WEBVTT

1
00:00:00.100 --> 00:00:14.840
[upbeat music] What's going on, y'all? Welcome back to Blockspace Live, brought to you by CleanSpark. Charlie, I'm not even an Elon fanboy, but I gotta say, that intro video got me pretty fired up, man.

2
00:00:14.960 --> 00:00:16.600
We're going to space, Colin.

3
00:00:16.680 --> 00:00:28.880
We're going- I don't know when, and, and we're going to space We're going to space, and we have been all week, because it's really SpaceX week at Blockspace and in the larger fintech world, because SpaceX IPO's tomorrow.

4
00:00:28.960 --> 00:00:39.800
So we've got yet another segment on SpaceX today. Charlie's going to be leading that, talking about satellites. And also, we're gonna be riding the SpaceX mega ramp, which is its valuation chart.

5
00:00:39.880 --> 00:00:49.300
We've got some supplements there to show y'all just why some people feel a little shaky about this IPO and why other people are saying you're dumb if you don't buy it. It's kinda damned if you do, damned if you don't.

6
00:00:49.730 --> 00:01:01.130
But following that, we will have a quick note on CoreWeave's $3.5 billion unsecured proposed note that just dropped on the wire this morning. Then we've got an interview with Rich Miller,

7
00:01:02.080 --> 00:01:11.720
longtime journalist in the data center space, who just took a tour of the Lake Mariner facility and Tera- uh, uh, Terrawulf's Lake Mariner facility, and he's got some pretty interesting takeaways.

8
00:01:12.080 --> 00:01:19.730
You'll be shocked at how many contractors are working on that site right now. Or if you know kinda what this entails when you're building one of these AI data centers, maybe not.

9
00:01:20.340 --> 00:01:25.960
Following that, we are going to be discussing... Uh-oh, censorship, Charlie. Uh-oh.

10
00:01:26.000 --> 00:01:40.420
Specifically, OpenAI banning two accounts that they believe they're linked to Chinese state agencies that were pushing anti-data center and anti- uh, American government LLM-produced content.

11
00:01:40.970 --> 00:01:46.580
And then following that, we are going into the token wars. Not crypto tokens.

12
00:01:46.920 --> 00:02:01.240
Specifically, OpenAI's decision to lower its token pricing, and why [laughs] Zerohedge thinks that this is doomsday for AI, despite the fact that usage for these tokens and adoption keeps spiking. That's right.

13
00:02:01.300 --> 00:02:16.080
Blockspace goes live every weekday at 1:00 PM Eastern, featuring quick hits on AI, emerging tech, a little bit of Bitcoin still, and some Bitcoin mining where it makes sense. This episode turns into a podcast.

14
00:02:16.120 --> 00:02:25.580
Wherever podcasts are found, make sure to subscribe to that feed. And if you're listening on CoinDesk, we are ending the CoinDesk feed. Head over to the Blockspace feed.

15
00:02:25.660 --> 00:02:35.460
Search Blockspace on whatever podcast player you are listening to and subscribe there, because this will no longer be on CoinDesk n- by the end of next week.

16
00:02:37.200 --> 00:02:53.560
This show is brought to you by CleanSpark, NASDAQ-listed, ticker CLSK. More on CleanSpark later on in the show. So Colin, let's talk about SpaceX again, and the thing is, uh, it's a retail madhouse.

17
00:02:53.660 --> 00:03:06.580
Let's just do the numbers really quick. So it IPOs tomorrow, I believe, unless something's changed. June 12th. Well, if- It's tomorrow... if Senator Warren has her way, the SEC would delay the IPO.

18
00:03:06.640 --> 00:03:15.800
Apparently, she put in a request for them to delay it, which is almost certainly not gonna happen. Wow. But I think just kinda shows you where we are in the current political landscape right now. Yeah.

19
00:03:15.840 --> 00:03:28.500
Are you ready to be very unpopular, Senator Warren? [laughs] Yeah. Because look, look at this. Uh, this is a retail madhouse. SpaceX IPO said to draw more than 70 billion in retail orders.

20
00:03:28.920 --> 00:03:42.440
Retail orders alone are basically enough to cover the entire raise, which is, I think, 75 billion. Retail will be allocated 20% of available shares for SpaceX. So Senator Warren, I get it.

21
00:03:42.460 --> 00:03:45.520
Gonna put some guardrails around things. Maybe comes from a good place.

22
00:03:45.860 --> 00:04:00.140
Rarely do I ever say that it does, but if you are in this for support and votes, I don't think this is the way to get them, because the people want a piece of SPCX. The people want a moonshot. Surprise, surprise.

23
00:04:00.470 --> 00:04:09.020
A moon- But that's really an incredible stat, the fact that the original raise amount of 75 billion could be covered solely by retail.

24
00:04:09.080 --> 00:04:18.600
And as we covered on the show this week, according to Reuters, the whole thing is 3X oversubscribed at, like, 250 billion. Really curious where that final number shakes out. Yeah.

25
00:04:18.740 --> 00:04:28.740
Maybe I can pull up a Polymarket s- uh, here in a second. But before I do that, I wanna talk about one of the core things, 'cause what are we gonna do? I, you know, I can't, really don't know much about rockets.

26
00:04:28.780 --> 00:04:30.660
I don't really know much about satellites.

27
00:04:30.700 --> 00:04:46.300
But I will say, as the s- as the SpaceX satellite is their vector and way they're going to shoot their AI modules into lower Earth orbit, it's of, uh, very in- you know, a lot of scrutiny right now.

28
00:04:46.680 --> 00:04:59.260
Because the SpaceX plan, and they unveiled this the other day, was to have these little satil- satellite modules which they rocket up into space and create a constellation of them, interlinked, pointing down to the ground and to each other.

29
00:04:59.340 --> 00:05:14.840
And so they kinda r- you know, rolled out a video of it, gave some dimensions and h- some high-level specs, and that got people on Twitter arguing about it. And, uh, the big question is, do these satellites work?

30
00:05:14.960 --> 00:05:28.580
I think it's almost all around the, uh, like, heat dispersion, uh, argument. And I'll pull up this viral tweet from Jerry Rig Everything, who, by the way, fantastic YouTuber.

31
00:05:28.680 --> 00:05:41.590
You can find him, uh, breaking and demolishing phones. I'm a longtime subscriber of this guy. I love him. He's a great engineer. But he put out this tweet that says, this, uh, um... And, and, you know,

32
00:05:42.680 --> 00:05:51.500
someone says, "Satellites work in space because they don't w- get that all, all that hot." Somebody subtweets him. Peter says, "The Soviets put nuclear reactors in space, you idiot."

33
00:05:51.530 --> 00:06:05.964
[laughs] And then Jerry Rig Everything says, "Nuclear reactors in space is funny, because those nuclear reactors were allowed to hit, internally hit 3,272 degrees Fahrenheit while radiantly shedding 150 kilowatts-Worth of heat.

34
00:06:06.244 --> 00:06:18.784
Elon plans on shedding the same amount of heat while keeping his processors cool at 140 degrees Fahrenheit. Jerry asserts that the radiator would need to be the size of a football field.

35
00:06:18.844 --> 00:06:25.614
So Colin, how do we square this circle? Is Elon gearing up for another massive disappointment?

36
00:06:25.794 --> 00:06:37.904
And this may be the biggest of all, because this is his big foray into public markets entirely around the value thesis of putting these spa- these satellites into- I-... space.

37
00:06:38.184 --> 00:06:52.224
I'm just gonna go out on a limb and say that I probably trust the engineering acumen of the team that built the first reusable rocket booster, rather than the YouTuber who has never built an aeronautics company.

38
00:06:52.904 --> 00:07:03.284
I know that might come to some people as a shock, but you know what this kinda reminds me of, Charlie, to tether it, so to speak, in our erstwhile careers as Bitcoin journalists.

39
00:07:03.724 --> 00:07:21.744
Reminds me of all the Tether truthers who thought that they had found some, you know, glorious insight into the one major flaw in the entire Bitcoin ecosystem, and completely discounted all of the professional opinions of the people actually working in that industry.

40
00:07:21.944 --> 00:07:35.744
I, I'm just gonna go out and say it a little more bluntly. I don't think, uh, this YouTuber has any insight that has not already been discussed in the planning process for SpaceX's satellites. Okay.

41
00:07:35.804 --> 00:07:49.344
So, yeah, I mean, so one of the, uh, one of the very fair criticisms of Elon is that he sells the moon, and he r- he regularly misses timelines, and he has these incredibly ambitious goals which he often doesn't hit.

42
00:07:49.364 --> 00:07:59.254
But on the other hand, he, he and his companies do produce incredible engineering feats, which are the cusp of, like, technological progress.

43
00:07:59.324 --> 00:08:12.424
So here's a reply, I haven't verified this, to Jerry's tweet saying that, uh, SpaceX's AI1, which is the satellite design, uses a double-sided deployable liquid radiator.

44
00:08:12.764 --> 00:08:18.524
Steven Boltzmann calculations confirm this area radiates the appropriate amount.

45
00:08:19.144 --> 00:08:32.164
So, uh, I think a lot of folks are trying to calculate radiation based upon, like, old engineering or improper, like, engineering, inaccurate engineering assumptions. Um, I can't do this.

