1
00:00:01,294 --> 00:00:03,434
Thanks for tuning in to Let's Get Digital.

2
00:00:03,434 --> 00:00:08,054
I'm Carrie Charles, your host, and happy to have you here with me today.

3
00:00:08,054 --> 00:00:10,754
We've got another great episode for you.

4
00:00:10,754 --> 00:00:13,334
I have with me Ron Mann.

5
00:00:13,334 --> 00:00:17,794
He is the Vice President of Compute Dynamics Modular.

6
00:00:18,414 --> 00:00:20,654
Great to see you, Ron, again.

7
00:00:20,654 --> 00:00:23,994
What, I saw you at Data Cloud a few weeks ago?

8
00:00:24,581 --> 00:00:26,266
Good to see you too, Gary.

9
00:00:26,269 --> 00:00:27,974
Thank you for the opportunity here.

10
00:00:28,108 --> 00:00:29,559
Yes, absolutely.

11
00:00:29,559 --> 00:00:32,442
We sat on a panel together and I was just so impressed.

12
00:00:32,442 --> 00:00:37,386
thought not only did you do a great job, but you just had so much information.

13
00:00:37,386 --> 00:00:41,308
I learned a lot from you and I thought let's do this on the podcast.

14
00:00:41,629 --> 00:00:45,332
But first, I want to know a little bit about you.

15
00:00:45,618 --> 00:00:48,096
How did you get to where you are?

16
00:00:48,096 --> 00:00:52,237
Tell me about your path to success, if you will.

17
00:00:52,950 --> 00:00:56,893
Well, it was a long, arduous journey, uh but it was been a lot of fun.

18
00:00:56,893 --> 00:01:01,556
So yeah, so I kind of started out in IT space at actually Compaq Computers.

19
00:01:01,556 --> 00:01:11,515
uh Actually, before that even, was doing some, I was at a subcontractor in all places,
Huntsville, Alabama, that was building computer circuit boards for Compaq.

20
00:01:11,515 --> 00:01:16,769
And that's how I got the role at Compaq back in the, let's just say early 90s.

21
00:01:16,769 --> 00:01:22,149
And uh back then when I hired on, I hired on is what they called a new

22
00:01:22,149 --> 00:01:24,080
uh products manufacturing engineer.

23
00:01:24,080 --> 00:01:29,945
So my job was to take the designs from product development and make sure they could be
manufactured in a way.

24
00:01:29,945 --> 00:01:33,577
And back then was PCAs, circuit assemblies, things like that.

25
00:01:33,757 --> 00:01:36,779
And back then we started out with what's called the system pro.

26
00:01:36,779 --> 00:01:42,373
So that was kind of the first server that was introduced and it actually derivative of the
desktop.

27
00:01:42,373 --> 00:01:45,586
So I was there when this whole server thing started, right?

28
00:01:45,586 --> 00:01:50,669
And Compaq wanted to kind of get into a new business and put a higher end.

29
00:01:50,757 --> 00:01:52,697
kind of desktop out there.

30
00:01:52,697 --> 00:02:01,437
So I came up with these servers and then that led to an infrastructure, what we call the
data center infrastructure group.

31
00:02:01,437 --> 00:02:03,877
Because back then we were selling a lot of these things.

32
00:02:03,877 --> 00:02:05,117
These are all towers back then.

33
00:02:05,117 --> 00:02:07,297
They weren't rack mount servers as you see them today.

34
00:02:07,297 --> 00:02:12,017
But we had a company that we were dealing with, a little small company at the time called
Microsoft.

35
00:02:12,057 --> 00:02:17,678
And they said, hey, we're using these servers, but you know, they're kind of bulky to put
them in bread racks and

36
00:02:17,678 --> 00:02:24,653
these tower things, could you like turn it on its side, maybe put it in some kind of a
rack thing that we could just do a little bit more consolidation.

37
00:02:24,653 --> 00:02:26,694
And so that's how the rack business got started.

38
00:02:26,694 --> 00:02:33,448
We actually looked at the Grainger catalog, found a company that could do racking uh and
there was an industrial rack.

39
00:02:33,448 --> 00:02:35,880
It really wasn't an IT rack at the time.

40
00:02:35,880 --> 00:02:39,422
And we made a bezel that was horizontal in that same server chassis.

41
00:02:39,422 --> 00:02:42,544
We started doing that and that became the rack business.

42
00:02:42,732 --> 00:02:44,743
And then that led to, what about backup?

43
00:02:44,743 --> 00:02:45,744
What about UPSs?

44
00:02:45,744 --> 00:02:48,265
What about oh power distribution, PDUs?

45
00:02:48,265 --> 00:02:49,035
Those didn't exist.

46
00:02:49,035 --> 00:02:50,264
No, those didn't exist back then.

47
00:02:50,264 --> 00:02:53,098
was all plug strips and whatever you think of, right?

48
00:02:53,098 --> 00:03:00,221
And so it kind of evolved into an entire uh kind of ecosystem of stuff that we were
supporting.

49
00:03:00,221 --> 00:03:01,472
And so it just wasn't out there.

