1
00:00:00,020 --> 00:00:03,400
Ejaaz:
You've all heard the headlines. AI is draining our water supply.

2
00:00:03,620 --> 00:00:06,580
Ejaaz:
Data centers are stealing drinking water from the communities.

3
00:00:07,060 --> 00:00:14,520
Ejaaz:
ChatGPT is literally drying up the planet. It sounds terrifying and it's also completely fake.

4
00:00:15,020 --> 00:00:17,980
Ejaaz:
Today we're going to bust one of the biggest myths on the internet and walk

5
00:00:17,980 --> 00:00:22,540
Ejaaz:
through the actual numbers which reveal something crazy that the world's largest

6
00:00:22,540 --> 00:00:27,620
Ejaaz:
data center uses about as much water as two burger joints.

7
00:00:27,960 --> 00:00:31,080
Ejaaz:
We're going to walk you through exactly how much water is needed for the biggest

8
00:00:31,080 --> 00:00:35,260
Ejaaz:
data center in the world, Colossus 2, and why the majority of water used is

9
00:00:35,260 --> 00:00:40,820
Ejaaz:
only a fraction of the water used by your local golf club that your friends go to every weekend.

10
00:00:41,020 --> 00:00:44,360
Josh:
So part of the feedback that we've seen as we become more of a presence,

11
00:00:44,520 --> 00:00:48,500
Josh:
a platform in the world of AI, is that there's a lot of narratives that try to take down progress.

12
00:00:48,740 --> 00:00:51,980
Josh:
And the newest and hottest topic has been the topic of water consumption,

13
00:00:51,980 --> 00:00:55,240
Josh:
to the point where when I talk about with my friends about what I'm working

14
00:00:55,240 --> 00:00:59,520
Josh:
on and what I'm interested in, that frequently comes up as the first rebuttal.

15
00:00:59,620 --> 00:01:02,920
Josh:
It's like, this is horrible for the environment. It's using so much water, so much energy.

16
00:01:03,060 --> 00:01:06,960
Josh:
And this episode, we're going to focus on the water, particularly around the

17
00:01:06,960 --> 00:01:11,600
Josh:
public perception versus comparing it to the reality and how far off it really is.

18
00:01:11,700 --> 00:01:14,380
Josh:
We're seeing on the screen a few headlines of people from Utah,

19
00:01:14,780 --> 00:01:17,800
Josh:
very viral publications and headlines that have been talking about this.

20
00:01:17,920 --> 00:01:20,360
Josh:
But the reality is, is it's simply not true.

21
00:01:20,540 --> 00:01:24,740
Josh:
So we're going to methodically and in a very fun way kind of dissect how wrong

22
00:01:24,740 --> 00:01:28,100
Josh:
this actually is, starting with the book that kind of spawned it all.

23
00:01:28,340 --> 00:01:31,680
Josh:
The author's name is Karen Howe. No pun intended.

24
00:01:31,880 --> 00:01:35,660
Josh:
She really, like Karen, Karen seems like a very fitting name as an author called

25
00:01:35,660 --> 00:01:39,820
Josh:
the Empire of AI, claiming a Google data center would use 1,000 times the water

26
00:01:39,820 --> 00:01:42,020
Josh:
of an 88,000 person city.

27
00:01:42,480 --> 00:01:47,580
Josh:
Studies projecting AI will consume 1.7 trillion gallons of fresh water by 2027.

28
00:01:48,420 --> 00:01:51,460
Josh:
These claims are a bit outrageous, but this has been the narrative that people

29
00:01:51,460 --> 00:01:55,780
Josh:
have been using as they discuss their rebuttals against why AI should exist.

30
00:01:56,000 --> 00:02:00,260
Josh:
And this simply is not true. And I guess this is kind of where we got skeptical.

31
00:02:00,460 --> 00:02:05,180
Josh:
We were like, okay, 1.7 trillion gallons, a thousand times more than an 88,000 person city.

32
00:02:05,360 --> 00:02:08,680
Josh:
These numbers are outrageously large, but they're being peddled around as if

33
00:02:08,680 --> 00:02:10,400
Josh:
it is truth in these large publications.

34
00:02:10,920 --> 00:02:15,060
Ejaaz:
The thing is, like, we work in the AI space, Josh.

35
00:02:15,220 --> 00:02:19,240
Ejaaz:
So trillion dollars, trillions, We're kind of like used to it at this point.

36
00:02:19,300 --> 00:02:19,780
Josh:
Trillion here, trillion there.

37
00:02:19,960 --> 00:02:24,020
Ejaaz:
Yeah, right. But what I wasn't used to is a trillion gallons of water.

38
00:02:24,400 --> 00:02:29,540
Ejaaz:
And the first thought I had to my head was, what is this compared to?

39
00:02:29,660 --> 00:02:35,820
Ejaaz:
Like, I can't quite comprehend how much water this is and how much water is

40
00:02:35,820 --> 00:02:38,820
Ejaaz:
used in other industries that aren't AI adjacent.

41
00:02:39,000 --> 00:02:44,560
Ejaaz:
So I started getting skeptical and I came across what I think now is the most

42
00:02:44,560 --> 00:02:50,740
Ejaaz:
important bit of journalism done on data centers by these guys at SemiAnalysis.

43
00:02:50,920 --> 00:02:57,060
Ejaaz:
They are a crazy team of researchers that kind of like dig into all the boring,

44
00:02:57,360 --> 00:03:01,940
Ejaaz:
nitty-gritty, hard data center stuff to bring us the facts about lots of different

45
00:03:01,940 --> 00:03:07,280
Ejaaz:
things, including how much water is consumed at the data center level.

