Latest / Wired to Build / Guest Intro - Matt Caldwell (Hyper Solutions)

Transcript

Guest Intro - Matt Caldwell (Hyper Solutions) · Wired to Build

0:00
Hey, what's going on, everyone? Welcome to Wire to Build, a
0:03
data-driven, technology focused podcast digging into how
0:06
building owners and operators use technology to manage the
0:08
full life cycle of their projects.
0:10
I'm Nick Carabella, your share runner and curator, and I'm
0:13
excited you're joining us today to talk about the people driving
0:16
the change in our industry. You Today I'm joined by Christy
0:19
Hunt. How are you doing?
0:20
Hey, Nick. Hey, everybody.
0:21
Thanks for having me as usual. Glad everybody's out there,
0:25
TuneIn. I'm glad you could make it.
0:27
I know it's allergy sees it kicking your butt right now, but
0:31
you know just a little bit hopefully you could heal you up
0:36
in the audio editing later. That's right.
0:39
But today's episode and it feels like kind of every conversation
0:43
we're having today where demand is outpacing supply and the data
0:48
center market from melting GPU's for AIM is generation to
0:53
increasing demands on power and all the equipment to support it.
0:57
Yeah, exactly. And it's, you know, it's not
0:59
just about scaling it up, it's about whether the supply chain
1:02
can even keep up with it. And that's why today's
1:05
conversation, I think, is going to be really relevant.
1:07
We're pulling back the curtain on how the data center industry
1:10
is evolving and what it takes to build faster, smarter and with
1:13
more transparency. So joining us today, our guest
1:16
is Matt Caldwell, Director of the AI and Hyperscale at Hyper
1:19
Solutions. Matt has spent his career
1:21
building some of the most complex infrastructure in the
1:24
digital world, from hyperscale data centers to cloud campuses.
1:28
Anna Hyper is focused on solving the supply chain bottlenecks
1:31
that are slowing the industry down.
1:33
So, Matt, welcome to the show. Thanks for having me.
1:35
And Nick, it's good to be on your podcast.
1:37
I know we've been talking about this for like a year now.
1:40
So glad we were able to sync our calendars and be able to chat
1:43
for a little bit of time here. I know it's with conference
1:48
season being in full spin, it's kind of nice that we get a
1:50
little bit of a summer break where we can return to the
1:53
office and get caught up. But it's always good to see you
1:56
out on the trail and connect with you at different shows.
1:58
So but glad to have you on the platform.
2:00
Thanks for having me. Yeah, Matt, tell us a little bit
2:03
about yourself. Give us an idea of your
2:06
background and how you found your way into this data center
2:09
and infrastructure space. Yeah, yeah.
2:11
So I've been in the data center space for over 20 years now, and
2:15
I can't say I sought it out. It kind of found me.
2:18
I graduated from undergrad from Uri, didn't really know what I
2:22
wanted to do with my life. So I thought I wanted to be a
2:25
lawyer. So took the Lsats, was applying
2:27
to law school and got a call from APC asking if I wanted to
2:30
sell batteries. So didn't really know a whole
2:33
lot about, you know, battery backups or data centers or
2:37
computers in any way, shape or form.
2:39
But by the time I would have gone to law school, I've been
2:42
promoted a few times and things look pretty good.
2:44
So continue to ride that wave. And APC got acquired by
2:48
Schneider Electric. I spent a total of about 12
2:51
years at APC. Schneider move me out to the
2:54
West Coast to manage all the enterprise government accounts
2:58
out there. Through that, got some exposure
3:00
to the other sides of the Schneider Electric portfolio, so
3:04
the control space, the power space, the energy services space
3:08
and doing some performance contracting for the federal
3:11
government. Got my MBA from UCI, kind of
3:15
took stock of where the industry is going and made the jump from
3:19
Schneider over to Siemens to manage Hewlett Packard globally.
3:23
So did that for four years. Did a grief stin at Albirio
3:28
Energy and then spent four years at train on the mechanical side
3:32
of the data center world. So that's all before about 8
3:36
months ago I got a call from Hyper Hyper Solutions asking me
3:40
if I was interested in joining their sales team.
3:43
So been there for less than a year now.
3:46
But things are great. The industry is exploding and
3:49
Hyper's scaling up with it. So it's an exciting time.
3:53
I had known Dennis Streeter, one of the Co founders of Hyper for
3:57
20 years. He also owns LDP, which is
3:59
Schneider Electric's largest manufacturer's Rep.
4:02
So he's doing a large amount of business, basically anything on
4:05
the West Coast of the US three phase for Schneider's going
4:08
through LDP and he he saw a need in the market a few years ago
4:13
for a manufacturer of quality electrical distribution
4:16
equipment and himself and his Co founder Vlad Gilgaroff started
4:22
Hyper. And really it's a it's we'll get
4:25
into this later in the podcast, I know, but it's, it's a
4:28
different model. It's a different business model
4:30
for for delivering equipment into the data center space.