50
00:03:01,472 --> 00:03:08,196
So we got into the rack business, got into the UPS business, the PDU business, just all
sorts of stuff.

51
00:03:08,196 --> 00:03:12,676
And that led finally into, which we'll talk about in length here, the modular business.

52
00:03:12,676 --> 00:03:14,896
And so it is kind of a step function, right?

53
00:03:14,896 --> 00:03:25,176
And it was a pretty successful business for Compaq at the time and eventually HP, but it
really all started because we were trying to support selling servers.

54
00:03:25,396 --> 00:03:27,796
What's the most economical way to sell a server?

55
00:03:27,796 --> 00:03:29,836
And that's how it kind of all started.

56
00:03:30,134 --> 00:03:35,656
So tell me about CDM and now that's the, how we refer to the company, right?

57
00:03:35,656 --> 00:03:38,050
Compute Dynamics Modular.

58
00:03:38,789 --> 00:03:47,109
copy dynamics is basically a group that or a company that we started to basically focus on
IT modularization.

59
00:03:47,209 --> 00:03:59,549
So there's a lot of people in the data center space that are talking about, you know, how
to do power modules, cooling modules, that type of stuff that all support a traditional

60
00:03:59,549 --> 00:04:00,529
stick built data center.

61
00:04:00,529 --> 00:04:05,109
But what we want to do here is support the actual IT elements.

62
00:04:05,109 --> 00:04:07,172
So we want to put IT in modules.

63
00:04:07,172 --> 00:04:14,452
and ship them as configured systems that you can basically put anywhere, especially for
the edge, as you hear a lot about things like that.

64
00:04:14,452 --> 00:04:20,692
And that actually started back in the compact days and HP days, so we were starting to do
that.

65
00:04:20,772 --> 00:04:24,292
But it's kind of evolved now to become more of a thing.

66
00:04:24,292 --> 00:04:35,052
so, yeah, we'll do a power module and things like that to support the IT modules, but
we're really here to support IT modularization, both lower, mid-tier power ranges, all the

67
00:04:35,052 --> 00:04:36,072
way up to AI.

68
00:04:36,982 --> 00:04:45,540
Okay, then so tell me a little bit more about like who you serve and just get into the
company a little bit deeper.

69
00:04:46,628 --> 00:04:49,289
So we have a lot of interesting uh customers.

70
00:04:49,289 --> 00:04:56,253
uh We deal a lot with the, what I call the IT, uh original equipment manufacturers, OEMs.

71
00:04:56,253 --> 00:05:01,376
They're interested in selling their IT and they want to be able to put what I call the
wrapper around that.

72
00:05:01,376 --> 00:05:04,038
And modularization is a good way of doing that.

73
00:05:04,038 --> 00:05:09,280
And they don't want to focus on the modules because they want to focus on the storage and
the servers and the AI and all that.

74
00:05:09,280 --> 00:05:16,130
So looking for trusted partners to help put together that final solution who do it for a
living, right?

75
00:05:16,130 --> 00:05:21,184
And that's interesting because when I was doing this at Compaq, that's the way I did the
business.

76
00:05:21,184 --> 00:05:26,107
We didn't want to develop UPSs and create a whole team to do UPSs.

77
00:05:26,107 --> 00:05:28,559
We would go hire a company that was expert in that.

78
00:05:28,559 --> 00:05:30,770
We'd say, here's our spec or here's our requirements.

79
00:05:30,770 --> 00:05:32,712
Can you build to this type of thing?

80
00:05:32,712 --> 00:05:41,758
And it was a great partnership because you think about it, know, at the time, HPE Compaq
had a huge channel, but they needed certain support elements that just weren't out there

81
00:05:41,758 --> 00:05:44,557
yet, you know, because we were developing servers that were...

82
00:05:44,557 --> 00:05:46,458
six, nine, 18 months out.

83
00:05:46,458 --> 00:05:50,363
We knew what the requirements were going to be and there wasn't the right support
structure.

84
00:05:50,363 --> 00:05:53,296
So I take that same philosophy now and say, okay, Mr.

85
00:05:53,296 --> 00:05:54,898
ITOEM, what are you coming up with?

86
00:05:54,898 --> 00:05:56,129
What are you doing?

87
00:05:56,129 --> 00:06:00,043
And how can we make sure that our infrastructure supports that when it

88
00:06:01,204 --> 00:06:04,631
So when you say modular, what exactly do you mean?

89
00:06:05,781 --> 00:06:15,035
In my terms, modular is basically anything that's done in a factory that can be shipped to
the site and commissioned or installed at the site.

90
00:06:15,035 --> 00:06:21,508
So you'd want to do as much as you can in the factory environment versus the field because
it saves you time and money.

91
00:06:21,548 --> 00:06:30,788
On average, uh you can save anywhere from 30 to 50 % doing modular in both time and cost
if you plan it correctly.

92
00:06:30,788 --> 00:06:31,288
Right.

93
00:06:31,288 --> 00:06:33,888
And so, and so think about the advantages of manufacturing.

94
00:06:33,888 --> 00:06:37,008
You have standard processes, standard quality controls.

95
00:06:37,008 --> 00:06:42,508
There's a lot of standardization you can put into these systems and solutions that when
they get to the site, then you're just hooking it up.