46
00:03:07,520 --> 00:03:12,580
Ejaaz:
And the revealings are super interesting. But what I love most about this,

47
00:03:12,580 --> 00:03:17,580
Ejaaz:
is, Josh, is that not only is it really informative, it's also hilarious.

48
00:03:18,180 --> 00:03:18,940
Ejaaz:
Because... Oh, it's great.

49
00:03:19,180 --> 00:03:20,420
Josh:
I love this post so much.

50
00:03:20,760 --> 00:03:23,680
Ejaaz:
Right. They've compared data center water usage

51
00:03:24,400 --> 00:03:29,560
Ejaaz:
to burger restaurants, specifically In-N-Out. So for some context here,

52
00:03:29,820 --> 00:03:35,440
Ejaaz:
they took the largest data center in the world, which is Elon Musk's Colossus 2.

53
00:03:35,580 --> 00:03:41,720
Ejaaz:
It's a data center based in Memphis, and it is the first data center to reach a gigawatt of compute.

54
00:03:41,840 --> 00:03:44,340
Ejaaz:
We actually mentioned this on yesterday's episode. If you haven't seen it,

55
00:03:44,620 --> 00:03:46,040
Ejaaz:
definitely go check that out. It's super cool.

56
00:03:46,440 --> 00:03:49,740
Ejaaz:
So we're talking about $18 billion worth of GPUs here, Josh.

57
00:03:49,880 --> 00:03:52,940
Ejaaz:
So as you can imagine, it's probably using a lot of water, right?

58
00:03:53,140 --> 00:03:56,800
Ejaaz:
So the math that they uncovered was the most revealing part.

59
00:03:56,980 --> 00:04:05,060
Ejaaz:
So Colossus 2 has an annual water footprint of 346 million gallons of water per year.

60
00:04:05,420 --> 00:04:08,840
Ejaaz:
Guess how much the average in-and-out store consumes per year?

61
00:04:09,400 --> 00:04:10,500
Josh:
A lot.

62
00:04:12,880 --> 00:04:16,580
Ejaaz:
147 million gallons per year. Per store.

63
00:04:16,860 --> 00:04:21,780
Ejaaz:
Per store, sorry, yeah, per store, yes. So that means that the largest data

64
00:04:21,780 --> 00:04:26,160
Ejaaz:
center in the entire world currently today that everyone's complaining about

65
00:04:26,160 --> 00:04:31,080
Ejaaz:
consumes the same amount of water as two and a half In-N-Out stores.

66
00:04:31,400 --> 00:04:33,120
Ejaaz:
Why aren't we protesting In-N-Out?

67
00:04:33,900 --> 00:04:37,800
Josh:
This is super fascinating to me. And it gets even better than this.

68
00:04:37,900 --> 00:04:39,260
Josh:
If you think that's funny, it gets even better.

69
00:04:39,420 --> 00:04:44,240
Josh:
So basically, the report looked at tokens per burger. So we were able to get

70
00:04:44,240 --> 00:04:46,960
Josh:
a metric for how you can actually justify output.

71
00:04:47,280 --> 00:04:49,360
Ejaaz:
Wait, wait, wait. Run that unit by me again.

72
00:04:49,560 --> 00:04:54,320
Josh:
We are breaking the news right here on Limitless. We are measuring water efficiency by tokens per burger.

73
00:04:54,540 --> 00:04:59,820
Josh:
And a single burger's water footprint is about 245 gallons of water.

74
00:04:59,960 --> 00:05:06,140
Josh:
That equals 2.7 billion AI output tokens, which roughly equates to one burger...

75
00:05:06,400 --> 00:05:12,260
Josh:
Being equivalent to using Grok 30 times a day for 668 years.

76
00:05:12,960 --> 00:05:15,700
Josh:
So the numbers are just absolutely astronomical.

77
00:05:16,540 --> 00:05:17,240
Ejaaz:
Wait, wait, wait.

78
00:05:17,400 --> 00:05:21,400
Josh:
Before I'm done, wait, one more thing, one final conclusion to this.

79
00:05:21,580 --> 00:05:25,220
Josh:
There are over 400 in and outs. There's so many of them.

80
00:05:25,480 --> 00:05:29,900
Josh:
And there's what, maybe 10 AI data centers? So the scale and the magnitude at

81
00:05:29,900 --> 00:05:33,140
Josh:
which they are wrong is just so astronomical.

82
00:05:33,380 --> 00:05:35,880
Josh:
I found it really funny like once you actually get into the numbers you realize

83
00:05:35,880 --> 00:05:38,260
Josh:
this is so not a problem that it's.

84
00:05:38,260 --> 00:05:42,020
Ejaaz:
Not even funny wait dude so this is you're telling me so this is incredibly click

85
00:05:42,020 --> 00:05:45,840
Ejaaz:
baity basically um the figures are astronomically

86
00:05:45,840 --> 00:05:48,460
Ejaaz:
wrong and if you compare it

87
00:05:48,460 --> 00:05:51,700
Ejaaz:
to just a casual restaurant we're talking about burgers only here by the way

88
00:05:51,700 --> 00:05:55,720
Ejaaz:
we're not talking about the fries sides and everything else just milkshakes

89
00:05:55,720 --> 00:05:59,200
Ejaaz:
god knows how much one of the milkshakes use we're just talking about the burgers

90
00:05:59,200 --> 00:06:03,460
Ejaaz:
including like their restaurant output and the supply chain for this so the

91
00:06:03,460 --> 00:06:06,280
Ejaaz:
fact of the matter is, this problem isn't really a problem.

