← Blockspace
Blockspace
Galaxy Digital and TeraWulf Q2 Earnings, TeraWulf CFO Interview, SpaceX Q2 Earnings and Orbital Data Centers
01:21:32
Loading summary
Transcript123 lines
- [00:00]
A
What's up y'?
- [00:00]
B
All?
- [00:00]
C
Welcome back to Blockspace Live presented by Clean Spark for day three of earnings week and our second day of financials to pour through today. We have Galaxy Digital and Terra Wolf just released their Q2 earnings this morning. Riots were expected today, but they actually rescheduled their call. So we'll have that on the calendar and let y' all know on the stream once we get a time pin down for that. So for today we've just got Galaxy and Ter Wolf at the top of the show. We're going to be going through numbers from their earnings, the actuals versus the expected for revenue and earnings per share, plus operational updates and some new data centers at least from Galaxy side of things in terms of bumping up their gigawatt expansion pipeline. We will lead with those two earnings calls recaps and then following our Terra Wolf one second up we have CFO of Terra Wolf Patrick Fleury, on to talk about some of the financing considerations he is juggling as he looks at massive expansion. Not just across Lake Mariner where we're also talking about Hawesville, we've got the Morgantown site, Lake Cayuga which is further out, but a lot to manage. So we're going to be grilling Patrick over just exactly how he's approaching managing the capital stack for Terra Wolf right now and what financing options are will be available to the company in the future as they tackle these expansions. After that we will also be covering SpaceX's most recent earnings, the first one post IPO. Charlie's got the receipts as well as some commentary on their far flung aspirations to put data centers in space.
- [01:41]
A
That's right, Block Space goes live every weekday at 1pm Eastern. We are COMPUTE's daily live show featuring quick hits on AI, data centers, markets and emerging technology. With the exception of today because we have esteemed guest Patrick Fleury, CFO of Terrible, on sort of bumping it up an hour. Hopefully we didn't disrupt your daily plans too much. If you like the live stream but you miss it, you can catch it in podcast form shortly after we wrap up here, wherever podcasts are found. If that's not enough for you and you want to get more Blackspace content, you can go to our website BlockSpace Media written content coverage get our newsletter delivered straight to your inbox every single day with stuff on the pod and off of it. Go to newsletter.blockspace media.com this show is brought to you by Clean Spark NASDAQ listed ticker CLSK More on Clean Spark later on in the show. Today we kicked it off with two quarterly earnings from both Terawolf and Galaxy. So you know what that means. Colin, it's time for the earnings update.
- [03:00]
C
The majority of the remainder of this podcast will be broadcasted in Morse code, just to let everyone know.
- [03:06]
A
Exactly. So have your AI transcribe Morse code. Shouldn't be too hard. That's a. That's a sonnet level task. Okay, so if you weren't paying attention, this is earnings week one of two that we're doing here at Block Space. We're on day three, Wednesday. Today we had Terrible from Galaxy Riot, originally had their earnings schedule for this morning. They have moved it. The new date is tbd. So we had Terrible and Galaxy Stream this morning. We'll get into those in a second. But looking ahead at the week, Thursday, tomorrow in the afternoon we have Clean Spark and Mara holdings back to back. And then Friday, Core Scientific. We've been having executives from these companies on the live stream for interviews during the week. We had Eric Ellingson of Fortitude Ionic Digital with Andy Stewart yesterday, Patrick Fleury of Terra Wolf today, and on Friday, asher Ganut of Hut 8. So Colin, I'm a toss it to you because you're the one who knows the numbers, I'm just the color. So what's happening with our first equity of the day?