96
00:06:42,508 --> 00:06:44,328
Think about how much time you spend.

97
00:06:44,468 --> 00:06:45,288
I'll give an example.

98
00:06:45,288 --> 00:06:48,008
Let's say you're doing something simple like a pump skid.

99
00:06:48,008 --> 00:06:48,248
All right.

100
00:06:48,248 --> 00:06:51,488
Say there's five pumps on this pump skid.

101
00:06:51,488 --> 00:06:53,828
Traditionally, you'd have those pumps delivered to the site.

102
00:06:53,828 --> 00:06:57,588
You'd have a electrician and a plumber come out and they'd hook up each individual one.

103
00:06:57,588 --> 00:06:59,556
If I can put five of those on a skid.

104
00:06:59,556 --> 00:07:07,036
and it's all hooked up and plumbed up and ready to go, then all I'm doing is connecting
that skid when it gets to the site, all five pumps, well then I've saved all sorts of time

105
00:07:07,036 --> 00:07:08,516
and effort for everybody.

106
00:07:08,696 --> 00:07:17,416
And that's really the essence of what modular is, is how to make it more cost effective
for both the field installation and what you can do in the factory.

107
00:07:18,344 --> 00:07:22,801
So what would you say that really changed the game in Modular?

108
00:07:23,300 --> 00:07:26,580
Oh, AI, no doubt about it.

109
00:07:26,600 --> 00:07:36,480
One of the things, and I've been doing modules a long time, and when we were first doing
this back in the day, and the reason why we started doing it is it was kind of an

110
00:07:36,480 --> 00:07:36,920
evolution.

111
00:07:36,920 --> 00:07:40,240
Because you think about it, when you first started doing racks, what would happen?

112
00:07:40,240 --> 00:07:48,260
You'd go to the data center site, the rack would be delivered empty, all these server
boxes would come, and storage boxes and whatever else, and you'd integrate all that on

113
00:07:48,260 --> 00:07:49,157
site, right?

114
00:07:49,157 --> 00:07:55,457
So we got the interesting idea at Compact, say, okay, what if we integrate that in the
rack and shipped it all integrated and fully connected?

115
00:07:55,457 --> 00:07:58,457
Now you have a rack full of stuff all ready to go.

116
00:07:58,457 --> 00:07:59,897
And so we made that a business.

117
00:07:59,897 --> 00:08:02,297
We called it Compact Factory Express and it was very successful.

118
00:08:02,297 --> 00:08:07,097
We had developed a rack that you can move around, you know, like on small little wheels.

119
00:08:07,097 --> 00:08:09,557
And there was a lot of design and development going into that.

120
00:08:09,557 --> 00:08:12,357
And then we got into, well, gee whiz, we can do it for a rack.

121
00:08:12,357 --> 00:08:13,777
What about a row of them?

122
00:08:13,777 --> 00:08:15,497
That's where the modularization came in.

123
00:08:15,497 --> 00:08:17,657
We put them in an IT module, right?

124
00:08:17,886 --> 00:08:19,767
And it was a fairly successful.

125
00:08:19,767 --> 00:08:25,798
sold a lot of that stuff to a lot of the big hyperscale guys at the time who were really
expanding quickly and they needed a quick way to get to market.

126
00:08:25,918 --> 00:08:34,281
But what uh made it interesting for us is the fact that we could do all that again in a
factory environment and get it out the door that way.

127
00:08:34,281 --> 00:08:42,903
So that kind of led to what we're doing now and why we like to integrate this stuff kind
of uh in that format.

128
00:08:42,903 --> 00:08:45,884
But that kind of started the whole thing with uh

129
00:08:45,956 --> 00:08:47,276
of how far you can take it.

130
00:08:47,276 --> 00:08:57,236
Now, what happened on the power side was that what we realized was when you start most
data centers today, I'm sorry, let me go back.

131
00:08:57,236 --> 00:09:07,676
Most data centers say three or four or five years ago before AI, pre-AI, where typically a
rack of IT was 10 kilowatts, I like to talk about kilowatts or watts per rack.

132
00:09:07,676 --> 00:09:13,604
So most data center racks are 10 kilowatts per or lower per rack, right, on average, which
meant

133
00:09:13,604 --> 00:09:18,244
that all data centers for the last 30 years were designed to handle that kind of a load.

134
00:09:18,304 --> 00:09:21,204
And so really the IT was taken out of the equation, right?

135
00:09:21,204 --> 00:09:22,844
So all data centers could handle that kind of load.

136
00:09:22,844 --> 00:09:30,924
And oh, if you did have a 20 kilowatt rack or you had some application that required that,
well, in-row cooling came in, rear door heat exchangers came in.

137
00:09:30,924 --> 00:09:39,184
There are ways to do what they call spot level cooling to bring that load from 20 or 30
kilowatts back down to 10 so the data center could take the rest of the heat.

138
00:09:39,184 --> 00:09:40,864
And it worked great for years.

139
00:09:40,864 --> 00:09:43,364
And I remember having these discussions.

