第029集 - 模块化数据中心将建设周期缩短至12个月（数据中心、能源）| Nico Bontigui, Jordan Nanos, Nigel Chiang, Eric Wen

你好，

Hello,

大家好。

everyone.

欢迎回到SemiAnalysis周刊。

Welcome back to SemiAnalysis Weekly.

今天我们要聊的是乐高式数据中心的狂野西部。

We are here to talk about the wild, wild west of LEGO Datacenters.

几周前，我们发了一篇关于模块化数据中心和建设的文章，所以今天我请来了Eric、Nigel和Nico，我们要深入探讨模块化是什么意思，有什么好处，瓶颈如何从建设时间转移到劳动力，以及其他各种问题，模块化建设与传统建设之间的权衡，存在哪些风险，以及有哪些大玩家在推进这件事。

A couple of weeks ago, we put out an article on modular data centers and construction, and so I've got Eric, Nigel, and Nico, and we're gonna dig into what modular means, what the benefits are, how the bottleneck has moved from construction time to labor and all sorts of other things, the trade offs between modular construction and traditional, what sort of risks exist, and who the big players are that are pursuing it.

那么，各位，欢迎来到节目。

So, guys, welcome to the show.

很期待深入探讨。

Excited to dig in.

接下来把话筒交给Nico，让他来详细说说。

Gonna pass the ball to Nico for the exactly, man.

尼科和奈杰尔是回归嘉宾。

Nico and Nigel are returning.

埃里克是第一次参加，这正好代表了SemiAnalysis的各个部门。

Eric's first time attending, and this is a nice cross section across SemiAnalysis.

我们有来自咨询团队的埃里克，来自核心研究的奈杰尔，以及来自数据中心能源和工业模型团队的尼科。

We've got Eric from the consulting team, Nigel from Core Research, and Nico from the Data Center Energy and Industrials model team.

所以，这次参与模块化数据中心相关工作的各方代表都齐了，也参与了这篇新闻通讯文章的撰写。

So, got a little taste of everybody who touches modular data centers and contributes to newsletter articles like this on this one.

是的，很高兴能来这里。

Yeah, excited to be here.

太好了，各位。

Awesome, guys.

好。

Okay.

那我们开始吧，先请尼科给我们介绍一下什么是模块化数据中心，它是什么意思，和传统的数据中心建设方式有什么不同。

So let's start by passing to Nico to give us the overview of what's a modular data center, what's it mean, what makes it different from the traditional way to build data centers.

好的。

Sure.

对。

Yeah.

我们就从这儿开始吧。

Let's start with that.

我觉得，对于什么是模块化数据中心，其实没有一个统一的定义，这也是我们决定写这篇文章的主要原因，因为每个人对模块化，或者有时候叫预制化，都有自己的理解。

I mean, I I think that is no single definition for, like, what is a modular data center, and that was the single reason why we decided to put out these these article in the sense that everyone has its own definition for what motorization or, you know, prefabrication sometimes call it like that, what it means.

所以我们基本上就是发布了那篇文章，其中很大一部分内容就是一个完整的分类体系，来回答当我们谈论模块化时，我们到底指的是什么。

And so essentially, we put out that note, which a big part of that, it's a huge whole taxonomy into what are we actually meaning or talking about when we talk about modular to answer that question.

不过，先从最简单的概念说起吧，就是模块化或者预制化。

But just to start off with like a pretty simple idea, modularization or prefabrication.

我们把文章命名为“乐高数据中心”的原因是，你不用传统的方式去建数据中心，传统方式基本上就是把所有设备运到现场，然后把所有工人，高峰期大概七千人，都拉到得克萨斯州阿比林那个地方。

And the reason why we named the article LEGO Datacenters is because instead of building your data center in a traditional way, where you essentially, you know, bring all the equipment to the site and then you bring all the workforce, all the seven k workers at peak to in the middle of Abilene, Texas.

然后才开始建数据中心。

And then you start building the center.

与其那样做，你就像拼老式乐高一样，一块一块地搭起来。

Instead of doing that, you are going to build it like your, you know, old LEGO, just pieces by pieces.

当然，这些块可以有很多种变化。

Of course, those pieces can vary a lot.

但想想看，与其把所有设备运到现场再组装、接线，你可以在工厂里完成一部分工作。

But think of instead of putting all that equipment and assembling, doing connections on-site, you are going to do part of the work off-site.

你会做一些模块化的部分，因为实际上你用的是模块，或者想象成带设备的金属外壳。

You're going to do some modular sections because you are actually using modules or think of metal enclosure with equipment at site.

这些在工厂里做好，然后运到现场。

You're going to do that in the factory, and then you're going to ship that into the site.

然后你把不同的模块，也就是数据中心的不同乐高积木，一个个堆起来。

And you are going to stack different modules, different LEGO pieces LEGO pieces of that data center.

最后再把它们全部连接起来。

And then you're essentially going to connect them all.

这应该是即插即用的。

It's supposed to be plug and play.

然后，你知道，我们可以聊聊为什么我说“应该”，但基本上，数据中心就是即插即用的搭建方式。

Then, you know, we can talk about why I say supposed, but, you know, essentially, it's a plug and play built of a data center.

简单来说，这就是我们所说的模块化或预制化。

That's, you know, essentially in super simple terms, what we mean with modularization or prefabrication.

对。

Yeah.

而且，当你提到模块化数据中心时，可能还有几种不同的类型，比如在描述模块化数据中心的建设方式时，可能有一整套不同的选项。

And maybe there's a few different types of modular data centers when you talked about, like, maybe there's a whole different group of options when it comes to describing how a construction happens with the modular data center.

比如，屏幕上我有一张图表，描述了从传统的现场搭建到今天的基准，重型滑橇、全模块化建设、集装箱式，或者就是字面意义上的海运集装箱，以及它们从场地准备和地基工程（灰色）到黄色结构外壳，再到蓝色机械和绿色调试，从最多两年到最短十二个月不等，这是给看图表的人参考的。

Like, on screen, I've got this chart describing stick build versus the baseline today, MBP heavy skids, full modular build, containerized, or it's like literally a shipping container, and how they can range from up to two years down to as little as twelve months when it comes to getting stuff on-site from site prep and groundworks in gray to the yellow structure shell, and then blue mechanical and green commissioning for those who are looking at the chart.

对。

Yeah.

Eric，也许你可以谈谈你所看到的采用模块化方法的人所获得的一些好处，以及这种加速实际上是如何帮助他们的。

Eric, maybe you can talk about some of the benefits for what you've seen of people that are going with the modular approach and how the speed up is actually helping them.

好的。

Yeah.

我觉得市场上我们看到的，大概有两个最大的好处。

I think there are, guess, two kind of biggest benefits that we have seen on the markets.

第一个可能就是劳动力。

The first one is probably labor.

劳动力方面，我们见过很多人来找我们咨询模块化设计，因为现在市场对劳动力的需求非常强劲。

Labor has we I mean, we have seen multiple people come to us asking about module designs because they are the market is super strong labor.