92
00:06:06,420 --> 00:06:09,640
Ejaaz:
It's actually massively overblown. But what I'm curious about,

93
00:06:09,740 --> 00:06:13,180
Ejaaz:
Josh, is like, okay, is this a realistic take?

94
00:06:13,320 --> 00:06:15,540
Ejaaz:
Like, I know that data centers use a lot of water.

95
00:06:15,720 --> 00:06:19,980
Ejaaz:
How are they using it? And like, how much of this is renewable versus is actually

96
00:06:19,980 --> 00:06:22,920
Ejaaz:
burned and evaporated into the air and we'd never get it back?

97
00:06:23,080 --> 00:06:24,100
Ejaaz:
Like, what's the comparison?

98
00:06:24,300 --> 00:06:27,100
Josh:
Yes, yes, yes. Okay, so the truth, they are using lots of water.

99
00:06:27,220 --> 00:06:30,440
Josh:
There's hundreds of millions of gallons per year that are going through these data centers.

100
00:06:30,740 --> 00:06:34,600
Josh:
Some of it gets lost, a lot of it gets preserved. And the way they do it is

101
00:06:34,600 --> 00:06:35,720
Josh:
there's two types of cooling.

102
00:06:35,840 --> 00:06:39,340
Josh:
There's dry cooling, and then there's adiabatic cooling, which is the process

103
00:06:39,340 --> 00:06:41,980
Josh:
where air cools down without exchanging heat with its surroundings.

104
00:06:42,200 --> 00:06:45,580
Josh:
So if you remember the old iPhone episode that we did that talked about vapor

105
00:06:45,580 --> 00:06:49,820
Josh:
chamber, you can imagine scaling a vapor chamber to the size of an industrial data center.

106
00:06:49,940 --> 00:06:52,380
Josh:
And that's roughly how the adiabatic cooling works.

107
00:06:52,520 --> 00:06:56,240
Josh:
So it evaporates the water, the water leaves the system, and that's where they

108
00:06:56,240 --> 00:06:58,240
Josh:
lose about 267 million gallons per year.

109
00:06:58,660 --> 00:07:02,180
Josh:
And then the second loss function is the flush and discharge.

110
00:07:02,380 --> 00:07:05,020
Josh:
So one thing that I learned in preparing for this episode is that,

111
00:07:05,500 --> 00:07:07,640
Josh:
I mean, there's a lot of mineral buildup in the water that they use.

112
00:07:07,780 --> 00:07:12,120
Josh:
And 67 million of those get discharged as waste water per year.

113
00:07:12,340 --> 00:07:13,580
Josh:
But those are the two ways they do it.

114
00:07:13,860 --> 00:07:19,620
Josh:
It's dry and it's adiabatic. And there are promises in 2027 and 2028 from a

115
00:07:19,620 --> 00:07:23,040
Josh:
lot of the major AI labs to decrease this waste to about 95%.

116
00:07:23,940 --> 00:07:27,920
Josh:
Currently, it sits about 90%. So 90% gets recycled, 10% gets lost.

117
00:07:27,920 --> 00:07:31,840
Josh:
Those numbers are going to increase incrementally until about 2030 when the

118
00:07:31,840 --> 00:07:33,120
Josh:
number is actually net positive.

119
00:07:33,680 --> 00:07:37,940
Ejaaz:
Okay, so it sounds like the classical way of thinking about how data centers

120
00:07:37,940 --> 00:07:43,140
Ejaaz:
get cooled is you run a bunch of water in pipes through all these different GPUs,

121
00:07:43,180 --> 00:07:48,960
Ejaaz:
and it removes the heat from these GPUs so they are able to perform at optimal levels, right?

122
00:07:49,480 --> 00:07:54,680
Ejaaz:
And part of this water evaporates, never to be seen again, and some of this

123
00:07:54,680 --> 00:07:58,960
Ejaaz:
water gets kind of recycled over and over again. And the adiabatic system,

124
00:07:59,080 --> 00:08:02,120
Ejaaz:
I think is what you said, is kind of a hybrid of both of these things.

125
00:08:02,260 --> 00:08:04,880
Ejaaz:
But most importantly, it's more of a closed loop system.

126
00:08:05,040 --> 00:08:07,660
Ejaaz:
So we've kind of got like the majority of the water being renewed.

127
00:08:07,840 --> 00:08:13,800
Ejaaz:
So there's a comparison between water being renewed and water being consumed,

128
00:08:13,900 --> 00:08:16,240
Ejaaz:
which means lost forever. Do I have that right?

129
00:08:16,960 --> 00:08:20,900
Josh:
Yeah, so there's, you could think of a closed-loop system, going back to that vapor chamber.

130
00:08:21,200 --> 00:08:25,580
Josh:
If a vapor chamber is 100% efficient at being closed-loop, where water reaches

131
00:08:25,580 --> 00:08:29,280
Josh:
the processor, it heats up, it evaporates. As it evaporates, it dissipates the heat.

132
00:08:29,580 --> 00:08:33,400
Josh:
This is that at scale, although with some sort of a loss function at the end,

133
00:08:33,460 --> 00:08:36,300
Josh:
where some of that evaporated water does currently exit the system.

134
00:08:36,460 --> 00:08:39,260
Josh:
It sits now at about 10%, and that's where that loss comes from.

135
00:08:39,400 --> 00:08:43,460
Ejaaz:
But I'm guessing it's not 1.7 trillion gallons of water a year. Yeah, it's a

136
00:08:43,460 --> 00:08:48,340
Josh:
Far cry from the 1.7 trillion number. And then there's a second aspect to this

137
00:08:48,340 --> 00:08:50,780
Josh:
that gets a lot of criticism, which is the actual power generation,

138
00:08:50,980 --> 00:08:54,300
Josh:
how much water gets used in the generation of energy through these turbines

139
00:08:54,300 --> 00:08:56,080
Josh:
that are natural gas, some solar.