- [04:19]
C
So Galaxy reporting this morning at what, like 8:30am ET? Bright and early for all the early birds. Looking at the 10Q headline here from Block Space, quote, Galaxy Digital reports narrower loss as Helios phase one comes online. Q2 earnings. Galaxy had a big quarter for their data center segment, announcing at throughout during the quarter the first commencement of Phase one at Helios and continued expansion for that core weave contract. The disappointment for Galaxy came from their cryptocurrency business and that's I think largely why the stock was selling off today. Charlie, it's down like 15% as of the time of recording. We won't be covering the cryptocurrency stuff as much. We'll be focusing mainly on the data center segment. But I did want to highlight those numbers regarding. Regarding why this quarter was somewhat of a disappointment from that cryptocurrency angle. So digital Asset revenue in Q2 2026 for Galaxy was 8.7 billion versus 10.35 billion in Q1, so about a 16% quarter over quarter change to the downside, their total revenue for the quarter was 8.557 billion, which just missed consensus estimates aggregated by FactSet at 8.85 billion, so about a 3% miss. Again, I think largely the revenue decline from the cryptocurrency segment is what is hurting them following this earnings call with the stock price. That being said, that's not where the bulk of their attention is right now. They're really hammering into this data center element. That being said, their adjusted earnings per share came in with a beat, a big beat. The estimate was about negative $0.20 per share, came in at about negative $0.09 per share. So a beat on earnings per share, a near miss on revenue and and digital asset income falling quarter over quarter. But data center revenue is up. Charlie Galaxy book 26 million in Q2 2026 for their data center segment versus 3.1 million in Q1, so plus 749%. It's not a huge number obviously, but the run rate for this revenue is much greater than what the headline from what they actually earned in Q2 states because they didn't have the full capacity online during Q. And that capacity is concentrated so far in Helios. Now, as we've covered on the show before, Helios is probably the most valuable M and A in the bitcoin mining space, data center space that we've seen to date. Galaxy purchased Helios from Argo Blockchain for I believe like 65 million back in, I want to say 2022 and somewhere
- [07:19]
A
in the depths that everybody just forgets. But it was. Funny.
- [07:27]
C
Yeah, it was back in the prior bear market, the bear market before this bear market and now that heliosite is a multi billion dollar asset for them. So phase one of Helios has been delivered 133 megawatts of critical IT load, 200 megawatts gross load to core weave by the end of the quarter. Under the 15 year lease rent scaled up throughout Q2 as capacity came online. Now it is fully outfitted. So rent has commenced for that phase one. Management guided that the full 133 megawatts now delivered phase one should generate $80 million per quarter in leasing revenue and a 90% plus projected or project level adjusted EBITDA margin starting in Q3 2026, which is a big step up from from the data center revenue booked in Q2. The exact number was 18.9 million rounded out to 20. And ultimately this is the bedrock of their data center play. They will continue to focus on it and get the other phases online throughout the rest of the year and into 2027. Phase 2 at 260 megawatts is under construction and data hall deliveries are expected in Q2 2027 and Galaxy on July 28th after the quarter closed closed a $3.5 billion private offering of senior secured notes due 2031 to fund phase two construction. So they have that senior secured note out there. The funding for this expansion has been secured and that will be the focus throughout the rest of the year and into Q2 2027. Now, that being said, there is a huge pipeline for Galaxy in addition to what they are doing with Helios, and Helios is part of that equation. They actually have more expansion at Helios to chew through. There's 830megawatts of already approved capacity at Helios, but management says discussions with prospective tenants are ongoing, plus 2 gigawatts more under study at Helios alone. Now, they are going to run into potentially some hiccups. As we reported yesterday, Greg Abbott has mandated an audit of all data centers in the queue for batch 0. This is the first batch under ERCOT's new rules for large load interconnections. And there are a host of questions up in the air right now as to what that will do to delay the timeline for approval for those projects in batch zero. And just how much will actually go through the queue once all of these audits are conducted and measured against Texas generating capacity and projected generating capacity over the coming years.
- [10:16]
A
Go ahead, Charlie. Yeah, I mean, and this is kind of a story. So the, you know, I believe it's
- [10:22]
B
the
- [10:25]
A
existing 1.63 gigawatts of Helios and is already approved. So that has gone through and it should not be affected by the batch geo moratorium. But the additional load requests, I think that you mentioned, the 700 megawatts and 900 megawatts are entering the queue. So if there's a delay on this current batch, then the subsequent batches, which I assume these are part of, also probably delayed. But at the same time it almost kind of makes anyone with existing approved interconnection, such as Galaxy already has that much more valuable. So relative to the field, if you've got, you know, gigawatts in the hand versus two in the queue, then that's, you know, that's a position of strength.
- [11:18]
C
So, yeah, and that's a good thing to point out and just to clear, clear up the numbers there, the 830 I was mentioning was unleashed capacity. And so Charlie said there's 1.3 approved at Helios. So Galaxy sitting on 830 megawatts at Helios that they have approved. Obviously it hasn't been built yet, but that will be prime real estate for any NEO cloud or compute operator that's looking to move in quickly. I wouldn't be surprised if we saw an announcement for that 830megawatts whip before the year's up, given how competitive the landscape is.