4:34
And you know, the data center industry as a whole is exploding
4:39
and hyper as a unique business model.
4:41
So we can talk about that over the course of this podcast.
4:43
I think you hit the nail that I wanted to talk to you today, to
4:46
you just even looking eight months back of like, that's kind
4:50
of where the AI boob really started to get its legs
4:52
underneath it. But before we start to even get
4:55
into today's excitement, I'd love to like learn a little bit
4:58
from just your history, the 20 years you've spent in this
5:01
space. How was it to work on data
5:05
centers pre today if you will? And how is things starting to
5:09
kind of reshape towards the the big changes we see with the AI
5:14
Reg? Yeah, so, so when anybody tells
5:16
you they've been in the data center space for 20 years, it it
5:19
doesn't mean they've been doing what they're doing today for 20
5:22
years, because what's happening today did not exist 20 years
5:25
ago. 20 years ago I was selling 100 kva U's into essentially
5:31
computer rooms or wiring closets and that was considered a data
5:34
center sale. My, my first year in the data
5:38
center space, I, I had a territory of Louisiana, southern
5:43
Texas and I, I sold A1 MW data center project into LSU with
5:49
the, the power, the inroad cooling hot out containment DCIM
5:53
complete turnkey solution. And I was getting high fives in
5:57
the hallway. Like that was a career win at
5:59
that point in my career. And, and everybody was super
6:03
excited about it. You know, get got called out on
6:06
the, on the all sales call that, that that year.
6:09
And today one MW is a rounding error.
6:12
And the reason for that is 20 years ago, if you were an
6:17
enterprise organization, a Fortune 500 and you needed a
6:21
data center, there's no option other than building it yourself.
6:24
You had to, you had to construct it, you had to hire IT
6:28
personnel, you had to hire facilities personnel to operate
6:31
it. That was, you're responsible for
6:33
everything. And somewhere along the way,
6:36
Digital Realty Equinax came along and said, hey, you know,
6:39
this isn't your core business. Why don't we set up a, a
6:43
business where we'll take care of all the power of the cooling.
6:46
You can rent out floor space, Rackspace cages within that
6:50
retail collocation model and enterprise organizations are
6:54
shifting their IT load from in house enterprise data centers to
6:58
retail Kolo. The next iteration would be
7:01
Amazon which realized hey, we have spare capacity, why don't
7:05
we just start leasing out our processing power in our existing
7:08
data centers and have an additional revenue stream,
7:11
right. So that has taken off and
7:15
through the cloud migration you see in this wholesale
7:18
collocation model take hold where those collocation
7:22
companies are no longer building retail collo to fill with small
7:26
tenants. It's built to suit large single
7:28
tenant data centers or basically the big collocation providers.
7:33
So if nobody ever said the word AI in their lives, we'd still be
7:39
talking about this boom in the data center market from the
7:42
migration of enterprise and retail Co location to wholesale
7:47
and cloud applications. And on top of that exponential
7:51
growth curve, you know, 2436 months ago, all of a sudden AI
7:56
got legs and layered on top of that exponential growth curve,
8:00
you had another one with AI and basically it blew the top off
8:03
the data center market. What we are seeing now in the
8:07
past 24 to 36 months is unprecedented and not only in
8:12
the data center market, but in human history.
8:15
Like there's never been this amount of dollars being spent in
8:19
the concentration of, of a small number of companies are actually
8:23
building. There's an argument to be made
8:25
that there's, you can count the number of actual end users on
8:28
your hand, right? 85% of the data center market is
8:32
being built by, you know, the big 5 or 6, you know, Amazon,
8:36
Microsoft, Meta, Google and a couple others.
8:39
And then you have maybe 30 to 40 wholesale code companies that
8:44
are building for them. And that's making up 8085% of
8:47
new construction in the data center space.
8:50
And when you look at the dollar figures and gigawatts needed to
8:54
power these data centers, that concentration and such a small
8:58
number of companies and decision makers is, you know, crazy if
9:02
you have to go back to like the Gilded Age to, to find a example
9:06
of a concentration of, of growth from such a small number of
9:10
players. And I think it gave me like a
9:12
little bit of a flashback to I, I was working on.
9:15
I didn't even realize I was have data center experience that goes
9:18
this far back until you just started told the story and
9:21
building out like a data Center for a life sciences client that
9:24
I had back when I still practice in architecture.
9:26
And it was like, we're doing it in the headquarters building,
9:29
raised floor and needed a specialty fire system and all
9:32
this extra special stuff to make the data Center for that
9:36
headquarters work. And you realize now looking at
9:39
how we do this today, it's like, why didn't we want to
9:43
consolidate those resources to these campuses that are
9:46
dedicated to doing such a thing? Because if anybody's ever
9:49
installed a fire system before, you know how many dollars that
9:51
can be attached to it. And so there's a actual
9:54
strategic advantage that makes sense to why we've grown this
9:58
way and to why, you know, data centers really are looked at as
10:01
infrastructure in many cases, rather than perhaps, you know, a
10:05
utility closet that's now, you know, or was being built 20
10:09
years ago at all of these headquarters.
