1
00:00:02,069 --> 00:00:06,473
is the sixth Democrat to
announce a run for governor

2
00:00:06,540 --> 00:00:10,777
with 43 and counting. Data
processing centers are

3
00:00:10,844 --> 00:00:13,747
popping up all over
Wisconsin. Like the

4
00:00:13,814 --> 00:00:16,149
Microsoft Data Center in
Mount Pleasant, where the

5
00:00:16,216 --> 00:00:20,087
tech giant is developing a
100 acre site. After buying

6
00:00:20,153 --> 00:00:24,892
it from Foxconn. Data
centers contain servers and

7
00:00:24,958 --> 00:00:27,861
drives to store, process,
and disseminate data.

8
00:00:27,928 --> 00:00:31,665
Companies like Microsoft,
Amazon, Google and Meta

9
00:00:31,732 --> 00:00:35,035
operate large scale data
centers. The Great Lakes

10
00:00:35,102 --> 00:00:37,905
region is a popular site
for them because of the

11
00:00:37,971 --> 00:00:41,108
large amounts of water.
Data centers need to cool

12
00:00:41,175 --> 00:00:43,911
the equipment. But what
about the electricity

13
00:00:43,977 --> 00:00:46,079
needed to run them? What
are the implications of

14
00:00:46,146 --> 00:00:51,185
that usage? We turn to Line
Roald, professor in the

15
00:00:51,251 --> 00:00:53,387
UW-Madison Department of
Electrical and Computer

16
00:00:53,453 --> 00:00:55,856
Engineering. And thanks for
being here.

17
00:00:55,923 --> 00:00:59,860
>> Thank you for having me.
>> So we know that

18
00:00:59,927 --> 00:01:02,563
everybody talks about the
millions of gallons of

19
00:01:02,629 --> 00:01:06,233
water that are needed to to
cool these data centers,

20
00:01:06,300 --> 00:01:08,902
but how much electricity do
they need to do all that

21
00:01:08,969 --> 00:01:11,538
processing like this.
Microsoft. 1 and Mt.

22
00:01:11,605 --> 00:01:15,175
Pleasant.
>> Yeah. So it varies quite

23
00:01:15,242 --> 00:01:18,979
a bit by data center and
the type of data center.

24
00:01:19,046 --> 00:01:21,014
But the Mt. Pleasant data
center I think, is supposed

25
00:01:21,081 --> 00:01:25,886
to use about 450MW of power.
And that is a lot. That's

26
00:01:27,054 --> 00:01:29,723
about the same amount as
all the households in

27
00:01:29,790 --> 00:01:33,026
Milwaukee together.
>> And so if you if you

28
00:01:33,093 --> 00:01:35,796
talk about 43 data centers,
obviously not all is large

29
00:01:35,863 --> 00:01:38,398
as that, but across the
state of Wisconsin, that's

30
00:01:38,465 --> 00:01:41,935
sucking a lot of power.
>> Yes it is. Yeah. And the

31
00:01:42,002 --> 00:01:43,904
projection is moving
forward with even more

32
00:01:43,971 --> 00:01:46,240
particularly of these very
large data centers coming,

33
00:01:46,306 --> 00:01:50,143
is that it will consume
really a lot of power that

34
00:01:50,210 --> 00:01:52,913
is going to change, really
the scope of use of

35
00:01:52,980 --> 00:01:56,116
electricity in Wisconsin.
>> So this growth of these

36
00:01:56,183 --> 00:01:58,752
centers seems explosive,
right. And all of a sudden,

37
00:01:58,819 --> 00:02:01,622
have you been surprised by
it?

38
00:02:03,457 --> 00:02:06,426
on data centers and their
energy use since about 2018.

