Decarbonizing the Cold Chain: Waste Less, Recover More

1
00:00:00,120 --> 00:00:04,740
- The era of the standard
data center is over.

2
00:00:04,740 --> 00:00:06,870
- Okay, that's a big statement, Dominique.

3
00:00:06,870 --> 00:00:08,670
- Scott, just think about it, right?

4
00:00:08,670 --> 00:00:12,090
The technology is evolving so fast,

5
00:00:12,090 --> 00:00:14,940
whether that's cooling
technology, the chip technology,

6
00:00:14,940 --> 00:00:17,700
even the demand, right, around AI.

7
00:00:17,700 --> 00:00:20,850
It's all changing so fast,
we just can't standardize.

8
00:00:20,850 --> 00:00:23,790
- Innovation always
happens when we view things

9
00:00:23,790 --> 00:00:26,340
as a problem that need a solution.

10
00:00:26,340 --> 00:00:28,740
And this is actually one of those moments

11
00:00:28,740 --> 00:00:31,380
where data centers are more complicated.

12
00:00:31,380 --> 00:00:34,260
We have to design them
differently, so it's ripe

13
00:00:34,260 --> 00:00:36,300
for sustainable innovation.

14
00:00:36,300 --> 00:00:37,770
- Wherever there is a challenge,

15
00:00:37,770 --> 00:00:39,210
there's always an opportunity.

16
00:00:39,210 --> 00:00:40,650
I love your thinking there.

17
00:00:40,650 --> 00:00:43,020
- So, if we've left
behind this standard way

18
00:00:43,020 --> 00:00:45,930
of doing data centers,
what's the new way forward?

19
00:00:45,930 --> 00:00:49,110
- Well, to find out, I
spoke to Oakley Roberts.

20
00:00:49,110 --> 00:00:52,590
He's the president of data
centers at Trane Technologies.

21
00:00:52,590 --> 00:00:54,210
Let's get into it.

22
00:00:54,210 --> 00:00:55,440
I'm Dominique Silva.

23
00:00:55,440 --> 00:00:58,800
- And I'm Scott Tew, and
this is "Healthy Spaces,"

24
00:00:58,800 --> 00:01:02,313
conversations at the frontier
of sustainable technology.

25
00:01:03,510 --> 00:01:05,670
- Yeah, I started working with
data centers back in 2019,

26
00:01:05,670 --> 00:01:08,643
and it was a different world then.

27
00:01:08,643 --> 00:01:12,630
It was before ChatGPT
was a household name,

28
00:01:12,630 --> 00:01:14,670
and things were very standardized.

29
00:01:14,670 --> 00:01:17,910
We're building, today,
data center structures

30
00:01:17,910 --> 00:01:20,910
that are designed to last decades of time,

31
00:01:20,910 --> 00:01:22,710
but the chips that we're putting aside

32
00:01:22,710 --> 00:01:24,870
of those data centers,

33
00:01:24,870 --> 00:01:27,600
they last a much shorter amount of time,

34
00:01:27,600 --> 00:01:29,340
and it's creating an environment

35
00:01:29,340 --> 00:01:30,840
of engineered flexibility.

36
00:01:30,840 --> 00:01:32,760
And the data center is
a physical creature.

37
00:01:32,760 --> 00:01:35,340
It has to sweat, breathe, reject heat,

38
00:01:35,340 --> 00:01:37,650
and you have to match the lungs

39
00:01:37,650 --> 00:01:40,530
to the local climate,
or the machine fails.

40
00:01:40,530 --> 00:01:45,360
So, if you think about London
with a milder, kind of,

41
00:01:45,360 --> 00:01:49,980
more human climate versus
Texas that's dry and hot,

42
00:01:49,980 --> 00:01:52,290
if you think about the thermal swings

43
00:01:52,290 --> 00:01:55,110
of the climate in Texas,

44
00:01:55,110 --> 00:01:58,860
it's a very diverse, you
know, thermal environment,

45
00:01:58,860 --> 00:02:01,170
and that's very different
than what you find in London.

46
00:02:01,170 --> 00:02:04,260
The result is you have to apply, you know,

47
00:02:04,260 --> 00:02:08,400
very different design
architectures to these locations.

48
00:02:08,400 --> 00:02:09,960
- Well, let's dive a little bit

49
00:02:09,960 --> 00:02:14,580
into this notion of different places,

50
00:02:14,580 --> 00:02:15,720
different answers, right?

51
00:02:15,720 --> 00:02:17,130
- We were kind of just
talking about climate,

52
00:02:17,130 --> 00:02:18,570
but it goes far beyond that.

53
00:02:18,570 --> 00:02:20,220
You move into the other factors,

54
00:02:20,220 --> 00:02:22,200
like regulatory environment,

55
00:02:22,200 --> 00:02:25,680
and, you know, just take,
like, Europe in general,

56
00:02:25,680 --> 00:02:28,890
but if you come back to
Texas, it's a big state,

57
00:02:28,890 --> 00:02:31,620
and a lot of these data centers in Texas

58
00:02:31,620 --> 00:02:36,420
are seeking opportunities
to help find their land,

59
00:02:36,420 --> 00:02:39,150
so open land, but also power.

60
00:02:39,150 --> 00:02:42,660
Power constraints is obviously
one of the driving factors

61
00:02:42,660 --> 00:02:44,550
in the infrastructure build-out.

62
00:02:44,550 --> 00:02:47,280
And there's a lot of
energy in Texas, you know.

63
00:02:47,280 --> 00:02:48,630
We're seeing the approach

64
00:02:48,630 --> 00:02:52,200
to "How do we solve the
power generation challenge?"

65
00:02:52,200 --> 00:02:53,490
is a different opportunity there

66
00:02:53,490 --> 00:02:55,560
because of the source that's available.

67
00:02:55,560 --> 00:02:58,680
And that creates different
design opportunities

68
00:02:58,680 --> 00:03:00,870
around cooling data centers.

69
00:03:00,870 --> 00:03:04,320
Those gas turbines that are
being used to help supply

70
00:03:04,320 --> 00:03:07,170
power put off high-temperature,
high-grade heat.

71
00:03:07,170 --> 00:03:09,750
And you can do things
like absorption chilling.

72
00:03:09,750 --> 00:03:11,450
You can actually capture that heat

73
00:03:12,360 --> 00:03:14,400
and use technology to create more cooling.

74
00:03:14,400 --> 00:03:16,650
What excites me, whether you're
talking about our example

75
00:03:16,650 --> 00:03:19,920
in London or Europe or in a
very different place like Texas,

76
00:03:19,920 --> 00:03:21,330
it's a system design challenge,

77
00:03:21,330 --> 00:03:22,867
and it's the opportunity to think about,

78
00:03:22,867 --> 00:03:24,870
"How do we reduce waste?"

79
00:03:24,870 --> 00:03:26,610
- Well, maybe let's
double-click a little bit

80
00:03:26,610 --> 00:03:30,270
on that whole cooling aspect, right?

81
00:03:30,270 --> 00:03:31,520
You've already alluded, right,

82
00:03:31,520 --> 00:03:34,350
to the thermal management complexity

83
00:03:34,350 --> 00:03:37,380
and the fact that we're
even looking to heat waste

84
00:03:37,380 --> 00:03:39,510
as a power source, ultimately.

85
00:03:39,510 --> 00:03:43,477
So, can you help me and our
listeners sort of break down,

86
00:03:43,477 --> 00:03:45,990
"What does the cooling process

87
00:03:45,990 --> 00:03:48,840
for a data center actually look like?"

88
00:03:48,840 --> 00:03:49,673
Right?