140
00:09:43,453 --> 00:09:48,965
at different forums probably 10 years ago about how, you know, it's going to get up to 20
and 30 and 40 kilowatts a rack.

141
00:09:48,965 --> 00:09:50,518
It's going to change the game.

142
00:09:50,619 --> 00:09:53,161
And it never did as fast as I thought.

143
00:09:53,161 --> 00:09:57,434
And the reason why was because the IT guys kept getting very innovative.

144
00:09:57,434 --> 00:10:04,770
They kept finding better ways to cool stuff with fans and heat sinks and better
utilization of power efficiency and things like that.

145
00:10:04,770 --> 00:10:09,434
So basically it kind of took the data center having to change out of the equation, right?

146
00:10:09,434 --> 00:10:12,566
And what we found when we were doing this back in the day is that

147
00:10:13,028 --> 00:10:19,208
To make modular cost effective, you typically had to be above 22 kilowatts a rack.

148
00:10:19,388 --> 00:10:23,108
And that made it more cost effective to be in a module than it did in a data center.

149
00:10:23,448 --> 00:10:26,228
I could go either way if it was below that, right?

150
00:10:26,708 --> 00:10:28,268
But, okay, fine.

151
00:10:28,268 --> 00:10:29,388
And so things were starting creeping up.

152
00:10:29,388 --> 00:10:30,928
All of a sudden here comes AI.

153
00:10:31,888 --> 00:10:35,628
135, 125 kilowatts a rack going up higher, right?

154
00:10:35,668 --> 00:10:36,828
Totally changed the game.

155
00:10:36,828 --> 00:10:41,308
There is not a traditional data center on the planet that could support that kind of heat
load, right?

156
00:10:41,308 --> 00:10:42,188
And so that...

157
00:10:42,188 --> 00:10:43,448
made a lot of sense for modules.

158
00:10:43,448 --> 00:10:45,790
Because you think about what you're doing in a data center environment.

159
00:10:45,790 --> 00:10:55,037
uh If you and I are in a conference room, and say it's built for say 20 people, just me
and you there, the AC could go out and we may not know it for an hour or two because it's

160
00:10:55,037 --> 00:10:56,068
over provisioned, right?

161
00:10:56,068 --> 00:10:59,099
There's more cooling capacity in that room than us.

162
00:10:59,160 --> 00:11:00,761
Consider us being IT relics.

163
00:11:00,761 --> 00:11:08,516
If you and I are in a car and we're driving in the Houston heat, the AC goes out of the
car, we're going to know it very quickly because the air is cooling us.

164
00:11:08,516 --> 00:11:09,727
It's not cooling anything else.

165
00:11:09,727 --> 00:11:11,714
And that's really the concept of a module.

166
00:11:11,714 --> 00:11:15,484
It's cooling exactly what it needs to cool, which is the IT, so it's more efficient.

167
00:11:15,484 --> 00:11:18,742
You're not cooling a lot of extra space, and that's what makes it efficient.

168
00:11:18,742 --> 00:11:19,774
uh

169
00:11:19,774 --> 00:11:24,953
So how does uh CDM do modular differently or different from others?

170
00:11:25,838 --> 00:11:26,528
Good question.

171
00:11:26,528 --> 00:11:38,414
uh The way I kind of look at it is uh not to begrudge anybody out there, what I see at
least today, what most modular providers do is they take, OK, I build these type of

172
00:11:38,414 --> 00:11:40,115
infrastructure solutions for data centers.

173
00:11:40,115 --> 00:11:43,947
How can I put them in a box and use it in a modular format, right?

174
00:11:43,947 --> 00:11:45,978
My in-row cooling, whatever I'm doing.

175
00:11:45,978 --> 00:11:51,170
And I'm not saying there's anything wrong with that, but it's basically taking a known
application or a known

176
00:11:51,257 --> 00:11:57,759
technique and you're just basically putting it in a different type of uh a box or a
different type of form factor.

177
00:11:57,819 --> 00:11:59,318
I have never looked at it that way.

178
00:11:59,318 --> 00:12:05,761
Well, how we look at it at CDM is because of our IT background, we look at it from an
application perspective.

179
00:12:05,761 --> 00:12:09,922
No one's in the data center business just because people want to sell hardware.

180
00:12:09,922 --> 00:12:11,817
Someone's trying to run an application.

181
00:12:11,817 --> 00:12:16,864
An application requires IT hardware, which requires the infrastructure around it to
support it.

182
00:12:16,864 --> 00:12:19,004
And so we always look at it from that perspective.

183
00:12:19,076 --> 00:12:28,036
So what we want to do in our applications is to be as efficient as we can with cooling and
power, no over or under provisioning so that you don't, so you can match it up.

184
00:12:28,036 --> 00:12:34,396
As an example, a lot of times you'll get to some of these module guys and they'll say,
okay, here's my solution is 500 kilowatts as an example.

185
00:12:34,396 --> 00:12:34,956
Well, I don't need that.

186
00:12:34,956 --> 00:12:35,756
I need 300.

187
00:12:35,756 --> 00:12:37,096
Sorry, this is what we got.

188
00:12:37,096 --> 00:12:43,736
I can match the cooling and power to your 300 and you're not going to be over or under
provisioned because of how we do our design work.

