1
00:00:00,000 --> 00:00:05,000
And so with hydro, it's really interesting. Like you have, it's basically direct to chip cooling.

2
00:00:05,000 --> 00:00:08,860
And this has been, you know, this is something that's existed in just data centers forever.

3
00:00:08,860 --> 00:00:13,780
But with, with, with mining, we're seeing like with this direct to chip cooling, specifically

4
00:00:13,780 --> 00:00:19,840
in a like liquid form, you kind of have like a mix of water and like a glycol mix. It's basically

5
00:00:19,840 --> 00:00:25,400
like shooting, you know, cold water basically on these chips and cycling them in real time.

6
00:00:25,400 --> 00:00:28,880
So you're getting like very precise cooling at the chip level.

7
00:00:29,920 --> 00:00:35,920
And so like that's really important because the quality of the chip basically degrades

8
00:00:35,920 --> 00:00:37,480
exponential with the temperature.

9
00:00:37,980 --> 00:00:38,380
HashrateUp.

10
00:00:38,660 --> 00:00:41,980
Hardware sales, advisory, hosting, and site brokerage.

11
00:00:42,400 --> 00:00:46,440
Find new and use basic deals through the website and the Telegram channel below.

12
00:00:46,980 --> 00:00:49,100
Make smarter decisions with HashrateUp.

13
00:00:49,320 --> 00:00:51,560
Welcome back everybody to HashrateUp.

14
00:00:51,780 --> 00:00:53,860
It's a very, very interesting time.

15
00:00:53,860 --> 00:00:58,920
very very interesting time sorry it's late for me there's gonna be a lot of these um in bitcoin

16
00:00:58,920 --> 00:01:05,760
mining right now and i have my good friend and supporter of the show and overall partner tom

17
00:01:05,760 --> 00:01:13,400
with me hi tom how are you hey jesse i'm good how are you uh yeah i'm doing okay i'm still smiling

18
00:01:13,400 --> 00:01:20,120
um we just sat here for quite a bit talking about the market and stuff and just decided to

19
00:01:20,120 --> 00:01:24,600
to riff it basically, which is often no pressure.

20
00:01:25,200 --> 00:01:30,140
Some of the best shows just end up coming from a point like that.

21
00:01:30,960 --> 00:01:35,060
For the first time ever, I believe, hash price sits below $30.

22
00:01:35,880 --> 00:01:41,580
So we are now well, well, well, well, well into the pain zone.

23
00:01:42,700 --> 00:01:45,100
Bitcoin price, let's just give a brief overview.

24
00:01:45,420 --> 00:01:46,280
I don't feel it anymore.

25
00:01:46,280 --> 00:01:58,220
um bitcoin price is at 64 000 hash price sits at 29 hash rates at 900 we're still looking at our

26
00:01:58,220 --> 00:02:04,720
difficulty adjustment of minus 13 which will bump that back up again but i'm sure as bitcoin is

27
00:02:04,720 --> 00:02:10,520
falling we're going to see even lower hash price in the meantime we're at 11 minutes and 30 seconds

28
00:02:10,520 --> 00:02:18,700
average block time let's quickly look at the time chain calendar at block height 935163 is

29
00:02:18,700 --> 00:02:24,900
when we were recording this and if my eyes don't see look at hash rate hashrate 20900 exahash

30
00:02:24,900 --> 00:02:35,620
yeah and then we are also 266 blocks too slow um we're in the middle of and i said this in the

31
00:02:35,620 --> 00:02:41,860
in a video before um a focus episode that i did where i just you know covered the hash price and

32
00:02:41,860 --> 00:02:49,540
the difficulty development right now i think we're looking at very um unprecedented times we've had

33
00:02:50,500 --> 00:02:56,420
um crazy times like this when the china mining ban happened and when bitcoin dropped from 32

34
00:02:57,300 --> 00:03:04,180
the enormous heights of 32 all the way down to two i believe um and i think we're doing we're

35
00:03:04,180 --> 00:03:05,940
We're seeing something like that.

36
00:03:06,100 --> 00:03:06,760
We're on our way.

37
00:03:07,420 --> 00:03:14,680
You know, when this 13% hits, I think we will be at around 25% overall over the last 12 weeks.

38
00:03:15,460 --> 00:03:17,640
And a lot of people are hurting right now, Tom.

39
00:03:17,780 --> 00:03:20,820
What are your thoughts at the moment looking at this?

40
00:03:22,860 --> 00:03:24,160
Yeah, I have a lot of thoughts.

41
00:03:24,160 --> 00:03:29,800
So, like, one is obviously the steepness of the Bitcoin price kind of drop.

42
00:03:29,980 --> 00:03:32,480
Like, we've been through things like this in the past.

43
00:03:32,480 --> 00:03:41,260
And I think a lot of there were a lot of discussions and like people thinking kind of previously like, oh, the four year cycle theory is dead.

44
00:03:41,380 --> 00:03:47,460
We have, you know, we're in this super cycle and the ups won't be as high, but the lows won't, you know, the drops won't be as high.

45
00:03:48,180 --> 00:03:55,460
I think we're like learning in real time that like, hey, this is still, you know, at least the price of Bitcoin is still a very volatile, a volatile thing.

46
00:03:56,060 --> 00:04:00,480
But I don't think the thesis, at least from my perspective, has really changed.

47
00:04:00,480 --> 00:04:07,440
right? Like, I nothing has changed in like the the code of this thing. Nothing has changed in like

48
00:04:07,440 --> 00:04:11,920
how the economics of mining work. Like it's the same thing. We're just seeing, you know, volatility

49
00:04:11,920 --> 00:04:16,060
in the price. And there's probably a lot of reasons as to like, why that is, but that's kind of like

50
00:04:16,060 --> 00:04:20,140
my first take. The other thing is like, yeah, like people are feeling a lot of pain right now. I think

51
00:04:20,140 --> 00:04:25,780
I think if you are in a position going into this, where maybe you had higher expectations,

52
00:04:25,780 --> 00:04:29,920
and maybe you were making decisions around like, you know, leverage and different things,

53
00:04:29,920 --> 00:04:33,020
it can be a bit of a scary time.

54
00:04:33,160 --> 00:04:36,880
I think for us, we live in the world of mining and hash price

55
00:04:36,880 --> 00:04:39,020
and how these systems work.

56
00:04:39,620 --> 00:04:45,880
And so I think it still is true that as long as you're not overpaying

57
00:04:45,880 --> 00:04:48,780
for the ASICs and your initial deployment,

58
00:04:49,460 --> 00:04:51,280
as long as you have good operating partners

59
00:04:51,280 --> 00:04:55,020
and you're on the right side of the efficiency curve,

60
00:04:55,020 --> 00:04:58,380
we could say, leaning towards higher efficiency machines

61
00:04:58,380 --> 00:05:01,440
relative to the average on the network.

62
00:05:01,740 --> 00:05:04,520
It's really just a game about staying online, right?

63
00:05:04,540 --> 00:05:07,440
We kind of see as hash price corrects

64
00:05:07,440 --> 00:05:08,720
when difficulty goes up,

65
00:05:09,140 --> 00:05:11,160
we'll see hash price correcting

66
00:05:11,160 --> 00:05:14,760
until machines don't turn off anymore, basically.

67
00:05:15,280 --> 00:05:16,960
And we don't know when that will be,

68
00:05:17,100 --> 00:05:19,600
but that's just how the economics kind of play out here.

69
00:05:20,120 --> 00:05:22,160
And so I think it's interesting

70
00:05:22,160 --> 00:05:23,440
to be talking about this right now

71
00:05:23,440 --> 00:05:27,180
because obviously there's a lot of people worry,

72
00:05:27,280 --> 00:05:28,080
there's a lot of things going on,

73
00:05:28,080 --> 00:05:31,480
There's also a lot of things going on just like in the world in general.

74
00:05:31,680 --> 00:05:36,780
Like we were talking about how gold as an asset class like is completely ripping, how

75
00:05:36,780 --> 00:05:40,680
the price of the dollar relative to, you know, other currencies is trading.

76
00:05:40,900 --> 00:05:43,460
So for us, like there's a lot of noise.

