engagement time spent by double digits.

在Meta和Facebook上，我记得视频观看时间全球增长了9%。

On Meta on Facebook, I think the number was it was up 9% globally time spent on video.

所以归根结底，我们看到人们更频繁地使用这个产品。

And so the bottom line is we're seeing people use the product more.

我们现在在Meta内部也听到，你知道，更多新产品要出来了。

We're hearing now internally at at Meta, you know, more newer, newer products coming out.

所以我们知道Instagram现在有20亿日活跃用户。

So we know Instagram's now at 2,000,000,000 d a DAUs.

Threads大概有5亿月活跃用户。

Thread is at, I think, 500,000,000 MAUs.

还有像Forum和Seller这样的新应用正在推出。

You have newer apps that are launching here with Forum and Seller.

所以我们应该会看到很多新产品陆续出来。

So we should start seeing a bunch of newer, products coming out here.

但归根结底，我认为因为整合了他们的LLM和AI，排名和推荐系统的改进正在推动这种更高的参与度。

But the bottom line is I think ranking recommendation improvements because of the incorporations of their LLM and AI are driving this greater engagement.

这感觉好像并没有体现在收入增长上，因为增速在放缓。你刚才提到了一点欧洲的情况，也许这就是你想说的重点。

It doesn't feel like that's showing up in the revenue growth per se because the deceleration was You talked about Europe a little bit, so maybe that's what your point to do here.

我听完之后，还是不太清楚为什么广告增长在放缓，而正如你所说，这些广告工具都在变得更好。

I walked away not having a clear picture of why ad growth is decelerating when all these ad tools are getting better, as you said.

是的。

Yeah.

我觉得，怎么说呢，有几个原因吧。

I mean, look, I think a few things.

以Meta这样的规模和体量，还能保持28%的增长，这已经相当了不起了。

To be of size and scale of Meta to be growing 28% is a pretty impressive feat.

我觉得，关于可能以25%以上的速度增长的说法，那是第三季度指引的上限，所以接下来的问题是，为什么不是更高呢？

I think the the comment of potentially growing at 25 plus percent, that's a high end of guidance for 3Q, that is the next sort of why wasn't it more?

而且，公司也提到了可能影响第三季度指引的三个关键问题。

And, you know, the company talked about three key issues that might have impacted their 3Q guidance.

其中一个就是展示次数的比较基数更高，这没问题。

One of them being tougher comps with impressions, fine.

另一个原因是，随着一些新的《数字市场法案》规则生效，我们在欧洲精准定位用户的能力受到了影响。

Another one being the ability to really target, users in in Europe as some of the new DMA rules come into effect.

这是我们新面临的一个问题，需要去应对。

That's something that's new that we need to deal with.

但归根结底，我们的规模和体量摆在那里，Meta在在线广告领域的许多同行，增长率都远达不到20%以上。

But the bottom line is we're at a size and scale that, you know, we're many of Meta's peers within the online advertising space not grow anywhere close to the 20 plus percent range.

而Meta在第三季度的增长率大约是28%或25%。

And Meta's growing 20 '8 20 '5, call it, in 3Q.

明白了。

Hear you.

还想了解更多。

Want to see more.

你提到了新产品，昨晚关于Meta的企业战略也有很多讨论。

You talked about new products, and there was a lot of talk last night about the enterprise strategy for Meta.

马克又来了那一套，说‘敬请期待’。

And Mark did his thing where he said, Stay tuned.

还有更多内容要公布。

There's more coming.

我们在电话里只能说这么多。

We can only say so much on the call.

但他们非常重视企业业务。

But they were leaning very heavily into enterprise.

他们没有提到编程工具，但我觉得他们是在往那个方向发展。

They didn't say anything about a coding tool, but it felt to me like that's where they're going with it.

我的意思是，企业业务对Meta来说一直是一条艰难的路，他们一直在努力攻克。

I mean, enterprise has been a long road for Meta that they've tried to tackle.

我有点好奇，你觉得两三年后他们会不会单列企业销售数据？

I'm sort of curious, do you see them breaking out enterprise sales, call it two, three years from now?

你知道，如果他们有编程工具，你觉得故事会往那个方向发展吗？

You know, if there's a coding tool, I mean, is that where you see this story going?

因为他们在这一块一直没什么起色。

Because it hasn't ever caught on for them.

对，这完全是另一回事，销售方式不同，购买方不同，买家不同，一切都不同。

Yeah, no, it's a totally different sale, a different purchase, a different buyer, different everything.