而且因为现在市场上劳动力成本非常高，再加上很难找到足够的劳动力来支撑大规模的数据中心建设，所以大家真的开始转向模块化数据中心，把它作为解决劳动力瓶颈的替代方案。

And because we the the rates of labor is super high right now in the markets and because of how people can find amount of labor to support the immense of data center build out, they're really turning to kind of monitor data center as the alternative solutions for that labor bottlenecks that we're seeing on the markets.

这是最大的一个好处。

That's the biggest one.

我觉得第二个，可能更重要的是确定性和模块化设计。

I think the second one, I think more importantly is just certainty and modular designs.

我觉得这和融资和资本有关。

I think that's related to financing and capital.

因为当我们看很多数据中心建设项目时，确定性是一个大问题。

Because when we look at a lot of data center build outs, certainty is a big thing.

即使是很多规模较小的供应商，他们也没有所有必要的设备来建造整个数据中心，因此他们也转向模块化，作为建造数据中心的潜在替代方案，以更快实现通电，同时也能确定何时能获得场外电力，以及需要多长时间。

Even a lot smaller kind of provider, they don't have all the necessary equipment to to build an entire data center, and therefore they turn to monitor also potential alternative solutions to build a data center to make it faster for time to power, but also have certainty to when do we get power off-site that's due amount of time.

我的意思是，所以我们看到很多不同的参与者，都在根据他们在劳动力和经济方面的限制，以及不确定性方面的限制，寻找不同的模块化解决方案。

I mean, so we see a lot of different players trying to look for different modular solutions based off where there are constraints in both labor and economic as well as where there are constraints of uncertainty.

通过引入模块化解决方案，可以为投资者和他们自己带来更快的通电时间，同时也能更确定电力何时能上线，因为整个方案是作为一个完整的包来交付的。

And by bringing kind of modular solutions, it can give both investor and themselves a quicker time to power and also more certainty to when that power can come online based off the complete, I guess, bringing a completed kind of package as a whole.

对，而且这不仅仅是那些随便搞的加密货币矿工在做这件事。

Yeah, and this isn't just random crypto miners that are doing this.

显然，几个月前我们发表那篇关于Beta的文章，讨论了他们的帐篷，那时候这件事就开始流行了。

Obviously, this became popular when we put out the article on Beta a few months ago and talked about their tents, Yeah.

然后在那篇文章里，我们也描述了AWS是怎么做这件事的。

And then in this article, we described how AWS is doing this as well.

所以，也许你们可以聊聊，有哪些不同的参与者在探索这个方向。

So, maybe you guys can talk a little bit about the different players that are exploring this.

好的，当然可以。

Yeah, absolutely.

没有。

No.

我是说，我本来想说，在讨论这些玩家之前，我更想先谈谈埃里克刚才提到的那些。

I mean, I was just gonna say that maybe before actually, you know, addressing the players, I would wanna come to what Eric mentioned more.

他提到了劳动力短缺，这当然是个大问题，光聊劳动力就能聊上一个小时。

He mentioned labor shortage, which, of course, is a huge we can spend an hour just talking about labor.

还有可预测性和质量。

Also, like predictability and quality.

但基本上，经过我们和不同的数据中心建设商、开发商聊下来，有一点很明确，就是没有一个清晰的先后顺序。

But essentially, one thing that's clear after, you know, all these conversations we've had with different DC Builders, developers, is that there's no a clear pecking order.

你知道，有些人做模块化是因为一个原因，因为他们确实喜欢这种可预测性，知道当你把模块插上去，它就会按承诺的那样运行。

You know, some folks are doing modularization because one reason, because they actually like this kind of predictability of knowing that when you are going to plug your modules, it's going to function as promised.

还有一些人这么做是因为，你知道，我们找不到足够的电工或焊工到现场来。

Then some others are doing because of, you know, we cannot get enough, electricians or welders coming to the site.

但还有第三个原因，我觉得，这可能是目前最重要的，那就是时间到电力，也就是建设时间。

But there's also a third reason, I would say, that it's the clear, I would say number one today, which is just essentially time to power, like building time.

基本上，就是缩短从重型设备进场做场地准备，到数据中心建成、可以出售或通电的时间。

Essentially, it's shrinking the time that it, you know, takes to build your data center since the day you get your heavy machines into the site to do the site works till the day you're, you know, ready for sale or ready for ready to energize.

这是很常见的做法。

And that's a common practice.

所以当我们考虑那些在做这件事的玩家，或者他们正在采取的不同策略时，归根结底就是怎么才能建得最快。

And so when we think of, you know, the players that that they're doing this or the different strategies these players that we're going to talk about now they're pursuing, essentially it all comes down to how I can build this as fast as I can.

怎么才能把通常需要十八到二十四个月的数据中心建设时间缩短呢？

How can I shrink these, you know, two years, eighteen to twenty four months that it usually took to build a data center?

有时候甚至超过三年，缩短到一年。

Sometimes, like, over three years, you know, to one year.

因为现在一切都跟算力挂钩。

Because now everything could translate to tokens.

而且，那十二个月的机会成本，就是没法产生那十二个月的算力收益。当然，我们最近在通讯里也聊过，以前我们说的是每兆瓦一千一百万到一千五百万美元。

And, you know, those twelve months of the opportunity cost of not being able to generate these twelve months worth of tokens, of course, we've, you know, talked about in our newsletter recent how if before we were talking about 11 to no, 11 to 15,000,000 per megawatt.

现在我们看到算力交易的价格远高于此，每兆瓦四千万美元，Anthropic 甚至在某些配置下超过每兆瓦一亿美元。

Now we're seeing compute deals and much higher than that, you know, 40,000,000 per megawatt, anthropic generating some configuration over a 100,000,000 per megawatt.

想想十二个月在代币生产时间上值多少钱。

Think of how much twelve months in token production time it's worth.

那么，有了这个，我们可以回到矿工的话题。

So, with that, we can go back to the miners.

是吗？

Yes?

对。

Yeah.

对。

Yeah.

我是说，确实。

I mean, yeah.

所以，具体来说，也许为了说明这一点，把这张图重新放回屏幕上，当你在这张图上看场地准备和地基工程时，五个月到四个半月对于集装箱化建筑和基准相比变化不大，但压缩最厉害的部分是机电安装。

So, just specifically, maybe to to make the point and bring this chart back on screen, when you look at site prep and groundworks on this chart, five months to four and a half months doesn't really change much for a containerized bill versus the baseline, but the part that compresses so much is mechanical and electrical fit out.

所以，你只需要做三个月的工作，而不是长达九个月的电气系统工作。

So, you're just doing three months of work instead of up to nine months of work of electrical work on the system.

那你们能不能聊聊，这对世界上那些正在建最大数据中心的大公司来说意味着什么？我觉得最上面那张图简直太惊人了，它显示了全国或全球最大工地上劳动力成本的快速上涨，比如得克萨斯州阿比林的Stargate项目，和Crusoe合作的那个，还有微软在威斯康星州的Fairwater站点。

So, can you guys talk about what this means for specifically the biggest companies in the world that are building the biggest data centers, I thought this chart at the top was just incredible where you show the rapid increase in the cost of labor at the biggest sites in the country or the world, which is like Stargate Abilene with Crusoe in Texas, as well as Microsoft's Fairwater site in Wisconsin.