89
00:03:49,673 --> 00:03:54,390
- As we got into the AI era
and the silicon and the servers

90
00:03:54,390 --> 00:03:58,470
and the compute technology
required to meet that demand,

91
00:03:58,470 --> 00:04:01,290
it changed the amount of
heat that's coming off

92
00:04:01,290 --> 00:04:03,090
of that due to these chips.

93
00:04:03,090 --> 00:04:06,090
And the chips then had to be
cooled in a different way.

94
00:04:06,090 --> 00:04:09,750
You couldn't just blow air
over a chip to cool it.

95
00:04:09,750 --> 00:04:11,490
You had to start to use liquid,

96
00:04:11,490 --> 00:04:13,800
and that revolutionized the way

97
00:04:13,800 --> 00:04:16,590
that data center infrastructure
is being designed.

98
00:04:16,590 --> 00:04:18,780
And so, the technology
that's being deployed

99
00:04:18,780 --> 00:04:20,460
in a modern AI data center today

100
00:04:20,460 --> 00:04:24,210
is primarily liquid-cooled,
direct-to-chip technology.

101
00:04:24,210 --> 00:04:27,510
So, you're actually taking
water and cooling it,

102
00:04:27,510 --> 00:04:29,400
and then it has to go
through a thermal system

103
00:04:29,400 --> 00:04:30,990
that arrives at the chip,

104
00:04:30,990 --> 00:04:33,360
where it then gets heated up by the chip.

105
00:04:33,360 --> 00:04:37,170
And then you have to reject
that heat to the atmosphere.

106
00:04:37,170 --> 00:04:38,460
One way to think about it

107
00:04:38,460 --> 00:04:42,660
is "chip to atmosphere"
type of thermal management.

108
00:04:42,660 --> 00:04:46,110
And that creates an
opportunity to really think

109
00:04:46,110 --> 00:04:49,320
about how you break that system
down to provide the best way

110
00:04:49,320 --> 00:04:51,540
to deliver liquid to the chip

111
00:04:51,540 --> 00:04:52,860
and get it out to the atmosphere.

112
00:04:52,860 --> 00:04:54,630
And so, even the design
is broken down now.

113
00:04:54,630 --> 00:04:57,510
So, if you're inside of
what's called the data hall,

114
00:04:57,510 --> 00:04:59,310
where these servers sit and
these chips are being cooled

115
00:04:59,310 --> 00:05:02,670
by liquid, we use part of
our system called a CDU,

116
00:05:02,670 --> 00:05:04,500
or cooling distribution unit,

117
00:05:04,500 --> 00:05:07,230
to manage what's called
the technical water loop.

118
00:05:07,230 --> 00:05:08,640
And it's very specific

119
00:05:08,640 --> 00:05:11,760
to optimize delivering the
flow rates to the chips.

120
00:05:11,760 --> 00:05:15,660
And that CDU gets connected
to another liquid cooling loop

121
00:05:15,660 --> 00:05:19,680
that's much larger,
called the facility loop,

122
00:05:19,680 --> 00:05:21,420
that's connected to a chiller.

123
00:05:21,420 --> 00:05:25,140
And all that is connected
to a heat rejection device

124
00:05:25,140 --> 00:05:28,470
that allows you to release
that heat to the atmosphere.

125
00:05:28,470 --> 00:05:29,340
As we were just talking about,

126
00:05:29,340 --> 00:05:30,487
you can actually use that to say,

127
00:05:30,487 --> 00:05:31,500
"How do we capture that heat

128
00:05:31,500 --> 00:05:32,640
and not release it to the atmosphere

129
00:05:32,640 --> 00:05:35,220
and try to do something
more useful with it?"

130
00:05:35,220 --> 00:05:37,080
So, it's very complicated,

131
00:05:37,080 --> 00:05:40,470
and it requires a very high
level of system engineering

132
00:05:40,470 --> 00:05:43,050
and optimization to make that work.

133
00:05:43,050 --> 00:05:45,000
And liquid doesn't cool everything.

134
00:05:45,000 --> 00:05:48,270
We still have to use traditional methods

135
00:05:48,270 --> 00:05:49,890
of what we call, like, air cooling

136
00:05:49,890 --> 00:05:51,240
because there's many other things

137
00:05:51,240 --> 00:05:52,830
besides the chip that produce heat.

138
00:05:52,830 --> 00:05:55,260
We have space that it needs heat removed.

139
00:05:55,260 --> 00:05:57,150
So, it's become a hybrid environment

140
00:05:57,150 --> 00:05:59,370
where some of that air cooling technology

141
00:05:59,370 --> 00:06:01,440
that's been around for a very long time

142
00:06:01,440 --> 00:06:03,960
is being combined with this
liquid cooling technology.

143
00:06:03,960 --> 00:06:07,140
- You know, the other day, I
had to explain the difference

144
00:06:07,140 --> 00:06:09,000
between air cooling and liquid cooling

145
00:06:09,000 --> 00:06:11,160
to a friend of mine who works

146
00:06:11,160 --> 00:06:13,980
in IT and doesn't understand
anything about cooling.

147
00:06:13,980 --> 00:06:15,780
And I was feeling particularly inspired.

148
00:06:15,780 --> 00:06:18,330
So, I said, you know, "You're
going to go to the beach.

149
00:06:18,330 --> 00:06:19,890
It suddenly gets very hot.

150
00:06:19,890 --> 00:06:22,620
You've been fanning yourself
for a very long time,

151
00:06:22,620 --> 00:06:24,900
but if you're not feeling cooler,

152
00:06:24,900 --> 00:06:25,733
what are you going to do?"

153
00:06:25,733 --> 00:06:27,300
And I say, "I'm going to go dip

154
00:06:27,300 --> 00:06:28,380
into the water, right, swim."

155
00:06:28,380 --> 00:06:29,220
I was like, "Aha!

156
00:06:29,220 --> 00:06:31,770
You have just uncovered
liquid cooling, right?

157
00:06:31,770 --> 00:06:34,380
But you can't stay under the
water for too long, right?

158
00:06:34,380 --> 00:06:35,400
You still got to peep up your head

159
00:06:35,400 --> 00:06:36,600
because you still got to breathe.

160
00:06:36,600 --> 00:06:37,590
So, what do you do?

161
00:06:37,590 --> 00:06:38,550
You fan yourself."

162
00:06:38,550 --> 00:06:40,950
So, anyway, I think they understood.

163
00:06:40,950 --> 00:06:42,900
- I could take your analogy in another way

164
00:06:42,900 --> 00:06:45,210
that's really relevant to
what's going on in data centers.

165
00:06:45,210 --> 00:06:48,420
So, what you described is
actually evaporative cooling.

166
00:06:48,420 --> 00:06:51,570
Water usage is a very important
topic within data centers

167
00:06:51,570 --> 00:06:54,030
because we have to minimize
the amount of water

168
00:06:54,030 --> 00:06:55,830
that a data center uses to help.

169
00:06:55,830 --> 00:06:57,990
In history, one of the ways

170
00:06:57,990 --> 00:06:59,550
that you actually cool the data center

171
00:06:59,550 --> 00:07:01,080
was just like what you described.

172
00:07:01,080 --> 00:07:02,820
What you're doing is spraying

173
00:07:02,820 --> 00:07:06,450
or cascading water over
these heat-exchanging devices

174
00:07:06,450 --> 00:07:07,710
to use that cooling effect.

175
00:07:07,710 --> 00:07:09,840
But that uses water because
that water evaporates,

176
00:07:09,840 --> 00:07:10,920
and then it's not back.

177
00:07:10,920 --> 00:07:12,270
It's not in the environment.