我觉得昨晚管理层说的很有意思，如果他们自己在开发内部编程工具，为什么不试着把它提供给其他人呢？

I thought the commentary from management last night that if they're building internal coding tools for themselves, why not try to offer that to others out there?

我觉得当API完全开放，更多人能接触到他们的新模型时，我们会获得更多洞察。

I think an unlock that we're gonna have more insights will be when the API is fully released and many others can can sort of access their newer models here.

所以这是即将到来的事情。

So that's something that's on the come.

说实话，从财报来看，有一点很突出，就是自有核心产品优先于企业业务。

One of the things that frankly coming out of the quarter, just to hear owned and operated core products take precedence over enterprise.

不过，这些核心产品要规模化总是需要时间的，对吧？

That said, it always takes a while for these core products to get to scale, right?

Threads 达到5亿月活用户，我觉得只用了三年，这简直难以置信。

It's taken believable unbelievably only three years, I think, for threads to get to 500,000,000 MAUs.

但我的重点是，考虑到对算力的需求，以及Meta可能在2027年开始供应算力，随着新产品的推出，这或许能缓解他们的自由现金流问题。

But my point being is it seems to me that given the demand for compute capacity and potentially Meta having supply starting in '27 as their newer products launch, this could help to alleviate some of their free cash flow problems.

所以，是的，我敢打赌两三年后，你会看到他们推出新的收入项目，但核心业务仍然是最重要的部分。

So, yeah, I bet in two to three years' time, you could see them roll out a new line item of revenue, but the core is still the most important part.

你觉得Threads算成功吗？

Was was Threads a success to you?

你怎么看？

What do what do you think?

我觉得算。

I think it is.

我觉得它证明了很多事情。

I think it proved a lot of things.

我觉得它证明了Meta不仅仅能做自己的Instagram、Facebook、WhatsApp这些应用。

I I think it proved that Meta can actually do more than just their own, call it, Instagram, Facebook, WhatsApp apps.

他们是靠内部团队自然发展起来的Threads。

So they built threads organically.

他们用很小的团队就做到了，而且很快就达到了规模。

They did it with a very small team, and it got to to scale very quickly.

今年才刚刚开始变现。

Monetization is just beginning this year.

所以可以说，从收入角度看，我们才刚刚看到投资回报的初期阶段。

So you could argue we're just in the early days of seeing that return on investment from a revenue perspective.

但关键的是，Meta 能够在其整个所谓的元宇宙或所有应用之间有效引流。

But the key thing here is the ability for Meta to really drive traffic across its entire, call it metaverse or- or all of its apps.

我觉得这才是关键。

I think that's the key thing.

所以，随着我们看到卖家或新应用推出，Meta 已经学会了如何通过 Threads 和其他产品来扩展这些产品。

So, as we see seller or form or newer apps launch, Meta has now learned how to scale these products with threads and others.

这种整合，我觉得真的很关键。

That integration, I think, is really key.

所以，你的观点是，如果三年前你问人们 Meta 推出全新产品的能力，这些产品不一定是对其他公司的模仿，我想答案可能是，Meta 到底推出了什么真正获得关注的新东西？

So, your view is that, I think if you would have asked people maybe three years ago on Meta's ability to launch net new products that were not necessarily copycat features of other companies, mean, answer, I think, probably at that point would have been, well, what is the new thing that Meta has launched that has really gotten traction?

你的观点是，他们推出了这些获得关注的新产品，这可能表明他们有能力推出像编程工具这样的东西，也能获得关注。

Your view is that they have launched these new products that have gotten traction, and that might be an indication of them being able to launch something like a coding tool that could also get traction.

我这样理解你的意思，对吗？

Is that what I'm reading from you?

对。

Yeah.

不，我觉得你说得对。

No, I think that's right.

而且我还要说，当你有36亿日活跃用户访问你的网站时，你就有巨大的规模和许可，能够为未来的其他新产品实现规模化。

And I would also say when you have 3,600,000,000 daily active people coming to one of your sites, you have tremendous scale and license really to be able to get to scale for other newer products down the road.

而且仅仅在Facebook内部，就有很多独立的业务。

And there are so many individual businesses just within Facebook alone.

比如看看市场，那里有机会。

Let's look at marketplaces, you know, that that have an opportunity.

最后我想说的是，Akash，关于商业代理这一块，我……

Last thing I would say, Akash, is to is, the business agent side, I.

嗯。

E.

现在，代理商能够直接通过Meta与消费者对话，这个功能已经火起来了。

The ability for agents to now converse directly with their consumers through Meta is something that's taken off.