39
00:02:06,493 --> 00:02:09,796
And back then it was
already clear that we are

40
00:02:09,863 --> 00:02:13,133
likely going to see an
increase in energy use from

41
00:02:13,200 --> 00:02:15,802
data centers just because
it's getting harder. It was

42
00:02:15,869 --> 00:02:19,306
getting harder to keep up
the efficiency gains that

43
00:02:19,373 --> 00:02:22,075
have allowed us to do ever
more computing with

44
00:02:22,142 --> 00:02:24,745
essentially kind of the
same amount of electricity.

45
00:02:24,811 --> 00:02:27,581
So for a long time,
electricity use was fairly

46
00:02:27,648 --> 00:02:30,350
flat. But then now. So you
were expecting that there

47
00:02:30,417 --> 00:02:32,819
would be an increase. But
of course, with AI and

48
00:02:32,886 --> 00:02:36,056
these large language models
and these very, very large

49
00:02:36,123 --> 00:02:39,159
models that need really,
really large data centers

50
00:02:39,226 --> 00:02:42,763
for training and inference,
it's really been quite

51
00:02:42,829 --> 00:02:45,265
explosive, I would say. So
it's a bit unprecedented

52
00:02:45,332 --> 00:02:48,001
for the grid to see this
amount of growth in this

53
00:02:48,068 --> 00:02:51,538
short time period.
>> In Wisconsin. Do we have

54
00:02:51,605 --> 00:02:55,509
the capacity to provide the
electricity?

55
00:02:55,976 --> 00:02:57,978
are hundreds and hundreds
of megawatt of power that

56
00:02:58,045 --> 00:03:00,814
is just kind of laying
around being, waiting,

57
00:03:00,881 --> 00:03:04,051
waiting to be used. And so
we don't necessarily have

58
00:03:04,117 --> 00:03:08,055
the capacity to meet all
that demand as it is today.

59
00:03:08,121 --> 00:03:11,558
And so we are going to need
new infrastructure, new

60
00:03:11,625 --> 00:03:14,094
transmission lines, new
power plants in order to

61
00:03:14,161 --> 00:03:15,896
meet that demand, because
it's also not like the

62
00:03:15,963 --> 00:03:18,332
regions around us
necessarily have that

63
00:03:18,398 --> 00:03:21,735
energy there.
>> So what about the cost

64
00:03:21,802 --> 00:03:25,239
of this energy demand? And
then that infrastructure of

65
00:03:25,305 --> 00:03:27,908
which you speak, does that
get passed along to the

66
00:03:27,975 --> 00:03:30,911
consumer? Like am I going
to see it in my electric

67
00:03:30,978 --> 00:03:33,881
bill?
How much of it and how that

68
00:03:35,182 --> 00:03:37,084
cost will be spread. So you
can imagine that when more

69
00:03:37,150 --> 00:03:40,554
people want something, the
price sometimes tend to go

70
00:03:40,621 --> 00:03:44,091
up, because there's maybe a
limit to the supply of that

71
00:03:44,157 --> 00:03:48,028
of electricity, but also
the way utilities do things

72
00:03:48,095 --> 00:03:50,831
is that they build
infrastructure to meet the

73
00:03:50,898 --> 00:03:54,668
demand of their customers,
and they have to file what

74
00:03:54,735 --> 00:03:57,204
is called a rate case with
the Public Service

75
00:03:57,271 --> 00:03:59,973
Commission in order to
determine how that cost is

76
00:04:00,040 --> 00:04:03,777
being covered. So they have
to ask for permission to

77
00:04:03,844 --> 00:04:06,013
spend that money to build
that new infrastructure,

78
00:04:06,079 --> 00:04:08,982
and they have to talk about
how they are going to

79
00:04:09,049 --> 00:04:11,318
consumers. And there's a
lot of open questions in

80
00:04:11,385 --> 00:04:14,955
exactly how that is going
to happen. And so I think

81
00:04:15,022 --> 00:04:16,924
both in Wisconsin but also
elsewhere in the country,

82
00:04:16,990 --> 00:04:21,495
there's a lot of questions
of how this might impact