189
00:12:43,736 --> 00:12:48,706
And you got to keep in mind, we've been doing this for a long time, probably longer than
most people have done it.

190
00:12:48,706 --> 00:12:50,328
So we know what works and what doesn't.

191
00:12:50,328 --> 00:12:54,143
So we're not using the customer as the guinea pig.

192
00:12:54,143 --> 00:12:56,736
know, hey, we just got in the modular space last year.

193
00:12:56,736 --> 00:12:59,340
You you're our second or third customer.

194
00:12:59,340 --> 00:13:00,011
We're still learning.

195
00:13:00,011 --> 00:13:02,734
You you don't get that from us because we know what works and what doesn't.

196
00:13:02,734 --> 00:13:07,740
What we're focusing on is the performance of the architecture to support what your
application is.

197
00:13:08,428 --> 00:13:09,780
Got it, got it.

198
00:13:10,122 --> 00:13:18,119
So where are you seeing the hottest demand and maybe give us a couple of use cases where
the edge makes the most sense.

199
00:13:18,681 --> 00:13:19,741
You know, it's interesting.

200
00:13:19,741 --> 00:13:24,215
uh You hear a lot about learning and AI learning, right?

201
00:13:24,215 --> 00:13:25,412
All the big guys are doing it.

202
00:13:25,412 --> 00:13:27,427
These big bloated data centers.

203
00:13:27,427 --> 00:13:39,375
You look at the Nvidia, like NBL72 platform, and they put a lot of information out there
on how that's supported and what's the best way to support it, because it's so dense.

204
00:13:39,415 --> 00:13:41,877
And so that's a big one.

205
00:13:41,877 --> 00:13:46,361
uh What we try to do in our space, uh

206
00:13:46,361 --> 00:13:59,321
from this perspective is we try to look at it a little bit differently is again, it gets
back to the application and what we're trying to support, but it's really about the

207
00:13:59,321 --> 00:14:02,993
application and what's the best way to support the end to end.

208
00:14:02,993 --> 00:14:06,996
So I don't know if that answered the question thoroughly or not, but that's kind how I
look at it.

209
00:14:07,136 --> 00:14:08,510
Yeah, no, that makes sense.

210
00:14:08,510 --> 00:14:13,506
How do you see modular data centers enabling inference workloads at the edge?

211
00:14:13,506 --> 00:14:15,357
Okay, so the inference side's interesting.

212
00:14:15,357 --> 00:14:18,437
yeah, to answer that question a little bit better.

213
00:14:18,437 --> 00:14:23,359
eh What we try to do is, when we first looked at this, we looked at learning what was
going on.

214
00:14:23,359 --> 00:14:25,680
We said, okay, there's a lot of people doing that.

215
00:14:25,680 --> 00:14:27,100
We've got a module that support that.

216
00:14:27,100 --> 00:14:28,760
We call it our learning module.

217
00:14:28,760 --> 00:14:31,781
You can do anywhere from 1.5 to three megawatts.

218
00:14:31,821 --> 00:14:34,679
And then we started looking at, okay, well, where's inference going?

219
00:14:34,679 --> 00:14:36,282
What's inference mean?

220
00:14:36,523 --> 00:14:41,514
And inference is interesting to me because if you go back a few years and people talk
about the cloud.

221
00:14:41,828 --> 00:14:43,608
And everybody asks, what is the cloud?

222
00:14:43,608 --> 00:14:47,508
And you'd hear four different answers from four different people, what the cloud meant,
right?

223
00:14:47,508 --> 00:14:49,588
And in first, I kind of put in that bucket.

224
00:14:49,628 --> 00:14:52,008
the reason why I do is just because it's unknown.

225
00:14:52,348 --> 00:14:55,628
I wouldn't call it the wild, wild west, but there's a lot of, it's a new market.

226
00:14:55,628 --> 00:14:57,228
It's emerging market, right?

227
00:14:57,228 --> 00:15:00,608
And I think inference has a lot of interesting applications.

228
00:15:00,608 --> 00:15:03,908
So that's where we're seeing a lot of the growth is in inference.

229
00:15:04,168 --> 00:15:08,252
And what I mean by that is it's not, it's AI, but it's not.

230
00:15:08,398 --> 00:15:14,892
Say for example, if you look at the NVIDIA's NVL72 platform, I think it's eight AI racks
and eight network racks, right?

231
00:15:14,892 --> 00:15:17,334
It makes up that solution for learning.

232
00:15:17,334 --> 00:15:22,618
What we're finding in Ference is people need one AI rack or two AI racks and a couple of
network racks.

233
00:15:22,618 --> 00:15:28,222
So it's still high density per rack, 130 plus, but it's not a lot of them, right?

234
00:15:28,222 --> 00:15:30,003
And they want to put them in different locations.

235
00:15:30,003 --> 00:15:34,406
uh Edge is all about, people ask what a hedge is all the time.

236
00:15:34,406 --> 00:15:36,047
And it really depends again who you talk to.

237
00:15:36,047 --> 00:15:37,848
If I talk to a large,

238
00:15:37,925 --> 00:15:44,025
hyperscale guy, he might say, well, my edge data center is 50 megawatts, right?