46
00:08:32.344 --> 00:08:49.124
I will share this one interesting angle, though. So I think a lot of people just forget that SpaceX has some, I believe, 40- 42,000, uh, Starlink satellites in low Earth orbit right now. And in fact, uh,

47
00:08:50.564 --> 00:09:01.904
here's a whole, like, thing on space.com explaining them. Because surprise, surprise, Colin, it's actually a big trend right now to, uh... It's, it's quite common for people to think that the satellites don't exist.

48
00:09:02.404 --> 00:09:09.984
It's actually almost like you go onto some of these- Wait, really? Yeah, yeah. We have, we have, we have Starlink truthers now? We have a... It's a lot.

49
00:09:10.024 --> 00:09:21.904
So much so that, like, uh, SpaceX has a lot of marketing devoted to, like, proving empirically that these satellites exist. And so, um, you know, here's a space.com,

50
00:09:23.004 --> 00:09:33.164
uh, you know, piece on it, and it's, it's pretty interesting. But there's, you know, 40... They wanna have 42,000. They've... I'm sorry, I was, I misspoke. They have 10,000 Starlink satellites up.

51
00:09:33.404 --> 00:09:42.204
And, um, there's some interesting data in here which I won't get into, but I also learned that the average lifespan of a Starlink satellite is about five years.

52
00:09:42.284 --> 00:09:54.714
So, uh, I don't know if that changes the math on the AI sta- uh, satellite constellation idea, but we'll have to see. Uh- Yeah, for sure. Yeah.

53
00:09:54.744 --> 00:10:08.664
And one note on that before I pull up just a, a chart or two before we get Rich on. I mean, Pinchon couldn't write this any better in terms of people denying that those satellites exist.

54
00:10:08.824 --> 00:10:21.074
I mean, COVID just totally accelerated us towards a post-truth future, man, in a way that I think most people are still not reckoning with. But we'll, we'll leave that there. Y- you mentioned getting Polymarket up.

55
00:10:21.224 --> 00:10:26.784
I did wanna share the odds for SpaceX's, uh, post-IPO market cap.

56
00:10:27.244 --> 00:10:41.304
Currently, the consensus, if we can call it that, is that it will settle, 50% are betting that it will settle between 2 and 2.5 trillion, 33% are saying 1.5 to 2 trillion, 13% saying 2.5 to 3 trillion.

57
00:10:41.464 --> 00:10:53.444
Is this, is this- Sh- shout out to the 1% that thinks the IPO is going to double after- Whoa... it debuts. Uh- I don't know. Can it double? Don't they, like, gap these things up or down?

58
00:10:53.584 --> 00:11:01.003
I don't really know the stock market well. I don't think circuit breakers exist on the upside. Oh. They only exist on the downside. No crying in the casino, Charlie. Oh, yeah.

59
00:11:01.084 --> 00:11:11.804
Um, and, and, and last note on this, just to give an idea of, of how, of why people are so torn over this thing, because we t- we brought it up on the last stream, right?

60
00:11:11.884 --> 00:11:18.344
Like, betting against Elon in the past typically hasn't been good. I mean, sometimes it was good if your timing was right.

61
00:11:18.384 --> 00:11:30.584
But when you look at this, I mean, does this seem like something that, that, that retail should be aping into right now? Right? I don't know what... Looks like a crypto chart, man. You always buy all-time high breaks.

62
00:11:30.884 --> 00:11:40.484
I'm, I'm telling you. But- So, so for people listening on the stream right now and without visual, this is SpaceX's monthly valuation from 2015 to, to current.

63
00:11:40.864 --> 00:11:48.504
It's now, it's currently valued based on the IPO pricing at 1.77 trillion, and it just absolutely hockey sticks upwards starting in t- 2025.

64
00:11:48.984 --> 00:12:01.584
It w- it was below half a trillion in, in, in the outset of 2025, and now it's, you know, 1- 1.77 trillion. People are thinking it's gonna get above 2 trillion on IPO day. So.

65
00:12:02.744 --> 00:12:12.468
I mean, we'll have, we will have to see. I- Yeah. I don't know what's scarier, being in or out of the market currentlyYeah.

66
00:12:12.848 --> 00:12:24.548
Uh, w- just, you know, w- just, just take solace in that y- whatever you do, you're gonna make the wrong decision. [laughs] So that, that's just how these things work. Whatever you do, the market will do the opposite.

67
00:12:24.568 --> 00:12:37.988
Schrodinger's stock. Yeah. [laughs] Um, I had a couple more SpaceX topics lined up, but I think I'm gonna leave it there 'cause tomorrow's the IP. I don't wanna belabor this point.

68
00:12:38.048 --> 00:12:48.528
We have Rich Miller in the wings, and we're gonna have a very interesting conversation with him, so we'll bring him up here in just a moment. But before we do that, a word from our sponsor, CleanSpark.

69
00:12:48.828 --> 00:13:03.728
[gentle music] We are CleanSpark, America's Bitcoin miner, a publicly traded company with the largest operating hashrate, powered entirely by self-operated infrastructure across four states.

70
00:13:05.628 --> 00:13:21.828
This is our proof of work. We are setting the standard for what's next. Learn more about the intersection of energy and Bitcoin at CleanSpark.com. All right. Rich Miller in the backstage.

71
00:13:21.888 --> 00:13:34.158
We are gonna bring him on up here. We haven't even had a chance to do a mic check, but we're gonna just run right in, and we'll see how this works. Rich. Hey, I can hear you. Welcome to the show. Awesome.

72
00:13:34.228 --> 00:13:44.528
It is great to be here. Thanks for joining, man. Really, really appreciate it. Um, before we hop into it, Rich, just give people a quick background in, in what you do at Data Center Richness. Sure.

73
00:13:44.668 --> 00:13:56.188
Uh, I've been covering the data center sector full time now for about 25 years through a series of publications, uh, uh, that I, I launched. Uh, Data Center Richness is the newest.

74
00:13:56.308 --> 00:14:08.068
We look at, uh, just about all aspects of the data center industry, particularly in, in terms of handling the scale that's going on, and frankly, building better data centers, because that's a, a big topic now.

75
00:14:08.188 --> 00:14:16.728
But do podcast, uh, YouTube, and also on Substack. We- we're basically in the same, uh, line of work now, Rich. It, it's, it's amazing.

76
00:14:17.088 --> 00:14:23.848
You know, we've been covering Bitcoin miners for really since it became an American industry, but they all pivoted to AI.

77
00:14:23.958 --> 00:14:34.728
And so, you know, my first question for you is, you've been covering data centers for decades, and it's absolutely the most important industry in the American economy right now.

78
00:14:35.588 --> 00:14:44.728
When did you identify this, like, crypto to AI shift? When did it pop up on your radar? And what did you, like, what did you think when you initially started watching this trend?

79
00:14:45.828 --> 00:14:54.228
So the interesting thing is, I started paying close attention to, uh, Bitcoin mining operations back in the early days of, you know, sort of Bitcoin enthusiasm.

80
00:14:54.848 --> 00:15:02.928
2014, 2015, there was a lot of ideas that, you know, uh, crypto miners would be data center customers.

81
00:15:03.008 --> 00:15:14.268
Pretty quickly figured out that that was not gonna be the case because data centers are built for enterprises and cloud with a lot of redundancy, a lot of, you know, cost overhead to stay online all the time.

82
00:15:15.028 --> 00:15:25.488
Uh, crypto doesn't need that, and they're, were always looking for, uh, the sort of lowest cost approach, and they wanted a lot of power, which would turn out to be important.

83
00:15:25.888 --> 00:15:39.068
So crypto built its own style of facilities, uh, and data centers, you know, the data center sector plowed ahead with enterprise and cloud. The moment when everything changed was in 2022 when two things happened.

84
00:15:39.808 --> 00:15:53.908
The first was that, uh, in the state of Virginia, uh, a little area in northern Virginia called Data Center Alley in Ashburn ran out of electrical power, and this was important because that is ground zero for the data center industry.

85
00:15:54.118 --> 00:15:59.388
It's where all the networks meet. It's where the most cloud and data center capacity was deployed.

86
00:16:00.028 --> 00:16:06.458
Because historically, the way that data centers have always chosen where they were gonna locate is by following the network.

87
00:16:07.018 --> 00:16:15.368
They wanted to get close to the big internet intersections and the major business markets, and everybody built data centers right around the, the inter- interconnection exchange.

88
00:16:16.288 --> 00:16:27.418
Uh, when they ran out of power in northern Virginia, suddenly people couldn't deploy all the capacity they wanted, and that shifted to a new strategy, follow the power.

89
00:16:28.068 --> 00:16:41.568
So that was the first thing that happened, and of course, then in November 2022, ChatGPT appeared on the scene and we, you know, it became pretty clear that AI was gonna become a really significant use case.

90
00:16:41.608 --> 00:16:52.048
And by, you know, mid-2023, it was also clear that they would, uh, uh, these customers would want very large, uh, chunks of capacity.

91
00:16:52.768 --> 00:17:13.897
Um, and so what kind of happened is that the data center sector moved in the direction of what the crypto miners were, were doing in terms of, uh, suddenly wanting to follow the power, find places that didn't necessarily have to be in a core network market, but had lots of land, and particularly access to a lot of power.

92
00:17:14.418 --> 00:17:26.328
And there were a lot of, uh, uh, crypto mining companies that had that at that time that were thinking about the, uh... At first I think they talked about it as HPC, but then the AI shift.

93
00:17:26.588 --> 00:17:30.288
So, uh, it's really been fascinating to, to watch.

94
00:17:30.328 --> 00:17:43.428
And I've learned about some of these companies watching, uh, some of you guys' uh, podcasts, uh, because, uh, there were a lot of new players suddenly appearing on the scene who very quickly have become consequential, uh, for AI deployments.