140
00:08:56,340 --> 00:09:00,700
Josh:
And the answer to that is, in the case of Colossus 2, 0%. There is no meaningful

141
00:09:00,700 --> 00:09:02,620
Josh:
water consumption of power generation at all.

142
00:09:02,860 --> 00:09:08,020
Josh:
The entirety of it comes from cooling the GPU system, which is closed loop and done with water.

143
00:09:08,140 --> 00:09:12,360
Josh:
So maybe we just go into where that 1.7 trillion number even came from.

144
00:09:12,460 --> 00:09:14,060
Josh:
Because this is the source of

145
00:09:14,060 --> 00:09:17,220
Josh:
a lot of the narrative that we've seen play out over the last few weeks.

146
00:09:17,480 --> 00:09:23,280
Ejaaz:
Exactly. So I'm showing all of you folks who are watching a scientific paper

147
00:09:23,280 --> 00:09:27,740
Ejaaz:
from UC Riverside titled Making AI Less Thirsty.

148
00:09:27,960 --> 00:09:30,040
Josh:
When you say scientific, you have to do air quotes.

149
00:09:30,140 --> 00:09:35,740
Ejaaz:
Yeah, sorry, sorry. Pseudoscience. From this university called Making AI Less

150
00:09:35,740 --> 00:09:40,860
Ejaaz:
Thirsty, uncovering and addressing the secret water footprint of AI models.

151
00:09:40,860 --> 00:09:46,700
Ejaaz:
Now, this paper is the source of a lot of the clickbait headlines and TikToks

152
00:09:46,700 --> 00:09:48,460
Ejaaz:
that you watch online. It's this one number.

153
00:09:49,000 --> 00:09:55,560
Ejaaz:
1.7 trillion gallons of water will be consumed by AI data centers alone by 2027.

154
00:09:55,740 --> 00:09:58,340
Ejaaz:
That's around the corner. We're talking about next year here, right?

155
00:09:58,620 --> 00:10:01,380
Ejaaz:
And so a bunch of people kind of like threw up their hands and started protesting

156
00:10:01,380 --> 00:10:04,300
Ejaaz:
data centers because they were like, that is so much water.

157
00:10:04,420 --> 00:10:07,060
Ejaaz:
We're not going to be, we humans aren't going to be left with enough water to

158
00:10:07,060 --> 00:10:08,560
Ejaaz:
consume ourselves. Right?

159
00:10:08,740 --> 00:10:14,520
Ejaaz:
But here's what this study actually says. It claims 1.7 trillion gallons of

160
00:10:14,520 --> 00:10:16,880
Ejaaz:
water are used for withdrawal.

161
00:10:17,220 --> 00:10:19,940
Ejaaz:
That's a fancy term of saying recycled.

162
00:10:20,300 --> 00:10:24,900
Ejaaz:
So imagine like the water being taken, used to cool down the systems,

163
00:10:24,900 --> 00:10:29,780
Ejaaz:
the GPs that we just mentioned, and then recycled again and again and again.

164
00:10:29,840 --> 00:10:32,200
Ejaaz:
So it's not net new water we're talking about here.

165
00:10:32,680 --> 00:10:36,640
Ejaaz:
Josh, do you want to know the actual consumption that is being like permanently

166
00:10:36,640 --> 00:10:38,640
Ejaaz:
removed, the ones that we should be trying to protest?

167
00:10:39,100 --> 00:10:42,280
Josh:
Yeah, well, I mean, based on that, it's what? It's at least a full order of

168
00:10:42,280 --> 00:10:43,640
Josh:
magnitude off than what is projected.

169
00:10:43,940 --> 00:10:51,300
Ejaaz:
No, it's 100 to 158 billion, not trillion, billion gallons of water. That's 10%.

170
00:10:51,780 --> 00:10:55,700
Ejaaz:
Actually, it's less than 10% if you take the lower bound of the reported number.

171
00:10:56,040 --> 00:11:01,020
Ejaaz:
So the water that's actually being used by the data centers is only kind of like 10 to 15% of that.

172
00:11:01,340 --> 00:11:04,000
Ejaaz:
And okay, so some people then go like, what about the drinking water?

173
00:11:04,200 --> 00:11:07,780
Ejaaz:
Like what percentage of that is being affected from the study that was being

174
00:11:07,780 --> 00:11:10,780
Ejaaz:
made? It's only 3% of the headline number.

175
00:11:11,160 --> 00:11:15,120
Ejaaz:
So everyone took 1.7 trillion gallons of water and assumed all of that water

176
00:11:15,120 --> 00:11:19,160
Ejaaz:
was being wasted, never to be seen again, used again, and it was pulling from

177
00:11:19,160 --> 00:11:23,020
Ejaaz:
other resources when that strictly isn't being true. So the point I'm making,

178
00:11:23,160 --> 00:11:24,160
Ejaaz:
and I'm going to reiterate it again,

179
00:11:24,630 --> 00:11:28,690
Ejaaz:
The water, the 1.7 trillion gallons of water isn't being consumed.

180
00:11:28,910 --> 00:11:33,710
Ejaaz:
Think of it like this. Like imagine diverting a river to run through a mill. We do that today, right?

181
00:11:34,070 --> 00:11:38,170
Ejaaz:
And it flows back into the river. That's called withdrawal. That's the 1.7 trillion

182
00:11:38,170 --> 00:11:40,270
Ejaaz:
number that I'm pulling out here.

