0:04
Hello everyone.
Search this video
0:05
Welcome back to SemiAnalysis Weekly.
0:07
We are here to talk about the wild, wild west of Lego data centers.
0:11
A couple of weeks ago we put out an article on modular data centers and construction.
0:16
So I've got Eric, Nigel, and Nico, and we're going to 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.
0:37
So guys, welcome to the show. Excited to dig in.
0:40
I'm going to pass the ball to Nico for the... yeah, exactly, man.
0:46
Nico and Nigel are returning.
0:47
It's Eric's first time attending.
0:49
And this is a nice cross section across SemiAnalysis.
0:52
We've got Eric from the consulting team, Nigel from Core Research, and Nico from the data center, energy, and industrials model team.
0:58
So you get a little taste of everybody who touches modular data centers and contributes to newsletter articles like this one. Yeah, glad to be here. Awesome, guys. Okay.
1:10
So let's start by passing to Nico to give us the overview of what a modular data center is, what it means, and what makes it different from the traditional way to build data centers.
1:22
Sure, yeah, let's start with that.
1:23
I think there is no single definition for what a modular data center is.
1:29
And that was the single reason we decided to put out this article, in the sense that everyone has their own definition for what modularization, or prefabrication, as it's sometimes called, actually means.
1:41
So essentially we put out that note, and a big part of it is a whole taxonomy of what we actually mean when we talk about modular, to answer that question.
1:54
But let's start off with a pretty simple idea.
1:57
Modularization or prefabrication.
2:00
The reason we named the article Lego Data Centers is that instead of building your data center in the traditional way, where you bring all the equipment to the site, and then you bring all the workforce, all 7,000 workers at peak, to the middle of Abilene, Texas, and then you start building a data center.
2:21
Instead of doing that, you are going to build it like your old Legos, piece by piece.
2:28
And of course those pieces can vary a lot.
2:30
But think of it this way: instead of putting all that equipment together and doing the connections on site, you are going to do part of the work off site.
2:40
You're going to modularize, because you are actually using modules.
2:40
Think of a metal enclosure with equipment inside.
2:49
You're going to do that in the factory.
2:52
And then you're going to ship it to the site.
2:54
You are going to stack the different modules, the different Lego pieces of that data center, and then you're essentially going to connect them all.
3:04
It's supposed to be plug and play.
3:06
We can talk about why I say "supposed," but essentially it's a plug and play build of a data center.
3:14
In super simple terms, that's what we mean by modularization or prefabrication.
3:21
Yeah, and maybe there are a few different types of modular data centers.
3:24
There's a whole different group of options when it comes to describing how construction happens with a modular data center.
3:35
On screen I've got this chart describing stick build, which is the baseline today, versus MEP heavy skids, a full modular build, or containerized, where it's 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
3:51
getting stuff on site, from site prep and groundworks in gray, to the structure and shell in yellow, then mechanical in blue and commissioning in green, for those who are looking at the chart. Eric, maybe you can talk about some of the benefits you've seen for people going with the modular approach, and how this speed-up is actually helping them.
4:03
Eric, maybe you can talk about some of the benefits you've seen for people going with the modular approach, and how this speed-up is actually helping them.
4:10
Yeah, I think there are two big benefits that we have seen in the market.
4:17
The first one is probably labor.
4:18
We have seen multiple people come to us asking about modular designs because the market is super short on labor.
4:27
Labor rates are super high right now, and people can't find the amount of labor needed to support the immense data center build-out.
4:38
So they're really turning to modular data centers as the alternative solution for the labor bottleneck we're seeing in the market. That's the biggest one.
4:49
The second one, and I think more importantly, is just certainty in modular designs.
4:54
That's really related to financing and capital.
4:57
When we look at a lot of data center build-outs, certainty is a big thing.
5:02
Even a lot of smaller providers don't have all the necessary equipment to build an entire data center, so they turn to modular as a potential alternative to build a data center faster, for time to power, but also to have certainty about when they would get power on site within a certain amount of time.
5:26
So we see a lot of different players looking for different modular solutions based on where their constraints are, both in labor and economics, as well as where their constraints are on certainty.
5:37
By bringing in modular solutions, it can give both investors and themselves a quicker time to power, and more certainty about when that power can come online, based on bringing in a complete package as a whole.
5:56
And this isn't just random crypto miners that are doing this.