95
00:17:44.288 --> 00:18:03.208
R- Rich, quick, quick th- question that, and, and for some context, I don't know what year this was, but when y- when he's talking, our listeners, about Data Center Alley, Virginia data centers account for, like, anywhere from, like, 25% of the entire state's electrical capacity at a given time.

96
00:18:03.828 --> 00:18:04.028
Right.

97
00:18:04.248 --> 00:18:17.348
Uh, but going back to the need to chase where the power is, did anything change about the networking, where they were now able to move further out the curve to Middle America, places like Texas, Ohio, Pennsylvania?

98
00:18:17.468 --> 00:18:21.508
Um, did anything change about the actual networking of these data centers that allowed them to do that?

99
00:18:21.548 --> 00:18:33.806
Or was it just purely necessity driven?It was purely necessity-driven that the thing they needed first was the power. Uh, 'cause you can always drag the network behind you, and that's what a lot of the, the big guys...

100
00:18:33.836 --> 00:18:42.276
If you're Google, if you're AWS, if you're Microsoft, uh, you can, you've got deep relationships with, uh, network providers.

101
00:18:42.596 --> 00:18:54.416
Most of them have dark fiber operations, so they can work pretty quickly to get, uh, fiber to a place where you wouldn't necessarily, uh, have the kind of connections that you have in northern Virginia.

102
00:18:55.076 --> 00:19:08.956
Uh, it, that doesn't mean you'd have the same level of connectivity there, particularly in terms of the number of networks you can reach, but, um, you know, the other thing is, you know, uh, some AI workloads need really low latency.

103
00:19:09.456 --> 00:19:13.636
Others seem to do okay with, you know, some of the reasoning models in particular.

104
00:19:14.216 --> 00:19:25.376
They're gonna take a little bit to answer your question, so the kind of latency that you would need for, you know, 911 services or emergency services and things like that or financial services, not the same.

105
00:19:26.896 --> 00:19:32.456
Rich, I'm curious about, like, the DNA of the different, uh, companies now.

106
00:19:32.576 --> 00:19:43.856
You have traditional data center builders and traditional tech and hyperscalers, and you've got the, I would say, historically a bit more Wild West Bitcoin crypto miners.

107
00:19:43.896 --> 00:19:51.016
But, uh, and I'm, I'm curious because, like, it seems like both industries are g- gonna have to come together to figure out how to talk to the other really well.

108
00:19:51.656 --> 00:20:06.356
From, from the perspective of, like, a Bitcoin crypto miner, is there something in the DNA of how those companies operate that you see that is different, that is an advantage in this new world of, like, rapid build-out speed to power?

109
00:20:07.036 --> 00:20:08.676
I think there's a couple things I'd highlight.

110
00:20:09.016 --> 00:20:29.076
One is that, um, data center operators are very used to, to dealing with enterprise clients, large, uh, clouds, so they've always pres- you know, have focused for a long time on really kind of being pretty corporate and, uh, or working towards strong credit ratings.

111
00:20:29.196 --> 00:20:39.336
And, uh, I think that's, that's an area that's been important for them and their clientele that's a little bit different as opposed to, uh, uh, crypto in, in most cases.

112
00:20:39.876 --> 00:21:02.916
Uh, I think the thing that's been important is that, um, there's all sorts of players now whose interests are aligned about, uh, bringing more AI capacity online, and some of those players, I think, have played a really important role in working with some of the, the crypto mining specialists who wanted to make that crypto to AI shift and had the power.

113
00:21:03.676 --> 00:21:14.736
Nvidia, uh, very clearly has worked closely with, with a, a couple of them to help them understand what the requirements are gonna be and help them to, uh, kind of get up to speed.

114
00:21:15.686 --> 00:21:19.516
Uh, and Dell, I think, has played an important role in that as well.

115
00:21:20.136 --> 00:21:36.496
And I, I think when everybody's sort of interests are aligned around getting that capacity online, uh, there's been some cooperation to help, uh, help some of these folks who are new to it, um, move up the, the, uh, the value chain a, a little bit more quickly.

116
00:21:36.876 --> 00:21:46.636
And I think, you know, uh, the other thing is that there's a lot of differences between the facilities, uh, that I think there's been a gap to bridge there.

117
00:21:46.796 --> 00:22:03.016
Um, I talked to Wes Cummins from, uh, Applied Digital on one of our podcasts, and he talked at length about there was a real learning curve, uh, to go from what they were doing with crypto and to deal with the kind of gear that, uh, Nvidia might be, uh, bringing in.

118
00:22:03.416 --> 00:22:04.916
Uh, but there are some corollaries too.

119
00:22:05.216 --> 00:22:22.316
Uh, lots of, uh, Bitcoin folks are very comfortable with liquid cooling, have seen it before, uh, probably, uh, to a higher degree than some of the people in the data center sector have, because it's been on the horizon for a long time, but it adds a whole maintenance and overhead thing that a lot of people have resisted.

120
00:22:22.796 --> 00:22:29.376
So it's really Nvidia that's shoving, uh, the entire sector into the liquid cooling, uh, um, sort of a frame of mind.

121
00:22:30.036 --> 00:22:46.616
You know, go- going, speaking to those, uh, challenges, Rich, I, I was in- initially very skeptical in 2023 when these Bitcoin miners started talking about making this pivot because, uh, and a handful of these companies have, quite frankly, never really been very proficient in mining Bitcoin.

122
00:22:46.656 --> 00:22:48.496
I mean, they're always mining it at a loss.

123
00:22:48.976 --> 00:23:01.576
Part of the reason why they could sustain their operations was during Bitcoin bull runs, they could sell their equity into the open market and then use that to cover their costs and expand much more rapidly than, let's say, a more financially prudent private miner would, would do.

124
00:23:01.676 --> 00:23:14.746
But as you noted in your articles, the, I think the biggest thing that they had an advantage here was that they were already sitting on the power agreements, and so then it was just a matter of what is the best model for them going forward?

125
00:23:15.086 --> 00:23:25.076
And it seems like the convergence on the powered shell model that the majority of them are going after, I believe out of the biggest one, Iron is the only exception, as in they're going fully integrated.

126
00:23:25.096 --> 00:23:32.556
But y- you got a chance to look at one of these constructions, uh, at TeraWulf's Lake Mariner facility m- recently.

127
00:23:32.616 --> 00:23:40.696
And so I would love to open this part of the discussion, if you could just tell our listeners what were some of your biggest takeaways from touring that site? It must have been pretty interesting.

128
00:23:41.696 --> 00:23:52.836
It really was, 'cause I've toured a lot of data center campuses, uh, over the years, and the first thing is that this is an entirely different kind of scale than we normally see.

129
00:23:52.956 --> 00:24:04.136
Uh, and it was an, an extraordinarily busy construction zone too. Uh, you know, you're used to seeing, uh, construction activity, but the thing about this is...

130
00:24:04.196 --> 00:24:11.976
And, and they've been public about their agreements with FluidStack. They're building two huge data centers for them at the same time.

131
00:24:12.576 --> 00:24:22.016
There is all sorts of, uh, simultaneous construction work going on on these, uh, data centers, which are 330,000 square feet.

132
00:24:22.496 --> 00:24:38.388
It's probably the size of, like, you know, three Walmarts, uh, uh, each building.And so there's, uh, a constant, uh, this constant frenzy of, of motion with, uh, backhoes and, uh, little trucks carrying construction equipment.

133
00:24:38.428 --> 00:24:46.728
There's huge cranes on site. Uh, there's the, the steel superstructures for the, the... are, are up, and they're framing out the buildings.

134
00:24:47.408 --> 00:25:03.928
Uh, and it's a lot of activity, uh, and to the, to the point where one of the biggest challenges I think [laughs] they have right now is directing traffic on the site as they're trying to, uh, move all the parts and equipment, uh, uh, into place at, at top speed.

135
00:25:04.228 --> 00:25:15.628
A- and the thing is, they're going 24 hours a day. They're working around the clock, uh, which is, uh, you know, pretty extraordinary. And, uh, I think they said they've got, you know, apart from...

136
00:25:15.788 --> 00:25:21.308
there's about 300 staff between TerraWolf and, and Fluid Stack, but then there's between...

137
00:25:21.348 --> 00:25:35.728
and in any given day, between 1,200 and 1,600, you know, contractors, construction workers who are on site and, uh, working very quickly to, to bring all the components, the power, the cooling.

138
00:25:36.348 --> 00:25:49.368
You have whole, uh, UPS and generator, uh, uh, skids that come in shrink-wrapped and are next to the data center ready to go. So it's a really interesting job site. Th- that...

139
00:25:49.408 --> 00:25:58.068
when reading your article of the tour, that's what really blew my mind, was the scale of the contractors on site. And, and two follow-up questions.

140
00:25:58.168 --> 00:26:08.418
So th- they're, they have contractors working night shifts, and I guess that's part of why this, this s- build out has been expedited. You mentioned that it's kind of breakneck speed. Mm-hmm.

141
00:26:08.418 --> 00:26:10.508
It's faster than you would expect for something like this.

142
00:26:11.008 --> 00:26:21.948
And the second question, those contractors, looking at the job creation angle, which is often hit on by these data center operators, those contractors almost certainly aren't local, right?

143
00:26:21.988 --> 00:26:27.498
Like, they're coming from around the country based on who they contracted for each specific construction job, I would imagine.