是的，我觉得这张图正好印证了一个观点，就是劳动力短缺，或者说很难找到七千名工人到这种基本上算是荒郊野外的地方来干活，这张图就是证据。

Yeah, I would say that this chart over here really ties to one of the points that is the labor shortage or the difficulties that finding seven ks workers coming to the middle of nowhere essentially as evidence.

也就是说，我们确实听到很多故事，比如‘哎呀，现场很难招到电工’。

In the sense that, yes, we hear stories of, oh, it's really hard to get the electricians on-site.

但然后你就想，好吧，我们能不能找到一些实际的、经验上的证据来证明这一点？

But then you actually, like, were like, okay, how actually can get some evidence, empirical evidence of this?

于是我们就去查了查，看看新招的电工，这些新入职的电工，他们的工资趋势是怎么样的。

And we looked into, okay, let's see how are the wages of new hires, these new electricians that are being hired are trending.

然后我们就找到了这张图，基本上就是你可以看到，在这些州、这些县，也就是那些最大的园区正在建设的地方，实际园区在高峰期需要三到四倍的工人，我们发现，这些新员工的工资，基本上翻倍了，在某些关键州甚至翻了三倍。

And we come across this chart, which is essentially that you can see that in these states, in these counties where the biggest campuses are being built, the actual campus that the demand between, as I said, three to four workers at peak, we we said that the, you know, wages for these new hires were, you know, essentially doubling, sometimes tripling in some key states.

Nigel，你对劳动力短缺有什么看法？

Nigel, what's your take on the labor shortage?

我猜你肯定从投资者那里，还有行业里各种公司，不管是模块化的还是其他的，都听说过这个情况。

I assume you hear about this from investors and maybe many different companies across the industry, modular and otherwise.

对，我觉得这是个很好的问题。

Yeah, I think it's a great question.

我觉得现在首先要指出的，也是客户、数据中心运营商、投资者等越来越认识到的一点，就是瓶颈往往不止一个。

I think the first thing to note now, and one that is increasingly being appreciated across customers, you know, data center operators, investors, so on and so forth, is that often different bottlenecks out there.

你知道，有电力，有算力。

You know, there's power, there's there's there's compute.

曾经，瓶颈在算力，还有内存。

There was once, you know, a bottleneck in this, and there was memory.

还有大家关心的，比如劳动力。

There was the the topic of of interest of and then, you know, labor.

劳动力是其中一个很难解决的制约因素，因为它的供应弹性真的很小。

Labor is one of the constraints that is really difficult to resolve because it's not like you know, the supply isn't very elastic at all.

所以即使说到电力，想想AI的经济性，也就是如果你能更快地上线一单位算力，就能获得巨大的溢价，这已经吸引了很多边际电力来源和电力形式。

So even when you talk about power, think the economics of of of AI, meaning the the the huge premiums that are there for undertaking, if you can bring up a unit of compute online quicker such that it has brought in a lot of your marginal your marginal power sources and power formats.

就像我们和能源团队在Core Research这里，我们写了很多关于往复式发动机如何成为边际电力来源之一的文章，而且大家也都知道燃料电池。

Like and and we we know again, with the energy team and and and and here at Core Research, we we talked about how we've written a lot about how risk reciprocating engines are one of the marginal formats that have come have come in, and everyone kind of knows about fuel cells as well.

但如果你考虑劳动力的问题，那真的很难解决，因为这些不是普通的电工，他们不能随便进来就建好一个数据中心，因为需要非常专业的电气布线。

But if you think about labor, it's it's really difficult to to to to solve that because this this aren't your, you know, garden variety electricians where where they can just come in and and and, you know, stand up a data center with all the specialized, you know, electrical wiring that that is that is that is needed.

这些是非常关键、任务关键的工作，而且你得让这些人去偏远的地方。

These are very critical, mission critical, you know, work that is being done, and you need to have these guys go out in the middle of nowhere.

你知道，这种工作有个术语叫现场劳动力。

You know, the the term for for this is called field labor.

而且正如你在图表上看到的，这些人的工资正在飙升。

And as as you see from that chart, the wage rates of these guys are just soaring.

所以正因为如此，我觉得大家开始意识到，解决这个问题的一个办法就是模块化。

So so because of that, I think I think it's it's being appreciated that one way to solve this is modular.

而且，正如我们在今天发布的一条推文中指出的，这实际上反映在股价上。如果你看整个数据中心工业领域，看建筑/工程采购和施工公司，也就是所谓的EPC公司，他们今年相对于设备OEM的表现，你会发现建筑公司的表现更好。

And as as, you know, we've kind of pointed out in in a tweet that went out today that, you know, this this actually kind of shows up in the in the share prices when you look at the entire data center industrial space, you look at how the construction slash, you know, engineering procurement and construction guys are, you know, what you call the EPC guys, how they have traded this year relative to the equipment OEMs, you see the the construction guys outperforming.

所以，如果你去看看我们的推特，我觉得我们有一系列不错的推文，里面展示了这些数据。

So, you know, if you check out check out our Twitter, I think we have a good kind of series of tweets where we cut the data showing that.

是的，这正好印证了我的同事尼科和埃里克一直在谈论的关于劳动力的问题。

And, yeah, it goes to show what my my colleagues, Nico and and Eric has been talking about on on labor.

确实。

Definitely.

对。

Yeah.

而且我觉得，可能再补充一点，我们经常听到瓶颈这个词，瓶颈这个词被频繁提及。

And I I think maybe just to put a a finer point on this, we hear a lot about bottlenecks, the term bottlenecks get thrown around a lot.

在我看来，瓶颈有两种不同的类型。

There are two different types of bottlenecks in my mind.

一种是行业层面的、高层次的限制，比如台积电能生产多少CoAs，三大内存厂商能生产多少内存。

There's industry wide, high level constraints, like how much CoAs can TSMC produce, how much memory can the big three memory vendors produce.

这是对整个行业能生产多少芯片的限制，然后还有局部瓶颈，这些瓶颈会限制某个项目在特定时间范围内上线，比如他们关心的几个月或几周内。

This is a constraint on the whole industry of how many chips you can produce, and then there are local bottlenecks that would restrict a given project from coming online in a certain number of months or weeks that they care about.

这是两回事，因为从某些方面来说，那些分阶段进行的项目，他们从全球供应链中分到的芯片配额可能无法超过某个数量，但他们有自己固定的配额。

These are two different things because in some ways, individual projects that have multiple phases, they may have their allocation of chips from the global supply chain that just can't be increased above a certain amount, but they have their allocation.

他们只是在等这些芯片到位并安装，而他们的瓶颈在于你提到的劳动力等问题。

It's waiting for them to come and install, and their bottleneck is things like labor that you're describing.

就是这样，比如像往复式发动机之类的，但如果我们从整个行业来看，一些在个别站点出现的瓶颈并不一定代表全球范围内的情况，当我们看到芯片随着时间的推移大量上线时。

Here it is, things like recip engines or whatever, but if we look at the industry as a whole, some of these individual bottlenecks that show up in individual sites are not necessarily representative of what's going on globally, worldwide, when we look at the kind of just massive amount of progress of chips being brought online over time.