5:58
Obviously this became popular when we put out the article on Meta a few months ago and talked about their tents.
6:07
And then in this article we described how AWS is doing this as well.
6:11
So maybe you guys can talk a little bit about the different players that are exploring this. Yeah, absolutely.
6:22
I was just going to say that maybe before addressing the players, I would add on to what Eric mentioned.
6:26
You mentioned the labor shortage, which of course is huge.
6:26
We could spend an hour just talking about labor.
6:35
Also predictability and quality.
6:38
But essentially, one thing that's clear after all these conversations we've had with different data center builders and developers is that there's no clear pecking order.
6:48
Some folks are doing modularization for one reason, because they actually like the predictability of knowing that when you plug in your modules, they're going to function as promised.
7:00
Some others are doing it because they cannot get enough electricians or welders coming to the site.
7:07
But there's also a third reason, which I would say is the clear number one today, and that is time to power, the building time.
7:15
Essentially it's shrinking the time it takes to build your data center, from the day you get your heavy machines onto the site to do the site works until the day you're ready for service, ready to energize.
7:30
And that's a common practice.
7:32
So when we think of the players that are doing this, or the different strategies these players are pursuing, which we're going to talk about now, essentially it all comes down to how I can build this as fast as I can.
7:45
How can I shrink these 18 to 24 months that it usually took to build a data center, sometimes over three years, down to one year?
7:56
Because everything translates to tokens.
7:59
Those 12 months are the opportunity cost of not being able to generate 12 months' worth of tokens.
8:07
Of course, we talked in our newsletter recently about how before we were talking about 11 to 15 million dollars per megawatt.
8:16
Now we're seeing compute deals much higher than that, 40 million per megawatt.
8:21
Anthropic, in some configurations, is generating over 100 million per megawatt.
8:24
Think of how much 12 months of token production time is worth.
8:30
So with that, we can go back to the miners.
8:35
Yeah, so just 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 build versus the baseline.
8:50
But the part that compresses so much is the mechanical and electrical fit-out.
8:55
So you're doing three months of electrical work on the system instead of up to nine months.
9:03
So can you guys talk about what this means specifically for the biggest companies in the world that are building the biggest data centers?
9:11
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, like Stargate Abilene with Crusoe in Texas, as well as Microsoft's Fairwater site in Wisconsin.
9:29
Yeah, I would say this chart really ties to one of the points, which is the labor shortage, or the difficulty of finding 7,000 workers to come to the middle of nowhere.
9:41
We hear stories that it's really hard to get electricians on site, but then we asked, okay, how can we actually get some empirical evidence of this?
9:56
We looked into how the wages of new hires, these new electricians being hired, are trending.
10:06
And we came across this chart.
10:06
You can see that in the states and counties where the biggest campuses are being built, the ones that demand between three and four thousand workers at peak, wages for these new hires were essentially doubling, sometimes tripling, in some key states.
10:33
Nigel, what's your take on the labor shortage?
10:36
I assume you hear about this from investors in many different companies across the industry, modular and otherwise.
10:44
Yeah, it's a great question.
10:48
The first thing to note, and one that is increasingly being appreciated across customers, data center operators, investors, and so forth, is that of the different bottlenecks out there, there's power, there's compute, which was once a bottleneck, and there's memory, which was the topic of interest, and then labor.
11:22
Labor is one of the constraints that is really difficult to resolve, because the supply isn't very elastic at all.
11:31
Even when you talk about power, the economics of AI, meaning the huge premiums that are there for the taking if you can bring a unit of compute online quicker, have brought in a lot of marginal power sources and power formats.
11:56
We at Core Research, again with the energy team, have written a lot about how reciprocating engines are one of the marginal formats that have come in, and everyone knows about fuel cells as well.
12:13
But if you think about labor, it's really difficult to solve, because these aren't your garden variety electricians who can just come in and stand up a data center with all the specialized electrical wiring that is needed.
12:37
This is very critical, mission critical work that is being done.
12:42
And you need to have these guys go out into the middle of nowhere.
12:46
The term for this is field labor.
12:49
As you see from that chart, the wage rates for these guys are just soaring.
12:55
So because of that, it's being appreciated that one way to solve this is modular.
13:07
And as we pointed out in a tweet that went out today, this actually shows up in share prices when you look at the entire data center industrial space, and how the construction, or engineering, procurement, and construction guys, what you call the EPC guys, have traded this year relative to the equipment OEMs.