10:11
So it's, it's really cool to see how the industry has shifted and
10:15
how we're even pulling our resources as businesses to
10:18
create a new infrastructure model that didn't exist before.
10:22
You know, we cared about the Internet and cats and laser
10:25
beams and all that other fun stuff.
10:28
I mean that that's really the most important aspect here.
10:31
More videos of cats with laser beams.
10:33
That's that's going to move society forward.
10:35
Can't hurt. Oh man.
10:36
I mean, I don't want to I don't want us to lose track, but I've
10:39
gone down, gone in deep to the Yeti and Bigfoot AI video rabbit
10:44
hole hole, which is well worth losing a couple minutes or hours
10:49
depending on how much you want to scroll if you want to kill.
10:52
So Ty. But anyway, get us back on track
10:55
just a little bit. Was there a moment where you
10:57
started to kind of see the shift from, you know, on site through
11:01
cola? Where did you start to see like
11:03
the new model start to need emerge and need to exist?
11:08
Was this where maybe the Amazon moment occurred and they
11:11
realized they just had surplus or?
11:12
I can give you a yeah, I can actually give you a good story
11:17
around the moment that I realized that Cloud is going to
11:21
going to take hold. So at the time I was with
11:24
Schneider and I was working on the data center consolidation
11:29
initiative with the Navy and it was run through Spay War, which
11:33
is kind of like the R&D arm of the of the Navy.
11:37
I got access to it wasn't a confidential PowerPoint, but it
11:42
was wasn't readily circulated. It was basically for the
11:44
Pentagon through the guy who was running that program.
11:48
He was showing it to me. And the Navy was looking to
11:50
consolidate all of their data centers down to 3 enterprise
11:53
data centers, basically San Diego, Charleston and New
11:57
Orleans than a few other like smaller like R&D stuff.
12:00
And the way that they're looking to do that, they're starting
12:03
with like over 500 data centers and looking to consolidate it
12:05
down to to three basically. And the slide that really stuck
12:10
in my mind is kind of seared in the back of my head is, you
12:15
know, a slide that said we're going to take all of the
12:17
critical equipment, the sensitive stuff that we don't
12:20
want the Chinese and the Iranians seeing and put into
12:23
these enterprise data centers. But all like the payroll, the
12:26
nonsense, the stuff that nobody really cares about, we're going
12:29
to put that in the public cloud. But it didn't say public cloud
12:32
is an arrow and it said AWS. And this would be probably 2010,
12:40
12:00-ish period, which if you think back to AWS in that point
12:45
in time, it was kind of like if you were a startup and couldn't
12:48
afford to build your own data center.
12:50
You went to, you went to Amazon, right?
12:53
Because they eventually you'd grow up and you build, build
12:56
your own data center. And that slide switched into my
13:00
head. I went from public cloud being a
13:04
stepping stone stone to building your own data center to an
13:08
actual application. If the, if the federal
13:10
government is willing to put 70% of their information on the
13:14
public cloud, this is going to take hold.
13:17
And the need for on site on Prem enterprise data center is, is
13:21
going away. That being said, there's still a
13:23
lot of legacy enterprise data centers out there.
13:26
I'm sure there's going to be comments here people saying
13:28
enterprise data isn't dead, you know, whatever.
13:31
Yes, occasionally somebody builds a new one and there's
13:34
existing data centers that are going to continue to operate.
13:37
But if you look at the scale of what's being built now, it is so
13:40
overwhelmingly the big hyper scalers and the wholesale Co
13:46
location market that the enterprise growth is a rounding
13:50
error to the amount of expansion.
13:52
Like there's no enterprise customer that's building a GW
13:55
campus. That idea of a GW campus was
13:58
ludicrous three to five years ago.
14:01
And now and once a week I'm hearing of a new one that's
14:03
that's popping up right somewhere in Texas or North
14:07
Dakota or Canada or wherever they're building another multi
14:10
GW data center. And when you extrapolate out all
14:15
the equipment, all the labor, all the financing, everything
14:18
that goes in a GW campus, it is eye watering the amount of
14:21
demand that's out there. And kind of Fast forward from
14:24
that point, you know, I guess I'd be 10 to 12 years ago now to
14:28
where we're at today. I'm getting forecast for the
14:31
next three to five years. And I I'm looking at the
14:34
equipment that we're providing and like, you know, how many of
14:37
these one MW PD us are you going to need in the next two or three
14:40
years? And the numbers I'm hearing are
14:43
2003 thousand 4000. I'm like these are gigawatts of
14:48
capacity coming down the pipe pipeline.