183
00:11:40,530 --> 00:11:45,250
Ejaaz:
So most of the power plant water is being reused over and over again.

184
00:11:46,210 --> 00:11:50,730
Ejaaz:
So that's only the first major kind of bit of pseudoscience that we needed to bust.

185
00:11:50,950 --> 00:11:54,050
Ejaaz:
But there's this second thing, and Josh, you mentioned it earlier from,

186
00:11:54,190 --> 00:11:58,550
Ejaaz:
ironically, Karen Howe, who wrote this famous book called...

187
00:11:58,550 --> 00:11:59,170
Josh:
Our favorite author.

188
00:11:59,470 --> 00:12:03,390
Ejaaz:
...called The Empire of Air. Let me get this up here so everyone can see this book.

189
00:12:03,550 --> 00:12:09,730
Ejaaz:
It has been rated over 1,300 times on Amazon, but more importantly,

190
00:12:09,930 --> 00:12:15,250
Ejaaz:
it has been quoted directly by The Economist, The New York Times, and so much more.

191
00:12:15,370 --> 00:12:21,890
Ejaaz:
For this specific stack, it says Google's data center will use 1,000 times the

192
00:12:21,890 --> 00:12:24,370
Ejaaz:
water of an 88,000-person city.

193
00:12:25,090 --> 00:12:31,650
Ejaaz:
Guess what? It was off by a factor of not 1,000 times, 4,500 times.

194
00:12:31,790 --> 00:12:35,830
Ejaaz:
Josh, can you run me down these stats? Because it's just insane, dude.

195
00:12:36,090 --> 00:12:39,130
Josh:
No, that stat actually makes me sick. It's horrible. So instead of comparing

196
00:12:39,130 --> 00:12:42,850
Josh:
it to Google, we're going to start with golf courses because that's just...

197
00:12:42,850 --> 00:12:45,270
Josh:
If you want to come at it, you got to start with the golf course.

198
00:12:45,410 --> 00:12:49,510
Josh:
The average golf course, 312,000 gallons of water.

199
00:12:49,750 --> 00:12:54,790
Josh:
Desert golf course is 1 to 2 million. A Google data center in Virginia is 400,000.

200
00:12:54,990 --> 00:12:57,230
Josh:
Now what does that 400,000 give you?

201
00:12:57,490 --> 00:13:03,790
Josh:
That powers Gmail, Google Drive, YouTube, the entire G Suite for billions of

202
00:13:03,790 --> 00:13:09,690
Josh:
people around the world with the equivalent of 1.2 golf courses worth of water for billions of people.

203
00:13:09,870 --> 00:13:14,030
Josh:
Google's global data center that powers 4 billion accounts,

204
00:13:14,910 --> 00:13:19,630
Josh:
equals 43 golf courses. And in the state of Arizona, there are 370.

205
00:13:19,890 --> 00:13:26,050
Josh:
So if you're comparing apples to apples here, that is the Lufx comparison. And it's just so off.

206
00:13:26,630 --> 00:13:30,450
Josh:
This essay that was reported, this book that was published, it's off by like

207
00:13:30,450 --> 00:13:32,490
Josh:
several orders of magnitude. It's not even close.

208
00:13:32,850 --> 00:13:36,750
Ejaaz:
I mean, I'm looking at some of the things that they got wrong because someone

209
00:13:36,750 --> 00:13:40,830
Ejaaz:
did a breakdown of this book that's been quoted so many times and that's behind all these headlines.

210
00:13:41,490 --> 00:13:46,650
Ejaaz:
The book reported 5 million liters as the city's annual water use,

211
00:13:46,810 --> 00:13:48,210
Ejaaz:
but that was misconstrued.

212
00:13:48,350 --> 00:13:51,970
Ejaaz:
What she actually meant was 5 billion cubic meters.

213
00:13:52,830 --> 00:13:54,870
Ejaaz:
For understanding here...

214
00:13:54,870 --> 00:13:57,190
Josh:
Liters and cubic meters are very different things.

215
00:13:57,270 --> 00:14:01,210
Ejaaz:
Very, very different things. So she confused liters with cubic meters,

216
00:14:01,370 --> 00:14:03,010
Ejaaz:
which is already 1,000x error.

217
00:14:03,270 --> 00:14:09,370
Ejaaz:
But then she said the data center would actually use 3% of the municipal water system, not 1,000x.

218
00:14:09,510 --> 00:14:14,850
Ejaaz:
That's where we led to the 4,500x factor off that she was.

219
00:14:15,050 --> 00:14:21,110
Ejaaz:
So basically, it is an absolute incorrect piece of reporting that has been spread

220
00:14:21,110 --> 00:14:24,670
Ejaaz:
by some of the most important and popular media publications that we've seen.

221
00:14:24,970 --> 00:14:30,170
Ejaaz:
Josh, I want to kind of take your golf course thing a little further because

222
00:14:30,170 --> 00:14:31,050
Ejaaz:
I actually quite like that.

223
00:14:31,450 --> 00:14:32,950
Josh:
Okay, we'll keep going on the golf courses.

224
00:14:33,110 --> 00:14:37,510
Ejaaz:
So what's the number one state that's been getting a lot of protests about data centers? Arizona.

225
00:14:38,850 --> 00:14:45,230
Ejaaz:
Let's just look at the golf courses in Arizona. Arizona has 370 golf courses.

226
00:14:45,710 --> 00:14:50,110
Ejaaz:
Each golf course consumes about 1 to 2 million gallons of water per day.