77
00:05:43,640 --> 00:05:46,920
We kind of just focus on the bare bones of mining economics.

78
00:05:46,920 --> 00:05:51,560
And I think like if you have high efficiency machines that you paid a good price for and

79
00:05:51,560 --> 00:05:54,640
you're running in good conditions, like you'll be able to withstand a lot of this.

80
00:05:54,640 --> 00:06:05,480
And then over time, like you're just stacking more Bitcoin if the price, you know, relatively drops, right? You're going to be stacking more in terms of like staying online.

81
00:06:05,480 --> 00:06:19,040
Yeah, it's a time where, and I just want to pull this up, you know, I've shown this before, like the way the Bitcoin hash price increases in Bitcoin terms as the difficulty keeps decreasing.

82
00:06:20,140 --> 00:06:21,720
I think it's what you said is very right.

83
00:06:21,820 --> 00:06:25,280
Like everybody just kind of has to decide what do they want.

84
00:06:25,980 --> 00:06:28,300
And it's a very weird thing to think about.

85
00:06:28,680 --> 00:06:31,680
What do you want your economic basis to be?

86
00:06:31,680 --> 00:06:38,140
you know it's a very fundamental question because i was telling you um that the the south african

87
00:06:38,140 --> 00:06:42,320
rand just increased in value versus the dollar and you were saying oh you know that's what we

88
00:06:42,320 --> 00:06:49,860
want as americans um my rent just went from like went up by 250 bucks you know which is roughly

89
00:06:49,860 --> 00:06:57,920
i want to say five percent seven and a half percent over the last six weeks just because of

90
00:06:57,920 --> 00:07:01,740
dollar and rand exchange rate.

91
00:07:03,300 --> 00:07:04,640
And at the same time, Bitcoin drops.

92
00:07:04,640 --> 00:07:05,560
Is your rank denominated into US dollars then?

93
00:07:05,660 --> 00:07:06,160
Is that what you're saying?

94
00:07:07,100 --> 00:07:07,660
No, no.

95
00:07:07,740 --> 00:07:10,460
But I converted because that's often what I earn.

96
00:07:10,860 --> 00:07:11,460
I see.

97
00:07:11,560 --> 00:07:12,600
I earn from overseas.

98
00:07:12,600 --> 00:07:19,760
So, yeah, it's just a very weird time where there is no clear anchor, you know?

99
00:07:20,520 --> 00:07:25,720
And you feel this like on a global stage, but then also, you know, starting to affect

100
00:07:25,720 --> 00:07:27,620
our personal lives, especially as we are.

101
00:07:27,620 --> 00:07:36,980
so exposed to to bitcoin volatility itself um but that's kind of where where this whole

102
00:07:36,980 --> 00:07:42,260
whole bitcoin thing helps me right because i that's essentially my long-term vision is

103
00:07:42,900 --> 00:07:50,340
you know i've i've made the choice long ago to to um value everything in sets right and we were

104
00:07:50,340 --> 00:07:57,060
talking about customers of yours that are now you know maybe slightly uh uh miffed or have negative

105
00:07:57,060 --> 00:08:02,740
sentiment because they bought machines maybe four weeks ago five weeks ago and now the machines are

106
00:08:02,740 --> 00:08:11,540
literally half the price or you know somewhere in between um and i i was just saying look if you

107
00:08:11,540 --> 00:08:17,620
price the machines in bitcoin terms they're still the same price they haven't changed in price right

108
00:08:17,620 --> 00:08:21,540
because the machine i think that's that's that's important right like to to understand that the way

109
00:08:21,540 --> 00:08:26,400
basic pricing works and the way a lot of these economics it's a lot of it is denominated in

110
00:08:26,400 --> 00:08:31,760
bitcoin so if you kind of look at you know the relative bitcoin that you spent versus now and

111
00:08:31,760 --> 00:08:36,260
the value of the machines yes in dollars it may be different but the value of bitcoin in dollars

112
00:08:36,260 --> 00:08:40,480
is different as well so if you're denominating in bitcoin you're actually like you know your

113
00:08:40,480 --> 00:08:47,440
your assets in those machines is actually like pretty much still the same exactly yeah and it's

114
00:08:47,440 --> 00:08:51,500
it's a it's because everything everything that i do you know when you when you make a large

115
00:08:51,500 --> 00:08:59,940
investment when you buy a car you know um i bought a car at this price range um i think over a year

116
00:08:59,940 --> 00:09:07,760
ago and afterwards the price just jumped up to the 100 or whatever and i was like damn what a bad

117
00:09:07,760 --> 00:09:14,920
decision right now i'm like hey the car is the same price again you know so it's like um it's

118
00:09:14,920 --> 00:09:21,740
just a matter of perspective and yes sure like we can buy less um for for the money that we have

119
00:09:21,740 --> 00:09:29,180
and what i feel for most is those people who um who get told hey save your money in sats save your

120
00:09:29,180 --> 00:09:34,300
money in sats um especially those and this is a bit off topic but here in south africa you know

121
00:09:34,300 --> 00:09:39,100
when we when we talk about using bitcoin and we just finished the adopting bitcoin conference

122
00:09:39,100 --> 00:09:45,880
last weekend and a lot of newcomers you know um that are there that for them maybe 500 is a lot

123
00:09:45,880 --> 00:09:50,920
of money right and they if they have it all in sets their halvings just got halved as well and

124
00:09:50,920 --> 00:09:56,020
they might not be able to afford that and that's often you know the volatility is what keeps Bitcoin

125
00:09:56,020 --> 00:10:03,880
from being this um great asset that we that we um think it will be at some point um yeah but then

126
00:10:03,880 --> 00:10:12,580
again, you find volatility everywhere, right? Gold market is what, 30, 34 trillion. And it just

127
00:10:12,580 --> 00:10:20,700
dropped by 20% in over the course of two weeks. Like it's nuts. It dropped more than the market

128
00:10:20,700 --> 00:10:26,440
cap of Bitcoin. And I think like when we think about mining, like we talk about this all the

129
00:10:26,440 --> 00:10:34,660
time, it's like the game is how do I stay online and keep my machines making, you know, generating

130
00:10:34,660 --> 00:10:40,100
Bitcoin in an economically viable way for as long as possible. So I'm able to get the most out of

131
00:10:40,100 --> 00:10:46,520
those machines. And I think like having options to be able to continue running the machines in,

132
00:10:46,560 --> 00:10:52,140
you know, different economic times with deals like profit share deals and, you know, redeploying

133
00:10:52,140 --> 00:10:56,120
machines instead of just sitting on them and scrapping them. Like, I think those options are

134
00:10:56,120 --> 00:11:02,100
emerging as the like reality of the economics kind of play out. And I think there's a world

135
00:11:02,100 --> 00:11:06,900
where like, if you have a machine, you can actually keep the machine running as long as

136
00:11:06,900 --> 00:11:11,280
it's useful life and still be benefiting from that. I think it's going to take a while to get

137
00:11:11,280 --> 00:11:14,300
there. But like, those are the types of things that like we always talk about and we're always

138
00:11:14,300 --> 00:11:19,760
working on is like, okay, given an efficiency of a machine, does this contract structure even make

139
00:11:19,760 --> 00:11:24,240
sense at what point do we need to consider you know turning this into a profit share structure

140
00:11:24,240 --> 00:11:29,520
or re redeploying it somewhere else um because you know customers still want to you know hold

141
00:11:29,520 --> 00:11:42,851
the exposure to the to the productive uh compute asset that is the asic um that that a really interesting thing and i think a lot of people just need to understand and have the right expectations when kind of engaging with the mining market Like

142
00:11:43,311 --> 00:11:48,091
what we're seeing is not anything new. It's just, you know, maybe it's new to some people.

143
00:11:49,291 --> 00:11:53,871
This episode is sponsored by Hashbranch, the easiest way to mine Bitcoin in the US.

144
00:11:54,311 --> 00:11:59,131
Sourcing hardware and rackspace usually is a mess. Bad deal structures, misaligned incentives,

145
00:11:59,131 --> 00:12:01,131
and no visibility after deployment.