13:33
You see the construction guys outperforming.
13:35
So check out our Twitter.
13:35
I think we have a good series of tweets where we cut the data showing that.
13:42
And yeah, it goes to show what my colleagues Nico and Eric have been talking about on labor. Definitely, yeah.
13:51
And just to put a finer point on this, we hear a lot about bottlenecks, and the term gets thrown around a lot.
13:59
There are two different types of bottlenecks in my mind.
14:03
There are industry-wide, high-level constraints, like how much CoWoS can TSMC produce, how much memory can the big three memory vendors produce.
14:12
This is a constraint on the whole industry: how many chips you can produce.
14:15
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.
14:24
These are two different things, because individual projects with multiple phases may have their allocation of chips from the global supply chain that can't be increased above a certain amount, but they have their allocation.
14:39
It's waiting for them to come and install.
14:42
And their bottleneck is things like the labor you're describing here.
14:45
It's things like recip engines or whatever.
14:48
But if we look at the industry as a whole, some of these individual bottlenecks that show up at individual sites are not necessarily representative of what's going on globally, worldwide, when we look at the massive amount of progress in chips being brought online over time.
15:06
Nico, maybe you can comment a little more on high-level, industry-wide tracking and differentiate between the local bottleneck of labor and the industry-wide demand that we keep seeing.
15:21
Yeah, Jordan, I think this is an extremely important point, because absolutely, labor is a big challenge.
15:27
It's one of the main constraints operators name when we talk to them: actually getting the labor on site.
15:33
But we shouldn't extract from that the message that we don't have enough electricians in the US.
15:42
It is just a really complex optimization problem, in the sense that in the US as a whole, at least in the next twenty-four months, with the capacity that is expected to be built, you probably have enough.
15:58
But the thing is, how much of that labor is not only capable of taking part in a mission critical and very complex project, but also how much of the labor is reachable in these areas?
16:12
Labor is more of a logistics problem than it is a structural problem of labor being inelastic.
16:24
And it's a really complex one.
16:27
What we are seeing now, other than the solutions that contribute on a net positive basis to reducing the man hours required, like modularization and what we covered in our article, is that there are other strategies, which is finding the right incentives and the right solution to this optimization problem, which is how we actually reallocate all of the available labor to the right locations.
16:59
We're doing more work on that.
16:59
The good thing at SemiAnalysis is that because of the work the data center team has done over three years, we actually know exactly where all of these sites are going to be built.
17:15
When we talk about 15 gigawatts, 30 gigawatts being built, we actually have the breakdown by specific county.
17:24
So we can take all that and say, okay, let's take this capacity, these megawatts, and convert that into labor demand.
17:33
Then we can see, okay, in this specific county we need this many labor hours, and we don't actually have them.
17:42
How do we manage to bring in the electricians that are living 300 miles away?
17:48
This is the conversation.
17:48
This is what is actually happening in the industry today.
17:53
Of course wages are a big part of this, and this is why we included the previous chart, but we've also heard some other things, per diems and everything, all the way to taco trucks in the parking lot, because the actual workers who are going to decide demand it.
18:14
So it's absolutely a local problem, and the solution is not easy, but Well, let's talk about the specifics.
18:26
Eric, maybe I can pass it to you to talk a little bit about specifically what some of these people are doing, and what makes a modular data center different.
18:34
You guys had a great taxonomy for this in the article.
18:36
System, shell, and site, the three layers of the stuff that's getting built.
18:44
Maybe you can break down what these three components are, and 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. Yeah, absolutely.
18:58
Before that, I just want to comment on what Nico said earlier.
19:01
I feel like there's always an upper bound to where labor can be.
19:04
You can't produce humans overnight.
19:07
It's even harder to actually train a human to be data center proficient, to get to a level of site ready.
19:17
So there's a mismatch between where data centers want to have power online today and when humans can come in to actually build a data center in the future.
19:25
And that mismatch creates huge demand for modular.
19:29
So modular is not just about cost and timing, it's more about availability, and how we get power online, the sooner the better.
19:39
But going into the specifics of how modular works, Jordan, do you want to pull up that chart real quick? Yep, got it. Coming up.
19:51
Essentially there are three layers of a data center build-out, as you see on the top.
19:57
There's site, shell, and system.