144
00:26:27.788 --> 00:26:40.588
There's a lot of that, and that's a, a really big issue for the data center sector in general. Uh, it's no secret that community, uh, relations is a real, uh, challenge right now for the data center industry.

145
00:26:41.248 --> 00:26:43.728
Uh, people have mixed feelings about AI.

146
00:26:44.228 --> 00:27:02.788
There's a lot of attention right now to the resources that, uh, these large data centers use, a- and they're way more efficient than they used to be, but at the same time, what all of these are doing is taking an enormous amount of, of, uh, economic activity and concentrating it at a single point in a community and in the network.

147
00:27:03.008 --> 00:27:11.688
So particularly, there's a lot of power used, uh, and that's a sensitive issue at a ti- a time when everybody's concerned about, uh, energy affordability.

148
00:27:12.248 --> 00:27:17.188
So the trade-off, there's a couple of trade-offs, uh, uh, real benefits of having data centers in the community.

149
00:27:17.688 --> 00:27:41.148
One has been the tax impact, which is often huge because of the amount of not even the building so much as the gear that goes in, uh, to these, uh, uh, facilities is so expensive that, like, the sales and use tax on those things, uh, you know, I think in, in, uh, northern Virginia that those places, it, it, uh, supports all sorts of things for the community because they get that tax break.

150
00:27:41.428 --> 00:27:51.948
The other thing is that it brings a lot of workers into the community. Now, the data centers themselves are highly automated, so they don't create a huge amount of jobs.

151
00:27:52.108 --> 00:28:02.588
So they're a little bit of an odd bird from an economic development standpoint, but from the... during construction, there are these huge armies of construction workers that, that come.

152
00:28:03.388 --> 00:28:14.748
The hyperscalers in particular are working hard to work first with the local community and try to get folks who are in state, but they bring in a lot of people from out of state.

153
00:28:14.788 --> 00:28:25.388
And, and one of the huge problems for this whole industry right now is there are not enough construction workers, not nearly enough to keep pace with all of the projects that are out there.

154
00:28:25.448 --> 00:28:34.588
Uh, there's a number of, you know, things that are delaying projects, but that's the, the, um, uh, ability to get folks on site, uh, is a real challenge.

155
00:28:34.668 --> 00:28:51.148
So you're correct in that a lot of them will come from, from off s- from out of the, uh, the area, which then creates, you know, uh, other challenges about, you know, uh, you know, some places they'll come in with the size of the campus build outs, now they're there for longer.

156
00:28:51.248 --> 00:29:00.448
It's not just, "Well, we build this one thing, then we go away." If people are building, you know, Facebook, uh, uh, might, uh, or Meta might build nine buildings on a campus.

157
00:29:01.348 --> 00:29:06.208
Uh, and as they do them one after the other, sometimes those people will be there for three to five years.

158
00:29:06.768 --> 00:29:15.068
So it's, uh, that's a big, uh, bad factor, but getting qualified people, uh, on site is a challenge, and so you people...

159
00:29:15.168 --> 00:29:26.708
you see companies pull from all over based on the, the r- the relationships that they have in the trades. We talk about, like, speed to power, 'cause that's, like, TerraWolf's tagline.

160
00:29:26.748 --> 00:29:37.598
That's a lot of industry players' tagline. Right. TerraWolf COO, I sa- I believe, said they believe they're, quote, "the fastest ongoing data center build in the US right now."

161
00:29:37.628 --> 00:29:49.208
And I'm kinda curious, like, why and what factors. I look at, like, zooming out really far, the Lake Mariner site, if I understand, it's a brownfield site. So it's- Yeah...

162
00:29:49.228 --> 00:29:58.038
uh, like, redevelopment, and that seems to be pretty key to all this, 'cause then you have power, you've got transmission, you probably have city and infrastructure nearby.

163
00:29:58.188 --> 00:30:08.588
What's the landscape of brownfield sites in the US? If you could even, like, speak to that and, like- Sure... are those all bought up and ev- everybody called dibs on them a- or uh, are there just...

164
00:30:08.658 --> 00:30:14.888
or different plays to make when we're talking about, like, site selection in the US? I think there's different plays to make.

165
00:30:15.348 --> 00:30:27.308
I think TerraWolf is a really interesting example of a company that said, you know, "We're gonna work with the, uh, you know, infrastructure of the old economy to build the new economy on, on top of it."

166
00:30:28.128 --> 00:30:38.628
Uh, Lake Mariner is along the shores of Lake Ontario, not too far from Buffalo. It was home to a coal plant for many years that was the leading employer in the community.

167
00:30:39.308 --> 00:30:54.036
Uh, and it had a substation there, and of course, uh, because the power was going out from there, it has all the, uh, the transmission lines, uh, connecting it to the grid.Now, uh, you know, talking to the TerraWolf folks, that's an opportunity for them.

168
00:30:54.696 --> 00:31:02.776
Uh, it's not like there's, there's no work to do to, to make the, the transition. They had to do, you know, studies, uh, interconnection studies with the grid.

169
00:31:03.756 --> 00:31:15.816
But what they told me was that that might take a year or two, whereas if you're trying to get power from a utility and get a new connection, like in New- in Virginia right now, that's seven years.

170
00:31:16.456 --> 00:31:29.076
Uh, other markets are, you know, the two to three years is, is standard. Uh, Texas, they're having sort of contests to, to, like, see who can get to the front of the line to, to connect. Uh, so

171
00:31:30.096 --> 00:31:40.136
any of the ability to use a, a brownfield site that already has either some kind of energy infrastructure on site, and more importantly, that has transmission lines to it.

172
00:31:40.176 --> 00:31:50.176
So old power plants, um, you know, TerraWolf is also working with, uh, a couple of aluminum smelting sites, which of course have tremendous power draws.

173
00:31:50.836 --> 00:31:54.136
Uh, and so they've got a lot of infrastructure coming to their sites.

174
00:31:54.996 --> 00:32:06.976
I do think a lot of the opportunities, uh, uh, are probably either in process or being scarfed up right now, but there's also a lot of creativity and people trying to think outside the box.

175
00:32:07.656 --> 00:32:26.716
Because what's happened with AI, and this, you know, has played out very well for the crypto to AI players, is that all the old assumptions have kinda gone out the window to try and bring new capacity online, and the data center sector has probably never been more willing to think about new ways to get things done.

176
00:32:28.776 --> 00:32:35.316
I've got one, Charlie. Okay. I, I wanted to hit, hit on the public relations nightmare angle really quickly one more time. Sure.

177
00:32:35.526 --> 00:32:47.436
I, I, I'm curious, I know that you talked to the TerraWolf team, and so I'm gonna kinda ask this in two ways. What did they say were the primary concerns for these AI builds?

178
00:32:47.456 --> 00:32:55.036
And did you get a sense of the community's reaction to this? Like, did you have any conversations with locals about their thoughts on the data center?

179
00:32:56.516 --> 00:33:06.916
Uh, I didn't have any, uh, discussions with the locals while we were there, 'cause it was kind of a, a, a quick in and out. But I've done, you know, I've, I've read a lot about the community and what's going on.

180
00:33:07.416 --> 00:33:21.056
So the, the sort of backdrop on that is TerraWolf came in first, uh, to do crypto mining on the site. That was, you can see the sort of, uh, evolution of their infrastructure as you go across this large campus.

181
00:33:21.736 --> 00:33:31.116
Uh, early on, the, the concern was about noise, uh, that neighbors were concerned about the noise from the, uh, uh, from the, the, the mining rigs.

182
00:33:31.536 --> 00:33:43.176
I think as, as some of that capacity has been shifted over to AI uses, that's not so much a concern. Right now, the big concern is water. You know, the- they, um,

183
00:33:44.196 --> 00:34:11.056
say repeatedly that they use a closed loop system, which is one where you fill it once with, with, uh, with water, uh, and then you just recirculate it and, uh, use an outside cooling loop to, and a heat exchanger so that you're not constantly drawing water, uh, and evaporating it, which, uh, was a strategy that data centers used to, to use to optimize for their lowest carbon impact.

184
00:34:11.576 --> 00:34:17.956
That was the, the best way to do that, was to, to use water. Now, there's a lot of pushback on water usage.

185
00:34:18.056 --> 00:34:27.436
They've got another facility, uh, nearby in, uh, in Lake Cayuga, near, near Ithaca, where there's a lot of controversies and a lot of pushback about that.

186
00:34:28.156 --> 00:34:39.486
And of course, New York State, there's enough, uh, there's enough discussion about it that the, the state legislature has a proposal, uh, to do a moratorium on data centers. I'm not sure how long it is.

187
00:34:39.516 --> 00:34:47.476
They're typically about a year to, to 18 months, and, uh, the governor might have a decision on that, uh, if that makes it to her desk.

188
00:34:47.976 --> 00:34:58.496
Uh, so there's, there is a lot of community pushback, but what the TerraWolf team told me was, was, "Look," they, uh, because they had a lot of, uh, you know, public information sessions.

189
00:34:59.056 --> 00:35:02.576
And lots of times these are really difficult, uh, discussions.

190
00:35:03.156 --> 00:35:13.976
Uh, the people who are, are coming to a town hall are concerned about what this huge project is gonna mean for your community, what's gonna change, particularly in, in rural areas.

191
00:35:14.456 --> 00:35:25.036
I think there's a, a, a lot of concern that a facility like this could change the character of, uh, their community. Uh, and so there's a lot of tough questions.