239
00:15:44,365 --> 00:15:54,545
If you talk to a guy who's trying to do, I don't know, automated driving, it might be half
a megawatt, 300 kilowatts, right, an edge.

240
00:15:54,545 --> 00:16:03,045
But really what it's all about is edge is all about removing latency, getting the source
as close as you can to the customer so you don't have any latency, right?

241
00:16:03,045 --> 00:16:07,214
An example I like to use is if you're using one of the maps,

242
00:16:07,214 --> 00:16:13,981
Google Maps, whatever you use, Microsoft Maps, you don't want to find out free access to
go, you should exit it because your data was late getting there, right?

243
00:16:13,981 --> 00:16:15,713
So that's really what it's all about.

244
00:16:15,713 --> 00:16:19,135
And so people are finding exactly right.

245
00:16:19,376 --> 00:16:24,821
And so that's what you're trying to, you're trying to really get that stuff set up in a
way that makes sense.

246
00:16:24,822 --> 00:16:26,583
And so that's where we see a lot of uses.

247
00:16:26,583 --> 00:16:31,780
And the other thing that's coming up is people are trying to find and utilize stranded
power, right?

248
00:16:31,780 --> 00:16:34,300
which is another big thing on AI data center.

249
00:16:34,300 --> 00:16:35,540
think about that.

250
00:16:35,540 --> 00:16:41,680
If you've got a stranded power of two or three megawatts here or there, put a little AI
module there, right?

251
00:16:41,680 --> 00:16:43,020
And use that power.

252
00:16:43,020 --> 00:16:52,040
I've got some customers who are talking to us about, hey, I've got my data center, it's
full, I can't put any more IT in it without building more on it, but I've got five

253
00:16:52,040 --> 00:16:54,740
megawatts of stranded power I can't use.

254
00:16:54,860 --> 00:16:56,060
Perfect for a module.

255
00:16:56,660 --> 00:16:57,440
And we can put them anywhere.

256
00:16:57,440 --> 00:16:59,982
We can put them on roofs, we can put them on parking lots.

257
00:16:59,982 --> 00:17:02,967
put them just you know name it we put them there so.

258
00:17:03,754 --> 00:17:07,877
So talent is a big, really hot topic in our industry right now.

259
00:17:07,877 --> 00:17:12,020
And the talent gaps, we just don't have enough people.

260
00:17:12,020 --> 00:17:19,726
What are you feeling with CDM when it comes to being able to get the people that you need?

261
00:17:20,260 --> 00:17:21,361
It's getting tougher.

262
00:17:21,361 --> 00:17:22,261
I won't deny it.

263
00:17:22,261 --> 00:17:26,654
uh We've kind of got a little bit of a two-edged sword in the modular space.

264
00:17:26,654 --> 00:17:29,172
First off, there's not a lot of people who understand modular.

265
00:17:29,172 --> 00:17:31,566
It's still fairly new, right?

266
00:17:31,686 --> 00:17:34,578
And uh especially IT modularization, right?

267
00:17:34,578 --> 00:17:37,950
It's one thing if you're doing an electrical skid, okay, now I'm taking these components.

268
00:17:37,950 --> 00:17:41,412
Instead of putting them individually in the data server, I'm going put them on a skid and
ship them out.

269
00:17:41,412 --> 00:17:43,613
It's the same basic stuff.

270
00:17:43,613 --> 00:17:49,497
In the IT modularization, there's a little bit more uh skill sets and differentiation to
it.

271
00:17:49,497 --> 00:17:58,160
how you do the cooling, how you stand that way, how you handle backup, how you handle
redundancy, all this stuff uh in a very small form factor, right?

272
00:17:58,320 --> 00:18:02,492
That's the other thing about modularization is that one thing affects five others, right?

273
00:18:02,492 --> 00:18:03,983
I gotta change this because it doesn't work.

274
00:18:03,983 --> 00:18:08,044
Well, then you just change five other things to make that thing fit because it's so
compact.

275
00:18:08,044 --> 00:18:10,545
And so that's kind of skill sets really hard to find.

276
00:18:10,545 --> 00:18:18,434
uh What I look for is the experience in a certain field like electrician, uh

277
00:18:18,434 --> 00:18:22,746
It could be plumbing, could be general assembly type stuff.

278
00:18:22,746 --> 00:18:31,770
And then are they, do they have the aptitude to kind of think out of the box on how to
apply stuff versus just the way they're doing it, right?

279
00:18:31,770 --> 00:18:32,610
And I've seen both.

280
00:18:32,610 --> 00:18:39,463
I've seen really great electricians, master electricians, they make great supervisors on
the floor because they really know how to get stuff done.

281
00:18:39,463 --> 00:18:41,524
They're used to multitasking and all that.

282
00:18:41,524 --> 00:18:45,764
But you got to have to make sure that you get the one that can kind of see out of the box
and go, yeah.

283
00:18:45,764 --> 00:18:50,844
That's the way the building code works, but this is not a building per se, it's an
electronic device.

284
00:18:50,844 --> 00:18:55,844
So you have to be a little bit more flexible with how you think about how you apply this
stuff.