Nico，也许你可以多谈谈高层的数据中心、行业范围的追踪，并区分一下当地的劳动力瓶颈和我们持续看到的行业范围需求？

Nico, maybe you can comment a little more on high level data center wide, industry wide tracking and differentiate between the local bottleneck of labor and the industry wide demand that we keep seeing?

对。

Yeah.

对。

Yeah.

不，Jordan，我觉得这一点非常重要，因为，我的意思是，劳动力是一个大挑战。

No, and Jordan, I think this is an extremely important point because, I mean, labor is a big challenge.

主要的限制之一，你知道，当我们和运营商交谈时，比如，实际上让劳动力到场。

One of the main constraints that, you know, when we talk to operators, name, like, actually getting the labor on-site.

但我们不应该从中得出这样的信息：哦，我们在美国没有足够的电工。

But we shouldn't extract from that the message that, oh, we don't have in The US enough electricians.

这只是一个非常复杂的优化问题，因为从整个美国来看，至少在接下来的二十四个月里，考虑到预计要建设的产能，你有。

It is just a really complex optimization problem in the sense that in The US as a whole, you probably at least in the next twenty four months with the capacity that is expected to be built, you have.

但问题是，这些劳动力中有多少不仅能够胜任关键任务和非常复杂的项目，而且在这些地区有多少劳动力是可用的？

But the thing is, okay, how much of that labor it's actual, not only capable of, you know, taking place in a mission critical and very complex project, but also how much of the labor is reachable in these areas?

比如，劳动力更多是一个物流问题，而不是一个真正的结构性问题，因为劳动力缺乏弹性。

Like, is it labor is more of a logistics problem than it is like an actual, you know, a structural problem of because of labor being inelastic.

这确实很复杂，但我们现在看到的是，真正能净减少所需工时的解决方案，比如模块化，以及我们在文章中提到的那些。

And it's a really complex one, but what we are seeing now that all the way then the solutions that really contribute on a, you know, net positive basis to reduce the amount of man hours required, like modularization and what we covered in our article.

还有其他策略，比如找到正确的激励措施和正确的优化方案来解决这个问题，也就是如何重新分配所有可用劳动力到各个地点。

There are other strategies, which is finding the right incentives and the right, you know, solution optimization problem to this solution to this problem, which is how we actually reallocate all of the available labor in the locations.

我们在这方面做了更多工作，实际上，SemiAnalysis的一个优势是，由于数据中心团队三年来的工作，我们确切知道所有这些站点将建在哪里。

We're doing more work on that to actually know, the good thing at SemiAnalysis is that because of the work that, you know, at the data center team we've done during three years is that we actually know exactly where are all of these sites going to be built.

当我们谈到建设15吉瓦、30吉瓦时，我们实际上有按具体县分的详细数据。

When we talked about 15 gigawatts, 30 gigawatts being built, we actually have the breakdown, by a specific county.

所以我们实际上可以利用所有这些，说，好的，让我们把这些容量、这些兆瓦，转换成劳动力需求。

So we actually can take all that and say, okay, let's take this capacity, these megawatts, and convert that into labor demand.

然后我们可以看到，好的，在这个特定的县，我们需要这么多工时，但实际上我们没有这么多。

And that we can, you know, see, okay, in this specific county, we need this much amount of this many amount of labor hours, and we don't actually have them.

那怎么才能把住在300英里外的电工们弄过来呢？

How do we manage to bring the people that are the electricians that are living 300 miles away?

而且，你知道，这就是我们正在讨论的，也是现在行业里真实发生的情况。

And, you know, this is our conversation that you know, this is what is actually happening in the industry today.

当然，工资是其中很重要的一部分，这也是我们之前放那张图表的原因。但我们也听到了一些其他的东西，比如每日津贴之类的，甚至还有停车场里的餐车，因为去现场的工人就是这么要求的。

And of course, wages, it's a big part of this, and this is why we included these these previous previous chart, but we've also heard some other things, some per diems and everything, all the way from like taco trucks on a parking lot, just because demanded by the actual workers that are going to the site.

所以这绝对是个地方性问题，解决方案也不简单。

So there are it's absolutely a local problem and solution is not easy.

但是，你知道，是的。

But, you know, there Yeah.

好，那我们来说说具体细节吧。

Well, let's let's talk about the let's talk about the specifics.

Eric，要不你来聊聊这些人具体在做什么，以及模块化数据中心到底有什么不同？

Eric, maybe I can pass it to you to talk a little bit about specifically what some of these people are doing and specifically what makes a modular data center different.

你们在文章里对这块的分类特别好，系统、外壳、场地，就是建设的三个层面。你能不能稍微拆解一下这三个部分分别是什么？还有，当我们看一个模块化站点，跟传统建造流程比，在系统、外壳、场地这三个层面到底有什么实质区别？

You guys had a great taxonomy for this in the article, systems, shell, and site, like the three layers of the stuff that's getting built, and maybe you can break down a little bit about what these three components are and what, when we look at a modular site and compare it to the traditional building process, what actually makes it different across system, shell, and site.

对，对，完全同意。

Yeah, yeah, absolutely.

对，在那之前，我想先回应一下尼科刚才说的，我觉得劳动力总有一个上限。

Yeah, before that, I just want comment on what Nico just said earlier, is that I feel like there's always an upper bound to where labor can be.

你不可能一夜之间造出人来。

You can't produce humans overnight.

要培训一个人达到数据中心熟练程度，甚至更难，要能上岗干活。

It's even harder for you to actually train a human to be, I guess, data center proficient to get to a level of site ready.

所以，数据中心今天想上线运营，和未来能有人来真正建造数据中心，这两者之间存在一种错配。

So, there's kind of a mismatch to where a data center wants to be powered online today and where a human can come in to actually build a data center in the future.

这种错配给云模块化带来了巨大的需求。

And that mismatch creates a huge huge count demands for a cloud modular.

所以模块化不仅仅是关于成本和时机。

So modular is not just about costs and kind of timing.

更重要的是资源的可用性，以及我们如何尽快让电力到位，越早越好。

It's more about availability of of of things and how we get kind of power aligns the sooner, the better on it.

不过我想具体说说模块化是怎么运作的，乔丹，你能把那张图调出来吗？

But I guess going into it a specific of how modular work, Jordan, do you want to pull up that chart real quick?

好的。

Yep.

没问题。

Got it.

马上。

Coming.

对。

Yeah.

基本上，数据中心建设分为三个层级，我们刚才在顶部看到了。

Essentially, are three layers of datacenter build outs that we have seen on the top.

就是系统、外壳、场地外壳和系统。

Their system, shell, the sites, shell and system.

第一层是场地，更多是土地平整和地基。

The first layer is the sites as more the land grading foundations.

我是说，这部分没法模块化，因为你得先平整土地、开挖地基，才能建数据中心。

I mean, that that part is not something that can be modular modularized because you had to physically grade the grade the cut the the foundations of the sites for it to be able to build the data center.