227
00:14:50,370 --> 00:14:55,610
Ejaaz:
So that's around 400 to 800 million gallons per day for an Arizona golf course,

228
00:14:55,810 --> 00:14:57,330
Ejaaz:
right? For all the Arizona golf course.

229
00:14:57,730 --> 00:15:02,630
Ejaaz:
If you compare that to data centers that consume 905 million gallons,

230
00:15:02,890 --> 00:15:06,550
Ejaaz:
or rather 0.12% of county water,

231
00:15:07,340 --> 00:15:10,660
Ejaaz:
That's like just in Arizona alone. So golf courses in Arizona,

232
00:15:10,880 --> 00:15:12,660
Ejaaz:
29 billion gallons per year.

233
00:15:12,880 --> 00:15:16,700
Ejaaz:
And a data center, the biggest one in Arizona, or actually collectively all

234
00:15:16,700 --> 00:15:21,700
Ejaaz:
of these, actually, my correction, is just 0.12% of that.

235
00:15:21,980 --> 00:15:27,820
Ejaaz:
900 million gallons per year. So it's obvious that there's just a lot of misinformation out there.

236
00:15:28,080 --> 00:15:32,680
Ejaaz:
And I think it's really important to just bust this wide open completely. It's just wrong.

237
00:15:32,900 --> 00:15:36,480
Ejaaz:
And part of me thinks that a lot of the hatred, if I'm being honest,

238
00:15:36,560 --> 00:15:43,140
Ejaaz:
Josh, I think comes from people kind of equating AI to enriching people that they don't maybe like.

239
00:15:43,320 --> 00:15:47,540
Ejaaz:
Like, I get it. Like, AI will be used as a tool to enrich billionaires even further.

240
00:15:48,280 --> 00:15:52,540
Ejaaz:
It's so wasteful. I don't use it. What people don't realize is the average activity

241
00:15:52,540 --> 00:15:56,160
Ejaaz:
that you do on a weekend or the average bit of food that you might consume,

242
00:15:56,380 --> 00:16:01,320
Ejaaz:
buying a burger, 668 years straight of using Grok 30 times a day.

243
00:16:01,600 --> 00:16:02,780
Ejaaz:
People don't do that that often.

244
00:16:03,120 --> 00:16:06,160
Ejaaz:
It's just, it's important to level set, in my opinion.

245
00:16:06,160 --> 00:16:09,900
Josh:
Yeah, there are a lot of valid criticisms. This is unfortunately not one of them.

246
00:16:10,080 --> 00:16:14,260
Josh:
But there are some instances in which it does, or it has in the past actually

247
00:16:14,260 --> 00:16:19,300
Josh:
affected localized effects where very small towns have actually felt an impact of this.

248
00:16:19,420 --> 00:16:21,820
Josh:
They date most recently back to 2022.

249
00:16:22,140 --> 00:16:27,240
Josh:
There actually hasn't been many sources recently that have determined that it

250
00:16:27,240 --> 00:16:28,900
Josh:
is making a problem, but there have been some instances.

251
00:16:29,060 --> 00:16:32,980
Josh:
The first one was in Oregon, where a Google data center It consumed 29% of the

252
00:16:32,980 --> 00:16:33,940
Josh:
town's local water supply.

253
00:16:34,180 --> 00:16:36,340
Josh:
There's another one from 2019 in Virginia.

254
00:16:36,780 --> 00:16:40,400
Josh:
Virginia, famously, that's where a lot of the internet data center runs.

255
00:16:40,520 --> 00:16:43,960
Josh:
I think a majority of the data center is run from these Virginia AWS servers.

256
00:16:44,380 --> 00:16:49,220
Josh:
It consumed 63% of Lodon County in Virginia in 2019.

257
00:16:49,700 --> 00:16:54,640
Josh:
But since then, there really hasn't been that much of an interference with the local water supply.

258
00:16:54,780 --> 00:16:58,760
Josh:
And there have been solutions proposed from the companies who are most responsible for.

259
00:16:59,360 --> 00:17:03,260
Josh:
Creating that strain. And that is Google, Amazon, Meta, and Microsoft,

260
00:17:03,260 --> 00:17:07,860
Josh:
who all committed to be water positive by 2030 using new cooling techniques.

261
00:17:08,060 --> 00:17:11,280
Josh:
So that vapor chamber that we were talking about earlier will be done at an

262
00:17:11,280 --> 00:17:16,880
Josh:
industrial scale and will actually be able to preserve 100% of the closed loop

263
00:17:16,880 --> 00:17:19,460
Josh:
water supply. And I think that's going to be a really big deal.

264
00:17:19,720 --> 00:17:23,660
Josh:
That paired with direct to chip cooling is also going to make a big difference.

265
00:17:23,680 --> 00:17:27,340
Josh:
If you remember our Vera Rubin episode, where we talked about how the new chips

266
00:17:27,340 --> 00:17:30,040
Josh:
are cooled, the actual cooling temperature, if I remember right,

267
00:17:30,160 --> 00:17:33,400
Josh:
it was like 115 degrees Fahrenheit that it could be.

268
00:17:33,560 --> 00:17:37,460
Josh:
So now you could actually cool these ships with hot water. It requires much less.

269
00:17:37,720 --> 00:17:40,060
Josh:
Air cooling is getting a little more interesting.

270
00:17:40,340 --> 00:17:43,660
Josh:
There's a lot of solutions coming along the way that will make this a lot more

271
00:17:43,660 --> 00:17:45,440
Josh:
resource or a lot less resource intensive.