146
00:12:01,691 --> 00:12:03,891
I have placed tens of thousands of ASICs

147
00:12:03,891 --> 00:12:05,491
in partnership with Hashbranch.

148
00:12:05,891 --> 00:12:08,191
What I like most is that they cut through the noise.

149
00:12:08,671 --> 00:12:10,331
Fair pricing, aligned terms,

150
00:12:10,571 --> 00:12:12,871
and access to hosting that actually works.

151
00:12:13,251 --> 00:12:14,931
Use code hashredup100.

152
00:12:15,411 --> 00:12:17,331
Visit hashbranch.com today

153
00:12:17,331 --> 00:12:20,491
to get $100 off of each miner that you order.

154
00:12:20,811 --> 00:12:22,891
Yeah, no, again, what you say is right.

155
00:12:22,891 --> 00:12:26,411
The two of us, we always look for opportunities

156
00:12:26,411 --> 00:12:30,911
to make people money while being incentivized,

157
00:12:31,051 --> 00:12:34,171
obviously, to have successful customers, right?

158
00:12:34,191 --> 00:12:35,151
I think it's very important.

159
00:12:35,611 --> 00:12:39,471
We don't charge 30%, 40% margins on these machines.

160
00:12:39,471 --> 00:12:43,091
We don't charge, overcharge on the hosting rate, I think.

161
00:12:44,471 --> 00:12:47,431
And so, you know, I just did a focus episode

162
00:12:47,431 --> 00:12:52,131
that will be released when we release this interview

163
00:12:52,131 --> 00:12:54,191
or this chat, I should say,

164
00:12:54,191 --> 00:12:59,491
where you do hydro hosting for two reasons.

165
00:12:59,631 --> 00:13:01,551
Maybe just to recap that quickly, right?

166
00:13:01,691 --> 00:13:03,531
This is what I'm very, very bullish on

167
00:13:03,531 --> 00:13:05,891
is we do hydro hosting for two reasons,

168
00:13:06,051 --> 00:13:08,991
a long financial lifespan that we have access to

169
00:13:08,991 --> 00:13:11,071
and then a long physical lifespan, right?

170
00:13:11,071 --> 00:13:12,251
There's no moving parts.

171
00:13:12,691 --> 00:13:14,191
The machine doesn't really get in touch

172
00:13:14,191 --> 00:13:15,431
with the environment as much.

173
00:13:15,511 --> 00:13:16,911
There's no air moving through it.

174
00:13:17,551 --> 00:13:20,031
You just have water cooling the machines.

175
00:13:20,611 --> 00:13:22,951
They run at 30, 40 degrees Celsius

176
00:13:22,951 --> 00:13:30,151
years instead of 60 70 80 depending on on ambient temperature and then you know if we sort of solve

177
00:13:30,151 --> 00:13:38,311
for the physical lifespan uh the way we we handle the financial lifespan for situations like this

178
00:13:38,311 --> 00:13:45,031
are profit splits you know where um we just make a portion of the hosting rate variable

179
00:13:45,031 --> 00:13:52,531
um which means that you break even now when hash price is at 30 dollars and you're mining at six

180
00:13:52,531 --> 00:13:58,911
cents, right? But if hash price is at $100, you're mining at 14 cents, right? Sure, that's giving up a

181
00:13:58,911 --> 00:14:07,571
lot of upside, right? But the downside protection, once the machine is near or at ROI, the downside

182
00:14:07,571 --> 00:14:14,931
protection is invaluable because this market is so volatile still, right? And so then the modeling

183
00:14:14,931 --> 00:14:22,251
is less relevant. If you can guarantee that the machine stays online longer, the modeling and the

184
00:14:22,251 --> 00:14:29,111
scenarios become have less of a weight right because the problem comes when you have no other

185
00:14:29,111 --> 00:14:35,291
option than to turn off not to lose money to not lose money right and at that point you're stuck

186
00:14:35,291 --> 00:14:41,331
and you're sitting on less bitcoin that you mined than you could have bought two and a half years

187
00:14:41,331 --> 00:14:49,371
ago but there's a question here tom also right just to critically um uh critique our thinking

188
00:14:49,371 --> 00:14:56,891
as well right if i was devil's advocate now i would say okay what guarantees that the the hosting

189
00:14:56,891 --> 00:15:02,891
provider that actually manages the machines and owns the power contracts and stuff how do we make

190
00:15:02,891 --> 00:15:09,371
sure that they that we select hosts that don't turn around and want to do ai once the contract

191
00:15:09,371 --> 00:15:14,891
period is finished because then the machine is out of there yeah no that's a great question and i

192
00:15:14,891 --> 00:15:20,731
And I think there's like, there's a lot of thoughts there because we have like, on one

193
00:15:20,731 --> 00:15:27,291
hand, we have this like massive explosion in just AI as a, as a technology, right.

194
00:15:27,291 --> 00:15:29,551
Mainly through like generative AI.

195
00:15:29,551 --> 00:15:32,771
And then we're seeing like agents and all these things spin up.

196
00:15:32,771 --> 00:15:37,551
So like, there's a massive kind of exponential growth in that as like a technology.

197
00:15:37,631 --> 00:15:41,931
And then from that is kind of as an asset class in general.

198
00:15:41,931 --> 00:15:47,471
Like I think, you know, the way we look at mining is like one type of compute.

199
00:15:47,671 --> 00:15:58,811
It just happens to be like the most, you know, liquid form of compute that you can basically sell perpetually on the market because the buyer of that compute is the Bitcoin mining network.

200
00:15:58,811 --> 00:16:06,791
Right. So with mining, it's like, hey, the economics might be different, but at least you don't have to solve for who's going to buy my compute.

201
00:16:06,791 --> 00:16:26,391
But at the same time, we have these massive tailwinds that are saying the investment, the global spend in CapEx for AI infrastructure, including the GPUs, the power infrastructure, everything it takes to run AI compute, specifically large training loads, inference, and other things, is in the trillions.

202
00:16:26,571 --> 00:16:28,651
That's what people are publicly saying.

203
00:16:28,751 --> 00:16:34,411
We're talking like $5 to $10 trillion by the next, at the end of the decade, that's what people are putting.

204
00:16:34,411 --> 00:16:53,431
So when you see like this massive market, there are like, you know, there are you think about it, you're like, hey, you know, if I'm building a data center or I'm, you know, if I have access to a lot of power, should I look into AI compute? And I think the answer is like, yeah, you should look into it. There are like a lot of times where the economics make a lot of sense.

205
00:16:53,431 --> 00:16:59,551
it's not easy though like it's not as simple as like you know I get some GPUs and then there's a

206
00:16:59,551 --> 00:17:04,051
there's a Bitcoin equivalent that can just monetize them perfectly the problem that's

207
00:17:04,051 --> 00:17:09,211
still unsolved at least to my knowledge is like clear liquidity for your compute that you're

208
00:17:09,211 --> 00:17:14,071
generating on the AI side so like even the even the groups that are considering this sort of like

209
00:17:14,071 --> 00:17:19,671
transition which I think we'll see a lot of don't get me wrong that it's a very difficult problem so

210
00:17:19,671 --> 00:17:24,111
So I think for that reason, it's going to take a long time to solve.

211
00:17:24,111 --> 00:17:29,291
Maybe people will probably solve it in different ways, but I think that is something we need

212
00:17:29,291 --> 00:17:35,391
to have a conversation around, which is like, hey, if a current host has a bunch of infrastructure

213
00:17:35,391 --> 00:17:41,171
and they decide to transition everything to AI and compute, that is something that could

214
00:17:41,171 --> 00:17:42,171
happen.