这也是我们现在在工地建设上看到的最大瓶颈之一。

And that's what we have seen as one of the biggest bottleneck right now as well on the site constructions.

真正能模块化的部分是外壳和系统。

The part where we do get modularization is the shell in the system.

外壳更像是上层建筑，包括屋顶、结构、外立面，以及所有防水密封的部分。

Shell is more of the upper bound, you see the roof, the structure, the skin, everything else that waterproofs the enclosure.

我们举的最大的例子就是Meta的帐篷，那真是成千上万顶，直接立在地面上。

The biggest, I guess, I guess, wide sample that we give is that, a meta shell, meta's tent, is literally thousands, tens of tens to which stands up out of the grounds.

我记得下面章节里有一张图。

I think we have a picture near it down in the sections.

对，就在那儿，没错。

Yeah, I think right there, yes.

这些帐篷是放在Prometheus New Albany的地面上的，MEP只是在基础建筑上盖一个非常轻便的帐篷外壳。

So these are tents that are put on the grounds for a Prometheus New Albany where MEP just puts a high, a very soft enclosure tents on top of the foundation building.

这个速度快，周转快，处理也快，但不太能承受重型的施工工作，不过它能做到尽早接入电力，所以还是能派上用场。

It's quick, it's fast turnaround, it's quick to process, but it doesn't stand very hard construction work, but it does the job where it can get tied to power sooner or later.

所以，建造起来非常非常快。

And therefore, it's very, very fast to build a construction out loud.

这就是外壳的分类方式。

So that's the kind of shell taxonomy for it.

再往前看，第二部分更多是关于设备制造。

If you go back, the second part is more of the equipment manufacturer.

如果对比传统施工和预制施工，我们发现预制施工比传统施工快了差不多29%。

If So you look at a traditional versus precast, we basically see it's almost 29% faster in terms of precast versus traditional.

这种节省主要来自并行施工：设备在工厂里制造，同时现场也在施工，设备造好之后，可以直接运到现场，即插即用，而不是像传统那样按顺序一步步来。

Where that kind of savings come from is the parallel construction where the equipment is built at a factory, the site itself is being constructed, and once the equipment is built, they can essentially be shipped to sites and plugged in play instead of doing in a traditional bilateral process.

从我们看到的案例来看，传统方式从头到尾大概需要十五周，你得先做基础建设，然后再装设备，最后再连接起来。

The world we've seen from initiate takes about fifteen weeks to do end to end, where you have to do the basic construction first, and then you do the equipments and you plug it together.

而并行处理的话，一边在工厂做，一边在现场做，等两边都完成了，再组装起来，接到现场进行调试，这就是模块化带来的节省。

You do a parallel processing where you do it both, one in the factory, one's on-site, Once everything done, you put it together and plug on sites for commissioning, and that's where we see the savings come from in terms of modularizations.

太棒了，兄弟。

Awesome, man.

好的，我现在脑子里有好几个想法，但具体来说，我在做ClusterMax的时候，很多买家在考虑模块化的时候，最大的顾虑就是它可能会带来一些风险。

Okay, so my mind's going into a few different places right now, but specifically, I think when I'm working on ClusterMax, a big thing that some of the buyers think about when it comes to modular is that it may introduce some risk.

当然，你可以更快，但在传统的数据中心建设中，你有冗余电力、冗余冷却、互联网连接以及可扩展网络系统规模等方面的保障。

Sure, you can go faster, but with a traditional data center construction, you have some guarantees around redundant power, redundant cooling, the internet connection, the size of the system that you can get on the scale out network.

那么，你能谈谈，比如说，在这些项目中，谁承担风险？

So, can you talk a little bit about, let's say, who owns the risk in these projects?

是第二天运营团队吗？他们接手站点，期望获得与传统站点完全相同的正常运行时间，并且要管理所有风险，比如设计上的变化？

Is it the day two operations team that's, like, taking on the site and expecting to have the exact same uptime as a a traditional site, and they're gonna manage all the risk of, like, you know, the the changes in the design?

还是由OEM来调整电力和冷却需求，或者努力跟上站点设计，以适应机架密度增加带来的重大变化？

Is it over to the OEM to kind of change power and cooling requirements or try to keep up with the site design to keep up with the big changes as the racks get more dense?

还是由EPC自己来承担，因为他们才是被签约来做这些工作的？

Is it up to the EPC themselves who are getting contracted to actually do this stuff?

Nico，也许你可以谈谈人们是如何看待这里的风险和权衡的。

Nico, maybe you can talk about how people are thinking about the risks and trade offs here.

对，当然可以。

Yeah, absolutely.

我的意思是，我们可以聊聊我们听到的一些不同风险和挑战，比如实际部署系统时发生的一些真实案例。

I mean, we can go through some of the different risks and, let's say, challenges that we've heard, like actual stories that we've heard that happen when deploying the systems.

而且很多问题都跟这个产品还处于非常早期的部署阶段有关。

And most many of them have to do that we are at really early stages when it comes to deploying this product.

所以遇到一些问题也是可以理解的。

So it's kind of understandable that we run into some problems.

是的。

Yes.

你知道，我觉得有一定比例的失败率是可以接受的。

You know, I guess some margin of failure rate, it's acceptable.

其中一个，就像我一开始提到的，就是模块本身的故障，这些模块本应该是即插即用的，但你插上去之后它却不工作。

One of them, as I mentioned at the beginning, is just pure, you know, failure of these modules that when are supposed to be plug and play, you plug them and they don't play.

就是有些东西没有按照预期那样运行。

And it's just, you know, something it doesn't work as as suspected.

然后基本上，你得把所有的服务人员都叫到现场来。

And then basically, you need to bring all your, you know, service and and service guys into the site.

你得等他们到现场去搞清楚到底哪里出了问题，比如为什么UPS不启动之类的。

You need to wait then to go to the site to understand what is wrong, why the, I don't know, the UPS is not turning on, whatever.

我们确实听过这样的故事，特别是有些机电承包商直接说，你知道吗？

And we've heard the stories like this of some MEP contractors specifically saying, you know what?

这些模块是现成的，但现在我们听到的故事是，他们半夜打电话给我们，我们得赶过去看看发生了什么。

These modules that are sold off the shelf, well, now we are hearing stories that, oh, they called us in the middle of the night and we need to go there to just figure it out what's happening.

在我看来，这种情况下，责任当然在OEM那边。

In my view, in that case, of course, the responsibility, on the OEM side.

你知道，就是那些人。

Know, these are the guys.

如果我们想到OEM，比如维谛、施耐德、Flex这些公司，他们卖这种整套设备，他们得确保自己的系统真的能用。

If we think of OEMs, think of the Vertivs, the Schneider's, Flex, all of these guys that are selling one, like this package are the ones that need to make sure that, no, their systems actually work.

然后，除了质量问题，我会想到物流方面。

Then I would go to other than actually quality issues, I would go to logistics.