20:01
The first layer is the site.
20:01
It's the land grading and foundations.
20:04
That part is not something that can be modularized, because you have to physically grade the foundations of the site to be able to build the data center.
20:16
And that's what we have seen as one of the biggest bottlenecks right now as well, on site construction.
20:22
The parts where we do get modularization are the shell and the system.
20:26
Shell is more of the upper envelope.
20:28
You see the roof, the structure, the skin, everything else that weatherproofs the enclosure.
20:32
One example that we give is the Meta shell, the Meta tent.
20:32
It's literally tens of tents which stand up out of the ground.
20:46
I think we have a picture later in the session. Yeah, right there, yes.
20:57
So these are tents that are put on the ground for Prometheus in New Albany, where Meta just puts a very soft tent enclosure on top of the foundation.
21:09
It's quick, it's a fast turnaround, it's quick to process.
21:09
It doesn't withstand very hard construction work, but it does the job, where it can get time to power sooner, and therefore it's very, very fast to build.
21:25
So that's the shell taxonomy.
21:30
If we go back, the second part is more of the equipment manufacturing.
21:38
So if you look at traditional versus precast, we basically see it almost 29% faster in terms of precast versus traditional.
21:39
Where that savings comes from is parallel construction, where the equipment is built at a factory while the site itself is being constructed, and once the equipment is built, it can essentially be shipped to site and plugged in, instead of a traditional serial process.
22:04
So where we've seen it initially take about 15 weeks end to end, where you have to do the basic construction first, and then you do the equipment, and you plug it together, you do the parallel processing, where you do both, one in the factory, one on site.
22:18
And once everything is done, you put it together and plug it up on site for commissioning.
22:23
That's where we see the savings come from in terms of modularization. Awesome, man.
22:32
Okay, so my mind's going a few different places right now, but 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.
22:48
Sure, you can go faster, but with 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.
23:02
So can you talk a little bit about who owns the risk in these projects?
23:09
Is it the day two operations team that's taking on the site and expecting to have the exact same uptime as a traditional site, and they're going to manage all of the risk of the changes in the design?
23:23
Is it over to the OEM to change power and cooling requirements, or try to keep up with the site design as the racks get more dense?
23:33
Is it up to the EPC themselves, who are getting contracted to actually do this stuff?
23:38
Nico, maybe you can talk about how people are thinking about the risks and trade-offs here? Yeah, absolutely.
23:44
We can go through some of the different risks and challenges that we've heard, actual stories that happened when deploying these systems.
23:55
Many of them have to do with the fact that we are at really early stages when it comes to deploying this product.
24:02
So it's kind of understandable that we run into some problems.
24:07
Some marginal failure rate is acceptable.
24:12
One of them, as I mentioned at the beginning, is just pure failure of these modules.
24:13
They're supposed to be plug and play, but you plug them in and they don't play.
24:13
It's just something that doesn't work as expected.
24:27
And then basically you need to bring all your service guys onto the site.
24:31
You need to wait for them to go to the site to understand what is wrong, why the UPS is not turning on, whatever.
24:42
We've heard stories like this from some MEP contractors that specifically say, you know what?
24:48
These modules that are sold off the shelf, now we are hearing that they call us in the middle of the night, and we need to go there just to figure out what's happening.
25:02
In my view, in that case, of course the responsibility is on the OEM side. These are the guys.
25:09
If we think of OEMs, think of the Vertivs, the Schneiders, Flex.
25:09
All of these guys that are selling this package are the ones that need to make sure that their systems actually work.
25:24
Then, other than quality issues, I would go to logistics.
25:30
That's a risk, because when we think of shipping these modules, these containers, which sometimes are huge, from a logistics point of view it's not an easy task.
25:45
What we are seeing actually is that part of the value proposition, from a supply chain point of view, for all of these players, whether OEMs or system integrators that are assembling and shipping these modules, is that they are trying to be close to the sites.
25:59
To have footprint close to where the data centers are actually being built, because it's going to save a lot of headaches.
25:59
Imagine shipping all around the world these huge containers with all of the expensive equipment.
26:17
We've heard the stories of a module that included some really precious electrical equipment tipping over in the middle of the road, and there you go, some UPS just completely broke down.
26:32
So in terms of logistics, that's one of the main challenges that operators are finding in real life.
26:42
It seems like it's going to be as easy as just shipping these modules.