215
00:17:42,171 --> 00:17:48,831
For our customers specifically, on one hand, we have very strong relationships and contracts

216
00:17:48,831 --> 00:17:52,931
with our operators. So like, even if that were to occur, it would occur like after our, you know,

217
00:17:52,971 --> 00:17:57,031
client contracts are over and we'd have a redeployment and a relocation kind of strategy

218
00:17:57,031 --> 00:18:01,991
already built in. We're, we're constantly forward planning, but I think realistically, like it's not

219
00:18:01,991 --> 00:18:08,711
going to happen in like a overnight, all capacity is now AI. It's probably a transition. And most

220
00:18:08,711 --> 00:18:15,311
likely I think the best solution is to have a mix of, of compute resources, you know, at any one

221
00:18:15,311 --> 00:18:20,551
location because like right you know for example like right now like if if if you converted everything

222
00:18:20,551 --> 00:18:26,131
to ai it might be hyper profitable today it could also be not super profitable in six to 12 months

223
00:18:26,131 --> 00:18:30,051
depending on how the economics play out so like you kind of want to hedge your revenue streams and

224
00:18:30,051 --> 00:18:35,591
i think that's going to play out at the operator level as well yeah so i think you know you bring

225
00:18:35,591 --> 00:18:41,351
actually a lot of value to operators because now you know besides just being able to self mine

226
00:18:41,351 --> 00:18:44,071
and being exposed to Bitcoin volatility

227
00:18:44,071 --> 00:18:45,611
or doing the AI

228
00:18:45,611 --> 00:18:47,311
where you have to look for your customer

229
00:18:47,311 --> 00:18:47,951
and all of that,

230
00:18:48,531 --> 00:18:50,131
all of the things that you just said,

231
00:18:50,331 --> 00:18:52,671
you also have an aggregator like yourselves

232
00:18:52,671 --> 00:18:54,851
and you just make a margin on the power.

233
00:18:56,351 --> 00:18:57,491
And then also I was thinking

234
00:18:57,491 --> 00:18:58,351
while you were speaking,

235
00:18:58,471 --> 00:19:00,011
how does uptime play into this?

236
00:19:00,431 --> 00:19:01,391
And does Bitcoin mining

237
00:19:01,391 --> 00:19:03,311
maybe move to those points in the grid

238
00:19:03,311 --> 00:19:06,591
where uptime is 85%, 80%?

239
00:19:06,691 --> 00:19:07,151
No, for sure.

240
00:19:07,151 --> 00:19:08,551
I wouldn't mind,

241
00:19:08,551 --> 00:19:15,911
And I wouldn't mind my miner being online 85% of the year if the power price reflected that opportunity.

242
00:19:16,851 --> 00:19:17,071
Exactly.

243
00:19:17,391 --> 00:19:22,731
So like with mining, because you kind of control for like who is the customer, right?

244
00:19:22,731 --> 00:19:24,731
You can always mine Bitcoin.

245
00:19:25,311 --> 00:19:28,451
You just need internet and power and a compute resource and ASIC.

246
00:19:28,931 --> 00:19:35,391
Because of that, the only trade-offs you really can make are uptime and base, call it power or hosting costs.

247
00:19:35,391 --> 00:19:37,051
Those are really the two variables you have.

248
00:19:37,051 --> 00:19:42,331
but the reality is that like when you you know most new machines we talked about this on the

249
00:19:42,331 --> 00:19:46,931
last podcast like we think of like the brand new machines generally speaking you want to start them

250
00:19:46,931 --> 00:19:50,991
at a high uptime environment because you want to you know monetize those as long as possible

251
00:19:50,991 --> 00:19:58,251
but the reality is like economically it might make sense to actually run efficient units at a lower

252
00:19:58,251 --> 00:20:02,711
uptime because your base costs can be lower so really what you're trying to optimize for is like

253
00:20:02,711 --> 00:20:08,771
the time online that is profitable. And like that can be, you know, maybe you're only running 83%

254
00:20:08,771 --> 00:20:13,871
of the time, but your base is much lower. So you're actually able to stay online. And so those

255
00:20:13,871 --> 00:20:21,111
are the things that happen with AI and kind of the whole, you know, the compute markets around,

256
00:20:21,111 --> 00:20:26,011
call it like GPU compute, we kind of coined this term industrial compute, like mining is one

257
00:20:26,011 --> 00:20:30,651
type, but you have kind of this, this other type that that's emerging with GPUs.

258
00:20:31,531 --> 00:20:36,991
Like uptime is like very interesting because in order to even be able to monetize them,

259
00:20:36,991 --> 00:20:42,871
well, you have to hit a strict requirement for like even selling that compute.

260
00:20:43,091 --> 00:20:46,271
Meaning like you need to basically be a tier three data center.

261
00:20:47,071 --> 00:20:51,851
A tier three data center is like, you know, I don't know all the technical kind of specifications,

262
00:20:51,851 --> 00:20:54,311
but it's like very high uptime, very high redundancy.

263
00:20:54,311 --> 00:20:57,571
We're talking like 99.999% uptime.

264
00:20:58,251 --> 00:21:07,171
A lot of times you need, I think on some of the compute marketplaces, you literally need access to two separate grid-tied substations for redundancy.

265
00:21:07,851 --> 00:21:11,551
And you need, I think, fiber as well as another backup.

266
00:21:12,351 --> 00:21:14,751
So there's a lot of infrastructure that's even required.

267
00:21:15,431 --> 00:21:18,431
And so, yes, I think mining, but there's energy everywhere.

268
00:21:18,431 --> 00:21:24,471
So I think with, you know, mining will basically be able to be spun up and economical in places that don't yet have that infrastructure.

269
00:21:24,591 --> 00:21:30,371
Over time, it will make sense to maybe bring that infrastructure in and then start to diversify revenue streams.

270
00:21:30,491 --> 00:21:40,271
But like this stuff moves at the speed of the physical world, which means like to build a data center and actually do this, we're talking, you know, months, if not years.

271
00:21:40,271 --> 00:21:47,691
Like a lot of the capacity and a lot of the big infrastructure that's getting, that's coming online this year was all contracted like three years ago.

272
00:21:47,691 --> 00:21:49,411
So that's what we're talking about.

273
00:21:50,451 --> 00:21:50,551
Yeah.

274
00:21:51,431 --> 00:21:54,451
Yeah, no, look, I just also before we started recording,

275
00:21:54,591 --> 00:21:58,871
we read the news courtesy of the friends at the MinerMag

276
00:21:58,871 --> 00:22:04,931
that there were 250 gigawatts of power requests

277
00:22:04,931 --> 00:22:07,051
in the ERCOT queue being reshuffled.

278
00:22:09,011 --> 00:22:13,551
So it's just such an interesting time in energy and AI

279
00:22:13,551 --> 00:22:14,431
and Bitcoin mining.

280
00:22:14,431 --> 00:22:20,351
and I just have a feeling that getting a machine today

281
00:22:20,351 --> 00:22:23,291
which costs the same amount of Bitcoin as it did four weeks ago

282
00:22:23,291 --> 00:22:29,451
might just be a really, really good decision in five years from now

283
00:22:29,451 --> 00:22:32,931
especially if it's still running then and you can guarantee that

284
00:22:32,931 --> 00:22:38,151
and you have access because the hostess that we deal with

285
00:22:38,151 --> 00:22:43,671
they don't qualify for AI if their uptime is 90%

286
00:22:43,671 --> 00:22:52,451
95%. They don't have that option, right? They have all that capital investment. And if the grid

287
00:22:52,451 --> 00:22:59,931
operator only offers them 90% uptime, which they have control over, the contracts that they sign,

288
00:23:00,551 --> 00:23:07,491
then AI isn't an option. Right, at least not today. And not with a massive, you know, retrofit

289
00:23:07,491 --> 00:23:20,701
and and capex spend and even like just the yeah it it is a it something that can happen and probably will in in in some places i don think it going to happen across the board Like everywhere that was mining is now AI

290
00:23:21,121 --> 00:23:22,781
I think that's what people kind of say.

291
00:23:22,861 --> 00:23:24,361
And that's the like simple answer.

292
00:23:24,461 --> 00:23:26,121
But I think it's going to be more complicated than that.

293
00:23:26,601 --> 00:23:41,081
The other thing I'll throw out there, and this is like just kind of a maybe an AI black swan or like something to kind of like just get thinking is we just like recently saw, I don't know if you saw the news, but like there's SpaceX is basically going to be merging with XA.