你知道，我的意思是，是的，那是个风险，因为你知道，当我们考虑运输这些模块、这些集装箱时，有时候它们是非常大的集装箱。

And you know, that's I mean, yeah, that's a risk because, you know, when we think of shipping these modules, these containers that sometimes are like, there are huge containers.

从物流的角度来看，这可不是件容易的事。

It's not from a logistics point of view, it's not an easy task.

我们实际上看到的是，从供应链的角度来看，所有这些无论是OEM还是系统集成商，他们在组装和运输这些模块时，都在努力靠近站点，并在数据中心实际建设的地方附近建立足迹，因为这样可以省去很多麻烦，比如，哦，现在我需要把所有这些巨大的集装箱和昂贵的设备运到世界各地。

What we are seeing actually is that for, you know, part of the value proposition from a supply chain point of view, from all of these, whether it's OEMs or it's system integrators that are assembling and shipping these modules, they are trying to be close to the sites and close and have footprint close to where the datasets are actually being built because it's going to save a lot of headaches when it comes to, oh, now I need to I know imagine shipped all around the world these huge containers with all of the expensive equipment.

我们当然听说过这样的故事，比如一个模块里装了一些非常珍贵的电动汽车设备，在路上翻了车，然后，你懂的，有些UPS就完全坏了。

We've heard of course the stories of, you know, a module that included some really precious electric car equipment like tripping in the middle of the road, and you know, there you go, some UPS is just completely, you know, that broke down.

所以在物流方面，我认为这是运营商在现实中遇到的主要挑战之一。

And so in terms of logistics, that's a huge, I would say one of the main challenges that operators are in real life finding.

看起来，运输这些模块似乎会像发货一样简单。

It seems that, you know, it's going to be just as easy as ship these kind of modules.

在这种情况下，我再说一次，有可能，是的，这取决于OEM和集成商，他们负责把货物运给客户。

And in this case, again, I'd say there's possibility, yes, that it's in the hands of the OEM and the integrators that are shipping this to the customer.

对。

Yeah.

这些都是目前全球最大的公司在推进这件事。

These of are the biggest companies in the world that are pursuing this right now.

AWS、Meta、Crusoe的Stargate项目、Hut 8、Navias。

AWS, Meta, Crusoe with the Stargate sites, Hut eight, Navias.

这些真的是最大的新云服务商和数据中心运营商，对吧？

These are the biggest neo clouds, really, and the data center operators, right?

Compass、QTS、Aligned，你们聊过所有这些公司都在追求模块化，而且Eric，你从这里面听到过一些恐怖故事，对吧？

Compass, QTS, Aligned, you guys talked about all of them pursuing modular and different Eric, you got some horror stories from this, right?

对。

Yeah.

对。

Yeah.

我不能透露具体是哪家供应商，但我们听说过在北弗吉尼亚，他们当时在运输模块化数据中心设备，卡车载着这些模块。

I mean, so I can't say the vendor, but we have heard about North Virginia where they were shipping about modules modular data center stuff, and the truck was shipping the modules stuff.

结果因为太重，加上高速公路颠簸，卡车直接翻在了路边。

Like, it basically flipped over on the side of the highway because of how heavy it is and how bumpy the highway is.

所以这对供应商来说是个大问题。

So, it costs a big issue for the provider.

而且，这带来的后果是，很多保险公司不愿意为这些数据中心提供保险，不是融资，而是保险，因为那些模块化设备太贵了。

Also, the consequences of that is that a lot of insurance are unwilling to finance these, not finance, but insure these data center because of how expensive these, I guess, modular equipments are.

比如有个地方，一辆卡车翻到高速公路边上，模块化设备清理起来特别麻烦。

At one site, for example, a truck slip over to the side of the highway, Modular is a huge mess to clean up.

而且他们还要支付巨额罚款，保险费用也很高。

There's also huge penalties for them to pay off on the insurance price.

所以我们接触的很多供应商都反映，他们很难找到保险公司愿意从物流角度给他们承保。

So what we see from a lot of providers that we talk to is that they have an issue trying to find insurance company to insure them on the logistic perspective.

因此，我们看到越来越多的模块化工厂建在离现场更近的地方，以消除物流上的限制。

Therefore, we're seeing more of modular factory getting closer to the site itself to remove that logistic constraints.

我觉得这里头有

I think there's

各种物流问题，对吧？

all sorts of logistics, right?

对。

Yeah.

再说到物流这块，我觉得，谈到供应链问题，如果想想整个行业正在向模块化发展的大趋势，你可能会看到这样一个世界，就是那些原始设备制造商，比如伊顿、施耐德电气这些大公司。

Further on the logistics point, I think, when you talk about supply chain issues, I think if you think about where the industry is going towards modularization more broadly, you could get you could see a world where your OEMs, you know so, again, these are Eton's and your Schneider Electric's of of the world.

他们会变得以模块化为主，或者至少模块化业务会比以前大得多。

They become predominantly modular or or, you know, or at least, you know, the modular part of the business is much bigger than before.

而且我觉得在这种情况下，这些公司会更容易受到供应链风险的影响，因为就像这个词本身说的，一个撬装设备或者一个模块里有很多零部件，然后你要把这些东西运给客户。

And I think in such a scenario, like, these companies are more exposed to supply chain risks because as as the word suggests, right, like, there are many components that go into a skid or, you know, a a modular a module, you know, that you will ship to the customer.

所以，如果缺一个零件，整个撬装设备就发不了货，而它的平均售价可比现在高多了，现在你还能继续发那些没受供应链影响的货。

So, you know, missing one component now means that you can't ship an entire skid out the door, which is much higher, you know, average selling price compared to the status quo where you could continue to ship the the other stuff that isn't affected by by supply chain issues.

这是第一点。

So that's one.

我觉得我们看到的另一件事，考虑到行业的发展方向，也很有道理，就是原始设备制造商和集成商都在投入大量资金，为满足需求所需的模块化产能做准备。

I think the other thing that we've been seeing, which makes a ton of sense given, you know, where the industry is going is that, you know, OEMs and and integrators, they are all investing quite significant sums to kind of prepare themselves for for the the capacity, the the modular capacity that is needed to to meet the demand.

所以，仅从这一点来看，我觉得这本身就带来了一定程度的风险，比如他们能不能在某个重要客户那里拿到份额，或者客户会不会引入第二供应商、第三供应商等等。

So, from that alone, I think I think that introduces some some level of risk, whether it be, you know, they are able to get get share at at a certain important customer, whether the customer would kind of bring in second source or third source, etcetera.

所以我觉得，整个建设过程本身，或者说它带来的，确实给OEM或模块化集成商带来了一些风险。

So I think, yeah, just the entire build out itself, like, introduces, or I guess it brings along with that some risks to the OEM or the modular integrators.

你们可能对这些OEM和供应链中的不同公司做过一些分析，觉得哪些公司更积极地推进模块化，目前看到了什么趋势？

And you guys maybe have done a bit of analysis on these OEMs and these different companies in the supply chain that you think are more aggressively pursuing modular and seeing where the trend is going so far?