192
00:35:25.516 --> 00:35:38.936
Uh, and uh, and the TerraWolf team just said, "Look, when the questions get tough, you have to be thick-skinned, you have to, and you have to listen more carefully when the conversation gets difficult."

193
00:35:38.976 --> 00:35:50.986
Because those are the things, ultimately, if you are gonna be a, a part of the community for a long time, that you, you, you can't just, you know, steamroll the, the local pop- population.

194
00:35:51.016 --> 00:35:54.436
You have to hear them, you have to hear their concerns.

195
00:35:54.476 --> 00:36:18.876
And historically, you know, uh, uh, for, uh, TerraWolf is just in, in one scenario that the data center industry as a whole has, uh, over the years, has really tried to just like, you know, I'm not gonna say steamroll is the right word, but they prefer to, to work with local officials rather than with the public, and that has had repercussions now.

196
00:36:19.476 --> 00:36:32.316
Uh, there's a reckoning that's, that's come. There's, uh, some, uh, polls showing that, you know, 50 to 70% of Americans aren't interested in having a data center in the community.

197
00:36:32.776 --> 00:36:43.244
But meanwhile, everybody's using data centers-All day long for just about everything, YouTube, Netflix, your online banking, your social media.

198
00:36:43.944 --> 00:36:50.964
I [laughs] ironically, the most of the, the community groups that are opposing data centers are using Facebook to organize and do it.

199
00:36:51.004 --> 00:36:56.004
And, you know, so it's like they're using AI to- data centers to fight AI data centers.

200
00:36:56.124 --> 00:37:13.784
But, but I mean, I think the thing that I, I thought was interesting about the TeraWulf, uh, team is, is they were like, "Look, when the conversation gets tough, you have to listen harder, and you have to convince people that you're gonna be a part of the community, that you have a vested interest in, in being there and that people are being heard."

201
00:37:13.824 --> 00:37:27.084
And, you know, that's, uh, that's not an easy discussion, but that's, that's what's going on right now. Uh, I've spent a while in oil and gas, so I'm familiar of, like, the criticism of people who are also using it.

202
00:37:27.164 --> 00:37:35.644
I kinda know the, it... But, but what's interesting about the data center pushback is it's pretty apolitical. It's cross-party, cross-socioeconomic.

203
00:37:36.144 --> 00:37:40.544
It is a, it, as you said, it does feel like it's a reckoning coming from the industry.

204
00:37:41.724 --> 00:37:53.244
Okay, I kinda wanna, like, zoom out as we kinda wrap this up, and, um, you, again, you say 25 years in the data infras- data center infrastructure world.

205
00:37:53.344 --> 00:38:03.624
You've seen it change, and, uh, Colin and I guess relatively new careers in the non-Bitcoin data center world. What are we missing?

206
00:38:03.684 --> 00:38:15.344
Like, what is the, like, what is, like, the, uh, what's, like, the Bitcoin, former Bitcoin crypto folks missing, or trends, uh, in the traditional data center industry?

207
00:38:15.804 --> 00:38:27.424
Maybe it's, like, constraints or bottlenecks we're not aware of. Um, what are blind spots, uh, from those of us former Bitcoin miners now piling into the AI HPC game?

208
00:38:28.484 --> 00:38:37.324
I think the first thing that a lot of these companies, you know, really are gonna need to think about in coming years as they scale up these huge operations is talent.

209
00:38:37.364 --> 00:38:52.304
We talked about the difficulty of, of finding people, and there's a limited supply of folks who can operate, uh, AI, uh, hyperscale data centers. A lot of those folks are already working in the industry.

210
00:38:52.764 --> 00:38:57.304
All, all of the companies that have traditionally been working in data centers are hiring like mad.

211
00:38:57.884 --> 00:39:17.024
So I think that's one of the, the things, uh, that's gonna come up a lot as an issue probably, is trying to find, uh, skilled staff who can understand not just the infrastructure you need, you know, to run a, a, a, a crypto mining campus, but the additional layers that are involved with, with dealing with AI customers.

212
00:39:17.724 --> 00:39:28.824
The, the other thing that I think is really important that I think a number of the key players have done really well at is sorting out the, the financing to scale.

213
00:39:28.844 --> 00:39:47.924
And the way that some of these players, I'm thinking, you know, TeraWulf and Iren and, and Hut8 have been able to enlist their, their customers, uh, who have enormous financial strength to, to support them in ways that can help them put deals together to get the capacity deployed.

214
00:39:48.524 --> 00:40:00.684
Uh, there's a lot of, I think, creative thinking going on there. And I, I think that's one of the ways where we're, uh, you know, uh, there's, there's innovation that's, that's moving both ways here.

215
00:40:00.864 --> 00:40:17.804
Um, and, uh, but I do think that, that, uh, uh, the data center industry is becoming stratified in a fascinating way because there's the crypto to AI players, most of whom are doing the, uh, you know, powered land or powered shell, uh, model.

216
00:40:18.304 --> 00:40:33.124
Then there's the NeoClouds who, uh, are actually, like, leasing out the, the, uh, the GPU space, which apparently now includes SpaceX. [laughs] Uh, and, and that's a whole other, uh, as- asset class within there.

217
00:40:33.544 --> 00:40:44.904
Then you have your traditional corporate colocation centers, and then there's the, uh, the, the folks who have Google and, and Meta. H- Microsoft are still building their own data centers as well.

218
00:40:45.604 --> 00:40:48.784
So, uh, it's a fascinating landscape right now.

219
00:40:49.344 --> 00:41:05.984
Uh, and, but it, it's pretty wild 'cause j- you know, to your point, two or three years ago when I looked at these companies expressing HPC and AI, uh, aspirations, I'm like, "You know, that's not likely to happen," because there, there's, there were big differences.

220
00:41:06.084 --> 00:41:16.044
But those, the, the gap has been narrowed pretty quickly when everybody's aligned in getting more capacity online. Yeah, it's almost never been a better time to be a nerdy data center guy.

221
00:41:16.054 --> 00:41:25.264
[laughs] So you spent 25 years putting in the work, and now, now you could be the expert. In every, every dinner party, you get to know every single thing about what's going on.

222
00:41:25.724 --> 00:41:33.604
Well, it's wild because now I don't have to explain what a data center is a- anymore. [laughs] Everybody knows. So it used to be you say, "Yeah, I write about data centers."

223
00:41:33.664 --> 00:41:44.604
They're like, "Well, can you come fix my printer?" You know, like... [laughs] We, I think, uh, I won't speak for Charlie, but I feel similarly because I would usually just tell people I'm a tech reporter.

224
00:41:44.834 --> 00:41:53.494
And then they're like, maybe if they knew what Bitcoin is, I would say like, "Oh, I do Bitcoin or Bitcoin mining stuff." And you know, some, half of the time people are like, "Oh, okay, that's really interesting."

225
00:41:53.524 --> 00:41:59.024
And the other half, they're just- Yeah... like eyes glazed over, have no idea what you're talking about.

226
00:41:59.284 --> 00:42:08.924
But now it's AI and people care about it, so- And they either want to skin you alive when you mention it- [laughs]... or they, they're really interested in what you have to say. You never know.

227
00:42:09.684 --> 00:42:18.704
[laughs] Well, hey, Rich, thank you so much for joining, man. We'll have to do this again sometime. Uh, keep up the great coverage and appreciate you hopping on. Yeah. Ab- absolutely.

228
00:42:18.764 --> 00:42:29.894
This has been a blast, and I really appreciate the invite. Thanks a ton, Rich. See you online. Love that guy. Recommend his blog, one of the most interesting highest alpha blogs day to day.

229
00:42:29.904 --> 00:42:32.904
One more time on that data center richness. Yeah. So go check it out.

230
00:42:33.124 --> 00:42:45.808
And before we move on to, uh, our ad here, Charlie, and then our next segment, I just wanted to double tap what he said and about theData center pushback and what you said specifically, this is, comes from a newsletter we did a few weeks back.

231
00:42:46.348 --> 00:42:52.568
Spring 2026 Gallup poll found that 71% of respondents oppose the construction of a data center in or near their community.

232
00:42:53.248 --> 00:43:05.868
58% of those are Democrats, 58%-- 48% of those are Republicans, and 30-- uh, 58% of Democrats responded as such, 48% of Independents, and 39% of Republicans strongly oppose the measure.

233
00:43:06.428 --> 00:43:12.228
So- Oh, oh yeah, and I mean, I'm, I'm boots on the ground here in Oklahoma, the reddest state in the union.

234
00:43:12.328 --> 00:43:26.368
I have my rural conspiracy, anti, uh, world money, anti-Davos, like, redneck guy, uh, saying that this is a way for the, the deep state to surveil them.

235
00:43:26.888 --> 00:43:35.548
And I have my, my yuppie, like, socialist-adjacent, like, coastal elite saying that it's destroying creative jobs.

236
00:43:35.668 --> 00:43:48.248
This, you know, horseshoe, uh, effect, or fish hook, if you will, trend is manifested in the AI data center pushback. 100%. And to your, like, you know, redneck,

237
00:43:49.488 --> 00:44:00.548
Davos-fearing friend, um, they're already surveilling you, dog. Yeah, they're already- They can see you [laughs] They're inside your phone right now. [laughs] Okay, we gotta roll on.

238
00:44:00.648 --> 00:44:12.228
We, uh, we got CoreWeave, we've got OpenAI, we've got token prices. Holy smokes, we gotta keep on rolling. Before we do that, a word from our sponsor, Luxor.