272
00:17:45,860 --> 00:17:51,620
Ejaaz:
Cool. So if I were to summarize some takeaways for this myth busting episode,

273
00:17:51,900 --> 00:17:54,100
Ejaaz:
there's a few that come to mind. Number one,

274
00:17:54,730 --> 00:17:58,530
Ejaaz:
The numbers just don't support the panic that people are putting out there.

275
00:17:58,710 --> 00:18:02,570
Ejaaz:
The fact is, AI data centers currently today, they might change later,

276
00:18:02,990 --> 00:18:09,870
Ejaaz:
Use a fraction of what golf courses, agriculture, the t-shirt that you're wearing consumes to produce.

277
00:18:10,090 --> 00:18:14,910
Ejaaz:
So even if we tripled AI water usage today, it would still pretty much be a rounding error.

278
00:18:15,230 --> 00:18:22,350
Ejaaz:
Number two, the context matters. You can't confuse 1,000 liters with 5 billion

279
00:18:22,350 --> 00:18:26,510
Ejaaz:
or whatever the number was, cubic meters of water.

280
00:18:26,770 --> 00:18:31,230
Ejaaz:
That is super important. And comparing water consumption for AI data centers

281
00:18:31,230 --> 00:18:34,510
Ejaaz:
and your average burger joint might just be the comparison that you need to

282
00:18:34,510 --> 00:18:37,510
Ejaaz:
kind of like set you straight and be like, okay, well, maybe this isn't that

283
00:18:37,510 --> 00:18:38,410
Ejaaz:
important going forward.

284
00:18:38,590 --> 00:18:42,630
Ejaaz:
And then the third thing, which I think is kind of underspoken about quite a

285
00:18:42,630 --> 00:18:47,430
Ejaaz:
lot is, I think, if I were to guess, we're moving towards a world where we end

286
00:18:47,430 --> 00:18:50,970
Ejaaz:
up actually using less water for data centers, right?

287
00:18:51,170 --> 00:18:55,150
Ejaaz:
Part of it is due to the different systems, like the adiabatic system that you mentioned, Josh.

288
00:18:55,470 --> 00:18:59,610
Ejaaz:
But also, I think a lot of these AI companies are going to start building water

289
00:18:59,610 --> 00:19:04,170
Ejaaz:
recycling plants to kind of push that 90% water renewability figure that we

290
00:19:04,170 --> 00:19:05,830
Ejaaz:
mentioned earlier much, much higher.

291
00:19:05,910 --> 00:19:10,330
Ejaaz:
I think Colossus 2 and Elon is doing that right now for Colossus 3,

292
00:19:10,390 --> 00:19:12,890
Ejaaz:
actually. They're building out a water recycling plant.

293
00:19:13,030 --> 00:19:16,550
Ejaaz:
So I think overall, this is a nothing burger, pardon the pun.

294
00:19:16,890 --> 00:19:19,790
Ejaaz:
And we're going to look back on this. Yeah, we're going to look back on this

295
00:19:19,790 --> 00:19:24,010
Ejaaz:
in the future and realize that we're consuming water in much faster ways in

296
00:19:24,010 --> 00:19:26,610
Ejaaz:
so many other industries that we aren't currently protesting.

297
00:19:26,870 --> 00:19:30,290
Ejaaz:
So keep quiet, eat your burger, and let the AI flow.

298
00:19:31,040 --> 00:19:34,840
Josh:
So if you had to guess what the next narrative would be, that has a negative

299
00:19:34,840 --> 00:19:37,780
Josh:
spin on it. Do you have any ideas? I think my answer is going to be energy.

300
00:19:38,020 --> 00:19:42,260
Josh:
I think they're going to start to converge on the correct argument,

301
00:19:42,520 --> 00:19:46,960
Josh:
which is the energy consumption on a localized scale, starting to actually impact

302
00:19:46,960 --> 00:19:49,960
Josh:
the cost per kilowatt of the average person's home.

303
00:19:50,440 --> 00:19:54,900
Josh:
And how does that get offloaded? Well, a lot more natural gas turbines, a lot more solar panels.

304
00:19:55,220 --> 00:19:59,020
Josh:
And the process of scaling that up is happening, but it's happening slower than

305
00:19:59,020 --> 00:20:00,860
Josh:
the scale at which they're consuming.

306
00:20:01,040 --> 00:20:06,020
Josh:
So when you take a gigawatt data center like Colossus 2, that is consuming the

307
00:20:06,020 --> 00:20:10,680
Josh:
equivalent collective output of San Francisco localized to a small town in Tennessee,

308
00:20:11,020 --> 00:20:13,060
Josh:
there are impacts there that are real.

309
00:20:13,880 --> 00:20:18,480
Josh:
It's just a matter of time until those kind of get uncovered and then get dealt with.

310
00:20:18,660 --> 00:20:22,300
Josh:
I mean, they are dealing with it quickly, but there is a real strain happening

311
00:20:22,300 --> 00:20:25,080
Josh:
on some grids that are localized to where these data centers exist.

312
00:20:25,080 --> 00:20:29,880
Ejaaz:
I agree with you. And I actually think the numbers that will be quoted on headlines

313
00:20:29,880 --> 00:20:34,940
Ejaaz:
about that specifically will actually be closer to home and to the point because

314
00:20:35,720 --> 00:20:40,920
Ejaaz:
It's simple enough to kind of scale new water techniques to kind of cool stuff down.

315
00:20:41,020 --> 00:20:44,460
Ejaaz:
Like we reported on a previous episode, I think like two weeks ago,

316
00:20:44,620 --> 00:20:46,940
Ejaaz:
that they're not even using cold water anymore.

317
00:20:47,080 --> 00:20:51,880
Ejaaz:
They're using warm water, 45 degrees Celsius or Fahrenheit, which is super warm.