294
00:23:41,101 --> 00:23:47,621
AI, so like kind of the Elon's like AI company and like SpaceX, which basically owns like

295
00:23:47,621 --> 00:23:52,021
80% of the market of, you know, sending payload to orbit and they have, you know, Starlink

296
00:23:52,021 --> 00:23:53,021
and whatnot.

297
00:23:53,021 --> 00:23:55,641
And you can read like how they're talking about this.

298
00:23:55,641 --> 00:24:00,061
They think that a lot of compute is going to move to space.

299
00:24:00,061 --> 00:24:04,461
Now, whether that happens or not, and that completely changes the economics.

300
00:24:04,461 --> 00:24:06,881
So like, there's a lot of things that could happen.

301
00:24:06,881 --> 00:24:12,541
like the build out for ai to you know we'll say on the on the ground in you know those economics

302
00:24:12,541 --> 00:24:18,721
could completely change if it's actually possible to do this at scale for you know cheap yeah outside

303
00:24:18,721 --> 00:24:23,181
of like i guess the planet which is a weird thing to say so so i don't know what are we going to see

304
00:24:23,181 --> 00:24:29,841
mining in space yeah yeah exactly solar mining you're far enough out i had a i had the guy on um

305
00:24:29,841 --> 00:24:36,201
the earth is for mining space is for ai yeah yeah well you know you you get exactly you have solar

306
00:24:36,201 --> 00:24:42,921
uptime 100 i had this guy on talking about mining in space no um no for sure but it's it's definitely

307
00:24:44,681 --> 00:24:50,041
yeah very interesting time right now i would like to um ask you tom how do you or just explain to

308
00:24:50,041 --> 00:24:55,161
people briefly you know let's say they have a fleet of machines um with other holsters somewhere

309
00:24:55,161 --> 00:25:00,921
and they're not not profitable be it k pros i don't know whatever it is how do people do this

310
00:25:00,921 --> 00:25:05,721
like what is their first move after they contact you what do you do with these machines and maybe

311
00:25:05,721 --> 00:25:13,641
briefly explain how the profit split mechanism works yeah for sure so like a a big part of our

312
00:25:13,641 --> 00:25:20,841
business like from the get-go was all around this idea of redeployment and it's basically like hey

313
00:25:20,841 --> 00:25:26,201
i have this hardware already um i bought it you know whatever price here's the efficiency of the

314
00:25:26,201 --> 00:25:33,161
hardware it's now out of contract at another facility or it is sitting on a pallet somewhere

315
00:25:33,161 --> 00:25:40,481
It's not being well monetized. What can you do? Right. And so we like we've been working on this problem for years now.

316
00:25:40,481 --> 00:25:52,161
And basically, like the way it works for us is like a customer will come to us and say, hey, like I have X amount of units of this making, you know, makeup and efficiency.

317
00:25:52,441 --> 00:25:59,421
We basically intake all of their hardware in terms of like, OK, how many units you have? What types of units? What are the efficiencies? Where are they?

318
00:25:59,421 --> 00:26:01,081
and what's the situation?

319
00:26:01,281 --> 00:26:03,341
When do you think you can actually move them?

320
00:26:04,121 --> 00:26:06,541
Once we have all that, we basically say,

321
00:26:06,601 --> 00:26:08,061
okay, what's the quality of the machines?

322
00:26:08,681 --> 00:26:10,621
Because a lot of people don't know this,

323
00:26:10,701 --> 00:26:13,661
but a lot of machines, when you originally bought them,

324
00:26:13,701 --> 00:26:14,581
they've been running somewhere.

325
00:26:15,081 --> 00:26:16,621
You don't know how well they've been taken care of.

326
00:26:16,621 --> 00:26:17,661
They might've been running,

327
00:26:17,781 --> 00:26:20,061
but maybe the filtration wasn't great

328
00:26:20,061 --> 00:26:21,341
and you have debris everywhere.

329
00:26:21,901 --> 00:26:24,141
So we try to get an understanding for the quality.

330
00:26:24,621 --> 00:26:26,901
This comes ideally through hash reports.

331
00:26:27,181 --> 00:26:29,261
So what's the operating status of the machines

332
00:26:29,261 --> 00:26:31,581
at least as close as you can give us.

333
00:26:31,721 --> 00:26:33,561
So like if they're running right now, great.

334
00:26:33,661 --> 00:26:35,661
If they were running, do you have any data on that?

335
00:26:36,241 --> 00:26:39,461
The second piece is like, okay, what's the quality of the machine?

336
00:26:39,541 --> 00:26:42,561
So this is kind of subjective, but basically like, are they dusty?

337
00:26:42,921 --> 00:26:44,801
Are they, you know, have they have corrosion?

338
00:26:45,021 --> 00:26:46,121
Like what does that look like?

339
00:26:47,381 --> 00:26:50,061
Once we have all that data, like it's pretty straightforward.

340
00:26:50,441 --> 00:26:52,941
On the back end, what we've done is we've contracted out

341
00:26:52,941 --> 00:26:57,161
and we have rack space that like basically matches the profile of these units.

342
00:26:57,161 --> 00:27:04,461
And so what happens is like we kind of talked about this earlier, you might have contracts that have low base rate, low uptime.

343
00:27:04,861 --> 00:27:07,901
You wouldn't put a new unit there, but you might put a K pro or an XP.

344
00:27:08,641 --> 00:27:13,841
That's like the first piece. So we have contracts in place that allow us to kind of redeploy older units.

345
00:27:14,321 --> 00:27:21,941
The second piece is the profit share component. So in these contracts, you kind of have the ability to, hey, I just want a low rate, low uptime.

346
00:27:21,941 --> 00:27:27,541
I'm cool with that. Or I actually want an even lower rate and then have a split of the kind of

347
00:27:27,541 --> 00:27:33,701
of the of the rewards less the costs with the operating party. That's like easy to describe.

348
00:27:33,741 --> 00:27:38,221
It's actually hard to pull off because there's a lot of like mechanisms around billing and how are

349
00:27:38,221 --> 00:27:43,721
you calculating this and how are you settling and all that stuff. Our software really orchestrates

350
00:27:43,721 --> 00:27:47,781
that. And then we have the contracts on the back end that basically is the rack space that can take

351
00:27:47,781 --> 00:27:51,981
your machine. So once we've kind of like, okay, we've, we know what types of machines there are,

352
00:27:52,021 --> 00:27:57,261
we get another understanding. Um, we have the rack space ready. Um, the next step is like,

353
00:27:57,261 --> 00:28:02,921
if the units are not in good quality and I'll say like, as an example, Hey, they've been sitting on

354
00:28:02,921 --> 00:28:07,801
pallets. I don't even have pictures. You know, some of them are broken. We don't know. We generally

355
00:28:07,801 --> 00:28:13,561
have a checkpoint, uh, before the redeployment. Um, sometimes this is at the site. So like sometimes

356
00:28:13,561 --> 00:28:19,161
a site is capable of doing the checkpoint sometimes we have to send them to a repair stopover to like

357
00:28:19,161 --> 00:28:24,761
get checked basically what that means is like they will take it in diagnose them be like okay

358
00:28:24,761 --> 00:28:29,321
you know x amount of machines are actually good to go the others need maybe a new power supply

359
00:28:29,321 --> 00:28:33,801
new hash board whatever and they do it like a cleaning basically to get them ready for the

360
00:28:33,801 --> 00:28:38,281
redeployment because a lot of operators that we work with like the last thing they want is units

361
00:28:38,281 --> 00:28:42,761
to show up that just don't work or junk because they they want to monetize at this time yeah

362
00:28:42,761 --> 00:28:48,761
yeah it's also an objective checkpoint right these repair centers they're being paid exactly

363
00:28:48,761 --> 00:28:55,241
the machines and they're an objective neutral party so there's also no no discourse after the

364
00:28:55,241 --> 00:29:00,521
fact where hash branch says no the units are shite basically and the customer says no they were all

365
00:29:00,521 --> 00:29:05,561
working when I sent them off what happened right so no you have somebody in the middle that says

366
00:29:05,561 --> 00:29:11,201
hey you know pay me I don't know 20 a machine and I will give you a report on it and give you a

367
00:29:11,201 --> 00:29:14,641
sheet that tells you exactly what the state of these machines is at the moment.