178
00:07:12,270 --> 00:07:13,290
And so, a different way

179
00:07:13,290 --> 00:07:16,590
of cooling yourself from
jumping in the water

180
00:07:16,590 --> 00:07:19,707
is a cooling vest that is a
closed-loop system, you know,

181
00:07:19,707 --> 00:07:22,200
and you could actually
connect it to a cooling source

182
00:07:22,200 --> 00:07:23,697
that circulates cold water, right?

183
00:07:23,697 --> 00:07:25,980
And there's different ways of doing that,

184
00:07:25,980 --> 00:07:28,380
but you're not evaporating water.

185
00:07:28,380 --> 00:07:30,870
So, in a data center, we're
doing a very similar thing now.

186
00:07:30,870 --> 00:07:34,020
We put primarily what's
called a dry cooler device

187
00:07:34,020 --> 00:07:36,300
instead of a water cooling device.

188
00:07:36,300 --> 00:07:37,320
And a dry cooler device

189
00:07:37,320 --> 00:07:38,940
is nothing more than just a closed loop.

190
00:07:38,940 --> 00:07:40,950
So, you're not evaporating water,

191
00:07:40,950 --> 00:07:44,030
you're just recirculating
water and then using the air

192
00:07:44,030 --> 00:07:46,410
or the environment to
actually just remove heat

193
00:07:46,410 --> 00:07:48,090
from that radiator device.

194
00:07:48,090 --> 00:07:50,220
That sounds simple, but it's
a very complicated system,

195
00:07:50,220 --> 00:07:53,490
but what it does is it
removes the need to jump

196
00:07:53,490 --> 00:07:56,100
into the pool to cool off.

197
00:07:56,100 --> 00:07:57,750
- Well, I still want
to jump into the pool,

198
00:07:57,750 --> 00:08:00,330
but I want to ask you another question.

199
00:08:00,330 --> 00:08:03,210
We opened up this conversation
talking about the fact

200
00:08:03,210 --> 00:08:07,560
that chip technology is
evolving very fast, right?

201
00:08:07,560 --> 00:08:09,090
And we're building data centers

202
00:08:09,090 --> 00:08:11,103
that are expected to be around for,

203
00:08:11,970 --> 00:08:14,637
let's just say, decades,
right, for the most part.

204
00:08:14,637 --> 00:08:17,790
The technology that goes into
these data centers can change,

205
00:08:17,790 --> 00:08:20,010
which will ultimately have an impact

206
00:08:20,010 --> 00:08:21,780
on how they are cooled, right?

207
00:08:21,780 --> 00:08:23,100
So, in your experience,

208
00:08:23,100 --> 00:08:25,860
how are Trane Technologies
helping operators prepare

209
00:08:25,860 --> 00:08:27,840
for a future that they don't really know

210
00:08:27,840 --> 00:08:29,430
what it's going to look like?

211
00:08:29,430 --> 00:08:32,490
- It comes back to this concept
of engineering flexibility

212
00:08:32,490 --> 00:08:36,870
and thinking through the system
in terms of what is possible

213
00:08:36,870 --> 00:08:38,130
and changing the variables

214
00:08:38,130 --> 00:08:40,590
without having to change
your capital investment.

215
00:08:40,590 --> 00:08:42,127
A couple simple ones you
could think about is like,

216
00:08:42,127 --> 00:08:46,410
"Well, how much liquid cooling
do I need versus air cooling?

217
00:08:46,410 --> 00:08:48,150
What's the temperature
of the liquid cooling?

218
00:08:48,150 --> 00:08:50,040
What's the mix of the air versus liquid?"

219
00:08:50,040 --> 00:08:52,410
All these things have to be
designed very specifically

220
00:08:52,410 --> 00:08:54,817
for today, but you can think
about tomorrow and just say,

221
00:08:54,817 --> 00:08:57,000
"Well, could I change it by 10%?

222
00:08:57,000 --> 00:08:58,500
Could I change it by 20%?

223
00:08:58,500 --> 00:09:00,450
How wide of a range could I have

224
00:09:00,450 --> 00:09:03,300
so that my capital investment

225
00:09:03,300 --> 00:09:05,607
has that flexibility in the future?"

226
00:09:05,607 --> 00:09:06,840
And so, we spend a lot of time

227
00:09:06,840 --> 00:09:08,460
thinking about that with our customers.

228
00:09:08,460 --> 00:09:11,430
And it starts when you're at day zero,

229
00:09:11,430 --> 00:09:12,990
like before you break ground.

230
00:09:12,990 --> 00:09:16,230
And what that turns into is
more hybrid architecture.

231
00:09:16,230 --> 00:09:19,860
So, it's more heterogeneous
in terms of the technology

232
00:09:19,860 --> 00:09:22,350
that we're using to
provide that flexibility.

233
00:09:22,350 --> 00:09:25,927
Think about, like, digital
and controls and then,

234
00:09:25,927 --> 00:09:28,320
"How do you even use
artificial intelligence

235
00:09:28,320 --> 00:09:29,700
to help artificial intelligence?"

236
00:09:29,700 --> 00:09:34,200
And we're building, like,
very high-performance machines

237
00:09:34,200 --> 00:09:38,010
and systems for our customers
and for these data centers

238
00:09:38,010 --> 00:09:41,100
so that they can operate most efficiently,

239
00:09:41,100 --> 00:09:42,930
so they can have this flexibility

240
00:09:42,930 --> 00:09:44,550
that's needed into the future,

241
00:09:44,550 --> 00:09:47,717
so it can meet those local
demands of Texas versus London.

242
00:09:47,717 --> 00:09:48,550
(upbeat music)

243
00:09:48,550 --> 00:09:51,630
- You know, Dominique,
the discussion with Oakley

244
00:09:51,630 --> 00:09:53,467
really struck me because
he used this term,

245
00:09:53,467 --> 00:09:55,230
"engineering flexibility."

246
00:09:55,230 --> 00:09:57,990
We're engineering today for very diverse

247
00:09:57,990 --> 00:10:00,870
situations in very unique places.

248
00:10:00,870 --> 00:10:04,860
Every place, every local
place we place a data center

249
00:10:04,860 --> 00:10:08,550
requires a engineering customization

250
00:10:08,550 --> 00:10:10,590
because there are different requirements:

251
00:10:10,590 --> 00:10:14,190
water, energy, climatic conditions.

252
00:10:14,190 --> 00:10:17,370
And so, we're engineering
at times for uncertainty.

253
00:10:17,370 --> 00:10:19,950
I mean, we can't predict
where tech is headed.

254
00:10:19,950 --> 00:10:22,290
We just know it's moving faster,

255
00:10:22,290 --> 00:10:26,370
and this pace of change
really makes this flexibility

256
00:10:26,370 --> 00:10:28,860
that Oakley spoke about
more and more important.

257
00:10:28,860 --> 00:10:31,500
- And do you know who
else is feeling the heat?

258
00:10:31,500 --> 00:10:32,850
Policy makers.

259
00:10:32,850 --> 00:10:35,220
I mean, amidst all of this rapid change,

260
00:10:35,220 --> 00:10:37,980
policy is also trying to catch up, right?

261
00:10:37,980 --> 00:10:39,630
Because all over the world,

262
00:10:39,630 --> 00:10:42,217
governments have spoken
about their ambitions:

263
00:10:42,217 --> 00:10:44,880
"We need to scale data center capacity."