26:48
In this case again, I'd say the responsibility is in the hands of the OEM, or the integrators that are shipping this to the customer.
26:59
Yeah, these are some of the biggest companies in the world that are pursuing this right now.
27:05
AWS, Meta, Crusoe with the Stargate sites, Hut 8, Nebius, these are the biggest neoclouds, really.
27:13
And the data center operators, right? Compass, QTS, Aligned.
27:16
You guys talked about all of them pursuing modular in different flavors.
27:20
And Eric, you've got some horror stories from this, right? Yeah, yeah.
27:26
I can't say the vendor, but we have heard about Northern Virginia, where they were shipping modular data center stuff, and the truck shipping the module basically flipped over on the side of the highway because of how heavy it is and how bumpy the highway was.
27:44
So it caused a big issue for the provider.
27:47
I think the consequence of that is that a lot of insurers are unwilling to insure these data centers because of how expensive the modular equipment is.
28:02
Once a truck flips over on the side of the highway, the module is a huge mess to clean up, and there are also huge penalties for them to pay on the insurance price.
28:16
So what we see from a lot of providers that we talk to is that they have an issue trying to find an insurance company to insure them from the logistics perspective, and therefore we're seeing more module factories getting closer to the site itself to remove the logistics constraints.
28:38
Yeah, and further on the logistics point, when you talk about supply chain issues, if you think about where the industry is going towards modularization more broadly, you could see a world where your OEMs, so again these are your Eatons and your Schneider Electrics of the world, become predominantly modular, or at least the modular part of the business is much bigger than before.
29:12
In such a scenario, these companies are more exposed to supply chain risks, because as the word suggests, there are many components that go into a skid or a module that you would ship to the customer.
29:32
So missing one component now means that you can't ship an entire skid out the door, which is a much higher average selling price compared to the status quo, where you could continue to ship the other stuff that isn't affected by supply chain issues. So that's one.
29:53
The other thing that we've been seeing, which makes a ton of sense given where the industry is going, is that OEMs and integrators are all investing quite significant sums to prepare themselves for the modular capacity that is needed to meet the demand.
30:18
So from that alone, I think that introduces some level of risk, whether it be that they are able to get share at a certain important customer, or whether the customer would bring in a second source or third source, et cetera.
30:39
So the entire build-out itself brings along with it some risks to the OEM or the modular integrators.
30:53
Yeah, and you guys have maybe done a bit of analysis on these OEMs and the different companies in the supply chain that you think are more aggressively pursuing modular, and seeing where the trend is going so far?
31:10
Sorry, Nico, but yeah, as you alluded to there, with modular it's very important to see the suppliers that have a strong position with the customers, because it is really the customers who are setting the pace, so to speak, of modular development and the modular capacity build-out.
31:39
So one company that everyone knows is at the forefront of this is Comfort Systems.
31:48
About two months ago, a little more than a month ago, we went out with a note on this.
31:54
Our analysis just started with a very simple question.
31:58
We know that these guys are at the forefront, or at least one of the early movers.
32:03
We know that they're in a good position with Google and Meta.
32:07
Let's see where that shows up in the numbers.
32:10
So just to share our screen a little bit, we did an analysis showing how much they have invested in capacity that is capable of producing modular.
32:29
To do that analysis, we used a peer, a peer that isn't really known for their modular progress, at least to date, as a baseline.
32:35
And you can just see the scale of capex that they have been engaged in for the purpose of modular.
32:50
So I think this is a very good visual representation of the sums that are being invested in modular, and the opportunity that is ahead of these guys, and what they're really going to capture.
33:07
And aside from Comfort Systems, there are also several other players that we at SemiAnalysis are keeping a close watch on, but I'll let Nico and maybe Eric touch on them. No, not much to add.
33:23
I think those are the key players in the space.
33:29
My point would be that in the article we tried to be very clear.
33:29
We've been discussing what AWS, Meta, and all the colos are doing in the modular space.
33:42
That's from a demand point of view, but also from a supply chain point of view: who's actually supplying these modules?
33:48
It helps to have a taxonomy for our audience to know, because we've been talking about Vertiv, Schneider, Eaton, but then Comfort, Sterling, Quanta, these are all EPC companies.
34:03
We can split this into two groups.
34:08
Two groups that are actually selling these modules to the customer.