285
00:18:55,844 --> 00:18:57,468
And that's really what I look for.

286
00:18:58,014 --> 00:19:00,009
So what do your training programs look like?

287
00:19:00,009 --> 00:19:00,971
Is it formal?

288
00:19:00,971 --> 00:19:02,584
Is it informal?

289
00:19:03,330 --> 00:19:04,641
Right now it's informal.

290
00:19:04,641 --> 00:19:12,507
It's more about, you when you're in a young company like we are, it's, you know, I also
look for people who can do multitasking, right?

291
00:19:12,507 --> 00:19:14,308
There's no such thing as a single task.

292
00:19:14,308 --> 00:19:20,973
I think as you get bigger, you want that, but I look for people who can wear four or five
different hats, you know, and then okay with it.

293
00:19:20,973 --> 00:19:25,726
I mean, it's an industry, I mean, it's a very good and exciting time to be in the
industry.

294
00:19:25,726 --> 00:19:31,205
I don't know there's how many times where you could actually change the industry, uh the
way we're doing it.

295
00:19:31,205 --> 00:19:32,845
But you have to have the right mindset.

296
00:19:32,845 --> 00:19:39,085
What I try to do is that I want people who come to work for CDM to be happy to be here.

297
00:19:39,085 --> 00:19:44,065
And what I mean by that is when you wake up on a Sunday morning, you're not going, oh
gosh, got to go work tomorrow.

298
00:19:44,425 --> 00:19:49,645
I want to think, oh my God, one more day of family and friends that I got to get back to
work, got this to do that.

299
00:19:49,645 --> 00:19:53,045
I want to be excited about doing that because you spend a lot of time there, right?

300
00:19:53,045 --> 00:19:57,105
A lot of it has to do with the atmosphere that you create, the environment that you
create.

301
00:19:57,205 --> 00:20:00,463
And also, I think giving people the bandwidth

302
00:20:00,463 --> 00:20:02,474
to make decisions, you know?

303
00:20:02,474 --> 00:20:04,556
I am by no means a micromanager.

304
00:20:04,556 --> 00:20:10,156
My job is to get the right people in the right place and remove roadblocks to make them
successful, right?

305
00:20:10,156 --> 00:20:14,603
And I'll give you an example of what I mean by what I call a good employee and an
exemplary employee.

306
00:20:14,603 --> 00:20:17,885
And that example would be, let's say I give you a task.

307
00:20:17,885 --> 00:20:22,789
I say, okay, ABC, we need to look at ABC and see which one works best, all right?

308
00:20:22,789 --> 00:20:28,052
And you come back you go look at ABC, you come back and say, okay, Ron, I looked at ABC,
now I'm really gonna work.

309
00:20:28,068 --> 00:20:32,048
based on my analysis, and you did a really good job on the analysis, what's next, right?

310
00:20:32,048 --> 00:20:32,868
That's a good employee.

311
00:20:32,868 --> 00:20:35,768
You've actually fulfilled what I asked you to do.

312
00:20:35,968 --> 00:20:42,828
An exemplary employee would come back and say, you know, Ron, I looked at ABC, none of are
gonna work, but I talked to Joe down the hall, he thought D might work.

313
00:20:42,828 --> 00:20:44,428
Look at this, what do you think?

314
00:20:44,428 --> 00:20:45,228
That's the difference.

315
00:20:45,228 --> 00:20:47,688
That's what I look for, is that kind of employee.

316
00:20:47,810 --> 00:20:49,491
Wow, I love that analogy.

317
00:20:49,491 --> 00:20:56,173
And what you said about uh one person wearing multiple hats is actually a future trend in
our industry.

318
00:20:56,173 --> 00:21:04,197
So if you look, you're gonna see each role is gonna have multiple tasks that they do, not
just one.

319
00:21:04,197 --> 00:21:08,899
So that's gonna happen all over the industry and we're gonna see it soon.

320
00:21:08,899 --> 00:21:11,020
So what's next for CDM?

321
00:21:11,020 --> 00:21:17,542
Next 24 months, let's say, what bets are you making for growth, for opportunities?

322
00:21:18,308 --> 00:21:20,908
Well, you know, we're going to continue on our trend.

323
00:21:21,008 --> 00:21:23,768
Half the battle on this stuff is this.

324
00:21:23,768 --> 00:21:25,508
I mean that in a nice way.

325
00:21:25,968 --> 00:21:27,108
It's competitive battle.

326
00:21:27,108 --> 00:21:28,488
It's a lot of people out there.

327
00:21:28,488 --> 00:21:36,868
Anytime you have a market like this is growing, and I think it's like 19 % CAGR over the
next five years, this market, you're going to have a lot of people entering the business.

328
00:21:36,868 --> 00:21:39,428
And there's going be a lot of people coming in.

329
00:21:39,428 --> 00:21:41,788
It'll be interesting to see who's still here in five years.

330
00:21:41,788 --> 00:21:45,408
But what we're trying to do is really develop those partnerships and relationships.