264
00:10:44,880 --> 00:10:46,200
In Europe, for example,

265
00:10:46,200 --> 00:10:48,300
I don't know if you saw
the headline, Scott,

266
00:10:48,300 --> 00:10:50,160
Europe earlier this year announced

267
00:10:50,160 --> 00:10:53,600
that they're planning to triple
their data center capacity

268
00:10:53,600 --> 00:10:55,530
in just the next seven years, right?

269
00:10:55,530 --> 00:10:56,690
Can you imagine that?

270
00:10:56,690 --> 00:10:58,530
- Well, I though it was
impossible for us to double

271
00:10:58,530 --> 00:11:01,470
that infrastructure, and
now we're talking tripling.

272
00:11:01,470 --> 00:11:04,860
- And so, we also want to
do that in the right way,

273
00:11:04,860 --> 00:11:08,400
and that's usually what
regulations and policies are for.

274
00:11:08,400 --> 00:11:11,970
So, how do we balance this
need to meet AI demand,

275
00:11:11,970 --> 00:11:14,340
to scale data center construction,

276
00:11:14,340 --> 00:11:16,890
but do it in the right
way, the sustainable way,

277
00:11:16,890 --> 00:11:18,810
the resource-efficient way?

278
00:11:18,810 --> 00:11:21,690
- And be careful not to let
regulations push us in a corner.

279
00:11:21,690 --> 00:11:22,523
- Right.

280
00:11:22,523 --> 00:11:24,720
And you know who knows a lot about that?

281
00:11:24,720 --> 00:11:26,130
My next guest.

282
00:11:26,130 --> 00:11:29,160
To answer the question of
how Europe is going to triple

283
00:11:29,160 --> 00:11:31,080
its data center capacity,

284
00:11:31,080 --> 00:11:32,790
I spoke to Michael Winterson,

285
00:11:32,790 --> 00:11:36,780
Secretary General of the
European Data Center Association.

286
00:11:36,780 --> 00:11:39,600
- Very potted history to
get to where we are today.

287
00:11:39,600 --> 00:11:43,410
Ursula von der Leyen, so
she first became president

288
00:11:43,410 --> 00:11:46,860
of the European Commission
in the beginning of 2020.

289
00:11:46,860 --> 00:11:51,120
And they leaked very early
on both the Green Deal,

290
00:11:51,120 --> 00:11:52,327
which we all know about,

291
00:11:52,327 --> 00:11:54,807
"Europe will be carbon neutral by 2050,"

292
00:11:55,800 --> 00:11:57,750
but they also leaked something

293
00:11:57,750 --> 00:12:00,210
that no one really read except us geeks,

294
00:12:00,210 --> 00:12:04,740
which was "A Europe fit
for the digital age."

295
00:12:04,740 --> 00:12:06,600
And they made a statement in 2020

296
00:12:06,600 --> 00:12:09,360
that this was going to be
Europe's Digital Decade.

297
00:12:09,360 --> 00:12:13,950
And I can tell you very
clearly that both we humans

298
00:12:13,950 --> 00:12:17,670
and them policymakers see that, so far,

299
00:12:17,670 --> 00:12:19,530
the needle has not moved.

300
00:12:19,530 --> 00:12:22,020
And in fact, since then, we've seen things

301
00:12:22,020 --> 00:12:25,680
such as Mario Draghi's
competitiveness report,

302
00:12:25,680 --> 00:12:27,240
where he's talking about Europe's failure

303
00:12:27,240 --> 00:12:28,830
to be productive like America,

304
00:12:28,830 --> 00:12:31,620
predominantly due to the
failure to adopt technology.

305
00:12:31,620 --> 00:12:35,130
And then, last year,
we heard from a senior

306
00:12:35,130 --> 00:12:38,460
executive vice president,
Virkkunen, who stated

307
00:12:38,460 --> 00:12:39,990
that she wanted to see the tripling

308
00:12:39,990 --> 00:12:42,810
of data center capacity in seven years.

309
00:12:42,810 --> 00:12:47,040
And so, the idea behind the
Cloud and AI Development Act

310
00:12:47,040 --> 00:12:51,810
is to define the ways in with which Europe

311
00:12:51,810 --> 00:12:56,310
and its member states
can subsidize, support,

312
00:12:56,310 --> 00:12:58,470
and develop this market space.

313
00:12:58,470 --> 00:13:00,450
So, each member state will be asked

314
00:13:00,450 --> 00:13:02,730
to write effectively a digital strategy

315
00:13:02,730 --> 00:13:06,630
and then, behind that, a
digital infrastructure plan.

316
00:13:06,630 --> 00:13:10,380
And then that will, in
theory, receive funding

317
00:13:10,380 --> 00:13:12,210
and support from the European Union.

318
00:13:12,210 --> 00:13:14,700
- Why do you think Europe needs this now?

319
00:13:14,700 --> 00:13:16,830
- If I ever want to get
a regulator riled up,

320
00:13:16,830 --> 00:13:19,470
I usually tell them
that their digital brain

321
00:13:19,470 --> 00:13:22,653
is about half the size of the
North American digital brain.

322
00:13:23,880 --> 00:13:25,440
That usually bothers someone.

323
00:13:25,440 --> 00:13:26,940
And then you explain, you say:

324
00:13:26,940 --> 00:13:28,553
- That would do it.

325
00:13:28,553 --> 00:13:29,610
(everyone laughs)

326
00:13:29,610 --> 00:13:32,130
- You explain that compute capacity

327
00:13:32,130 --> 00:13:34,890
in Europe is roughly half the size

328
00:13:34,890 --> 00:13:37,560
of North American
installed compute capacity.

329
00:13:37,560 --> 00:13:41,850
And American businesses adopt
technology faster and quicker

330
00:13:41,850 --> 00:13:43,560
and have benefited from it

331
00:13:43,560 --> 00:13:46,440
more than European businesses have.

332
00:13:46,440 --> 00:13:49,200
And so, Europe sees that they

333
00:13:49,200 --> 00:13:51,540
fundamentally need to change the game,

334
00:13:51,540 --> 00:13:54,330
and that has multiple issues.

335
00:13:54,330 --> 00:13:56,493
One is cost of infrastructure,

336
00:13:57,390 --> 00:14:00,120
where Europe is still more
expensive on a variety of fronts,

337
00:14:00,120 --> 00:14:02,490
including raising capital and energy,

338
00:14:02,490 --> 00:14:06,180
two very important issues
for us data center operators.

339
00:14:06,180 --> 00:14:08,970
But the other one is actually making sure

340
00:14:08,970 --> 00:14:12,600
that the technology is
available and can be used.

341
00:14:12,600 --> 00:14:16,380
And in a world of geopolitical
issues, you realize

342
00:14:16,380 --> 00:14:21,380
that Europe is actually can't
depend upon remote technology.

343
00:14:21,720 --> 00:14:25,410
And then the next thing,
which has become a major,

344
00:14:25,410 --> 00:14:27,210
major issue, and we're
seeing it quite publicly,

345
00:14:27,210 --> 00:14:29,250
is this whole issue of energy.

346
00:14:29,250 --> 00:14:31,560
And we've been putting up our hand

347
00:14:31,560 --> 00:14:33,787
for over two years now, saying,

348
00:14:33,787 --> 00:14:37,950
"We are now being told
to wait seven to 10 years

349
00:14:37,950 --> 00:14:39,810
to connect to an electricity grid."

350
00:14:39,810 --> 00:14:42,450
So, that's private
industry taking its capital

351
00:14:42,450 --> 00:14:47,100
and investing that capital in
sustainable energy generation.

352
00:14:47,100 --> 00:14:49,620
And we can't connect our solar farms

353
00:14:49,620 --> 00:14:51,120
and our wind farms to the grid,

354
00:14:51,120 --> 00:14:53,490
and we can't connect our
data centers to the grid.