368
00:29:15,541 --> 00:29:18,721
BitX, solomining.de, got it.

369
00:29:18,861 --> 00:29:20,661
Premium, made in Germany.

370
00:29:20,821 --> 00:29:21,621
You want to go fancy?

371
00:29:22,101 --> 00:29:25,661
Go for NerdX or NerdQX++ for a lot of power.

372
00:29:26,041 --> 00:29:29,301
Worldwide shipping, check out solomining.de today.

373
00:29:29,781 --> 00:29:30,101
Exactly.

374
00:29:30,301 --> 00:29:33,001
And like ideally you get that before they leave the original facility.

375
00:29:33,241 --> 00:29:34,621
You kind of, they've taken good care of it.

376
00:29:34,681 --> 00:29:35,201
They have the data.

377
00:29:35,361 --> 00:29:36,481
It's pretty, it's pretty simple.

378
00:29:36,481 --> 00:29:42,301
but we have a lot of cases where like I said units are just in storage containers we don't even know

379
00:29:42,301 --> 00:29:46,461
how they've been treated so that checkpoint is really important to like one just like give you

380
00:29:46,461 --> 00:29:51,921
a reality check because like even if you get these units plugged in without a checkpoint if only

381
00:29:51,921 --> 00:29:57,401
you know 70% of them even hash you have no idea what's going on so that's like an important step

382
00:29:57,401 --> 00:30:02,801
and yeah like it's a you know it's it's it's a verified like repair center or someone that we've

383
00:30:02,801 --> 00:30:06,881
worked with who's like done this with us. They don't really have skin in the game to like, you

384
00:30:06,881 --> 00:30:12,201
know, on either side, they're just like, Hey, like, this is the reality. You know, we get diagnostic

385
00:30:12,201 --> 00:30:16,181
reports and photos and all that stuff, really trying to just like streamline the process. So

386
00:30:16,181 --> 00:30:21,881
once you've kind of given us the intake data, we've kind of made a decision, like, should it get

387
00:30:21,881 --> 00:30:28,121
checked at an at a third party? Or does the site have a process? Then we go through the redeployment.

388
00:30:28,121 --> 00:30:33,521
And like, even that is kind of complex because the way, you know, these are physical servers moving.

389
00:30:33,521 --> 00:30:45,121
So like, as an example, I always talk about each one of the, um, each one of these ASICs is about 35 pounds, like a standard, um, uh, standard like air cool unit called like an S19 K pro or XP.

390
00:30:45,661 --> 00:30:46,881
It's like 35 pounds.

391
00:30:46,881 --> 00:30:47,121
Right.

392
00:30:47,141 --> 00:30:51,381
And you can only fit maybe, you know, 20 to 30 of them on a pallet.

393
00:30:51,821 --> 00:30:58,061
Um, and so we're talking about like, that's, you know, right there, we're talking about, you know, thousands of pounds, you know, maybe like.

394
00:30:58,121 --> 00:31:02,601
a thousand to fifteen hundred pounds of servers depending a lot of these deployments like it might

395
00:31:02,601 --> 00:31:07,261
be five to ten pallets and so we're talking like on the order of you know tens of thousands of

396
00:31:07,261 --> 00:31:11,681
pounds of servers that need to like go through this process and they need to physically move

397
00:31:11,681 --> 00:31:16,581
across you know we only operate in the U.S. so they're moving across um across the country

398
00:31:16,581 --> 00:31:21,241
sometimes just takes time so like this the logistics of getting the deployment going

399
00:31:21,241 --> 00:31:25,841
ideally you can just snap your fingers and they're online in another spot but the reality is like

400
00:31:25,841 --> 00:31:30,241
they're going to come in batches those batches are going to have serial numbers associated the

401
00:31:30,241 --> 00:31:34,921
site needs to understand you know what's going on with those it takes time to ramp those up

402
00:31:34,921 --> 00:31:40,741
um but yeah once they're going and you're kind of in this better structure then like we've seen

403
00:31:40,741 --> 00:31:48,321
like machines just stay online longer because the economics make sense yeah no exactly perfect

404
00:31:48,321 --> 00:31:55,701
thank you so much yeah it's a cool problem before we yeah it's absolutely um and there's

405
00:31:55,701 --> 00:32:02,761
it seems endless demand for it i've seen somebody today advertise uh m30 units stacked 80 on a

406
00:32:02,761 --> 00:32:08,321
pallet i'm like no thank you yeah that's gonna that's gonna tip over in transit and then people

407
00:32:08,321 --> 00:32:13,661
are gonna be pointing at each other and that bottom row is probably feeling a lot of pain

408
00:32:13,661 --> 00:32:19,981
yeah um yeah no last time we talked a bunch about type one and type two facilities also

409
00:32:19,981 --> 00:32:29,661
and going back to um the the hydro idea and again please if you haven't look at my little uh

410
00:32:30,301 --> 00:32:37,021
focus episode on on the hydro hosting um where i go into a lot more detail on this right right the

411
00:32:37,021 --> 00:32:42,541
the idea is to have a machine deployed and then be able without touching it and without moving it

412
00:32:42,541 --> 00:32:48,221
to just change the business model to keep it more profitable um and change it to a type two facility

413
00:32:48,221 --> 00:32:54,861
without having to physically move the machine um what else tom do we have anything else or do we do

414
00:32:55,901 --> 00:32:59,821
let me let me talk let me talk a little bit about hydro um because i think it's really really

415
00:32:59,821 --> 00:33:05,821
interesting like yeah at the end of the day as the the asics themselves so everyone talks about um

416
00:33:06,381 --> 00:33:10,941
an asic i'm you know i bought an asic all this stuff what you're buying is or like what you have

417
00:33:10,941 --> 00:33:16,941
is like an ASIC server really. And like, it's basically just a chassis with a bunch of, um,

418
00:33:16,941 --> 00:33:21,741
hash boards and each hash board has, you know, uh, dozens of these individual ASIC chips.

419
00:33:22,461 --> 00:33:28,461
And so like the reality is like with these chips, as the efficiency increases over time,

420
00:33:28,461 --> 00:33:33,021
and we have OEMs releasing new variants, like at some point you're starting to hit like physical

421
00:33:33,021 --> 00:33:38,141
reality in terms of like how you can actually dissipate the heat that comes off of these chips.

422
00:33:38,141 --> 00:33:43,561
And so with hydro, it's really interesting. Like you have, it's basically direct to chip cooling.

423
00:33:43,561 --> 00:33:47,421
And this has been, you know, this is something that's existed in just data centers forever.

424
00:33:47,421 --> 00:33:52,341
But with, with, with mining, we're seeing like with this direct to chip cooling, specifically

425
00:33:52,341 --> 00:33:58,381
in a like liquid form, you kind of have like a mix of water and like a glycol mix. It's basically

426
00:33:58,381 --> 00:34:03,961
like shooting, you know, cold water basically on these chips and cycling them in real time.

427
00:34:03,961 --> 00:34:07,421
And so you're getting like very precise cooling at the chip level.

428
00:34:08,081 --> 00:34:14,481
And so like that's really important because the quality of the chip basically degrades

429
00:34:14,481 --> 00:34:15,961
exponential with the temperature.

430
00:34:15,961 --> 00:34:21,421
So what that means is like if the temperature is stable in a good, you know, cool threshold,

431
00:34:21,921 --> 00:34:24,181
the chip is stable for its life.

432
00:34:24,381 --> 00:34:26,301
Like, you know, that's how it works.

433
00:34:26,561 --> 00:34:30,721
As you kind of ratchet up temperature, the failure rates go exponential, meaning like

434
00:34:30,721 --> 00:34:33,181
you're going to fail more and more often just because there's more heat.

435
00:34:33,181 --> 00:34:38,181
And so with hydro, because it's such a more efficient cooling mechanism,

436
00:34:38,181 --> 00:34:45,181
one, you get more stable chipsets for longer, which means that you can run these things for longer over time.