14:51
And that's where we'll, we'll probably get later in the
14:54
discussion, but that's where, you know, the industry is
14:58
changing at such a rapid pace. Doing things the same way that
15:02
we're doing them 5 or 10 years ago just doesn't make sense.
15:06
You can't scale to the level that you need to build a GW data
15:11
center using the same business model that you used when A10 GW
15:16
data center was considered huge, right?
15:19
Things, things are changing. The, the landscape around the
15:23
data center is changing and we've reached a point where
15:27
we've tapped out the utility, right?
15:29
Like 5-10 years ago was just a given.
15:33
There was a time period like, oh, maybe you have to install a
15:35
substation or whatever to get the power off the minion or
15:38
whatever. Now it's like if you want a
15:40
significant amount of power off the grid, good luck.
15:42
You know, everybody's pretty much resigned themselves to the
15:45
fact that any of these large scale campuses, you got to look
15:48
at onslie generation as a bridge to probably nuclear 5710 years
15:54
down the road. But that's that's the end goal.
15:57
Everybody I think agrees. And nuclear at some point is
15:59
going to be powering most of the data centers in North America,
16:03
if not the world. We just got to kind of get to
16:05
that point and there's regulatory issues and there's
16:09
adoption. Not in my backyard type issues,
16:11
but eventually we're going to get there.
16:13
It's just a matter of time. There's no way we're powering AI
16:18
and cloud computing for any significant amount of time off
16:21
of fossil fuels. That's, or even renewables for
16:24
that matter, Other than other than nuclear.
16:26
Yeah, there's, there's just so many challenges there.
16:29
You know, when you talk about the, the challenges they're
16:31
facing in this growth and, and then changing the way of
16:35
thinking and, and an approach you're, you're really passionate
16:38
about fixing this, this problem and, and fixing supply chain
16:43
transparency. Why?
16:45
Why are you so passionate? I I've got sucked in.
16:49
I've drank the Kool-aid on the, on the data center space.
16:51
Like I've, I, I accidentally landed here, but I've been doing
16:55
it for 20 years and I've basically been eating, drinking,
16:59
sleeping, dreaming data Centers for, for 20 years now.
17:03
So where we're at now is unprecedented in our lifetime,
17:08
right? Maybe like the Internet boom of
17:12
the 80s, early 90s, but that missed me as far as my age,
17:17
right? Like I wasn't, I wasn't old
17:19
enough to really participate in that.
17:21
So this is my kind of career defining, OK, You know, I'm 20
17:26
years into my career. I know enough to be dangerous.
17:28
I've got a pretty decent personal network in this space.
17:32
And this happens to be the time that the data center industry is
17:36
taking this hockey stick curve up on the growth cycle.
17:41
I don't know if I'm answering your question here or not, but
17:43
it's funny because three years ago I'd be at a family holiday
17:48
party and people would ask me for the 10th time that what is
17:52
it you do again? And I'm like, I'd do data
17:54
centers. And they're like, what is the
17:55
data center again? I'm like, you know, turn off
17:58
your phone, whatever it's not doing.
18:00
That's what's going on in the data center, right?
18:02
And like, I get it. And then six months later,
18:04
Easter, they're like, what is the data center again?
18:07
And now it's flipped. Now it's like, oh, you know, I
18:10
saw Kevin O'Leary or saw Jim Cramer talking about data
18:13
centers. That must be so exciting that
18:15
you're you're a part of this now.
18:17
So I like, why am I excited about it?
18:19
Like I've always been excited about it.
18:21
But now, like the world is kind of catching up to where I've
18:25
been all along and the world is starting to get excited about
18:27
data centers. And I just have, I was here the
18:29
whole time. I'm like, hey, hey guys, I was
18:31
here like glad, glad that you have visibility into what I've
18:35
been doing. And you're welcome for all of
18:37
your cat videos with laser beams and Facebook and Instagram and
18:40
all that stuff that you didn't realize I was working on the
18:42
whole time. Well, I think that's like the
18:44
funny part of this space is, you know, as As for end users, which
18:49
is basically all of us of the digital infrastructure.
18:52
And we take for granted the amount of effort of information
18:56
transformation, the work that these buildings do and the work
18:59
that we have to do in order to power everything that's going.
19:03
Because at the end of the day, everybody, it kind of summarizes
19:06
down to the text message that you get or the photo that you
19:09
get from a loved one. And like that's kind of the the
19:11
tangible touch that these that most people have with digital
19:16
infrastructure. But there's so much that goes on
19:19
behind it and there's so many dollars that go on behind making
19:22
any of this possible. It's hard to think that the
19:26
capacity of being able to send one of those photos from me
19:30
across the country to California or wherever requires quite a bit
19:35
of infrastructure and complexity because it's so easy to us as
19:39
lay people to do those activities.
19:42
You know, I think what's really exciting now just outside
19:44
looking in and kind of riffing off of your excitement is just,
19:48
you know, the speed at which like, and you nailed it on the
19:52
head of like it used to be 1 MW. Now if we're talking about a MW,
19:56
that's almost it's, it's practice.