355
00:14:53,490 --> 00:14:57,450
So, we cannot deliver
sustainable IT in Europe

356
00:14:57,450 --> 00:14:59,137
because grids are basically saying,

357
00:14:59,137 --> 00:15:01,350
"Sorry, we're shut for business."

358
00:15:01,350 --> 00:15:02,520
And we've been talking about this

359
00:15:02,520 --> 00:15:05,100
with the European
Commission for a while now.

360
00:15:05,100 --> 00:15:08,370
And so, it's driving us as an industry

361
00:15:08,370 --> 00:15:12,840
to look at other markets besides
our primary five markets.

362
00:15:12,840 --> 00:15:16,050
And so, we're aggressively
pursuing growth strategies

363
00:15:16,050 --> 00:15:20,430
in the Nordics and in
Italy, Spain, Portugal.

364
00:15:20,430 --> 00:15:24,330
Poland is now coming up,
Switzerland and Belgium.

365
00:15:24,330 --> 00:15:27,000
So, all of these markets are
now becoming very interesting

366
00:15:27,000 --> 00:15:30,870
to us because they
represent access to energy.

367
00:15:30,870 --> 00:15:34,050
And, in a certain way,
that's actually good

368
00:15:34,050 --> 00:15:37,440
because we're bringing the
technology closer to the markets

369
00:15:37,440 --> 00:15:38,940
that are using that technology.

370
00:15:38,940 --> 00:15:43,170
- And I'd love to also talk
to you a little bit about that

371
00:15:43,170 --> 00:15:48,170
because building a data center in Norway

372
00:15:48,300 --> 00:15:49,710
is going to look very different

373
00:15:49,710 --> 00:15:53,310
from building a data
center in Spain, right?

374
00:15:53,310 --> 00:15:57,360
And we know in Europe, when
we like to create regulations,

375
00:15:57,360 --> 00:16:00,450
it's usually with the intents
of trying to standardize.

376
00:16:00,450 --> 00:16:01,541
- Yes.

377
00:16:01,541 --> 00:16:04,020
- What's your take on that
for the data center industry?

378
00:16:04,020 --> 00:16:05,070
Is it possible?

379
00:16:05,070 --> 00:16:09,120
- So, to be fair, in a certain sense,

380
00:16:09,120 --> 00:16:12,270
this is a yes and a no type
answer to your question.

381
00:16:12,270 --> 00:16:13,320
At a very high level,

382
00:16:13,320 --> 00:16:16,740
there are some basic concepts
that we can standardize,

383
00:16:16,740 --> 00:16:18,480
but you are absolutely right

384
00:16:18,480 --> 00:16:21,480
that there are going to
be specific differences

385
00:16:21,480 --> 00:16:25,110
in geographies in that you
pick certain design criteria

386
00:16:25,110 --> 00:16:27,330
at the very highest of level:

387
00:16:27,330 --> 00:16:28,773
construction materials,

388
00:16:30,000 --> 00:16:33,873
levels of redundancy in mechanical
and electrical equipment,

389
00:16:34,860 --> 00:16:36,570
cooling methodologies,

390
00:16:36,570 --> 00:16:39,750
such as adopting free-air
cooling and variable cooling.

391
00:16:39,750 --> 00:16:42,150
So, a lot of the basics in how you run

392
00:16:42,150 --> 00:16:44,550
a data center can be standardized,

393
00:16:44,550 --> 00:16:48,900
and then you adapt to the
geographical situation.

394
00:16:48,900 --> 00:16:50,670
When we run data centers in Dubai,

395
00:16:50,670 --> 00:16:52,770
we spend a lot of time humidifying

396
00:16:52,770 --> 00:16:54,360
because it doesn't matter

397
00:16:54,360 --> 00:16:55,824
how your air conditioning's working

398
00:16:55,824 --> 00:16:58,200
if humidity levels are zero.

399
00:16:58,200 --> 00:16:59,790
And that actually is the same problem

400
00:16:59,790 --> 00:17:01,110
in Norway in the wintertime.

401
00:17:01,110 --> 00:17:02,520
And then the other side is in,

402
00:17:02,520 --> 00:17:04,410
we are trying to adopt models

403
00:17:04,410 --> 00:17:09,090
where we transfer our heat to a demand.

404
00:17:09,090 --> 00:17:13,650
And we typically see that demand drops

405
00:17:13,650 --> 00:17:15,210
as you get into warmer climates.

406
00:17:15,210 --> 00:17:18,510
However, one of our members
just announced a partnership

407
00:17:18,510 --> 00:17:21,960
with a heat network that's
being built in Milan.

408
00:17:21,960 --> 00:17:24,240
And so, there is going
to be an exchange of heat

409
00:17:24,240 --> 00:17:27,300
directly to a heat network,
which benefits both

410
00:17:27,300 --> 00:17:29,430
because not only are we obsessed

411
00:17:29,430 --> 00:17:31,410
with becoming carbon neutral by 2030,

412
00:17:31,410 --> 00:17:33,150
these 100-year-old heat networks

413
00:17:33,150 --> 00:17:35,490
have to decarbonize themselves between now

414
00:17:35,490 --> 00:17:37,560
and 2040, something like that.

415
00:17:37,560 --> 00:17:40,500
So, they're also desperately
seeking new ways to get heat.

416
00:17:40,500 --> 00:17:43,230
- Well, we've spent a bit of
time talking about heating,

417
00:17:43,230 --> 00:17:46,320
but this heat only
exists because, in fact,

418
00:17:46,320 --> 00:17:47,580
we have a need to cool.

419
00:17:47,580 --> 00:17:49,920
I think, specifically in
the data center space,

420
00:17:49,920 --> 00:17:51,930
the speed of innovation has been unlike

421
00:17:51,930 --> 00:17:53,190
what I've ever seen, right?

422
00:17:53,190 --> 00:17:55,020
Even in the built industry.

423
00:17:55,020 --> 00:17:57,270
And it also means we've ended up

424
00:17:57,270 --> 00:18:00,660
with a very wide and complex range

425
00:18:00,660 --> 00:18:02,670
of different cooling solutions

426
00:18:02,670 --> 00:18:04,830
that data centers can use today.

427
00:18:04,830 --> 00:18:06,000
What's your take on this,

428
00:18:06,000 --> 00:18:08,340
and how do you see this playing out?

429
00:18:08,340 --> 00:18:10,890
- I think some people think
of data centers being built

430
00:18:10,890 --> 00:18:14,040
for liquid cooling versus data centers

431
00:18:14,040 --> 00:18:15,810
being built with air cooling.

432
00:18:15,810 --> 00:18:20,810
I think, in most cases,
it's going to be a hybrid,

433
00:18:21,210 --> 00:18:23,670
and data centers are
going to have to build

434
00:18:23,670 --> 00:18:26,220
with both circuits available,

435
00:18:26,220 --> 00:18:28,020
and we see this happening.

436
00:18:28,020 --> 00:18:31,140
And in most cases, we are
actually able to share

437
00:18:31,140 --> 00:18:35,040
a single cooling circuit and break out

438
00:18:35,040 --> 00:18:38,490
both liquid and air cooling
systems on the same loop.

439
00:18:38,490 --> 00:18:40,770
That's the dream that
we're moving towards,

440
00:18:40,770 --> 00:18:43,080
and many of us are
heading in that direction.

441
00:18:43,080 --> 00:18:44,580
- I think it's really interesting

442
00:18:44,580 --> 00:18:48,660
that we talked a lot
about standardization,

443
00:18:48,660 --> 00:18:50,730
and then our conversation sort of morphed

444
00:18:50,730 --> 00:18:52,890
into the need for flexibility,

445
00:18:52,890 --> 00:18:56,910
but these aren't different
concepts that can coexist.