437
00:34:45,181 --> 00:35:06,572
You have the ability to overclock them without losing efficiency which is important because if you can control the temperature you don lose as much efficiency as if you were overclocking an air unit And the last thing is they way more power dense because you can control heat better So that means you get a lot more in a smaller volume of space

438
00:35:06,572 --> 00:35:10,932
And so you can be pulling a lot more power, and it just is a much more efficient layout.

439
00:35:11,112 --> 00:35:18,692
So we love hydro, and we see a lot of OEMs shifting and exploring more hydro-cooled options.

440
00:35:19,152 --> 00:35:35,892
But yeah, I think it's really interesting. This is something that's existed forever, but we're seeing it start to be really kind of the standard. I think air-cooled will always exist because you can't set up high. If you're just going to send the unit somewhere and run it in a chicken coop or something, it probably won't spin up hydro.

441
00:35:35,892 --> 00:35:45,312
But I think like a majority of my, my, my bet would be a majority of mining is going to be hydro, maybe not immediately, but like the production, like there's going to be more hydro.

442
00:35:46,072 --> 00:35:48,332
But that takes time to build out capacity.

443
00:35:48,332 --> 00:35:50,832
So it's kind of a, yeah.

444
00:35:51,752 --> 00:35:57,352
So gun to your head, how long does, do these hydro machines last and what temperatures are you seeing?

445
00:35:58,372 --> 00:35:58,672
Yeah.

446
00:35:58,872 --> 00:36:01,732
So I'll answer both questions.

447
00:36:01,732 --> 00:36:09,792
So on the temperature side, what we've been doing for the last probably like four months or so is we've actually been trying to benchmark.

448
00:36:09,972 --> 00:36:11,652
And I'll have more data that I can share.

449
00:36:11,732 --> 00:36:12,952
I mean, we can do another episode.

450
00:36:13,452 --> 00:36:15,932
We've been kind of trying to benchmark all OEM hydro setups.

451
00:36:15,932 --> 00:36:19,852
And so we currently have every, and OEM is like the manufacturer.

452
00:36:19,852 --> 00:36:32,932
So we currently have just about every one of the latest hydro-cooled variants all running at, you know, the same kind of, we'll say the same geography, kind of trying to control for as much as possible.

453
00:36:33,592 --> 00:36:36,612
And we're just, one thing is we're just seeing stable temps across the board.

454
00:36:37,152 --> 00:36:42,052
I'll give you more detail kind of as the data comes out, like which chips are performing better or whatever.

455
00:36:42,852 --> 00:36:44,852
But like, it's really interesting to see.

456
00:36:44,972 --> 00:36:48,092
So we'll have some benchmarks around the manufacturers there.

457
00:36:48,092 --> 00:37:12,092
And then the other question was like, gun to my head, you know, how long would these things last? I would say, given the data we're seeing in terms of just stability of temperature, the physical life of these machines is going to be like, six plus years in terms of like, they can actually run for that long, probably longer, but the economic life is totally dependent on, you know, the contract that you that you have in place.

458
00:37:12,092 --> 00:37:16,012
um so that's a different thing but i think like my you know gun to my head and like you know

459
00:37:16,012 --> 00:37:21,492
don't hold me to it i think it's the wear and tear and the quality is basically like there's

460
00:37:21,492 --> 00:37:26,112
no wear and tear basically um assuming you're you're controlling for things like uh you know

461
00:37:26,112 --> 00:37:30,792
the pressure and all these things in the fluid but i think these things will last a long time

462
00:37:30,792 --> 00:37:34,372
uh you just have to find you know have the right place to put them and run them for the long term

463
00:37:34,372 --> 00:37:41,852
so what is the highest um probability of failure what do you think has the highest probability

464
00:37:41,852 --> 00:37:50,652
which part of failing in those six years or after those six years on what do you do if that happens

465
00:37:50,652 --> 00:38:00,772
yeah it's a great question i think i think it comes down to like basically the so when you're

466
00:38:00,772 --> 00:38:04,572
moving this liquid through through a loop continuously you need to like make sure that

467
00:38:04,572 --> 00:38:09,332
the flow rate is sufficient to basically pull the heat out and do it well so like if you have

468
00:38:09,332 --> 00:38:15,012
anything that happens with the the pumping mechanism where you're not the flow rate drops

469
00:38:15,012 --> 00:38:20,652
like you're not moving the hot liquid out fast enough um you can see like chip level failures

470
00:38:20,652 --> 00:38:24,592
right because the heat will pick up so i think like controlling the pumping mechanisms and making

471
00:38:24,592 --> 00:38:29,012
sure the flow rate is is is good is like that's kind of an operator thing but like that's one

472
00:38:29,012 --> 00:38:34,892
way that these could fail um i think the other way is like it's it's it's somewhat similar and

473
00:38:34,892 --> 00:38:40,012
kind of just affects the flow rate but like each of these machines has a specific mix of it's super

474
00:38:40,012 --> 00:38:45,452
interesting there's a specific mix of glycol to water ratio that you need in the in the fluid that

475
00:38:45,452 --> 00:38:50,652
goes through and it's slightly different per manufacturer they might recommend a percentage

476
00:38:50,652 --> 00:38:54,812
of glycol to water that's different than another manufacturer now the other thing you have to

477
00:38:54,812 --> 00:39:00,892
control for is like the ambient temperature because if you have more water to glycol ratio

478
00:39:00,892 --> 00:39:05,472
and at lower temperature, you might be more susceptible to like that stuff freezing or not

479
00:39:05,472 --> 00:39:11,092
flowing as well as quickly. So there's all these like, uh, almost like fluid mechanics and stuff to

480
00:39:11,092 --> 00:39:15,132
get, to get, to get right. But, um, I guess like maybe your question was like, what fails most

481
00:39:15,132 --> 00:39:18,972
likely? It's probably like something like that, like messing up the flow rate or the glycol mix.

482
00:39:19,272 --> 00:39:24,792
Um, obviously power surges can always happen. Um, yeah. What about the PSU?

483
00:39:24,792 --> 00:39:31,672
yeah yeah the psu um probably like i would say like of the thing that may fail quicker is probably

484
00:39:31,672 --> 00:39:36,792
the power supply that those things are easily swapped out um really what you want to make sure

485
00:39:36,792 --> 00:39:41,872
is that your chips are good if your chips are good like you're good yeah and that the operator

486
00:39:41,872 --> 00:39:47,072
is is good right because now you just said all of these machines have different kind of cooling

487
00:39:47,072 --> 00:39:53,172
requirements so now when all of these machines are in one container chained up in a loop right

488
00:39:53,172 --> 00:39:57,372
You need to have a partner that really knows what they're doing.

489
00:39:57,652 --> 00:40:01,792
And it sounds like you found exactly that, which is obviously very good news.

490
00:40:02,712 --> 00:40:03,472
Yeah, exactly.

491
00:40:03,832 --> 00:40:06,472
The other thing that's interesting is the form factor of these.

492
00:40:06,472 --> 00:40:15,392
Like we're seeing a trend to going towards a traditional like cabinet, like a U style form factor for these hydro cooled machines.

493
00:40:15,832 --> 00:40:18,372
So you'd like single U, two U, three U.

494
00:40:18,472 --> 00:40:22,372
They're all the same form factor horizontally and you can stack them on top of each other.

495
00:40:23,172 --> 00:40:31,752
But some infrastructure is not designed to take those yet because a lot of the earliest hydro machines that came onto the market were like, we call them shoebox form factors.

496
00:40:31,752 --> 00:40:38,112
They're like way different. So like we find a lot of interesting times where people will, oh, you have hydro capacity.

497
00:40:38,612 --> 00:40:42,632
Yeah, you have hydro capacity. Great. I bought a bunch of these. Can I send them?

498
00:40:42,732 --> 00:40:47,412
And like no one's thinking, will it actually fit? Because that's a whole different thing.

499
00:40:47,412 --> 00:40:51,592
And with air, it's kind of like, do we have enough CFM running through?

500
00:40:52,092 --> 00:40:54,132
Do the PDUs work and the cables work?

501
00:40:54,172 --> 00:40:56,272
And as long as we're clearing enough air, you're good.

502
00:40:56,852 --> 00:40:58,652
With hydro, it's just a little more complex.