331
00:21:45,688 --> 00:21:48,252
Because I really think that's the opportunity is

332
00:21:48,289 --> 00:21:58,943
is working with the ITOEMs, working with the NVIDIAs and the DELs and the HPs of the world
and these different uh industrial manufacturers to build the batteries and the PDUs and

333
00:21:58,943 --> 00:22:00,179
the power distribution.

334
00:22:00,179 --> 00:22:01,520
And because it's an ecosystem.

335
00:22:01,520 --> 00:22:08,725
And what we discovered even years ago when we were doing this at Compaq is that there's no
one company that can do it all.

336
00:22:08,725 --> 00:22:10,607
You have to have partnerships.

337
00:22:10,607 --> 00:22:14,510
And so we're looking for those long-term partnerships that can help us develop that
solution.

338
00:22:14,510 --> 00:22:17,476
So I really think over time the solution is going to change.

339
00:22:17,476 --> 00:22:21,276
how we're doing modularization today is not going to be the way it's going to be tomorrow.

340
00:22:21,276 --> 00:22:23,176
And I'll use just a simple example.

341
00:22:23,176 --> 00:22:30,776
If you look at what we're doing today, we're taking a box, which is a computer, we're
putting it in another box, which is a rack, we're that in another box, which is a module

342
00:22:30,776 --> 00:22:31,816
or a data center.

343
00:22:31,816 --> 00:22:33,836
So it's a box in a box in a box.

344
00:22:33,836 --> 00:22:34,916
Why are we doing that?

345
00:22:34,916 --> 00:22:39,336
We're doing that because the server came from the desktops and it was a box, right?

346
00:22:39,336 --> 00:22:43,172
But over time, think about if you can start eliminating those boxes, you start doing...

347
00:22:43,172 --> 00:22:51,272
Some of the things that the server guys did in the early days with the blade servers where
you have things that plug into different components and you could kind of remove some of

348
00:22:51,272 --> 00:22:53,732
that infrastructure and simplify it, right?

349
00:22:53,732 --> 00:22:55,872
Especially as the power density goes up.

350
00:22:55,872 --> 00:22:57,532
I think that's where it's going to go.

351
00:22:57,532 --> 00:23:02,592
But I think again, it's a system of partnerships who have expertise in those areas.

352
00:23:02,592 --> 00:23:05,192
You hear a lot about 800 volt DC, right?

353
00:23:05,192 --> 00:23:08,692
That's being talked about a lot for better, more efficient power distribution.

354
00:23:08,692 --> 00:23:10,876
Well, that takes time to develop, right?

355
00:23:11,240 --> 00:23:13,400
And so I think what you're going to see is a step function.

356
00:23:13,400 --> 00:23:16,340
And so what's up for us is we're going to sell today what we've got.

357
00:23:16,340 --> 00:23:24,700
We've got some very good solutions, but we want to innovate with the industry and be able
to innovate as fast as the IT guys.

358
00:23:24,700 --> 00:23:31,740
And I think that's the paradigm shift is that no longer can the data center become static
and the IT guys just do what they do and you just go with the data center.

359
00:23:31,740 --> 00:23:39,660
You have to be able to change with the IT as the IT comes out, as they go higher density,
as they need different types of cooling and power requirements.

360
00:23:39,829 --> 00:23:47,559
then the infrastructure, the opportunity for us in this side of the space is to innovate
with them and become their new best friends.

361
00:23:48,619 --> 00:23:49,320
I love it, Ron.

362
00:23:49,320 --> 00:23:52,455
How do we reach you learn more about CDM?

363
00:23:53,038 --> 00:23:57,852
So we are at CDmodular.com, I think CD-modular.com on our website.

364
00:23:57,852 --> 00:24:00,835
And you can get all the information there on what we're doing, how we're doing it.

365
00:24:00,835 --> 00:24:05,618
ah You can reach out to me, Ron, at CD-modular.com as well.

366
00:24:05,819 --> 00:24:08,281
But yeah, it's all there on the website.

367
00:24:08,281 --> 00:24:13,626
We are going to a lot of different shows and speaking at lot of different events, so you
can kind of see us there too.

368
00:24:13,626 --> 00:24:18,350
ah So we'll be at Gartner in uh December, so we'll be there.

369
00:24:18,350 --> 00:24:20,952
If anybody wants to come chat, we'll be there for that.

370
00:24:21,230 --> 00:24:25,273
You know, just uh different venues like that, but we're more happy to hear from you.

371
00:24:25,273 --> 00:24:28,448
even if we don't, you know, there's not a fit right now for business.

372
00:24:28,448 --> 00:24:35,955
I'm more than happy to talk about where Modular is going or what we can do, what you need
to be cognizant of if you're looking at these types of solutions.

373
00:24:36,183 --> 00:24:37,865
Thank you, Ron, for coming on the show.

374
00:24:37,865 --> 00:24:39,151
I've learned a lot.

375
00:24:40,023 --> 00:24:40,677
Well, no problem.

376
00:24:40,677 --> 00:24:41,811
I really enjoyed it.

377
00:24:42,325 --> 00:24:43,465
You take care.

378
00:24:43,758 --> 00:24:44,658
Thank you.

379
00:24:47,284 --> 00:24:49,412
Alrighty, that was awesome!