318
00:20:51,960 --> 00:20:54,820
Ejaaz:
I think it's like 90 degrees Fahrenheit to cool these systems down.

319
00:20:55,080 --> 00:20:58,960
Ejaaz:
I think it's a different game with energy where we actually do have limited constraint.

320
00:20:59,140 --> 00:21:04,000
Ejaaz:
It takes so much more work and expertise to scale that. and we're going to have

321
00:21:04,000 --> 00:21:08,360
Ejaaz:
to tap into like town supply or city supply. So I agree with you.

322
00:21:08,760 --> 00:21:11,580
Ejaaz:
I'm looking forward to kind of like unpicking that one in the future.

323
00:21:11,700 --> 00:21:14,840
Josh:
Yeah, we'd be good at this. We should design the next PSYOP against our own industry.

324
00:21:15,020 --> 00:21:21,140
Josh:
I think that would be much better to consult with the experts prior to doing this next time.

325
00:21:21,300 --> 00:21:21,680
Ejaaz:
That's hilarious.

326
00:21:21,780 --> 00:21:25,700
Josh:
But yeah, I guess that concludes the Mythbusters episode on the first one that

327
00:21:25,700 --> 00:21:30,300
Josh:
we'll be dealing with, which is the water consumption and the new metric that

328
00:21:30,300 --> 00:21:32,240
Josh:
is burger or tokens per burger.

329
00:21:33,500 --> 00:21:38,260
Josh:
And in the case of our tokens per burger metric, the cost is very low.

330
00:21:38,400 --> 00:21:40,160
Josh:
And I don't think this is anything to actually worry about.

331
00:21:41,880 --> 00:21:45,160
Ejaaz:
That was the end of the episode, but I'm actually curious whether you guys enjoyed

332
00:21:45,160 --> 00:21:47,880
Ejaaz:
this. We hope you guys learn something new from this.

333
00:21:48,760 --> 00:21:51,640
Ejaaz:
Josh and I kind of like went back and forth on this, whether we should do this episode.

334
00:21:51,740 --> 00:21:55,880
Ejaaz:
We realized like a myth-busting series could be really cool because there's

335
00:21:55,880 --> 00:21:58,900
Ejaaz:
just a lot of myths and false claims out there and we face it every day.

336
00:21:59,020 --> 00:22:01,860
Ejaaz:
We try and unpack it, spend all of our time figuring this stuff out.

337
00:22:01,980 --> 00:22:03,720
Josh:
And I love that show growing up. It was great.

338
00:22:03,820 --> 00:22:07,620
Ejaaz:
Yes, same. Actually, dude, maybe we need to, we need to come back with some

339
00:22:07,620 --> 00:22:10,820
Ejaaz:
glasses for the next myth-busting episode. maybe a trench coat,

340
00:22:10,960 --> 00:22:13,280
Ejaaz:
like a fedora, if we really lean into it.

341
00:22:13,440 --> 00:22:14,740
Josh:
I'll write in my magnifying glass.

342
00:22:15,020 --> 00:22:17,920
Ejaaz:
Exactly. But yeah, if you found this informative, if you found this interesting,

343
00:22:18,200 --> 00:22:23,040
Ejaaz:
and you aren't subscribed to us, which apparently is around 70% to 80% of you,

344
00:22:23,320 --> 00:22:24,980
Josh:
Please do so. It's a percentage that is far too high.

345
00:22:25,140 --> 00:22:29,040
Ejaaz:
It's way too high. It is actually almost as high as these false headlines that

346
00:22:29,040 --> 00:22:31,160
Ejaaz:
we keep seeing about water usage for data centers.

347
00:22:31,520 --> 00:22:36,660
Ejaaz:
So if we're describing you currently, it takes two seconds. Please subscribe.

348
00:22:36,820 --> 00:22:37,660
Ejaaz:
Please turn on notifications.

349
00:22:37,860 --> 00:22:41,580
Ejaaz:
If you're listening to this on Spotify, Apple Music, or wherever the hell you

350
00:22:41,580 --> 00:22:43,960
Ejaaz:
listen to this on, please also do the same and give us a rating.

351
00:22:44,140 --> 00:22:49,120
Ejaaz:
It helps us out massively and puts our videos out to way more people so that

352
00:22:49,120 --> 00:22:54,840
Ejaaz:
we get more eyeballs on this and we can keep producing better videos for you. I think that's it, Josh.

353
00:22:55,060 --> 00:22:57,700
Josh:
Anything from you? Yeah, and just a small reminder about the newsletter.

354
00:22:57,880 --> 00:23:01,360
Josh:
Today, we just dropped a new piece that coincided with the roundup,

355
00:23:01,480 --> 00:23:04,660
Josh:
which was a weekly roundup of the five most important, noteworthy things that

356
00:23:04,660 --> 00:23:05,560
Josh:
you want to be informed on.

357
00:23:06,880 --> 00:23:09,420
Josh:
We post that twice a week, once every Wednesday, one's every Friday,

358
00:23:09,580 --> 00:23:10,760
Josh:
one's a thought piece, one's a recap.

359
00:23:11,080 --> 00:23:14,720
Josh:
So you can join 100,000 other people who are also subscribed to getting the

360
00:23:14,720 --> 00:23:16,480
Josh:
info prior to these episodes dropping.

361
00:23:16,860 --> 00:23:20,180
Josh:
And I think that concludes it. That just wraps it up. So thank you so much for

362
00:23:20,180 --> 00:23:22,780
Josh:
watching as always. And we will see you guys in the next episode.

363
00:23:23,420 --> 00:23:23,980
Ejaaz:
See you guys.