这个功能大约两个月前，也就是六月初才刚推出。

That just launched in the about two months ago in early June.

现在已经有100万中小企业在使用它了。

So we now have a million SMBs on it.

好的。

Okay.

那我们来谈谈盈利能力吧，这正是投资者关注的核心。

So let's talk about profitability then, which was sort of the heart at what investors are reacting to.

自由现金流，我觉得今年这个季度还不到去年同期的十分之一。

So free cash flow, I think less than a tenth of what it was this quarter last year.

我是说，Ron Josey，低到什么程度才算太低？

I mean, Ron Josey, what would be too low?

我的意思是，他们可能开始烧钱了。

I mean, they might start burning cash.

这些信号是不是在说，我们已经到了一个临界点。

Are the signals that, Hey, we've hit a threshold.

我们有点走得太远了。

We're going a bit too far.

对。

Yeah.

我的意思是，根据我们的模型，我们预计他们今年、明年，甚至到2028年都会烧钱。

I mean, in our model, we think they'll burn cash actually this year, next year, and into 2028.

这很大程度上取决于他们的愿景，以及他们在所有即将推出的项目上的资本支出。

And a lot of that is predicated on their vision and their CapEx spend for everything coming down the pike.

所以当我们看到今天股价的表现时，我能理解。

So as we get to this so I understand what the stock is doing today.

我能理解大家对收入和AI的担忧，以及什么时候能看到投资回报。

I understand the concerns about revenue and AI and when do we see the ROI.

但这是第一个季度，管理层真正谈到了这些较新的产品，并展示了对未来一到三年、三到五年的愿景。

But this was the first quarter when we actually heard management talk about these newer, call it, products and to see a vision for the next one to three, three to five years.

所以你问的自由现金流和盈利能力的问题。

So your question about free cash flow and profitability.

你看。

Look.

归根结底，我们需要看到增长真正落地。

At the end of the day, we need to see that growth come through.

而市场显然对第三季度25%的增长有些消化不良。

And the market clearly is having some digestion issues with the 25% growth for 3Q.

短期来看，只要我们能持续看到稳定的增长。

Let's call it shorter term as long as we can see consistent continued growth.

我们刚才提到企业业务可能成为一个收入来源，但更重要的是，核心的自营业务，无论是商业代理、编辑工具、图像工具还是其他什么，如果能带来更大的增长，那就会推动投资回报。

And we just talked about potentially enterprise being a revenue stream, but more importantly, the core owned and operated business, whether it be business agents, whether it be their editing tool or image tool or whatever it is, driving greater growth, that's gonna drive the ROI.

所以他们的资产负债表非常健康。

So their balance sheet is incredibly healthy.

他们手里有现金。

They've got the cash.

他们能撑过一年、两年甚至更久的负自由现金流。

They can withstand a year or two or more of negative free cash.

但我们真的需要开始看到成果了。

But we really need to start seeing the results here.

这就是我们一开始说的，参与度依然很惊人，Instagram 在时间花费上增长很快，核心 Facebook 也是如此。

And this is where we started the conversation saying the results, our engagement continues to be impressive with Instagram growing, as fast as they are from a time spent perspective, even on core Facebook as well.

所以这就是我们的愿景。

So really that's our vision.

我们就是这样看待这家公司的。

That's how we look at the company.

马克在电话会议上提到的那些算力报价，我们收到了很多，而且价格都高于我们投入的成本。

These compute offers that Mark talked about on the call, we're getting a lot of offers at a premium of what we've invested in the compute.

你觉得他们会接受这些报价吗？

Do you think they take those offers?

我觉得他们会。

I think they do.

你觉得呢？

Would You think

他们会接受吗？

they do?

好的。

Okay.

他有点含糊其辞。

He a little wishy washy.

他说，留着吧。

He said, have them.

他其实不太看好这些。

He he wasn't really excited about them.

他就像在说，它们就在那儿。

He was like, they're there.

我们宁愿不要——他没这么说。

We'd rather not I he didn't say this.

但是

But

对。

Yeah.

我的意思是，你看。

I mean, look.

我一点都没搞错。

I make no mistake.

我觉得归根结底，他们是为了自己的自营业务扩充了产能，就这么简单。

I think at the end of the day, they built up their capacity for their owned and operated business, full stop.

而我们一直在等那个自营业务真正起飞。

And we're waiting for that owned and operating business to really ramp up.

而且，你知道，我们讨论这个已经有一阵子了，但我们几年前就听过谷歌CEO说过，先普及，后盈利。

And, you know, we've been talking about this for some time, but we we've heard Google's CEO years and years ago talk about ubiquity first, revenue later URL.