我觉得，抱歉，抱歉尼科，但确实，正如你提到的，模块化很重要的一点是要看那些在客户那里有强势地位的供应商，因为实际上是这些客户在设定模块化发展和产能建设的速度。

I think, sorry, sorry Nico, but yeah, I think, you know, as as you kind of alluded to that, I think with modular, it's very important to kind of see the suppliers that are that have a strong position with the customers, because it is really these guys, you know, the customers who who are setting the pace, so to speak, of of modular development and and the modular capacity build out.

所以，有一家公司大家都很清楚走在前沿，那就是Compass Systems。

So so so, you know, one company that everyone kind of knows is in the forefront of this is is Compass Systems.

大概两个月前，或者一个多月前，我们发布了一份关于这个的研报。

And I think about two months ago, a little more than than a month ago, we went out with a note on this.

我们的分析从一个非常简单的问题开始。

And, you know, our analysis just started with a very simple question.

我们知道这些家伙是前沿的，或者至少是早期的推动者之一。

So we know that these guys are at the forefront or at least one of the the early movers.

我们知道他们在谷歌和Meta那里有很好的位置。

We know that they are in good position with Google and Meta.

我们来看看这在数据上是怎么体现的。

Let's see where that shows up in the numbers.

所以，如果稍微分享一下我们的屏幕，我们做了一些分析，看看他们在能够生产模块化产品的产能上投入了多少。

So if just to share our screen a little bit, we we kind of looked at we did kind of an an an analysis showing, okay, how much have they invested in a capacity that is that is capable of producing modular.

为了做这个分析，我们找了一个在模块化进展上目前还不太知名的同行作为基准。

And to do that analysis, we kind of use a peer a peer that isn't really known for the modular progress and at least to date, and we use that as a baseline.

你可以看到他们为了模块化投入的资本开支规模有多大。

And you can just see the scale of of of CapEx that they have been engaged in for for the purpose of modular.

所以我觉得这是一个非常好的直观展示，显示了在模块化上投入的资金规模，以及这些公司面前的机会，还有他们真正想要抓住的东西。

So I think this is I think this is a very good, like, know, visual representative representation of of the sums that that are being invested in in modular and the the opportunity that is hit ahead of these guys and what they're really gunning to to capture.

对。

Yeah.

对。

Yeah.

而且我觉得，除了Compass Systems之外，还有其他几个玩家，我们SemiAnalysis也在密切关注，不过我会让Nico和Eric来提问他们。

And I think, you know, aside from comfort systems, there is also several other players that we we at SemiAnalysis are keeping a close watch on, but I'll let Nico and maybe Eric question them.

不，奈杰尔。

No, Nigel.

我觉得这些就是这个领域的关键玩家。

I mean, I think those are the key players the space.

我的意思是，在文章里我们尽量说得很清楚，我们一直在讨论AWS、Meta，还有所有那些超大规模数据中心公司在模块化领域做的事情。

I mean, my point would be that, in the article we tried to be very clear, we've been discussing what AWS, Meta, all the cobas are doing in the modular space.

这既是从需求的角度来看，也是从供应链的角度来看，比如到底是谁在供应这些模块。

That's, you know, from a demand point of view, but also from a supply chain point of view, like who's actually supplying these modules.

有一个分类体系对我们受众来说也很有帮助，因为我们一直在讲维谛、施耐德、伊顿，还有康明斯、斯特林和昆塔。

It helps to have, you know, a taxonomy for, you know, for our audience also to to know because we've been talking about Vertiv, Schneider, Eaton, but then Comfort, you know, Sterling Quanta.

这些都是EPC公司。

These are all EPC companies.

其实我们可以把这些分成两组。

I mean, it's just that we can split this into, you know, two groups.

两组真正向客户销售这些模块的公司。

Two groups that are actually selling these modules to to the customer.

首先，有那些OEM厂商，比如维谛和伊顿。

First of all, have the OEMs, you know, the Veritivist Netters.

这些公司已经在向客户销售UPS、开关设备、变压器和整流器了。

These guys are the guys that are already selling the UPS, the switchgearity transformers, the rectifiers to the customers.

他们意识到——维谛是第一个意识到的——其实可以把所有设备装进自己的集装箱或金属外壳里，然后直接运出去。这不仅对客户有用，就像我们刚才在电话里讨论的那些原因，而且对我们自己来说也更好。

They realize, and and Verdiv was the first company to realize that, oh, we can actually put all of our stuff in our container, in our metal enclosure, and ship them, it's not only going to be useful like, be interesting for our customers because of all the reasons we've been discussing in this call, but also for for us, it's much better.

我们有机会扩大每兆瓦的可服务内容。

We have the opportunity to expand our addressable content per megawatt.

而且仅仅因为包含了维谛、伊顿或其他品牌的内容，就需要这种曝光。

And just because of we are just including Vertiv, Originator, or Eaton content need that exposure.

这就是OEM那一类。

That's the OEM group.

但还有第二类。

But then there's also this second group.

我们称之为EPC集成商，基本上就是那些不直接向客户销售UPS和开关设备，但擅长设计和工程化这些机械和电气系统的公司。

We call it EPC integrators, essentially companies that are not the companies that are selling the UPS switchgear to customers, but have the expertise into design, engineer these systems, these mechanical and electrical systems.

奈杰尔刚才提到了舒适系统。

And Nigel was talking about comfort systems.

他们是北美最大的机电承包商之一，和最大的超大规模数据中心运营商合作。

They are one of the main MEP contractors in in in North America working with the biggest hyperscalers.

他们有这方面的专业能力。

They have that expertise.

他们做的事情是，和这些客户合作，不提供自己的UPS，也不提供自己的开关设备，而是负责集成、组装，并帮助客户进行设计和工程。

And what they are doing is they are working with these players, not supplying their own UPS, they're not supplying their own switchgear, but they are integrating, assembling, and helping their customers, designing and engineering the customers.

所以，你知道，这两类公司实际上就是负责把模块集成并销售给客户的那批人。

So that's you know, those are the two groups that are actually, let's say, integrating and selling the modules to the customers.

对。

Yeah.

事实胜于雄辩。

Proof is in the pudding.

这确实有效。

It's working.

这些站点已经上线了。

These sites are up.

有人在用它们。

There's people using them.

这不是理论上的，不是说他们将来可能会这么做。

It's not like this is theoretical that they should maybe do this in the future.

他们现在真的在建。

They're actually building them right now.

我觉得，如果你想想这三种不同的方式，OEM主导、系统集成商主导，还是运营商主导。

It seems to me, like, if you think about those three different approaches where the OEM is leading it versus the system integrator is leading it versus the operator is actually leading it.

比如，自建运营商主导，像AWS或Meta就说，我要这么做。

Like, the the self build operator led, you know, AWS or Meta just says, I'm gonna do this.

这显然是让供应链上所有人都明白的方式。

It's, like, obviously, the way to make everybody figure it out in the supply chain.

但也许结论是，那些去年或两年前没考虑模块化的人，将来建站点时会受益于整个供应链的成熟，因为所有供应商都已经服务过最大的客户，现在他们也得服务那些做新东西的新玩家，对吧？

But maybe the conclusion is that people who weren't considering modular last year or two years ago are now gonna get the benefits of this whole supply chain being built out when they go to build a site in the future because, well, all the suppliers have already had to serve the biggest customers, and now they're gonna have to serve the new guys doing new stuff too, right?