19:58
It's silly with everything being notification.
20:01
When you're talking to my MW now, we're talking about a rack,
20:04
we're not talking about a data center.
20:06
Exactly. And then you're like, I close my
20:08
eyes and I look back at like learning like my computer course
20:11
in high school. They're like, you know, it used
20:13
to be computers for the entire size of a room and how like how
20:18
much that's changed. And now that we have
20:19
supercomputers that fit in our pockets and data centers that
20:23
are huge again, it's really fascinating to just think about,
20:28
you know, all of the infrastructure and how it works
20:30
together and the challenges that they have to kind of solve for
20:35
with meeting that demand. And so now you've painted this
20:38
picture, right? You have the the Fang players,
20:40
the big 5, the mag said we whatever you want to call it all
20:43
looking for tower and land. And I think, you know, we hear
20:48
this a lot too equipment and you know, why does the way that used
20:54
to work no longer work for today and how what what is the vision
20:59
for scaling into? Yeah.
21:01
So I mean, take a look at the landscape of the equipment being
21:06
sold into the data center space. And I, I have a unique
21:08
perspective because I've worked for the large OEMs on the
21:11
electrical side, I've worked for large OEMs on the mechanical
21:15
side. I've been in the control space.
21:16
I've you know, I've, I've got a pretty good cross section of the
21:20
industry to kind of generalize a little bit.
21:22
So don't read too much into anything I say.
21:24
It's just kind of more of a generalization.
21:27
So you have these big manufacturers that are
21:30
delivering UPS's electrical distribution gear chillers, all
21:35
the equipment that makes up a data center and they're large
21:39
and their business model requires them to be 90% plus
21:45
utilized. So when you and also I can't
21:49
leave out that they, they have large capacity agreements now
21:53
with these large hyper scalers, right.
21:55
So where they used to have a production line and they're
21:58
looking for smaller players that they can kind of push, spread
22:03
out that operational load and account for, you know, little
22:07
bottlenecks or, or lumps in there and their supply and, and
22:10
smooth that out. Now they've got these large
22:13
capacity agreements where a Google and Microsoft and Amazon
22:17
will go into any large OEM and say, basically, I want to buy
22:20
your capacity for the next three to five years.
22:22
And some companies buy into that.
22:24
Some companies are like, no, you know, we, we can only take a
22:27
certain amount. We need to diversify our risk
22:30
and have have at least several companies, but for the most
22:33
part, those small chunks that used to make up that large
22:36
manufacturing line of demand have turned into these very
22:39
large chunks. So when you go to any of the
22:43
large OEMs and you say what's your lead time, what's your
22:46
capacity? Basically what you're asking is
22:48
I need to cherry pick build slots that are available in your
22:52
existing production line. And to their credit, most of
22:56
these large OEMs see the the increased demand for data
23:00
centers and they're building additional capacity to support
23:03
it. But if you listen to any
23:05
interview with Andy Jassi or or the CEO of Microsoft or whatever
23:09
like their, their demand outstrip supply 12:50 like that,
23:13
the the wave of the demand is overstripping supply.
23:17
And if you're, if you're a large company and you're looking to
23:20
increase your capacity, that's a large capital investment that's
23:24
going to take time and resources to, to grow into right?
23:28
That's not available tomorrow. O that's opened up a gap in the
23:33
industry where you have people from outside the data center
23:35
industry being like, hey, you know, chillers maybe are a
23:38
little too complex, but a panel board's fairly simple.
23:41
OK, you know, I've got 40,000 square feet of production
23:44
capacity. I can, I can deliver X amount
23:47
of, you know, switchboards or PD us or whatever.
23:52
I, you know, I have the capability of doing that.
23:54
The problem there is the scale of the the projects that we're
23:59
talking about. They may have capacity today,
24:01
but they're one PO away from being tapped out, right?
24:03
They get one big order, they're done.
24:06
And yes, they can build another factory.
24:09
And a lot of them are like, yeah, you know, if we're going
24:11
to double our capacity, we're going to do this.
24:12
Great. But that's a capital investment.
24:14
You got to put a shovel in the ground.
24:15
You got to come up with investment dollars to to justify
24:19
that and, and not all of the, the people on the finance side
24:22
believe that this is in the bubble and it's not going to
24:24
pop. So you got to make a compelling
24:26
case on why this big capital investment that you know is a 15
24:30
year, 20 year, 30 year capital expense is going to just be
24:36
there. You know that demand is going to
24:38
be there when you build it. So that is where, and I'm kind
24:42
of dovetailing now into to what I'm doing at hyper.
24:45
That's where what Dennis and Vlad put together kind of solves
24:50
that middle ground between the two and think like a network
24:54
effect similar to an Airbnb or an Uber applied to the
25:00
manufacturing space for electrical distribution
25:02
equipment in the data center space.