对吧？

Right?

所以我们还在等这些新产品上线。

And so we're waiting for these new products to launch.

等它们上线了，我也不确定它们是否真的需要Meta已经扩展出来的那么多算力。

As they launch, you know, I don't know if they necessarily need all of the capacity that Meta has grown to grown into.

所以我觉得真正让人眼前一亮的是，一方面他们看到的对算力的需求，另一方面是谷歌和其他公司签的合作伙伴关系，以及相关的金额。

And so what I think was the eye opening event that happened was, a, the demand they're seeing for potentially their compute b, we've seen partnerships that Google signed with others and and the the dollars that are associated with it.

我们刚才也聊到了一点自由现金流。

And we just started talking a little bit about free cash flow.

如果这些算力在短期内能帮助缓解一些自由现金流的压力，或者能进入企业市场，卖算力，或者他们的前沿模型能成为最先进的，那就有意思了。

If this compute capacity on a short term basis can help alleviate some of the free cash flow challenges and potentially go into an enterprise and and maybe sell compute or if their frontier model can be that state of the art, that's interesting.

但别搞错，36亿日活用户，核心产品仍然是我们关注Meta的重点。

But make no mistake, the 3,600,000,000 daily active people, the core products are still really the core of what we're looking for at Meta here.

我觉得那也是他们的北极星。

I think that's still their north star as well.

好的。

Okay.

那最后一个问题问你。

So Last question for you.

我们的联合执行主编马丁·皮尔斯昨晚写道，马克可能会借鉴萨蒂亚·纳德拉的做法，稍微控制一下成本。

I mean, our co executive editor, Martin Pierce, wrote last night that Mark might consider taking a page out of Satya Nadella's playbook and maybe putting a lid on costs a little bit here.

看起来他的策略更有野心一些。

I mean, he's a little bit more ambitious with his strategy, it seems.

听起来你不同意这个看法。

Sounds like you wouldn't agree with that.

听起来你对现在的投资和资本支出节奏是认可的。

Sounds like you're okay with the investment and the pace of CapEx right now.

我跟你说，只要产品战略能按计划推进，而且这是很长一段时间以来，我们第一次看到CEO，博士。

I'll tell you, as long as we can see that product strategy play out, and this was the first quarter in a while where we've seen CEO, Doctor.

伯格谈到这些即将推出的多个新产品，无论是新应用，还是带订阅的Meta One，还是商业代理、排名或推荐改进，等等等等。

Berg, talk about these newer multiple newer products that are on the come, whether it be newer apps, whether it be Meta One with subscriptions, whether it be business agents or ranking or recommendation improvements, and the list goes on.

当然还有Muse模型。

Muse models, of course.

而且我觉得再过几个月，MSL还会推出更多产品。

And and I think we're months away from more products coming out of MSL.

只要我们看到这些，并且能看到未来的价值，那我觉得他们就应该继续投资，你知道，我们希望——希望这个词不太对——但我们会据此来估算回报，并让他们对此负责。

As long as we see that and we can actually see what the value could be going forward, then I think they should be investing as, you know, and we hope to get that hope is a wrong word, but we would then model the returns and hold them to that.

但坦白说，如果我们处在一个规模效应不断放大的世界里，而且计算、能源和智能这三项关键资源确实很难找到，如果Meta能把它们整合起来，那么坦白说，我觉得Meta就应该为此投资。

But frankly, if we are in a world where, you know, scale begets scale and the three key resources that really are very hard to find between compute, energy, and intelligence, and if Meta can put that together, then frankly, I think Meta should be investing to that.

他们目前有足够的资产负债表来支撑这些投资。

They've got the balance sheet for now to do it.

太好了。

Great.

好的，Ron，非常感谢你来做客。

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

这位是花旗集团董事总经理、资深互联网分析师Ron Josey，正在TITV接受采访。

Is Ron Josey, Managing Director, Senior Internet Analyst at Citi here on TITV.

谢谢你的邀请，Akash。

Thanks for having me, Akash.

微软昨晚公布业绩后，股价大涨。

Microsoft shares jumped after its results last night.

Azure的增长加速到了43%。

Azure growth accelerated to 43%.

这次公司终于告诉我们Azure的规模了，营收突破了1000亿美元。

The company actually told us for once how big Azure is, crossing $100,000,000,000 in revenue.

关于这个话题，我们请来了RBC Capital Markets的软件股票研究董事总经理Rishi Jaluria。

For more on that, I want to bring on Rishi Jaluria, Managing Director of Software Equity Research at RBC Capital Markets.