446
00:18:56,910 --> 00:18:58,890
- The standardization is we all agree

447
00:18:58,890 --> 00:19:00,840
the need to do a hybrid data center,

448
00:19:00,840 --> 00:19:03,000
and we all agree that there
are two or three ways to do it.

449
00:19:03,000 --> 00:19:06,457
And then we engage the suppliers
in our industry, saying,

450
00:19:06,457 --> 00:19:08,760
"We may build this way,
option A, option B,

451
00:19:08,760 --> 00:19:12,300
or option C, depending
on cost and location."

452
00:19:12,300 --> 00:19:14,190
And so, those vendors are taking

453
00:19:14,190 --> 00:19:16,800
that standardized optionality,

454
00:19:16,800 --> 00:19:17,940
if you want to call it that,

455
00:19:17,940 --> 00:19:19,500
and they're designing their products

456
00:19:19,500 --> 00:19:21,210
to fit that, which is great.

457
00:19:21,210 --> 00:19:24,990
That's a novel way of us working together,

458
00:19:24,990 --> 00:19:26,310
which we didn't in the past.

459
00:19:26,310 --> 00:19:30,150
In the past, our cycles of
innovation were very slow

460
00:19:30,150 --> 00:19:33,780
because we weren't sharing
information across each other

461
00:19:33,780 --> 00:19:37,020
at the supplier level or
at the customer level.

462
00:19:37,020 --> 00:19:38,760
And even the engineering companies

463
00:19:38,760 --> 00:19:40,500
that were in the middle doing the design

464
00:19:40,500 --> 00:19:42,480
and the build weren't collating

465
00:19:42,480 --> 00:19:44,310
the information and sharing it.

466
00:19:44,310 --> 00:19:46,260
So, we're accelerating the design cycle

467
00:19:46,260 --> 00:19:49,260
through this idea of
allowing for optionality,

468
00:19:49,260 --> 00:19:51,060
but quickly picking best practice

469
00:19:51,060 --> 00:19:52,620
and then adopting it across the board.

470
00:19:52,620 --> 00:19:56,100
- And hopefully, this new
model of collaboration

471
00:19:56,100 --> 00:19:57,390
and partnership will inspire

472
00:19:57,390 --> 00:19:59,130
other industries as well, right?

473
00:19:59,130 --> 00:20:01,830
The could be very fragmented.
- So, you say "hope";

474
00:20:01,830 --> 00:20:02,790
this is real.

475
00:20:02,790 --> 00:20:06,240
And in fact, we get a lot
of stick in the press,

476
00:20:06,240 --> 00:20:09,270
but one of the major good
news stories that's happening

477
00:20:09,270 --> 00:20:11,910
is that where we are innovating

478
00:20:11,910 --> 00:20:15,783
in areas of mechanical, electrical
and heating and cooling,

479
00:20:16,980 --> 00:20:21,300
we are forcing our vendors
to adopt new technologies,

480
00:20:21,300 --> 00:20:25,500
and we are effectively funding
the research and development

481
00:20:25,500 --> 00:20:27,810
of these products on our backs,

482
00:20:27,810 --> 00:20:30,780
which they then incorporate
into technologies

483
00:20:30,780 --> 00:20:31,980
that they take across industry.

484
00:20:31,980 --> 00:20:33,840
And this has already happened.

485
00:20:33,840 --> 00:20:37,500
- Now, Michael, I'm going
to appeal a little bit here

486
00:20:37,500 --> 00:20:40,380
to your sense of vision, right?

487
00:20:40,380 --> 00:20:42,330
And as someone who has
been in the industry

488
00:20:42,330 --> 00:20:44,940
for quite a few years now,

489
00:20:44,940 --> 00:20:47,970
if you have to imagine
what's going to happen,

490
00:20:47,970 --> 00:20:51,030
especially in terms of
data center capacity

491
00:20:51,030 --> 00:20:53,010
in the next seven years,

492
00:20:53,010 --> 00:20:54,120
what would be your best guess?

493
00:20:54,120 --> 00:20:55,620
Where do you see this is going?

494
00:20:55,620 --> 00:20:59,040
- So, luckily, it's not
just reading tea leaves,

495
00:20:59,040 --> 00:21:00,450
though I am drinking a cup of tea.

496
00:21:00,450 --> 00:21:01,764
So, if you want, I can do that.

497
00:21:01,764 --> 00:21:02,597
It's convenient.

498
00:21:02,597 --> 00:21:06,330
No, we are running an
annual research report.

499
00:21:06,330 --> 00:21:10,080
What the data basically says
is that from two years ago

500
00:21:10,080 --> 00:21:12,210
to this last year's report,

501
00:21:12,210 --> 00:21:14,460
we saw money being invested in the ground,

502
00:21:14,460 --> 00:21:16,410
not just some of the high-level stuff

503
00:21:16,410 --> 00:21:18,390
of "I'm going to build a
five-gigawatt something-or-other,"

504
00:21:18,390 --> 00:21:21,570
but real money announced being deployed.

505
00:21:21,570 --> 00:21:24,690
We saw that amount of
capital double in 12 months

506
00:21:24,690 --> 00:21:28,230
from about 80 billion euros
to about 170 billion euros.

507
00:21:28,230 --> 00:21:30,750
And so, I think there was
this real clear signal

508
00:21:30,750 --> 00:21:34,290
to the world that Europe is
ready to invest in technology.

509
00:21:34,290 --> 00:21:36,600
And if you wanted to sell in Europe,

510
00:21:36,600 --> 00:21:38,670
you better be here in Europe.

511
00:21:38,670 --> 00:21:39,660
- All right.

512
00:21:39,660 --> 00:21:41,340
Well, Michael, since you
mentioned at the beginning

513
00:21:41,340 --> 00:21:43,500
how much you thoroughly enjoy

514
00:21:43,500 --> 00:21:46,050
influencing policymakers and regulators-

515
00:21:46,050 --> 00:21:47,500
- Yes.
- I want to give you

516
00:21:48,346 --> 00:21:49,179
the platform one more time

517
00:21:49,179 --> 00:21:54,179
and ask you, what is that
one thing you just wish

518
00:21:54,780 --> 00:21:56,970
that policymakers understood?

519
00:21:56,970 --> 00:21:58,470
What would make your job easier?

520
00:21:58,470 --> 00:22:01,350
- I think policymakers understand this,

521
00:22:01,350 --> 00:22:06,350
but they are currently being pushed to act

522
00:22:06,990 --> 00:22:10,290
by, I'm going to say,

523
00:22:10,290 --> 00:22:14,880
a journalistic story that's
not fundamentally true.

524
00:22:14,880 --> 00:22:17,910
So, the whole thing today is that somehow

525
00:22:17,910 --> 00:22:20,370
data centers are taking
all the electricity

526
00:22:20,370 --> 00:22:22,800
and raising energy prices, oh,

527
00:22:22,800 --> 00:22:25,290
and that we're taking all the water.

528
00:22:25,290 --> 00:22:28,530
The reality is our innovation
in cooling technologies

529
00:22:28,530 --> 00:22:31,350
is such that we actually
take very little water,

530
00:22:31,350 --> 00:22:33,060
less than golf courses,

531
00:22:33,060 --> 00:22:35,550
and we reuse that water,
we recycle that water.

532
00:22:35,550 --> 00:22:37,020
In the case of electricity,

533
00:22:37,020 --> 00:22:39,570
we represent a huge growth opportunity.