25:04
And the way that they accomplished that is they built
25:08
a digital platform called hyperspace.
25:11
And basically what Hyperspace is, is it is a single source of
25:15
truth where all of the engineering takes place and all
25:18
of the processes take place. And that has to be buttoned
25:21
down. That has to be solid as a rock
25:24
for this to work. Because what we do is we go to
25:29
third parties, we go to call them satellites, right?
25:32
So we look at on Jason Industries that are doing high
25:35
tech manufacturing like aerospace, automotive, robotics
25:40
that have stranded capacity in North America.
25:43
So think of a factory that's 30 years old.
25:47
They have a lot of qualified technicians working there, but
25:50
they don't have visibility into the data center space.
25:53
And Hyper will go in there and lock in long term capacity
25:57
agreements, almost like a franchise agreement with those
26:00
manufacturing facilities and lock in 20, thirty, 35% of their
26:07
production capacity, train their employees on the process that we
26:11
use. And then Hyper owns the
26:14
engineering, it's all our own UL listings.
26:17
Hyper owns the process which every step along the way is
26:21
logged and Hyper owns the quality control.
26:24
So there's a Hyper employee that goes through and make sure that
26:28
the equipment was built to hyper standards and then test the
26:32
equipment and then puts it on truck and delivers it.
26:34
We're going to we're going to touch on transparency.
26:37
So in order for that to work, that whole process needs to be
26:41
transparent to hyper, right? We need visibility to every step
26:45
along the way. We need to be able to see we're
26:47
all so backing up a second. Hyper owns the supply chain for
26:52
all of the critical components. So think Breakers, Transformers,
26:55
control panels, all those pieces of equipment that are in high
27:00
demand. We have capacity agreements with
27:02
all the major manufacturers of those, those pieces of
27:05
equipment. So Seaman Schneider, ABB, Eden,
27:08
we have direct relationships there.
27:10
So when an order gets placed, Hyer will order that we have
27:14
visibility to when the critical components arrive on site, we
27:17
have visibility to when the build rocess starts.
27:19
We see every step along the way, how long that step took, who
27:24
signed off on it, the every step in the process.
27:27
We can see when the build is complete, we can see when the
27:30
testing was done, all of that, and then it goes off, you know,
27:33
for delivery, all that is done through AQR code, right?
27:37
So when that piece of equipment is born, it gets AQR code that
27:40
follows it throughout that cycle.
27:42
And then even in operation, after you know, it's delivered
27:45
and it's operational on site, you can take a, a snapshot of
27:48
that QR code and see all the serial numbers of the components
27:52
within that, all the history of that piece of equipment.
27:55
Now, what's unique to Hyper is we have visibility to all of
27:59
that and we share that visibility with our clients.
28:02
So think of a Domino's Pizza box tracker, right?
28:07
You have an app, you place an order for pepperoni pizza with
28:11
olives and anchovies. I don't know what you're into,
28:13
Nick, that sounds like something you'd like.
28:16
You place your order, you can see when it, when it gets into
28:19
the oven, you can see when it's out of the oven.
28:21
You can see when it's when it's out for delivery, Amazon, right?
28:25
You order a $3 tube of toothpaste.
28:29
You can see exactly where it's at in the process when it's
28:32
getting your door, a picture of it when it gets to your door.
28:35
It is ridiculous that for a $3 tube of toothpaste you got that
28:40
level of visibility. But when you order $100 million
28:43
worth of PDU equipment and you say, is this going to arrive on
28:46
site? You make a phone call and the
28:49
answer is I'm going to have to check on that for you.
28:53
And and that was actually that it's kind of funny story around
28:56
that I'm working at train and I'm pretty happy with what I'm
28:59
doing there. And Dennis Struder calls me and
29:02
he goes, Matt, when, when a customer calls you and asks if a
29:07
piece of equipment is going to arrive on site, what is the
29:09
process? And this is not unique to train
29:11
this, this could be any large OEM.
29:13
And I go, but I say, I'm going to look into that and make some
29:16
phone calls and we have some internal meetings and I get back
29:18
to the customer and you know, I tell them like, Yep, it's
29:23
probably going to arrive on this this date, right?
29:25
And, and he goes, what if instead of them calling you,
29:30
they had an app on their phone that they could pull up that
29:34
would tell them exactly where it's at in the production cycle.
29:36
And it's say, yellow, green, red.
29:39
If, if it's delayed, they'll see it in real time.
29:42
If it's not going to arrive on site at the date that we was
29:45
promised, they'll be able to see that.
29:47
And that's a conversation. But for the most part, it's
29:49
going to be green. And it gives them a, a certain
29:51
level of comfort that, OK, not only am I being told that it's
29:55
going to arrive on time, but I can see exactly where it's at in
29:57
the production process and I can see exactly what's going on.