Rishi，欢迎再次来到我们的节目。

Rishi, welcome back to the show.

很高兴你能来。

It's great to have you here.

谢谢邀请。

Thanks for having me.

很高兴再次回来。

Great to be back.

好的。

Okay.

你觉得微软这次财报的亮点是什么？

What was the headline from Microsoft for you?

是的。

Yeah.

对我来说，亮点是这季度表现非常强劲。

To me, headline is a really strong quarter.

我们看到Azure的增长在持续加速。

We continue to see acceleration in Azure.

我们还看到AI在应用层面的变现，特别是Copilot，而且他们在资本支出方面还保持了克制。

We continue to see the monetized AI at the application layer, especially with Copilot, and they're somehow maintaining responsibility with CapEx.

虽然他们明显在增加资本支出，但下一年度仍然能保持正向自由现金流。

They're obviously increasing it, but they're still gonna remain free cash flow positive this this coming year.

所以基本上，他们具备了实现强劲季度所需的所有要素。

So basically, all the ingredients they needed to put together to have a strong quarter.

而且，和Meta那边的情况不一样，Meta听起来好像还要烧一阵子钱，但Amy Hood说，他们今年根本不会烧钱。

And, I mean, unlike the story at Meta, where it sounds like they're going to be burning cash for a little while here, Amy Hood is saying, We're not going to burn cash at all this year.

对，这让我挺意外的。

Yeah, that was surprising to me.

在他们发布财报之前，我上调了资本开支预测，现在算上所有成本，苹果对苹果的资本开支差不多和我预期的一致，或者稍微高一点，但他们一个季度都不会烧钱，对吧？

Ahead of their earnings, I had raised my CapEx forecast, and now the fully burdened apples to apples CapEx is roughly in line or maybe slightly above where I'd expected, but they're not gonna burn cash in a single quarter, right?

至少，根据Amy的说法是这样的。

At least, you know, given where Amy has talked about it.

当然，情况可能会变，但我觉得这真的很让人意外。

Obviously, can change, but I think that's really surprising.

你知道，我一直被问到关于资本需求的问题，但我看不到这种需求，对吧？

You know, I've gotten questions around any needs on capital, and I don't foresee that, right?

我不觉得他们会需要发债、发股、可转债、SPV或者类似的东西。

I don't foresee that they're going to have to do debt or equity or converts or SPVs or anything like that.

那么，Azure的1000亿美元，比你预想的高还是低？

So $100,000,000,000 for Azure, was that higher or lower than you thought it was going to be?

那个数字更高。

That was higher.

更高，好的。

Higher, okay.

我原本以为他们的Azure增长率会在41%左右。

I thought they were gonna be doing a 41% type Azure number.

结果他们Azure的固定汇率增长率达到了43%。

They did 43% constant currency Azure.

嗯，我是说那个1000亿的Azure数字，我觉得这是我们第一次拿到这个数字，对吧？

Well, meant the the 100,000,000,000 for Azure, I mean, is the first, I think it was the first time we got that number, right?

是的，他们给了我们很多披露信息。

Yeah, they've given us a lot of disclosures.

所以，粗略算下来，和我的预期基本一致，因为他们这些年已经给了我们很多零散的信息，可以拼凑出来。

So yeah, the rough math was roughly in line with where my numbers are, just because they've given us the bits and pieces over the years to get there.

明白了。

Got it.

好的。

Okay.

那你看，我想跟你稍微深入聊一下，因为现在我们知道了Azure的规模，也知道了增长率，而且今晚有AWS的数据，上周有谷歌的数据，我就想问问，你有没有做过一些建模，看看这些业务在规模上会不会有趋同的趋势？

So look, I want to unpack this with you a little bit because now that we have a picture of the size of Azure and we have the growth rate, you know, we have AWS tonight, we had Google last week, I was wondering, have you done any modeling on your end sort of to see where these businesses could converge at all in size?

我是说，GCP现在增长超过80%，而且我觉得，你知道，一个季度250亿美元，我也不太确定。

I mean, have GCP that's growing 80 plus percent, and I think, you know, at $25,000,000,000 in the quarter, you know, I don't know.

我们有没有做过类似的建模？

Have we done any of this modeling?

是的，是的。

Yeah, yeah.

不，不。

No, no.

我想提醒你注意两点，这会让直接比较变得有点困难。

I would caution maybe two things, right, that makes it a little bit harder to go apples to apples.

第一是数据之间的可比性不足。

One is the lack of comparability in the numbers.