239
00:44:18.068 --> 00:44:24.588
This episode of Blockspace Live is brought to you by Luxor's Commander, Bitcoin miner management software for enterprise operations.

240
00:44:25.428 --> 00:44:33.458
Luxor's Commander gives you real-time fleet monitoring, bulk remote f- uh, fl- bulk remote commands across your fleet, and Intelligent Miner.

241
00:44:33.888 --> 00:44:41.558
Intelligent Miner is an automated profitability engine that runs every five minutes and adjusts your power settings to live hash rate and energy markets.

242
00:44:41.968 --> 00:44:55.088
ERCOT back tests show 10% improved profitability with Intelligent Mining versus binary mining. Commander Pro is $100 per megawatt or a 25 basis point pool fee adder, about half the price of competition.

243
00:44:55.108 --> 00:45:08.068
But you can also try it for free for 60 days. So if you wanna learn more and get started, go to luxor.tech/commander. All right, Charlie- Let's talk-... we are moving on to the debt side of things.

244
00:45:08.128 --> 00:45:17.688
We just talked about building a data center. Now we're gonna get a little bit into how these get financed- Yeah... specifically. And this story has my favorite term I've heard in finance yet. I know, yeah.

245
00:45:17.788 --> 00:45:26.808
I- So I'll let you drop it. I'll let you drop it. I, I thought that you would like that. So this story hit this morning. Let me get it up here on the screen.

246
00:45:26.868 --> 00:45:43.568
CoreWeave is, wants to issue a 3.5 billion senior unsecured note due 2032, and this note will be issued in dollar and euro-denominated notes, and this is a practice known as a reverse Yankee.

247
00:45:44.088 --> 00:45:56.048
Whoa. Which that's, yeah. Y- y- you always gotta love when you expand the lexicon with something you didn't expect. Because I was curious about this. I've not seen any Bitcoin miner do a euro-denominated note.

248
00:45:56.428 --> 00:46:03.428
This is not uncommon, and it, and it, it becomes more common when there is a discrepancy in, in real rates between the continents.

249
00:46:03.848 --> 00:46:08.988
So right now, you know, European central banks have dropped their rates at a time when the US has actually hiked.

250
00:46:09.028 --> 00:46:18.068
So this allows, uh, CoreWeave, if, as I understand it, and please, uh, finance professionals, give me a ping if I step in the mud here.

251
00:46:18.088 --> 00:46:30.628
It allows CoreWeave to actually issue, um, the notes to investors that are looking for a greater return over what they can get from the current, uh, rates with Treasuries and the Eurozone.

252
00:46:30.728 --> 00:46:37.258
But there's not much else to say about this in terms of the notes themselves, because CoreWeave has not priced these yet.

253
00:46:38.008 --> 00:46:47.228
As we covered on the show recently, you would have to think that they would have the opportunity to, uh, get better pricing than some of their prior notes.

254
00:46:47.888 --> 00:46:56.428
Because as we will see from our friend Rittenhouse Research right here, as we m- actually referenced on one of the shows earlier this week, CoreWeave credit spreads have plummeted this year.

255
00:46:56.848 --> 00:47:07.528
Their cost of capital has historically been improving as the company IPO'd and as it's shown itself to be a competent operator. And, you know, depending on who you ask, right?

256
00:47:07.648 --> 00:47:15.868
D- I believe SemiAnalysis ranks CoreWeave as the number one Neocloud in terms of cloud performance. I think they're the, I think they're the only one in the S tier.

257
00:47:15.968 --> 00:47:30.028
I could be wrong about that, but they're most r- but, but they do, um, at le- at least annual, I believe maybe quarterly analysis on the performance of these different Neoclouds, and CoreWeave is always sitting up high and, high and pretty.

258
00:47:30.688 --> 00:47:48.128
But I was curious, Charlie. I wanted to see what their current debt stack looks like, and CoreWeave's total debt currently is 25.15 billion. Their projected debt over the foreseeable future is 68.47 billion.

259
00:47:48.188 --> 00:48:01.808
That's plus 43 billion. That comes from their own presentation that they included in the 8-K for this press release. So if we look at the current debt load, that gives them a debt multiple of 6.9%,

260
00:48:02.908 --> 00:48:14.188
given the fact that their last 12-month adjusted EBITDA in Q1 was 3.64 billion. Now, some people might be looking at that and saying, "That's pretty high."

261
00:48:14.228 --> 00:48:36.968
But if you look at their projected, uh, last 12-month revenue for all of the contractual revenue that they have currently booked, that will be 18.76 billion once it starts booking all that revenue, which gives it a debt multiple in- of 3.6X, and that includes their forecasted 68.47 billion.

262
00:48:37.168 --> 00:48:54.036
But I think that just goes to show you, man, like-We always talk about these numbers. No matter where you sit in the stack, it is insanely capital intensive to operate within this sector,

263
00:48:55.116 --> 00:49:12.106
right? I mean, the l- the, the least capital intensive is building the powered shells, and that still runs you, like, $10 million per megawatt. But, you know, I, I just, I, I, I'm, I'm pausing here just because- I-...

264
00:49:12.116 --> 00:49:19.326
it's hard for me to even think- Yeah, it, it-... about losing that much capital... it's a bull market for a bunch of sectors of the American economy.

265
00:49:19.836 --> 00:49:32.496
Energy, uh, semiconductors, data center fabrication and build out, electricians, construction workers. You know, who's speaking for the creative finance bros? Because that sounds like the most bull market ever.

266
00:49:32.576 --> 00:49:45.046
Can you imagine being the guy who does- gets to do a reverse Yankee 3.5 billion? The investment bankers who are operating in this industry are having the, the run of their lives. Oh my God.

267
00:49:45.046 --> 00:49:53.036
They're also probably not sleeping- Yeah... I would imagine. Yeah. You know? Also, if you're an investment banker who's not sleeping, carve out 15 minutes of your day, slide into our DMs.

268
00:49:53.076 --> 00:50:02.776
We wanna in- interview you, and we want you to explain what the next reverse Yankee is going to be. [laughs] Or triple sow cow. I don't know what these are called. A triple sow cow, double Yankee split.

269
00:50:02.996 --> 00:50:13.216
All right, last note on this before we move on, Charlie. I wanted to highlight that cost of credit for CoreWeave really quickly, because if you look at the worst

270
00:50:14.116 --> 00:50:27.236
note in their debt stack and the one that matures the quickest, it is the all in rate based on where SOFR is now, plus the 9.62% that they had as the flat rate for the loan. It's 13...

271
00:50:27.256 --> 00:50:42.316
It's roughly 13.25%, and that was a GPU-financed loan that they took out in 2023 before they IPO'd. And now, you know, I was, like, playing around with Claude. Grain of salt here, people.

272
00:50:42.596 --> 00:50:53.476
It's like, well, okay, well, what will they get for this unsecured loan? The rate will probably be a little bit bigger since it's not secured by anything. Uh, it, it kinda threw out a ballpark of 7, 8%.

273
00:50:54.356 --> 00:51:08.476
So, and you can also look at one of their most recent, um, secured notes, which it matures in August 30th. Or, sorry, A- August, uh, 2030. The all in rate with SOFR is 7.63%. So, I mean,

274
00:51:09.556 --> 00:51:18.046
y- you've seen just an incredible compression of the credit, uh, or, or the interest rates that they, uh, they're taking on and their cost of capital.

275
00:51:19.356 --> 00:51:25.776
And it goes back to what we've been saying with the number of raises we covered earlier this week, like with Keel, also with Hut.

276
00:51:26.416 --> 00:51:40.775
All of these companies are starting to see some, uh, some pretty competitive rates, especially when you consider asset-backed financing for Bitcoin miners back in the day, which didn't really exist for very long, but was just m- eye-watering levels.

277
00:51:40.816 --> 00:51:52.296
So that, that's the last thing I just wanted to highlight. Uh, yeah, wild. Um, again, finance eludes me. A little bit over my head.

278
00:51:52.616 --> 00:52:00.656
Give me, give me something a little more that I can understand, such as, like, obscure, uh, spending script paths on Bitcoin. Zero- But-... knowledge proofs.

279
00:52:00.696 --> 00:52:09.236
Yeah, ze- well, speaking of, uh, script paths and, uh, obscure, uh, technical achievements on Bitcoin, let's get a word from our sponsor, Ligos.

280
00:52:09.276 --> 00:52:25.236
[static sound] Hedge funds are getting liquidated because everyone is buying the SpaceX IPO, and no one's buying Bitcoin. [laughs] It's not just Bitcoin's price drying up, though.

281
00:52:25.296 --> 00:52:32.496
Big hedge funds, big whales, lending desks are reeling as Bitcoin's price chops in the low 60K range.

282
00:52:33.016 --> 00:52:42.676
Counterparty risk is rampant, so it's more important than ever to know who controls your Bitcoin, and with Ligos Finance, that is always yourself. They are our preferred lender here at Blockspace.

283
00:52:42.736 --> 00:52:56.176
They use Bitcoin native smart contracts to make sure you always hold the keys to your coins when taking out a loan with them. No bridging, no rehypothecation, no nonsense. Just competitive rates as low as 10% APR.

284
00:52:56.296 --> 00:53:13.016
Go to ligos.finance to learn more. Okay, this next story, and we'll be pretty quick with this one, um, I have under the category AI censorship, 'cause I think we've seen, uh, in this past week, a tale of two censorships.