534
00:22:39,570 --> 00:22:42,210
We're not the cause of a grid problem.

535
00:22:42,210 --> 00:22:45,870
We are a symptom of an
underinvestment in the grid.

536
00:22:45,870 --> 00:22:50,370
We represent less than 9% of the increase

537
00:22:50,370 --> 00:22:52,830
on the grid that needs to happen

538
00:22:52,830 --> 00:22:54,540
for all of transportation,

539
00:22:54,540 --> 00:22:56,160
all of heating and cooling,

540
00:22:56,160 --> 00:22:57,930
heavy industry, light industry,

541
00:22:57,930 --> 00:22:59,310
three times larger each.

542
00:22:59,310 --> 00:23:02,280
So, we're not the cause of a problem.

543
00:23:02,280 --> 00:23:05,880
We are showing that a underinvestment
in grids over the past

544
00:23:05,880 --> 00:23:09,930
40 or 50 years is going to
fundamentally make it impossible

545
00:23:09,930 --> 00:23:12,150
for regulators to
achieve their Green Deal.

546
00:23:12,150 --> 00:23:15,570
They need to be making a
radical investment in grids.

547
00:23:15,570 --> 00:23:17,250
And I understand it's hard.

548
00:23:17,250 --> 00:23:18,960
These grids are regulated.

549
00:23:18,960 --> 00:23:20,400
It's hard to raise capital.

550
00:23:20,400 --> 00:23:22,890
It's hard to deploy the capital at pace,

551
00:23:22,890 --> 00:23:26,520
but the speed with which these
grids need to be grown to,

552
00:23:26,520 --> 00:23:28,540
not just for us, but for everybody,

553
00:23:28,540 --> 00:23:31,380
is two or three times faster

554
00:23:31,380 --> 00:23:33,450
than what the grids are
currently forecasting.

555
00:23:33,450 --> 00:23:35,850
- You wish that policymakers
would understand

556
00:23:35,850 --> 00:23:40,410
the urgent, critical
need to invest even more

557
00:23:40,410 --> 00:23:44,700
in our grid capacity, our grid resilience:

558
00:23:44,700 --> 00:23:45,533
- Absolutely.

559
00:23:45,533 --> 00:23:49,500
- Rather than trying to stall
the developments of industries

560
00:23:49,500 --> 00:23:51,150
that are contributing to the growth.

561
00:23:51,150 --> 00:23:53,820
- Yes, because currently
we are being asked

562
00:23:53,820 --> 00:23:56,310
to swallow more cost, which will slow down

563
00:23:56,310 --> 00:23:58,530
innovation, slow down investment.

564
00:23:58,530 --> 00:24:03,360
So, yes, it's either you want
this Green Deal and you want

565
00:24:03,360 --> 00:24:07,740
Europe to have a competitive
sustainable market,

566
00:24:07,740 --> 00:24:10,410
or you are going to be unwilling to do it

567
00:24:10,410 --> 00:24:12,930
to the timeframe that
because you're not willing

568
00:24:12,930 --> 00:24:14,190
to invest in these grids.

569
00:24:14,190 --> 00:24:15,023
We're here to help.

570
00:24:15,023 --> 00:24:15,856
We have capital.

571
00:24:15,856 --> 00:24:16,890
We have technology.

572
00:24:16,890 --> 00:24:19,200
We are willing to take some of the risk,

573
00:24:19,200 --> 00:24:21,360
but this is an all-of-society problem.

574
00:24:21,360 --> 00:24:22,661
It's not a data center problem.

575
00:24:22,661 --> 00:24:23,760
(upbeat music)

576
00:24:23,760 --> 00:24:25,020
- You know what really struck me

577
00:24:25,020 --> 00:24:30,020
about Michael's sort of reframe
of the data center situation

578
00:24:30,090 --> 00:24:32,910
is that, given all the pushback,

579
00:24:32,910 --> 00:24:35,400
there's actually a lot
of opportunity here.

580
00:24:35,400 --> 00:24:36,600
There's a lot of pressure.

581
00:24:36,600 --> 00:24:38,310
I mean, we have the speed pressure,

582
00:24:38,310 --> 00:24:41,520
we have the scaling,
tripling, for instance.

583
00:24:41,520 --> 00:24:43,410
We have the demand for AI,

584
00:24:43,410 --> 00:24:46,050
and there's also communities
that have expectations.

585
00:24:46,050 --> 00:24:47,400
And what that's really doing

586
00:24:47,400 --> 00:24:50,520
is forcing some true innovation.

587
00:24:50,520 --> 00:24:54,990
Innovation in cooling and how
we reuse and think about heat,

588
00:24:54,990 --> 00:24:57,720
how we improve the grid,
all of this simultaneously.

589
00:24:57,720 --> 00:24:59,550
I mean, the interesting thing here

590
00:24:59,550 --> 00:25:01,800
is that solving the data center problem

591
00:25:01,800 --> 00:25:05,820
will help us solve some
other sustainability issues.

592
00:25:05,820 --> 00:25:08,280
That gives us a gain, an overall gain.

593
00:25:08,280 --> 00:25:11,670
These innovations can be
applied to other areas

594
00:25:11,670 --> 00:25:13,380
of our economy, of the world.

595
00:25:13,380 --> 00:25:17,550
Things like hospitals, factories,
schools all stand to gain

596
00:25:17,550 --> 00:25:18,900
because of the innovations

597
00:25:18,900 --> 00:25:20,430
that we're in the middle of right now

598
00:25:20,430 --> 00:25:21,870
because of data centers.

599
00:25:21,870 --> 00:25:23,970
- So, both Michael and Oakley

600
00:25:23,970 --> 00:25:26,460
actually landed on the same point.

601
00:25:26,460 --> 00:25:30,270
The answer is all about making
it a hybrid of technologies.

602
00:25:30,270 --> 00:25:32,910
There's no such thing
as one-size-fits-all.

603
00:25:32,910 --> 00:25:35,790
It's also about designing smarter systems,

604
00:25:35,790 --> 00:25:38,160
spoiler alert for the next episode.

605
00:25:38,160 --> 00:25:41,130
It's an opportunity to design
intelligent AI controls

606
00:25:41,130 --> 00:25:43,680
that respond to requirements in real time,

607
00:25:43,680 --> 00:25:46,590
and this opportunity is huge.

608
00:25:46,590 --> 00:25:48,870
To be fairly honest
with you, I don't think

609
00:25:48,870 --> 00:25:52,080
there's ever been a more
exciting time to be an engineer.

610
00:25:52,080 --> 00:25:54,540
And on that thought,
let's go back to Oakley

611
00:25:54,540 --> 00:25:57,600
to hear what speed really
means for his teams.

612
00:25:57,600 --> 00:25:59,310
And that includes engineers,

613
00:25:59,310 --> 00:26:02,737
but also customers, that
whole data center ecosystem.

614
00:26:02,737 --> 00:26:03,570
(graphics swoosh)

615
00:26:03,570 --> 00:26:05,220
- It's an AI race.

616
00:26:05,220 --> 00:26:06,450
You hear that a lot.

617
00:26:06,450 --> 00:26:08,480
People say that, and it is a race.

618
00:26:08,480 --> 00:26:12,090
The race is about speed to deployment,

619
00:26:12,090 --> 00:26:14,850
and the race is to get to the finish line

620
00:26:14,850 --> 00:26:16,950
because at the finish line

621
00:26:16,950 --> 00:26:21,420
is a tremendous demand from all of us,

622
00:26:21,420 --> 00:26:26,190
humans, businesses, everyone
that's demanding access to AI.