30:00
We're still kind of refining exactly how much information is
30:03
the right amount of information. It's not that we're trying to
30:05
hide anything. It's more that information
30:08
overload, right? Like we have all the information
30:11
coming in, sharing that with the client isn't necessarily
30:13
productive because that's just too much then too much
30:17
information. But just a basic, OK, you know,
30:20
here's a step that it's in, here's here's what's going on,
30:23
you know, thumbs up or hey, we ran into a problem.
30:25
Here's our our process to rectify that.
30:28
Is it going to impact the delivery date or not?
30:30
That's really the information that as a client would benefit
30:34
you and your construction meetings and, you know, making
30:38
sure that everything's going to arrive on site because COVID
30:41
taught me, you know, during that process, customers, and this
30:44
isn't really the case anymore. But during that really rough
30:47
time where shipping dates were slipping across the board,
30:51
customers are pretty transparent.
30:52
They're like, look, if it's going to be delayed, that's OK.
30:55
I just need to know two months ahead of time, not two days.
30:58
Because once I've got the cranes in place, I've got the
31:02
contractors lined up to to be there, there's no turning back
31:05
at that point. If the equipment doesn't show up
31:07
at that point, that's a problem. I think we're as an industry, I
31:10
think we're, we're pretty much across the board.
31:13
Beyond that, there's still issues with shipping here and
31:15
there, but it's not not the same bottleneck that we saw in 2022
31:20
ish. I think you touched on a really
31:22
interesting point of the a couple of a couple of points
31:26
there. One is the expectations that
31:29
come in from life as a consumer, right?
31:32
Your my consumer use of an app to see where my tube of
31:36
toothpaste is, let alone order a very strange pizza.
31:40
We can talk about pineapple on pizza separately.
31:42
That could be a whole other thing we get into you.
31:44
But but there are consumer their expectations that you get and
31:47
that you have as a consumer to who uses a phone, who uses apps
31:52
to get that kind of real time update.
31:54
And I know many of us come in to work and think, gosh, why can't
31:58
I just see and have that same impact, the same level of
32:02
information. But then the other part of it
32:05
that you touched on is that what we love to call the three legged
32:08
stool of people, process and technology.
32:11
You've got all of those different pieces.
32:13
And as you build out a team to support the mission that that
32:19
you're trying to achieve, what have you learned and what what
32:22
surprised you as you kind of put those three legs in the stool
32:25
together of people processing technology and managing
32:29
expectations? Not even and beyond just a post
32:32
COVID world, but just managing the expectations of we've got to
32:35
have some better visibility. We've got to have some better
32:37
process. Surely there's a tool.
32:40
What's been surprising to you on that journey?
32:41
Yeah. So interesting on that.
32:44
So again, we have multiple facilities that were that were
32:48
manufacturing this equipment and we have the same process at
32:51
every single one of them, right. So we have visibility when
32:54
you're making a one MW PDU, Step 3 in that process should take 33
33:00
minutes, let's say, right? On average it's 33 minutes with
33:03
a variance of say two or three minutes on either side, right.
33:06
So we have the ability to see, OK, this technician at This site
33:12
is consistently doing it in 50 minutes.
33:16
Does that person need additional training to get back down to?
33:20
Why is it taking them longer than everybody else?
33:22
It should be taking 33 minutes, right?
33:24
But on the other side, let's say it's taking one person 25
33:28
minutes, but the quality is still there.
33:32
OK, let's dig into that. What is that person doing that's
33:34
different that if we can learn from that individual now we can
33:38
extrapolate across everybody, we can change our process to
33:42
improve the speed and quality that we're delivering.
33:45
A great example of that is one of our manufacturing facilities
33:48
of the, the RPPS figured out that if they built this tilt, I
33:55
don't know it's exactly called, but it tilts the the RPP on its
33:59
side, right tilt table, right. So you put the RPP in it, you
34:02
turn it 90°. It allows them to work on it
34:05
easier. They don't have to reach up or
34:07
reach down. It's all that waste level,
34:09
right? So they do all the work that
34:11
they need to tilt it back up. It goes to the next process.
34:13
So we've taken that and we've incorporated that into the
34:17
process for other manufacturing facilities that that were
34:20
operating right. So there, there's a lot of lot
34:24
to be learned when you have a standardized process that's
34:27
automated and digital across a networked manufacturing platform
34:34
that you're not going to get necessarily when you have a
34:37
large OEM and you've got a manufacturing facility in
34:41
Colorado and you got a manufacturing facility in
34:44
Wisconsin. There's a lot of tribal
34:47
knowledge of, hey, this group of people has been doing it for 20
34:50
years their way. And even if you try to go in
34:53
there and say, hey, there's a better way of doing it, you're
34:56
going to meet resistance from, well, this is the way we've
34:58
always done it. It's been working, it's good.
35:00
We're going to keep doing it that way.
35:02
And I'm not saying they do a bad job, like you're going to get
35:05
quality, but it's good. There's going to be variations
35:08
between the different sites, right?
35:09
And a great story is a personal friend of mine.