所以我们知道，比如谷歌把Google Workspace，也就是我们熟悉的G Suite，也算在那个数字里。

And so we know, you know, for example, Google includes the Google Workspace or what we might know as the g Suite in that number.

所以，显然，微软的Azure数字里不包括Office。

So, you know, obviously, Microsoft's not including Office in their Azure numbers.

所以这是让它们不太能直接比较的一个方面。

So that's one thing that makes it a little bit less apples to apples.

另一个就是这些公司怎么核算AI训练收入。

The other thing is just how all these companies are accounting for AI training revenue.

你记得，微软至少没有把OpenAI的训练收入算进Azure数字里。

As you recall, Microsoft has not included at least OpenAI training revenue in that Azure number.

可能Anthropic的一些收入确实算进去了。

You know, maybe some of the Anthropic stuff does show up in there.

我觉得亚马逊确实把Anthropic算进去了。

You know, I believe Amazon does include Anthropic in there.

谷歌那边，我有点不太清楚。

Google, I'm a little bit unclear.

所以确实有点缺乏可比性，但我觉得从方向上看，你的理解是对的。

So there's a little bit of a of a kind of a lack of comparability, but I think directionally you're thinking about it right.

比如，如果我们按它们目前的增长轨迹来画图，你会发现未来几年它们都会达到一个差不多的规模。

Like, we kinda plot these together on their current trajectory, you have a point where they all tend to hit a similar sort of size in the next several years on that current growth trajectory they're on.

那你确实认为谷歌会变得和Azure一样大，甚至有一天超过AWS吗？

And so you do expect that Google will be as big as Azure and maybe even AWS one day?

是的。

Yeah.

需要说明的是，这是基于当前增长轨迹的假设，对吧？

To be clear, that's assuming the current growth trajectory, right?

而且，你知道，我的同事负责报道谷歌，他可能有自己的看法。

And, you know, my colleague covers Google and probably has his own viewpoints on that.

但你看，我觉得至少从数学上看，让Azure达到和AWS一样的规模并不难。

But, you know, look, I think at the very least, getting the math for Azure to be the same size as AWS, it's not hard to get there.

对吧？

Right?

而且，你知道，一直以来大家的看法都是AWS是老大，Azure是落后的那个，或者说是第二名。

And and, you know, it it's always been the viewpoint of AWS as the leader, Azure as the laggard, or or maybe the number two.

现在我觉得，到了这个阶段，大多数观点都认为微软和亚马逊在基础设施方面实力相当，无论是传统的云计算还是AI基础设施。

And now I think at this point, you know, most of the viewpoints are Microsoft and Amazon are similarly strong positions when it comes to infrastructure, both for traditional cloud compute and for AI infrastructure.

对。

Right.

所以昨晚也有一些关于计算供需动态的问题。

So there were questions last night as well about the supply demand dynamics of compute.

他们显然在大力投资建设数据中心。

They're obviously investing heavily in their data center build.

你对Amy Hood关于他们如何考虑产能过剩及其风险的评论满意吗？

Were you satisfied with Amy Hood's commentary on how they're thinking about overcapacity and the risks of that?

满意。

I am.

对吧？

Right?

比如微软，你知道，和其他公司相比，之前确实犯过错误，试图削减资本支出，而且他们也承认那是他们犯下的错误，你知道，他们因此丢了一些订单，或者不得不和很多第三方供应商合作，所以他们不会再犯同样的错误了。

Like, Microsoft, you know, in contrast to the others, did make the mistake to try to dial back on CapEx a while ago, and, you know, they've admitted that that was a mistake that they made, that, you know, they left deals on the table or they had to work with a lot of third party vendors, so they're not gonna repeat that mistake.

现在，微软有一个逃生阀，我觉得这让他们处于一个独特的位置，就是他们有很多第三方合作，对吧，和甲骨文、Core Web、Nebius，还有甲骨文。

Now, the one escape valve that Microsoft has, and I think puts them in a unique position, is they have a lot of third party deals out there, right, with Oracle, with Core Web, with Nebius, you know, and cover Oracle.

如果微软真的过度建设了产能，他们可以调整和第三方的合作，把那些产能收回来，内部消化。

And what could happen in a situation where Microsoft somehow overbuilds capacity is they have the ability to dial that back, what they're doing with the third parties, and bringing that in house.

他们公开说过，目标是把所有第三方产能都收回来，内部消化。

They have talked publicly their goal is to bring all the third party capacity back in house.

所以他们有一点缓冲空间，你知道的，特别是考虑到这种可灵活调整的架构，他们可以投入很多。

And so they have a little bit of an escape valve, you know, especially given this fungible architecture that they can spend a lot.