也许你可以聊聊这个过程，比如，我特别好奇工厂测试是怎么运作的。

Maybe you can talk a little bit about like, like, the process in the fact like, one thing I'm kinda fascinated by is how factory testing works.

你讲过那个集装箱从卡车上滚下来的故事。

I mean, you tell the story about the shipping container rolling off a truck.

很多机架一直装在卡车里运到数据中心，通常你得在数据中心现场做一大堆集成工作。

Like, many racks have been in trucks all the time getting shipped to data centers, and generally, what happens is you have to do a whole bunch of integration work on-site in in the data center.

不管你在工厂里怎么测试，到了现场总得重新测试和调试。

And as much as you try to test stuff in the factory, it's always gonna have to be retested and commissioned on-site.

你觉得现在这种情况会不会有大的改变？比如有了那些靠近建设现场的枢纽，可以在那里生产所有设备，这样是不是能降低质量风险，也就是说，缩短系统真正上线的时间，我是说IT系统，不只是电力和电气设备？

Do you see this meaningfully changing now where there are hubs where you can produce all of this equipment that's near where a lot of these sites are being built, and, you know, it is gonna de risk quality, in other words, like the time it takes to actually bring the systems online, as opposed to, like, meaning the IT systems as opposed to just the power and electrical equipment?

我觉得，当然，说到调试，预制化确实能帮助调试。

I mean, I think, of course, talking about commissioning and how prefabrication helps commissioning.

这也是个大问题，一个复杂的话题，因为有很多误解，比如我们确实知道有些故事说，他们在工厂里就完全完成了调试，你知道，五级调试全在工厂做。

It's also a big question, a complex topic in the sense that there's a lot of misunderstanding out there in the sense that we actually know who stories that, oh, they are running commissioning fully, you know, the five levels of commissioning fully at the factory.

现场什么都不用做。

You don't need to do anything on-site.

这不是真的。

That that's not true.

比如，工厂里能做的调试工作其实很多。

Like, you can there's so much commissioning of the commissioning process that you can do at the factory.

但是，基本上，一直到调试的第五级，现场还是需要做一些调试工作的。

But those, you know, essentially, all the way to l five of commissioning, you need to do some level of commissioning on-site.

这是第一点。

That's one.

第二点是，实际的调试时间真的取决于具体情况。

Then two is that the actual commissioning time, it's really dependent.

我们试图展示我们的中间值或者指导值，比如从传统建造到完全集装箱化的数据中心，施工时间能缩短多少。

And we try to show kind of our midpoint or our guidance of you know, how much the construction timing can be shortened from stick built to, you know, fully containerized data center.

但请记住，调试是过程中非常依赖项目的一部分，时间可以从三个月到八个月不等。

But bear in mind that the the commissioning, it's a part of the process that's super dependent on the project, and it can run all the way from three months to eight months.

所以，给出一个通用的答案并不容易，不可能直接说，用这种模块就能完全避免调试，或者把调试周期从五个月压缩到三个月。

So it's not as easy as to give a general it's not possible to give a general answer into, oh, by doing this, using this kind of module, you can directly avoid commissioning or or squeeze the commissioning period from five months to three months.

最后一点我想说的是，我们听说过，你知道，有些运营商会采用不同的调试方式。

And last point I would say is that we've heard, you know, we've known operators doing different kinds of commissioning.

我觉得，他们可能只是跳过了一些步骤，或者对流程不太信任，然后就说，好吧，我不做调试了。

And I say, you know, that maybe they are just avoiding some steps or less trusting trusting the process and, okay, I'm not going to do commissioning.

我直接开机，直接启动我的GPU。

I'm going to turn on my turn on my my GPUs.

然后如果出了问题，那好吧，算我的错。

And then if something happens, well, you know, my bad.

但我们确实听说过这种情况。

But we've heard of that.

这其实不完全是模块化带来的后果。

And that's not really a full, you know, consequence of modularization.

以前就已经有这种情况了。

That was happening already.

所以，不管运营商是完整走调试流程，还是跳过一些步骤，我们都可以聊聊为什么他们会选择跳过某些步骤。

So whether an operator is doing the full commissioning process of just skipping some steps, And, you know, we can discuss why operators are choosing to skip some steps.

但很多时候，归根结底还是电力问题。

But again, many times, it all comes down to tend to power again.

你明白吗？

You know?

而且如果液体负载组的交货周期很长，在美国就是这样，我可不想等那么久。

And if long there's lead time for liquid load banks, there's been in The US, I'm not going to wait for that.

我会跳过CDU的调试。

I will skip the commission of my CDUs.

我会跳过那一步。

I'll skip that.

有道理，

Makes makes

对，有道理。

sense, yeah.

但这种情况不会发生。

Doesn't happen.

这样做的后果就是，你知道，质量问题就会冒出来。

The consequence of that is that, you know, quality issues then arise.

但再说一次，这并不完全是做这些套件或模块化直接导致的后果。

But again, this is not a full, a consequence, a direct consequence from doing these kits or modularization.

对，我想补充一点，我们跟供应商聊的时候，也让我们挺意外的，就是L5调试现在成了模块化故事里最大的瓶颈之一。

Yeah, I think I'm gonna add to that, when we talk to vendors, was surprising to us too that L5 commissioning is becoming the one of the biggest constraints of the modular story right now.

如果从宏观来看，测试其实分两部分。

If you think of high level, there's two parts of testing.

一个是工厂测试，一个是调试测试。

There's the factory testing, there's the commission testing.

工厂测试就是在工厂里做的，证明每个模块在工厂里能独立运行。

Factory testing is like the one that you do in factory where you prove each module works independently in the factory itself.

但真正到了现场，你得证明整个系统能在不同的工作流程和不同的故障条件下协同工作。

But really on-site, you have to prove the entire system work together under different workflow and different failure condition.

比如说，假设市电掉电了，那其他的UPS是不是能按正确的顺序一起启动起来。

So let's say, guess, low if utility power drops, for example, do the rest of the UPS kind of lines up together in the right sequence.

所以实际上，就像我们听到的那样，有些东西在工厂测试了一百次都没问题，但到了现场可能就不行了。

So really something like, if you when we have heard that, if for something that's tested, like 100 times of faculty work, it might not work on sites.

所以运营风险真的很高，因此对很多人来说，验收条件变得非常困难。

So the operational risk is really high and therefore, alpha condition has been super difficult for a lot of people.

这也是我们从别人那里听到的，整个过程中需要克服的一个问题。

And that's one of the things that we have heard from people that there's something that people needs to overcome as a result of the entire process.

有道理。

Makes sense.

好的。

Okay.

在结束之前，我想鼓励所有正在收听的朋友们。

As we move to wrap, I just wanna encourage everybody that's been listening.

如果你对模块化数据中心感兴趣，一定要读读这篇文章。

If you're interested in modular data centers, you should totally read this article.

这篇文章是几周前发布的。

It came out a few weeks ago.