34:12
First of all you have the OEMs, the Vertivs and Schneiders.
34:15
These guys are the ones that are already selling the UPS, the switchgear, the transformers, the rectifiers to the customers.
34:23
They realized, and Vertiv was the first company to realize, that we can actually put all of our stuff in a container, into a metal enclosure, and ship it.
34:34
It's not only going to be interesting for our customers because of all the reasons we've been discussing in this call, but also for us. It's much better.
34:44
We have the opportunity to expand our addressable content per megawatt, just because we are including Vertiv or Schneider or Eaton content in that enclosure. That's the OEM group.
34:57
But then there's also this second group.
35:00
We call it EPC integrators.
35:00
Essentially, companies that are not the companies selling the UPS and switchgear to customers, but that have the expertise in designing and engineering these mechanical and electrical systems.
35:16
Nigel was talking about Comfort Systems.
35:18
They are one of the main MEP contractors in North America, working with the biggest hyperscalers.
35:24
They have that expertise.
35:25
And what they are doing is working with these players, not supplying their own UPS, not supplying their own switchgear, but integrating, assembling, and helping their customers with design and engineering.
35:39
So those are the two groups that are actually integrating and selling the modules to the customers.
35:50
Yeah, and the proof is in the pudding. It's working, right? These sites are up.
35:56
There are people using them.
35:57
It's not like this is theoretical, that they should maybe do this in the future.
36:03
They're actually building them right now.
36:04
It seems to me that 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, the self-build, operator-led approach, where AWS or Meta just says, I'm going to do this, is obviously the way to make everybody in the supply chain figure it out.
36:25
But maybe the conclusion is that people who weren't considering modular last year or two years ago are now going to get the benefits of this whole supply chain being built out when they go to build a site in the future, because all the suppliers have already had to serve the biggest customers, and now they're going to have to serve the new guys doing new stuff too, right?
36:43
Maybe you can talk a little bit about the process in the factory.
36:50
One thing I'm kind of fascinated by is how factory testing works.
36:57
You tell the story about the module rolling off a truck.
37:02
Many racks have been in trucks getting shipped to data centers all the time.
37:08
Generally what happens is you have to do a whole bunch of integration work on site in the data center.
37:13
As much as you try to test stuff in the factory, it's always going to have to be retested and commissioned on site.
37:21
Do you see this meaningfully changing now, where there are hubs where you can produce all of this equipment near where a lot of these sites are being built, and it is going to de-risk the quality, in other words, the time it takes to actually bring the systems online, meaning the IT systems, as opposed to just the power and electrical equipment?
37:50
Talking about commissioning and how prefabrication helps commissioning, it's also a big question, a complex topic, in the sense that there's a lot of misunderstanding out there.
38:03
We actually hear stories that they're running commissioning, the five levels of commissioning, fully at the factory, and you don't need to do anything on site. That's not true.
38:14
There's a lot of the commissioning process that you can do at the factory.
38:19
But all the way to L5 of commissioning, you need to do some level of commissioning on site. That's one.
38:28
Then two is that the actual commissioning time is really dependent.
38:33
We try to show our midpoint, or our guidance, on how much the construction timing can be shortened from stick built to fully containerized data center.
38:46
But bear in mind that commissioning is a part of the process that's super dependent on the project.
38:52
It can run all the way from three months to eight months.
38:55
So it's not possible to give a general answer that by using this kind of module you can directly avoid commissioning, or squeeze the commissioning period from five months to three months.
39:12
The last point I would say is that we've heard of operators doing different kinds of commissioning.
39:21
Maybe they are just avoiding some steps, or trusting the process less, and saying, okay, I'm not going to do commissioning.
39:30
I'm going to turn on my GPUs, and then if something happens, well, my bad. We've heard of that.
39:39
And that's not really a consequence of modularization.
39:46
That was happening already.
39:47
So whether an operator is doing the full commissioning process or just skipping some steps, and we can discuss why operators are choosing to skip some steps, again, many times it all comes down to time to power.
40:06
If there is a long lead time for liquid load banks, as there has been in the US, I'm not going to wait for that.
40:15
I will skip the commissioning involving my CDUs. I'll skip that. That does happen.
40:22
The consequence of that is that quality issues then arise.
40:26
But again, this is not a direct consequence of doing these skids or modularization.