623
00:26:26,190 --> 00:26:30,990
And so, that race is about
speed and speed to deployment.

624
00:26:30,990 --> 00:26:34,260
And how you get there is a big driver

625
00:26:34,260 --> 00:26:37,590
of how we think about system design.

626
00:26:37,590 --> 00:26:39,270
And the way that you take all those things

627
00:26:39,270 --> 00:26:40,860
and actually make it happen faster

628
00:26:40,860 --> 00:26:43,230
is through things like
modular manufacturing.

629
00:26:43,230 --> 00:26:45,660
Recently, we acquired a
company called Stellar Energy,

630
00:26:45,660 --> 00:26:47,340
which is a company that's a leader

631
00:26:47,340 --> 00:26:49,800
in modular manufacturing for data centers.

632
00:26:49,800 --> 00:26:51,810
So, you're able to package
all this technology

633
00:26:51,810 --> 00:26:52,927
instead of thinking through, like,

634
00:26:52,927 --> 00:26:56,580
"Well, how do I deploy that
in a construction project

635
00:26:56,580 --> 00:26:59,730
onsite in Texas or onsite in Michigan

636
00:26:59,730 --> 00:27:01,710
or onsite in Virginia,"

637
00:27:01,710 --> 00:27:05,130
which are very different environments.

638
00:27:05,130 --> 00:27:07,770
And when you move it into a
modular manufacturing approach,

639
00:27:07,770 --> 00:27:09,600
which means that when
you're designing the system,

640
00:27:09,600 --> 00:27:11,650
you're actually designing it to be built

641
00:27:12,540 --> 00:27:15,300
in a controlled environment

642
00:27:15,300 --> 00:27:19,530
that can be replicated
consistently over time and fast.

643
00:27:19,530 --> 00:27:21,000
On the thermal side, we take

644
00:27:21,000 --> 00:27:22,920
all that system engineering expertise.

645
00:27:22,920 --> 00:27:26,220
We're able to put it into
a package that we can tweak

646
00:27:26,220 --> 00:27:28,080
and nudge for these different sites

647
00:27:28,080 --> 00:27:31,140
and then ship those modules
to the data center site

648
00:27:31,140 --> 00:27:33,090
and then put them together
and get them running.

649
00:27:33,090 --> 00:27:35,910
And it creates a differentiated
speed to deployment.

650
00:27:35,910 --> 00:27:39,420
- It's really important,
not only tackling this issue

651
00:27:39,420 --> 00:27:40,417
of speed of deployment,

652
00:27:40,417 --> 00:27:42,300
"We got to go faster,"

653
00:27:42,300 --> 00:27:45,210
but I'm guessing, and correct
me if I'm wrong, Oakley,

654
00:27:45,210 --> 00:27:49,200
it also sounds like it's much
more iterative as well, right?

655
00:27:49,200 --> 00:27:51,540
- There's a convergence of
a couple of these topics,

656
00:27:51,540 --> 00:27:52,797
and the future's a little bit closer to us

657
00:27:52,797 --> 00:27:56,046
than in this data center
industry because of the speed.

658
00:27:56,046 --> 00:27:58,290
It's also about, "How do we
get speed to deployment?"

659
00:27:58,290 --> 00:27:59,730
Things have to be iterative.

660
00:27:59,730 --> 00:28:03,630
You have innovation around the chip sets

661
00:28:03,630 --> 00:28:06,727
and the silicon that's coming
out within the timeline of,

662
00:28:06,727 --> 00:28:09,740
"Hey, I've located and identified a site

663
00:28:09,740 --> 00:28:11,610
that I'm going to develop
into a data center."

664
00:28:11,610 --> 00:28:13,500
If that's kind of like day zero

665
00:28:13,500 --> 00:28:15,420
and the finish line is the data center

666
00:28:15,420 --> 00:28:17,673
is running and operating and producing,

667
00:28:18,810 --> 00:28:22,560
what happens in between those
two dates is very dynamic

668
00:28:22,560 --> 00:28:25,920
because the technology can
change within those dates.

669
00:28:25,920 --> 00:28:28,320
The regulations can
change within those dates.

670
00:28:28,320 --> 00:28:29,820
The way the community wants to work

671
00:28:29,820 --> 00:28:31,470
with the data center can change.

672
00:28:31,470 --> 00:28:36,000
So, there's so many things
changing within what used to be,

673
00:28:36,000 --> 00:28:37,590
like, that was very rigid.

674
00:28:37,590 --> 00:28:39,420
In between the start and
finish, that's pretty fixed.

675
00:28:39,420 --> 00:28:40,440
Like, things don't change, right?

676
00:28:40,440 --> 00:28:43,350
Things change outside
of those two bookends.

677
00:28:43,350 --> 00:28:44,520
And that's not the case anymore.

678
00:28:44,520 --> 00:28:48,180
So, that flexibility is so important now,

679
00:28:48,180 --> 00:28:49,650
and it comes back to, like,

680
00:28:49,650 --> 00:28:51,990
we're talking about modular manufacturing.

681
00:28:51,990 --> 00:28:56,970
Well, if you've designed
a system to be modular,

682
00:28:56,970 --> 00:28:58,470
if something changes,

683
00:28:58,470 --> 00:29:01,890
it's much easier to tweak, nudge, adjust

684
00:29:01,890 --> 00:29:06,240
than if you're trying to do a
one-off construction project.

685
00:29:06,240 --> 00:29:07,770
It's a different way of, you know,

686
00:29:07,770 --> 00:29:09,000
to modify and change things.

687
00:29:09,000 --> 00:29:12,240
So, innovation pace is so fast here

688
00:29:12,240 --> 00:29:14,200
that you have to be super engaged in it,

689
00:29:14,200 --> 00:29:15,900
but you have to be very flexible

690
00:29:15,900 --> 00:29:17,760
within that, even within that process,

691
00:29:17,760 --> 00:29:19,440
because the future actually arise

692
00:29:19,440 --> 00:29:20,973
inside of those two bookends.

693
00:29:21,900 --> 00:29:23,910
- Okay, that is a wrap.

694
00:29:23,910 --> 00:29:25,890
And I love where this landed.

695
00:29:25,890 --> 00:29:29,490
The era of the standard
data center really is over,

696
00:29:29,490 --> 00:29:33,060
and what replaces it is
engineering flexibility.

697
00:29:33,060 --> 00:29:35,070
It's consistent enough to repeat.

698
00:29:35,070 --> 00:29:39,810
It's modular enough to adapt
wherever it's being built.

699
00:29:39,810 --> 00:29:42,210
And the part that excites me,

700
00:29:42,210 --> 00:29:47,130
doing that fast is also our
chance to do it sustainably.

701
00:29:47,130 --> 00:29:49,260
We're not waiting for
the future to arrive.

702
00:29:49,260 --> 00:29:53,760
We're building it in every
climate alongside our customers.

703
00:29:53,760 --> 00:29:56,130
I want to say a big
thank you to our guests,

704
00:29:56,130 --> 00:29:58,710
Oakley Roberts and Michael Winterson.

705
00:29:58,710 --> 00:30:00,120
- And thank you for listening.

706
00:30:00,120 --> 00:30:02,640
This has been "Healthy
Spaces" with me, Scott Tew,

707
00:30:02,640 --> 00:30:04,740
and my co-host, Dominique Silva.

708
00:30:04,740 --> 00:30:06,900
We're back in two weeks
with another episode,

709
00:30:06,900 --> 00:30:10,020
so please be sure to like and subscribe

710
00:30:10,020 --> 00:30:11,140
so you don't miss out.

711
00:30:11,140 --> 00:30:13,723
(upbeat music)