到了某个时候，如果他们可能过度建设了，他们也不会让产能闲置。

At some point, if they may be overbilled, they're not gonna be sitting in an unused capacity.

对。

Right.

对。

Right.

昨晚还提到了另一件事，就是萨蒂亚谈到了在开源和闭源之间走钢丝，他之前也写过这方面的文章。

The other thing that came up last night was Satya talked about towing the line between open source and closed source, and he's written about this as well.

这其实挺有意思的，对他们来说是个棘手的问题，对吧？

This is kind of an interesting it's just kind of a thorny issue for them, right?

因为他们从闭源热潮中获益良多，但现在他却站出来大声疾呼，基本上就是说世界应该是开放的。

Because they've benefited so much from the closed source boom, and yet now he's coming out swinging, basically saying, The world should be open.

你应该能控制自己的模型之类的。

You should have control over your own models and stuff like that.

那你对他昨晚那些额外评论怎么看？

So what did you make of his added comments last night on that?

对。

Yeah.

听着，这跟塞蒂一贯的做法非常一致。

Look, it's very consistent with what I would have expected of Sethi.

如果你还记得一年半前那个“深度探索”大事件的时候，塞蒂其实是在给我们介绍吉文悖论，他谈到了蒸馏的概念以及为什么它如此重要，而且他们是首批真正在Azure上支持托管DeepSeek的平台之一，这样就能解决数据主权和很多隐私问题。

If you remember a year and a half ago when we had the big deep seek moment, Sethi was actually introducing to all of us, right, Jeevan's Paradox, and he was talking about the concept of distillation and why that's so important, and they were one of the first platforms to actually support hosted deep sea, you know, on Azure, so there's data sovereignty and gets through a lot of privacy concerns.

我听到这个并不意外。

I'm not surprised to hear that.

如果我想想，你知道，微软能给很多客户提供的价值主张，尤其是在应用层，其实就是把多模态的优势发挥到极致——我不需要什么都用GPT-5.6或者Fable 5，有些可以用旧一点的模型，有些可以用小型的、中型的语言模型，还有些可以用开源的。

If I think about, you know, what is the value proposition that Microsoft can offer a lot of customers, especially at the application layer, it's really getting the best of multimodality that I don't need to use, g p t five dot six sole or or Fable five for everything, that some can be done with older generation models, some can be done with small English, medium language models, and some can be done with, yes, open source.

所以我觉得这跟塞蒂亚会说的话非常一致。

And so I think this is very consistent with what kind of the talk track I would have expected out of, Sethia.

当然，他们跟OpenAI之间显然还有很强的经济联系，这也是他们第一方模型开发的主要动力。

To fair, they still have obviously this very strong economic relationship with OpenAI, and that's what's powering a lot of their first party model development.

但我觉得，从长远来看，支持多模态确实非常符合微软的利益，坦白说也符合客户的利益。

But I think it is very much in Microsoft's interest, and candidly in the customer interest, to have that multi modality over time.

嗯，这让我想到，从客户利益的角度看，我觉得萨蒂亚·纳德拉想说的是开源是未来。

Well, and it sort of gets to it feels like from a customer interest point of view, I feel like Satya Nadella wants to say open source is the future.

但从投资者的角度看，闭源必须继续是故事的一部分，因为就像你说的，OpenAI对他们的业务贡献太大了。

But from an investor perspective, closed source has to remain part of the story because, as you said, OpenAI contributes so much to their business.

对，我觉得这不是一个非此即彼的情况。

Yeah, and look, I think it's not an eitheror scenario.

我的想法是，最终我们会进入一个八二分的世界，大约80%的AI工作负载可以用旧模型、非前沿模型或开源模型完成，而20%可能需要当时的前沿模型。

My mental model is ultimately, you know, we're going to hit an eightytwenty world, where about 80% of AI workloads can be done with older models, non frontier models, or open source, and 20% probably require whatever the frontier model at that time will be.

如果我们按每个AI工作负载的美元价值来看，市场总量仍然集中在那个丰厚的20%里，而不是在长尾部分。

Now, if we think about this in like dollars per AI workload, the TAM is still in that kind of fat 20% rather than in the long tail.

我觉得这会是现实情况。

And I think this is gonna be the ground reality.

这不是非此即彼的选择。

It's not an eitheror.

而是上述所有方案的混合。

It's gonna be a mix of everything above.

而且，这就是你听到微软在发布Critique和Counsel时所说的。

And, you know, that's where you heard Microsoft when they announced and launched Critique and Counsel.