35:14
She works for a company, she leads procurement, and her
35:16
engineering staff was a little sceptical of this whole model.
35:19
So we challenged her and her engineers to go to multiple
35:24
sites and find a difference between the PDU that was being
35:28
built in Wichita, KS and the PDU that was being built in Grand
35:33
Rapids, MI. And they went through with a
35:35
fine tooth comb down to the zip ties where the zip ties were
35:40
located, how the zip ties were cut.
35:44
They could not find a single difference between those two
35:47
pieces of equipment that were built in two separate facilities
35:50
by two different teams. And the reason for that is
35:54
because they have basically a Lego instruction guide.
35:58
It's you do this step, this step, this step, it's all on an
36:00
iPad. They do the step, it logs who
36:04
did it, how long it took, and it's just a very, very
36:07
standardized process. This is incredible and I and
36:10
we're like fighting like the Dickens.
36:11
Not to get into the details here, which I'm glad.
36:14
We're kind of could put the bow on the first part of this
36:16
episode and really capture kind of the passion and opportunity
36:20
that's in this space. And so before, before we close
36:24
and get excited for the second part of our conversation where
36:26
we're really going to start to dig into the values of
36:28
transparency, how to build scalable manufacturing teams,
36:33
how to navigate what decentralized looks like.
36:36
I want to, you know, tease out with our last question of every
36:39
getting to know you episode of, you know, what makes you wired
36:43
to build? And is there a story that
36:45
perhaps embodies this mindset from for you?
36:49
I'm sure like a lot of the Hypermission kind of alliance
36:52
with wiring, you know how we secure things, but you know
36:56
where where are you at in your wired to build?
36:58
Your yeah, I mean, I think that goes back to my first days in
37:01
the industry. I, I kind of landed at APC at
37:05
the perfect time where Neil Rasmussen was rolling out
37:10
infrastructure, which at the time was modular, scalable, UPS
37:14
in row cooling, hot aisle containment, DCM before anybody
37:17
knew what that was. And it was revolutionary.
37:21
Nobody was doing that. Nobody was talking to the IT
37:23
staff about what they wanted for physical infrastructure.
37:28
It was all running the Gray space, a big Gray box in the
37:31
mechanical room, you know, crack units at the perimeter, the data
37:36
center and the IT staff didn't have any control over that.
37:39
So by going in and putting a UPS that was lined up in match with
37:44
all the black racks for an infrastructure solution, it
37:48
looks sexy. It's, it's hard to say you like,
37:51
Oh my God, that data center looks sexy, but it looks sexy.
37:54
It was, you know, everything fit together the way that it was
37:57
meant to. You had, you know, 2 rows
37:59
back-to-back with a common hot aisle plexiglass insert for the,
38:03
the hot aisle and you could do super high density.
38:06
I mean, talking 1215 kW per rack.
38:09
I mean, back in 2003, that was, that was pushing the limits.
38:13
And I remember hearing like 20-30 kW per rack and thinking
38:17
that was, that was, that was crazy.
38:20
And obviously with the rack densities that we're we're
38:22
talking about now, that seems naive of us to think, but you
38:27
know that that that really sparked the interest and you
38:29
know, ABC got acquired by Schneider.
38:31
I went to Siemens, I work for train and they're all great
38:34
companies. They built the high quality
38:35
products, but I've been waiting for the opportunity to get back
38:39
to working at a small company where I smaller company where I
38:44
have influence over the direction of the company and
38:47
we're doing really cool stuff that's to benefit our customers
38:51
in the industry as a whole. Like I really truly believe that
38:54
what I'm doing now at Hyper is changing the way that data
38:59
centers are designed, built and operated.
39:01
That's what we did at ABC 20 years ago.
39:04
It kind of leveled off for a while.
39:06
But you know, like I said right now, last 2436 months we've seen
39:11
incremental change and hyper is really jumping in to meet that
39:16
meet meet that demand and change the way that data center
39:21
equipment is built and deployed. And I I think that what we're
39:25
doing today, it has to be adopted by the industry at whole
39:29
in order for us to meet the the growth expectations that we're
39:32
going to see in the next three to five years.
39:34
And just I think that's what I, at least for me and for so many
39:38
folks I talked to was so exciting about this space is
39:40
it's the the Iterate innovate loop is so fast on data centers
39:45
from going from the the one kilowatts to the one megawatts
39:48
to the one gigawatts where the entire team, whether you're an
39:52
owner, an end user, a constructor, a creator of parts,
39:57
all of the different pieces that fit into this.
40:01
Becomes so important to not only do to meet the need today but
40:05
figure out how to be innovating for tomorrow and so a lot of
40:09
food to think on between now and two weeks from then.
40:12
The next episode is going to be coming out.
40:14
But really you want to thank you for joining us for this
40:15
conversation today and excited to pick up on Part 2 and and a
40:20
little bit here. See you in two weeks.
40:22
Thank you.