285
00:53:14.096 --> 00:53:19.646
OpenAI, notably with the lead story, just banned

286
00:53:20.676 --> 00:53:34.116
PRC, that's People's Republic of China, linked influence operations, uh, which they identified are targeting AI debates in the US, which is interesting, 'cause hearkening back to our conversation with Rich earlier.

287
00:53:34.296 --> 00:53:45.516
There are people using AI to make anti-AI memes. Very ironic. But in this case, it looks like it was, uh, much more targeted and more state sponsored.

288
00:53:45.616 --> 00:54:01.896
So ChatGPT, OpenAI, has identified two clusters of these accounts and banned them. Uh, so I'll say, you know, a lot of the listeners listening to this might be like, "Well, yeah, of course." This is still censorship.

289
00:54:02.036 --> 00:54:08.076
This is pro-American, pro-Western censorship, and, uh,

290
00:54:09.356 --> 00:54:21.736
uh, I think this is an interesting kind of case study for this week, which maybe not didn't do the rounds that much on social media as much as the other case of censorship, which is Anthropic.

291
00:54:22.556 --> 00:54:34.156
So Anthropic dropped their Mythos model. If you're not... If you were under a rock, Mythos is the one that was, you know, their super powered model which is gonna hack everything.

292
00:54:35.016 --> 00:54:52.492
I'm, I'm oversimplifying, but basically it's this one that they let a lot of key folks, industry players try out first because it could be dangerous, so they could armor their systems.Well, they dropped, Anthropic dropped Fable, which is the Mythos model with guardrails around it, and guess what?

293
00:54:53.052 --> 00:55:27.192
There were, uh, a little too many guardrails, and it became the subject of s- of criticism to the level of there were biologists, researchers, like white hat people, who basically found they were, um, de facto almost banned from using this model because what happens is you start typing in, um, things which are a little bit too close to, like, how to bioengineer, uh, something that Anthropic thinks is too dangerous, and then you get a message which says, "We can't let you do this anymore."

294
00:55:27.252 --> 00:55:37.592
It's, uh, very much a I can't do that, Dave, message. But this time it's coming from, uh, you know, moral thought leader, effective altruist-driven Anthropic.

295
00:55:38.312 --> 00:55:49.122
Now, Anthropic- Notable, notable philanthro- philanthropic company. Yeah. You know? N- notably never used for anything for military application. Yeah, yeah.

296
00:55:49.172 --> 00:56:01.612
Not used for advanced targeting in military systems- Bro, the-... in Venezuela and Iran... bro, the Venezuela thing was three months ago. That is ancient history. Doesn't mat- yeah. Um, so here's the Wired article.

297
00:56:01.692 --> 00:56:08.892
There were a bunch of other stories like this. Anthropic walks back policy that could have, quote, "Sabotaged AI researchers using Claude."

298
00:56:09.952 --> 00:56:20.672
Well, I, I saw something, and I don't know if it was substantiated, but also they started kind of banning people who were using the model to try to figure out how to build their own models on the backs of it.

299
00:56:20.932 --> 00:56:28.732
I don't know if that's true, though. Yeah, so that actually is a thing. So this has happened before for Anthropic's models, I assume for other people.

300
00:56:28.792 --> 00:56:47.232
But Anthropic had a notable one where a lot of Chinese frontier models, I forget if it was DeepSeek or Kimi, but regardless it was that class of comp- of, of operators, they were basically able to divine how Opus worked from Opus.

301
00:56:47.412 --> 00:56:58.682
So you tested a bunch- Mm-hmm... you throw enough stuff at it, and you're able to figure out its weights, and you're able to basically kind of recreate that and kind of build a cheaper version of Opus. What's funny is,

302
00:56:59.932 --> 00:57:10.412
at a high level, this is not any different than the Chinese looking through open American patents- Right, and then reverse engineering things... to actually reverse engineering. Yeah.

303
00:57:10.612 --> 00:57:18.712
And, and maybe it's not as good, but, uh, you can get something which basically is kind of a Rolex, it's just not technically a Rolex.

304
00:57:18.812 --> 00:57:28.132
Right, and that's a part of what a lot of people think happened with DeepSeek as well, right? Yeah. Got why they were able to spin it up so quickly and why it was so perform- uh, performed so well.

305
00:57:28.672 --> 00:57:32.712
I, I, I wanna hit a few things on this, Charlie, before we move to our next segment.

306
00:57:33.172 --> 00:57:47.912
First, I wanna share this tweet from Sam Lyman of Bitcoin Policy Institute, head of research, we had him on the show recently, to actually talk about this very th- phenomenon of Chinese actors infiltrating the discussion on data centers in the US and trying to steer the conversation.

307
00:57:48.252 --> 00:58:02.592
But, but this is an example of, of some of the anti-data center, uh, cartoons that were produced. And so if you've ever seen one of these around, that's probably, probably what you're, uh...

308
00:58:02.752 --> 00:58:14.152
I mean, it, it might be related to these efforts. It's the most boomer coded stuff ever, though. I mean, look, look at, look at how... It's just so boring and so on the nose.

309
00:58:14.192 --> 00:58:26.492
You stated this, but this is- Why are our electric bills so high? Because we stopped building energy, that's why. Anyway, um- We're de- we've spent the past 15 years de-growthing sun. Yeah, exactly.

310
00:58:26.652 --> 00:58:28.592
And so there are a few others here.

311
00:58:28.732 --> 00:58:44.932
Um, the AI industry's booming, but the costs are being borne by ordinary people as a bolt of lightning splits the sky above, and there's an AI brain just sitting in the middle of a, a city surrounded by what look like boom boxes but are actually data center racks.

312
00:58:44.952 --> 00:58:55.272
Yeah, those are giant racks of B300s. And then [laughs] this is my favorite. When Trump declared a tariff war, China responded with fiscal discipline and rare earth minerals.

313
00:58:55.712 --> 00:59:09.482
It's Trump, uh, ripped in a America First wrestling outfit, swinging a hammer, except the hammer is in the air, and it's not making contact with the actual wall that says global future.

314
00:59:10.372 --> 00:59:19.972
And then there's just stocks and markets and allies, just signs in the background of what looks like a Mad Max-esque apo- post-apocalyptic city. Anyway, pretty low effort.

315
00:59:20.072 --> 00:59:29.272
I expect a little bit more from the CCP, although that being said, this is the kind of boomer slop you see on Facebook and on X these days, so I guess it doesn't really matter that much.

316
00:59:30.182 --> 00:59:36.012
[laughs] Um, the last thing I will note on this, Charlie, before we close out, unless you have anything else to say.

317
00:59:36.072 --> 00:59:54.632
One thing that I thought was interesting from the debriefing from OpenAI on this, first of all, they called the operation Data Center Bandwagon, and I wanna say they really missed an opportunity to just ma- to, for a good pun there in making it Banwagon, not Bandwagon.

318
00:59:54.911 --> 01:00:04.552
[laughs] Because actually a bandwagon is something that you join when you're, you know, late to the party, and you're jumping on it late a- a- and, and riding a wave of enthusiasm.

319
01:00:05.072 --> 01:00:12.332
But you don't actually have any sort of, you know, claim to supporting the thing, when it's actually the opposite going on here.

320
01:00:12.341 --> 01:00:23.812
Anyway, one of the insights from the debriefing on Operation Bandwagon comes from Ben Nimmo, OpenAI's principal investigator, who said, quote, "I want to be really clear here.

321
01:00:23.852 --> 01:00:33.632
This was not a case of an influence operation creating a debate. The debate already existed. This was an influence operation from China trying to interfere in it."

322
01:00:34.132 --> 01:00:45.362
And I think this is the real chicken and egg problem about the foreign interference. I, I don't think it necessarily matters, though. Like, obviously there was always going to be debate from Americans about this.

323
01:00:45.412 --> 01:00:53.812
I mean, we have a huge de-growth mindset on, in, with- within specifically the left in the US.And now that's starting to be an anti...

324
01:00:53.892 --> 01:00:59.732
There's an anti-establishment sentiment on the right that's also pushing this anti-data center rhetoric.

325
01:01:00.072 --> 01:01:12.992
So the debate was almost certainly gonna foment, but injecting needless criticism and trying to further foment it, I think is just as much of a problem as if it had been started by a foreign adversary in the first place.

326
01:01:13.032 --> 01:01:18.652
And the results from this, in 2025, local opposition, this is from a TMZ article, I believe.

327
01:01:19.512 --> 01:01:28.932
Local opposition blocked or delayed dozens of the US data center projects representing more than 150 billion in potential investment, and that was in 2025.

328
01:01:29.812 --> 01:01:41.032
That, before a lot, I mean, you know, the CapEx cycle was still booming then, but it's hit a fever pitch so far in 2026. And also, the rancor has hit a fever pitch against the data centers as well.

329
01:01:41.172 --> 01:01:48.132
So I thought that stat was worth highlighting. It is truly ma- it, it-- I mean, there's multiple narratives happening at the same time.

330
01:01:48.272 --> 01:01:53.492
One, you can have a much smaller team exert much greater effective influence in this.

331
01:01:54.012 --> 01:02:08.252
Now that we are basically in a post-reality online landscape, uh, boomers have lost the ability to decipher what is real and what's not. We will soon lose the ability to decipher what's real and what's not. Uh,

332
01:02:09.732 --> 01:02:25.032
and, uh, so how does this affect the landscape of us debating about our capabilities to produce real and, uh, fake, uh, things, and ideas, and images? Who knows? Uh, we're real.