他们当时在强调多模态，以及如何实时比较不同模型，选择最佳答案，选择最适合的工具。

They were leaning into that multimodality and how do you kind of compare the different models in live, in real time, and kind of use the best answer, use the best tool for the job.

我认为这正是企业界未来要发展的方向。

And I think that's exactly the direction we're going be headed in enterprises.

对。

Right.

我还有几个问题想问你。

A couple more questions for you.

所以，Copilot 这个季度确实增长了不少。

So Copilot was I mean, it saw some considerable growth in the quarter.

我们在节目里也一直在聊，增长是一回事。

Another thing that we've been talking about on the show is growth is one thing.

但客户是否真的从中看到了价值，并且用起来有效果，那是另一回事。

Whether or not customers are seeing the value from it and it's working for them is another question.

在这方面，你从客户那里听到了什么？

What are you hearing from customers on that level?

是的，很高兴你问这个问题。

Yeah, look, I'm glad you asked that.

答案是，客户终于开始对 Office Copilot 给出更多正面的反馈了。

And the answer to that is customers are finally starting to give more positive feedback to Office Copilot.

对吧？

Right?

一年前，反馈肯定要复杂得多，这么说算是客气了。

A year ago, the feedback was definitely a lot more mixed to to to be charitable.

但现在，自从Sethius接手这个产品，或者说从去年十一月开始，加上后续的发布，反馈已经好多了。

But now I think since Sethius stepped into the product or back in November and the subsequent releases, feedback has been a lot more improving.

所以不只是付费用户增长快，上个季度他们增加了1000万付费Copilot用户，关键是人们真的开始从中获得实际价值了。

So it's not only that you have this accelerating paid ads where they added 10,000,000 paid Copilot users last quarter, it's the fact that people are actually starting to get real value out of it.

对吧？

Right?

而且，你知道，我觉得这会体现在使用指标上。

And, you know, I think it's gonna show up in the usage metrics.

会体现在参与度指标上。

It's gonna show up in the engagement metrics.

上个季度他们分享了一个数据，Office Copilot的使用情况跟Outlook差不多。

They shared a metric last quarter that engagement with Office Copilot was on par with Outlook.

我觉得我们会从微软那里听到越来越多这样的数据点，就是为了说明人们不只是把这个当摆设，而是真的从中获得了实际价值。

I think we're gonna hear more and more of those sort of data points from Microsoft just to illustrate that people are not just using this as not just shelfware, but people are actually getting real value out of it.

好的。

Okay.

那Xbox呢？

And what about Xbox?

我的意思是，这项业务现在正处于转型期。

I mean, this is a business in transition right now.

你觉得他们应该把这块业务卖掉吗？

Do you think they should sell off that business at all?

真心觉得不应该。

Genuinely don't.

其实我理解那里的挑战，而且我——

Actually I I I get the challenges there, and I'm

是因为你玩Xbox吗？

Is that because you play Xbox?

我觉得现在的电子游戏对我来说太复杂了，毕竟我年纪大了。

I I I think video games have gotten too complicated for someone of my age these days.

我小时候的游戏简单多了。

They used to be a lot simpler when I was a kid.

但是，你知道，我认为这是因为他们拥有非常强大的游戏资产组合。

But, you know, the the reason I think is because they have such a a great combination of gaming assets here.

对吧？

Right?

不仅仅是实体Xbox主机。

It's not just a physical Xbox.

对吧？

Right?

如果你还记得，他们显然完成了动视的交易。

If you remember, they obviously had the Activision deal.

他们有很多第一方游戏作品。

They have a lot of first party properties.

对我来说，我觉得最终的愿景，以及游戏最终能发展到的地方，就是那个Game Pass产品。

And to me, I think the ultimate vision and ultimately where gaming can go is in that Game Pass product.

对吧？

Right?

它实际上会变成，打个不太恰当的比方，游戏界的Netflix。

That this effectively becomes, for lack of a better analogy, the Netflix of gaming.

要达到那个目标还需要一段时间。

It's gonna take a while to get there.

你知道，这里面有硬件方面的因素，还有，嗯，要贴近用户，还有移动端策略，这些都需要花点时间才能实现，并推动用户采用。

You know, there's kind of a hardware and, you know, meet the customer where they are and mobile strategy that takes a little bit of time to get there and drive that adoption.

你还得降低延迟。

You also have to get latency down.

但对我来说，那才是真正能扭转游戏业务命运的未来。

But that to me is really the future of what turns around the fortunes of the gaming business.

对。

Right.