40:34
Yeah, what I'm going to add is that when we talk to vendors, it was surprising to us too that L5 commissioning is becoming one of the biggest constraints in the modular story right now.
40:46
If you think of it at a high level, there are two parts of testing.
40:49
There's the factory testing, and there's the commissioning testing.
40:52
Factory testing is the one that you do in the factory, where you prove each module works independently.
40:59
But on site you have to prove the entire system works together under different workflows and different failure conditions, right?
41:05
So let's say utility power drops, for example.
41:05
Do the rest of the UPS line up together in the right sequence?
41:13
What we have heard is that even for something that's been tested hundreds of times in the factory and worked, it might not work on site.
41:21
So the operational risk is really high, and therefore L5 commissioning has been super difficult for a lot of people.
41:29
That's one of the things that we have heard from people, that it's something they need to overcome as part of the entire process. Makes sense.
41:40
Okay, as we move to wrap, I just want to encourage everybody that's been listening, if you're interested in modular data centers, you should totally read this article.
41:45
It came out a few weeks ago.
41:46
It's not necessarily breaking news the way we've been covering some stuff recently on the podcast, but it is a total, full overview of the whole industry.
41:57
I'm sharing on screen the modular vendor map that the guys put in this article, which shows just the vast number of players that these guys are tracking across the entire industry.
42:08
116 products, according to this chart.
42:11
And a lot of names that you might recognize from other businesses that are just getting into data centers now because of the great opportunity that modular data centers provide.
42:19
So that's what I'd like to leave it on.
42:22
Do any of you guys have comments or thoughts that we can leave the audience with as we close up?
42:28
Yeah, the one thing that I would leave is that a lot of people came to me asking, hey, is modular really about cost?
42:34
I think the argument is that it's not just about cost.
42:36
We actually see the cost benefits as not as much as what people think.
42:40
It's about eight percent, from what we have seen.
42:42
The bigger benefits are what we've talked about the whole podcast, right?
42:47
Labor, time to power, availability.
42:47
What's driving the whole modular decision is the entire industry that's shaping it.
42:55
And yeah, subscribe to our industrial model to find out more about the labor problem and more. Good one.
43:04
Yeah, my take on the vendor map and the whole state of the supply chain would be that one thing to look at in the coming twelve months is probably lead times, and actual manufacturing capacity.
43:17
Yes, this is great, we're saving labor, we're taking all the construction to the factories, but that means that we actually need footprint.
43:28
We need manufacturing plants.
43:31
We are industrializing the construction process of a data center.
43:35
That takes footprint, that takes production lines.
43:39
It's completely changing the model of what it takes to build a data center.
43:44
And what we are seeing, even if we are still in the early days of this transition, is that some of the companies at the leading edge of supplying these modules are already running into supply constraints.
43:56
They don't have the space, they don't have the factory workers.
44:02
They had a lot of electricians in the construction companies, but they don't have the factory workers.
44:07
So this complete transition is going to mean some friction.
44:16
And we at SemiAnalysis are closely looking at what the state of the supply chain is.
44:22
What if we run into not electrician shortages, but a shortage of factory workers that are actually working the production lines?
44:33
You still need electricians to assemble these modules, even if it's not on site.
44:40
Yes, a lot of productivity gains, but the labor problem here is not fully solved just by moving everyone into a factory.
44:54
You'd have to produce more people.
44:59
We are seeing companies, some of the leaders, with lead times for these products that are over a year.
45:06
If you want to tweak the module that they are selling off the shelf just a little bit, then the lead time is eighteen months.
45:14
We are already talking about the lower range of the time that we were showing it takes to build a data center in the traditional way.
45:26
So if you actually are looking at a 12 to 18 month lead time for these modules, then you need to start thinking, am I actually getting anything from these?
45:38
And that is why in our vendor map we were showing 160 players in the space.
45:47
When the top guys, the leading companies, are fully booked, the industry is going to go to these other new entrants, these integrators that have capacity, that have the expertise.
46:00
And it is a huge opportunity for a lot of new players, public companies coming from the construction space that have the integration expertise, coming in to play in the modularization and prefabrication space.
46:16
It's a huge opportunity for them.
46:19
All right, Nigel, anything from you as we close?
46:22
No, I was just going to say that there's more to come from the industrial team here at Semi Analysis. Exciting times, guys. All right, great job.
46:33
Thanks for joining the pod today. Take care, everyone. Thanks, guys.