83
00:04:21,562 --> 00:04:24,198
consumer bills. And frankly,
to be to to be honest with

84
00:04:24,264 --> 00:04:27,901
you, I am not entirely sure
where where the Public

85
00:04:27,968 --> 00:04:30,337
Service Commission, for
example, will land with

86
00:04:30,404 --> 00:04:33,040
that, how the utilities
will do things, and also

87
00:04:33,106 --> 00:04:34,842
what kind of contracts
these companies are able to

88
00:04:34,908 --> 00:04:37,044
get with the utilities. A
lot of that will determine

89
00:04:37,110 --> 00:04:39,847
how energy bills will or
might change.

90
00:04:39,913 --> 00:04:43,016
>> What would happen in
peak times like summer. Can

91
00:04:43,083 --> 00:04:46,987
the grid in Wisconsin
manage that load, and are

92
00:04:47,054 --> 00:04:49,590
there concerns about
blackouts or that kind of

93
00:04:49,656 --> 00:04:52,993
thing?
during peak times in summer,

94
00:04:54,294 --> 00:04:56,697
it can be challenging to
meet the demand as it is

95
00:04:56,763 --> 00:05:00,934
today. So we have seen in
past years that there has

96
00:05:01,001 --> 00:05:05,906
been warnings by Nordyke,
who is the regulatory body

97
00:05:06,773 --> 00:05:10,043
that has been saying that,
okay, there could be

98
00:05:10,110 --> 00:05:14,014
challenges meeting peak
during summer peak hours.

99
00:05:14,081 --> 00:05:16,984
And so what we might be
seeing is that there is a

100
00:05:17,050 --> 00:05:19,686
need for more demand
response. I mean, you might

101
00:05:19,753 --> 00:05:22,022
be familiar with, for
example, energy rush hours

102
00:05:22,089 --> 00:05:24,825
where the utility says,
okay, I'm going to control

103
00:05:24,892 --> 00:05:27,160
the thermostat in your home
and make it a little bit

104
00:05:27,227 --> 00:05:29,897
warmer during those peak
hours to reduce the amount

105
00:05:29,963 --> 00:05:33,100
of electricity that we need
in order to cool the house.

106
00:05:33,166 --> 00:05:37,171
During summer when it's
very warm. And so of course,

107
00:05:37,237 --> 00:05:39,106
this happens. If it happens,
it happens at a household

108
00:05:39,173 --> 00:05:41,775
level. There are very small
changes in energy demand,

109
00:05:41,842 --> 00:05:45,078
and the effect comes from
many, many people adjusting

110
00:05:45,145 --> 00:05:48,248
their temperatures at the
same time. So we could see

111
00:05:48,315 --> 00:05:51,084
more of that. But we could
also see that data centers

112
00:05:51,151 --> 00:05:53,086
actively participate in
this and that they are able

113
00:05:53,153 --> 00:05:57,090
to reduce their energy
demand during those peak

114
00:05:57,157 --> 00:06:00,027
to do so, we are going to
need less infrastructure.

115
00:06:00,093 --> 00:06:02,162
There would be less stress
on the grid, and there

116
00:06:02,229 --> 00:06:06,867
would be a lower chance
that there would be large

117
00:06:06,934 --> 00:06:10,304
Blackouts, typically.
Hopefully we will not see

118
00:06:10,370 --> 00:06:13,073
that because utilities are
quite good at predicting

119
00:06:13,140 --> 00:06:16,677
energy demand and
particularly large data

120
00:06:16,743 --> 00:06:19,179
possible to know a little
bit in advance how much

121
00:06:19,246 --> 00:06:22,816
they will use, so hopefully
we would not get there. But

122
00:06:22,883 --> 00:06:25,819
but there's definitely an
increased stress on the

123
00:06:25,886 --> 00:06:29,223
grid that could be
grid that could be