503
00:40:58,832 --> 00:41:01,532
But the trade-off is they're going to last longer.

504
00:41:04,192 --> 00:41:06,712
Tom, let's do a round of rapid fire for the shorts.

505
00:41:06,732 --> 00:41:06,892
Cool.

506
00:41:07,572 --> 00:41:07,932
You ready?

507
00:41:08,892 --> 00:41:09,312
Let's go.

508
00:41:09,432 --> 00:41:09,612
Cool.

509
00:41:11,112 --> 00:41:11,952
Short and sweet.

510
00:41:11,952 --> 00:41:16,692
One thing hydro hosting does better than any other cooling method.

511
00:41:17,412 --> 00:41:25,092
kind of already answered it but yeah go uh dissipate heat stable temps and just high uptime

512
00:41:26,052 --> 00:41:32,452
why not immersion okay immersion is interesting because from like a physics perspective actually

513
00:41:32,452 --> 00:41:38,612
immersion has a better like theoretical uh kind of it can cool better theoretically because you

514
00:41:38,612 --> 00:41:45,492
just have more surface area to to pull heat from um in reality it's a lot more i think it's a more

515
00:41:45,492 --> 00:41:51,452
complicated system because you have massive amounts of fluid that needs to be kind of changed

516
00:41:51,452 --> 00:41:56,652
out. You also have this idea of like, oh, I can take an air-cooled machine, strip the fans off,

517
00:41:56,792 --> 00:42:00,832
strip the grates off and dunk it in fluid. That doesn't always work because an air-cooled machine

518
00:42:00,832 --> 00:42:07,792
is designed for airflow, not liquid flow. So we've seen a lot of failures there. I think there are

519
00:42:07,792 --> 00:42:13,192
immersion-specific types of machines that can go in immersion-specific containers, and that does

520
00:42:13,192 --> 00:42:18,552
great but i do think in our experience like it just actually costs more to get right um so like

521
00:42:18,552 --> 00:42:25,592
you know theoretically it's there and practically it's yeah practically it's not there i i just let

522
00:42:25,592 --> 00:42:30,632
me add something there because um and now we're wearing off the rapid fire but it's fine we'll

523
00:42:30,632 --> 00:42:39,912
get back there um my editor can do his magic um the you run the risk of the oil becoming conductive

524
00:42:39,912 --> 00:42:46,152
that cannot happen with hydro right um so if you don't if you have an operator that doesn't really

525
00:42:46,152 --> 00:42:51,972
know what they're doing they're not keeping the air clean and the basins are open or whatever and

526
00:42:51,972 --> 00:42:59,852
oil um gets sort of fed with particles then that oil becomes conductive at some point anyway let's

527
00:42:59,852 --> 00:43:03,272
check on it's a lot cleaner you need a lot cleaner to set up whereas like with hydro it's a

528
00:43:03,272 --> 00:43:07,392
a closed loop direct to chip so you're only controlling your control you have to control

529
00:43:07,392 --> 00:43:09,232
less volume basically.

530
00:43:10,252 --> 00:43:10,312
Yeah.

531
00:43:11,192 --> 00:43:11,932
All right.

532
00:43:12,152 --> 00:43:14,672
Fixed price contracts or flexible pricing?

533
00:43:15,012 --> 00:43:16,812
Which works better long term?

534
00:43:17,152 --> 00:43:19,052
So let me rephrase this.

535
00:43:19,232 --> 00:43:20,692
If you had to choose today

536
00:43:20,692 --> 00:43:22,132
between fixed pricing

537
00:43:22,132 --> 00:43:23,892
or a profit split from the get-go,

538
00:43:24,012 --> 00:43:24,672
what would you choose?

539
00:43:26,032 --> 00:43:28,392
In this market today,

540
00:43:28,932 --> 00:43:32,012
I think it depends on the efficiency

541
00:43:32,012 --> 00:43:32,752
of the hardware.

542
00:43:33,592 --> 00:43:35,692
But I would say like a profit split

543
00:43:35,692 --> 00:43:37,692
is a great option right now

544
00:43:37,692 --> 00:43:38,972
because you're hedging your downside

545
00:43:38,972 --> 00:43:42,372
and we're in a market where things are dropping.

546
00:43:43,192 --> 00:43:45,292
I think both are great.

547
00:43:45,392 --> 00:43:46,972
I'd say right now, probably profit split,

548
00:43:47,112 --> 00:43:49,572
but realistically, depending on the type of machine,

549
00:43:50,272 --> 00:43:52,072
it might still make sense to do fixed rate,

550
00:43:52,152 --> 00:43:54,152
albeit maybe at a lower uptime or something like that.

551
00:43:55,052 --> 00:43:57,132
Worst contract close you've ever seen

552
00:43:57,132 --> 00:43:58,212
in a hosting contract?

553
00:43:59,292 --> 00:44:01,332
Oh my gosh, that's a good one.

554
00:44:01,332 --> 00:44:10,332
um i think gpt wrote that one yeah that's a really really good one i think um

555
00:44:11,012 --> 00:44:16,692
worst contract clause i've ever seen i'll have to think about that one i think one is like basically

556
00:44:16,692 --> 00:44:24,352
the unlimited ability for the operator to just like own the machines with very you know like

557
00:44:24,352 --> 00:44:29,812
low barrier to that kicking on so like as an example hey if we if we if you don't pay something

558
00:44:29,812 --> 00:44:34,192
in seven days like we own your we own your machines now like there's there's very nuanced

559
00:44:34,192 --> 00:44:40,052
things where like you have risk of like you know someone else taking title to your machine with

560
00:44:40,052 --> 00:44:45,232
you know unreasonable things so i think that's like that's one um yeah there's probably a lot

561
00:44:45,232 --> 00:44:52,092
though so that's probably the worst if you had to choose extend life via a profit split or sell

562
00:44:52,092 --> 00:44:56,692
the machine and you're not allowed to say depends on the market

563
00:44:56,692 --> 00:45:05,752
for me probably extend life through a profit split because i'm just i can mine for longer

564
00:45:05,752 --> 00:45:11,212
and then i have the option to sell when the market is is better for selling so that's probably

565
00:45:11,212 --> 00:45:16,652
what i would say all right but it depends on the market screw you

566
00:45:16,652 --> 00:45:26,192
yeah that's it we're done thank you tom this was a pleasure once again um i'm glad we did this and

567
00:45:26,192 --> 00:45:26,992
I didn't go to bed.

568
00:45:28,252 --> 00:45:29,912
Guys, you already know where to find Tom.

569
00:45:30,332 --> 00:45:32,172
I did a previous episode with him

570
00:45:32,172 --> 00:45:34,152
not too long ago. Please feel free

571
00:45:34,152 --> 00:45:35,752
to check him out in the subscription

572
00:45:35,752 --> 00:45:37,692
and the subscription in the description.

573
00:45:38,372 --> 00:45:40,052
Leave a sub, though. Leave a comment

574
00:45:40,052 --> 00:45:41,672
and a like and all of that good stuff

575
00:45:41,672 --> 00:45:43,532
so that sponsors like Tom

576
00:45:43,532 --> 00:45:46,112
can see that people value this content.

577
00:45:46,292 --> 00:45:48,112
It's very important for them so that they can

578
00:45:48,112 --> 00:45:49,032
show them good data.

579
00:45:50,412 --> 00:45:51,652
Guys, thank you so much.

580
00:45:51,932 --> 00:45:53,912
Stay strong in this tough

581
00:45:53,912 --> 00:45:55,112
hash price environment.

582
00:45:55,112 --> 00:46:02,492
let me just check where are we at the end of this recording we are at 28 now we used to be at 29 44

583
00:46:02,492 --> 00:46:09,332
remember those good times and yeah difficulty adjustment downwards is coming guys stay online

584
00:46:09,332 --> 00:46:15,732
if you're interested in getting your feet wet and getting into hosting now i still think is a good

585
00:46:15,732 --> 00:46:22,972
time and yeah if you have any questions let me know tom anything else quick uh no thanks for the

586
00:46:22,972 --> 00:46:26,212
time cool see you next time cheers bye