333
01:02:25.792 --> 01:02:33.632
You can trust us to cover this [chuckles] fairly and- Are we? Are we? Are we but specters of the machine?

334
01:02:34.122 --> 01:02:39.952
[laughs] It, it-- I mean, maybe by the end of the year, we'll, we'll turn into, like, VTubers, and so you'll never be able to tell.

335
01:02:40.012 --> 01:02:51.932
Um, okay, let's go to our last, uh, story, which is, uh, the closest we'll get to crypto today because it's about tokens. About tokens, but not- But it's not about tokens that go up.

336
01:02:51.972 --> 01:03:03.352
It's about tokens that go down, and this time it's a good thing. That was well done. Yeah. It was- A good little turn of phrase there. It's not the tokens you're thinking of, at least not crypto.

337
01:03:03.361 --> 01:03:13.732
This is coming from Zero Hedge. Quote, "It's over." [laughs] And the headline is, "OpenAI mulls significant cuts to what it charges for tokens: Wall Street Journal." Now, a few things here.

338
01:03:13.792 --> 01:03:24.432
This has not been corroborated or confirmed yet. But the fact of the matter is, is this is, this is an acceleration, if it's true, of a trend that's actually been going on for some time.

339
01:03:25.092 --> 01:03:42.112
So the reason why Zero Hedge is saying it's over is that there's a f- there's this, you know, sense from the AI naysayers or those who say that we're in a bubble and that this is all crazy, that there's not enough demand for these LLMs, or somehow the economics don't make sense.

340
01:03:42.692 --> 01:03:53.732
And if you look at annual run rate as reported by Anthropic and OpenAI, Anthropic lapped OpenAI with 30 billion as of April, uh, to OpenAI's 24 billion.

341
01:03:54.272 --> 01:04:05.752
So the revenues are clearly there, and if we look at the usage, it's also there. But first, I want to get up this chart that shows how token prices have been trending downward.

342
01:04:06.252 --> 01:04:09.242
For those of you who listened to the show yesterday, this is basically...

343
01:04:09.272 --> 01:04:23.092
This is the chart that materializes James, um, MacAvity's thesis that since compute is a commodity industry, it's going to move towards the lowest cost of production, and the marginal cost should decrease.

344
01:04:23.492 --> 01:04:37.052
That's exactly what we've seen. This is LLM, this is the LLM price index from Bench LLM that shows the cost per a million tokens from March '23 to present. In March '23, it was $60 per million tokens.

345
01:04:37.112 --> 01:04:48.391
That was just across one model. It was ChatGPT. It was the only model out at the time, I believe, right? Um- I forget. And I believe, I believe that's true.

346
01:04:48.412 --> 01:05:04.052
And now if we go to April 26, the blended average of 28 models tracked is $3.30 per million tokens. That is an insane decrease. Yeah, it's- Eighty to 95% drawdown- Yeah...

347
01:05:04.092 --> 01:05:21.772
in a matter of three years. I mean, that's spectacular. But and yet, at the same time, Colin, maybe you have this queued up. Use is going up. Use is going up. Co- a very empirical example of Jevons paradox.

348
01:05:21.812 --> 01:05:36.032
The cost of a good or utility or really something we produce goes down, demand for it and use goes up because people find it more useful because it's cheaper. This is from OpenRouter, and that's a caveat on the data.

349
01:05:36.092 --> 01:05:45.172
This is not total model usage. In fact, you'll see DeepSeek is overrepresented here, which I thi- I thought was really interesting.

350
01:05:45.212 --> 01:05:58.452
It makes me wonder how, how many people, uh, in Southeast Asia or in China are actually using OpenRouter, or just how many Americans are differentiating- No, no, it's, it's 'cause DeepSeek is just the, the best bang for your buck right now from a- I see...

351
01:05:58.472 --> 01:06:05.132
frontier. Yeah. Okay, got it. Thank you. But anyway, all that being said, this is AI, uh, model rankings.

352
01:06:05.152 --> 01:06:15.532
This shows the weekly usage of models across OpenRouter, which allows you to interact with multiple models simultaneously, and it's just straight up and to the right. I mean, look at this.

353
01:06:15.592 --> 01:06:32.372
Total in June 16th, 2025, 2.25 trillion tokens used. Uh, June 8th, 2026, 45.1 trillion. I mean, that's, that's an insane increase. And so

354
01:06:33.272 --> 01:06:45.592
I, I wanted to show that to push back against this idea that lowering the price necessarily means they're struggling for users. Maybe OpenAI is. I'm not gonna comment on that.

355
01:06:45.652 --> 01:06:57.248
They have been s- losing market share to, to Claude recently, and to Anthropic with Claude. That, that much is clear to me.But that doesn't mean that demand is waning across the sector.

356
01:06:57.808 --> 01:07:13.448
I- if you look at the data, it's the exact opposite. And you can also see this in GPU rental prices. Now, th- this is, this is, uh, not the most complete showing. This is, uh, hyperscaler GPU prices.

357
01:07:13.468 --> 01:07:28.748
But we see a bump in May. Uh, we see H100 GPUs go from roughly eighty-four-- eight, eight dollars and forty cents to eight dollars and sixty cents over the last couple of weeks, or, uh, over the last month, excuse me.

358
01:07:29.228 --> 01:07:39.338
And if we look at NeoClouds, we can see an even greater increase with Blackwells absolutely surging from, like s- what is this?

359
01:07:39.808 --> 01:07:54.208
Five dollars and thirty cents per hour to, uh, to about six dollars and seventy cents currently over the last month. So by all indications, this is not, to me, a demand story.

360
01:07:55.028 --> 01:08:03.588
Or it is a demand story, per what you were saying about Jevons paradox. They are competing for users, but there's not, doesn't-- there are no indications that there's lackluster demand.

361
01:08:03.628 --> 01:08:20.798
We just have a lot of competition in the market currently. Yeah. It's a really exciting time to be a token [laughs] if, if you're not in crypto [laughs]. Um, yeah. It's, uh, it's, it's, it's very clear.

362
01:08:20.928 --> 01:08:29.428
I, I, you know, I, I will say, the only token I really like talking about is Bitcoin. So I'm glad to talk about, uh, uh, computer tokens.

363
01:08:29.988 --> 01:08:38.048
Um, I'll show maybe this last, uh, thing, which is SemiAnalysis did a really interesting, um, study on, like,

364
01:08:39.028 --> 01:08:48.448
uh, how many tokens you can get out of the free subscription tier or the various paid subscription tiers to the two leading, um, ChatGPT and Claudes.

365
01:08:49.428 --> 01:08:59.908
And, um, [clears throat] like, if you're not aware, your twenty dollar a month Claude subscription is heavily subsidized. You know, roughly ten to one.

366
01:09:00.088 --> 01:09:07.388
So if you, if you use all twenty bucks of that, uh, of that plan, you're probably using two hundred dollars worth of, uh, tokens.

367
01:09:08.098 --> 01:09:22.777
And so SemiAnalysis did a really interesting study on this, um, basically trying to, you know, estimate, uh, the margins that these companies are getting, and then how many, like, true token usage you can get out of each plan, and, like, how much you could spend.

368
01:09:22.818 --> 01:09:33.807
And then do a little, um, stuff on, like, the, uh, margin by subscription, assuming different utilization levels between Claude and GPT. Um, really fascinating study.

369
01:09:33.888 --> 01:09:40.728
Um, I don't think I'm gonna g-get into this because, uh, I think that this story is not yet fully told.

370
01:09:40.848 --> 01:09:53.168
I think we need to figure out, uh, what the average utilization of the average plan is for, uh, users of these platforms, and then we can get some really interesting numbers.

371
01:09:53.788 --> 01:10:01.438
That will probably, I anticipate, this information being revealed as these companies march towards IPO.

372
01:10:02.228 --> 01:10:18.368
So, um, that will, I-- as weird as it is, like, this specific, like, number of, like, how much utilization these companies are having of their different tiered models, um, will be very revealing as to their true economics.

373
01:10:18.448 --> 01:10:34.488
Y- you're basically saying, you know, there's a clear difference between someone who has a twenty dollar tier for O- ChatGPT or Claude and uses it to, like, ask for dinner recipes versus a power user who is getting right up to that, that token limit every single day.

374
01:10:34.528 --> 01:10:34.768
Yeah.

375
01:10:34.808 --> 01:10:55.188
Imagine the user who has, uh, the f- you know, twenty bucks, the twenty dollar pro Claude subscription, who doesn't really need it, they could probably just get by with free, versus the two hundred dollar twenty X Claude who, who maxes out every five-hour rolling window with hundreds of agents that they coordinate under Fable.

376
01:10:55.218 --> 01:11:09.568
I did that yesterday, but also, I don't want to live the Uberman sleep schedule, where I have to wake up every five hours and one minute to re-trigger my, uh, army of bots. Okay. That's it for today.

377
01:11:09.708 --> 01:11:22.398
Thank you for listening to Blockspace. If you're on CoinDesk, we are leaving the CoinDesk feed imminently. Starting next week, you will no longer be able to listen to CoinDesk. So head over to the Blockspace feed.

378
01:11:22.808 --> 01:11:37.268
Search Blockspace in whatever podcast player you're using right now, and subscribe there. Thank you so much for listening. This show is brought to you by CleanSpark, Nasdaq listed ticker CLSK. I'm Charlie. I'm Colin.

379
01:11:38.388 --> 01:11:43.968
And we will see you tomorrow. [outro music]