- [11:49]
A
And we we haven't seen anyone at like at a top tier operator level have trouble leasing these things yet. Right?
- [12:00]
C
Yeah. Demand is insatiable. And the other thing that I will note with regards to the batch zero equation, one thing that was pointed out by Morgan Stanley in a recent research Note regarding Batch 0 and the interconnection process is that Galaxy and other peers, I think they listed Cypher as well for the sites that they have operational that they're trying to tack on extra megawatts to those could be classified as baseload, which would help them ease through the process much more quickly than if they weren't considered baseload because that is load that you can curtail and add back to the grid when capacity is at a shortfall. So a lot of questions regarding that. We're going to have some content over the next few weeks covering this pause in the audit in Texas with batch zero. Probably one of the bigger stories this year, if not the biggest story with regards to the whole AI Capex cycle. But a few more notes before we move on to Terra Wolf here. Charlie. Galaxy announced that it is scoping out two new sites called Caspian with 700 megawatt potential and Celine with 900 megawatt potential in Texas. So also potentially hamstrung by the delays to the batch zero and interconnection process. But this has a this ups their power pipeline to 5.7 gigawatts when we also account for Merlin. That's the McGregor Industrial Park, 500 acres that Galaxy has purchased and is looking to roll out an initial 74 megawatts, notably 1 megawatt less than the threshold, to be considered a large load in Texas. So Galaxy playing their cards right here saying we're going to sneak in underneath the threshold, get this thing going and, and then see how we can expand from there. But 5.7 gigawatts is massive. That's a lot of power to deliver and there's going to be a lot of sweat and toil between now and when that actually comes online. That'll be multi year pro. These will be multi year projects. Notably, they didn't really say much about Caspian and Seline. There was no Other supplementary information, it seems like they're just scoping these sites now. These are not sites they're actively developing, nor from what I can see in the disclosures, have they actually purchased anything yet. So that's important to note a few other things before I toss it to you for closing thoughts. Charlie. And we move on. Looking at data center segment liability that grew to 1.55 billion against 2.54 billion in assets. Just goes into the capex story. Leverage is building alongside these build outs. Nothing surprising considering the debt raises. Importantly though, data centers now represent 36% of Galaxy's $2.7 billion equity base. That's tied with digital assets, which is the not saying it's the lion's share isn't even fair. It is the revenue for the company right now and with their treasury and corporate at 28%. So digital assets and data centers are now tied with regards to their value in the equity base. And I think that more than anything kind of spells the future out for Galaxy. You know, there's they're going to hold on to the cryptocurrency financial, financial services angle. No doubt it's what made the business what it is today. But going forward like two, three, four years from now, I believe that the data center segment will be the bulk of the business that people care about unless something changes within Bitcoin between that time and then, you know, Galaxy is well positioned to capitalize on another bull market.
- [15:36]
A
So yeah, so zooming really far out, my takeaway on this, and this is not a unique takeaway, I think this is what everyone's saying here. Galaxy Digital, which has been one of the prominent like crypto asset digital asset trading firms and institutions, is not really defined by that anymore. And in fact, the data center component of Galaxy has now is now tied roughly for 36% of Galaxy's $2.7 billion equity base. Digital assets now 36, about 36%, same amount, and then the treasury and corporate at about 28%. And that's because data center has grown relative to the rest of the portfolio in equity. So Galaxy is basically their growth is happening in data centers. And that's great because that's a very bullish place to be and the sector for growth for pretty much the entire American economy right now. So that's my takeaway from the earnings like the super high level takeaway.
- [16:47]
C
I think we can cap that there
- [16:49]
A
and we will go on to the second earnings of this morning from Terra Wolf. But before that, a word from our
- [16:57]
C
sponsor, CleanSpark we are CleanSpark, America's Bitcoin miner. A publicly traded company with the largest operating hash rate powered entirely by self operated infrastructure across four states. This is our proof of work and we are setting the standard for what's next. Learn more about the intersection of of energy and bitcoin@cleanspark.com Alrighty Charlie, let's hop on over to Terra Wolf, see what's going on over there. Another big quarter for a company that we cover in terms of deal flow and expansion, but Wolf down slightly on the day. There was a huge pop leading up to the earnings but it has reverted back to roughly where it was at the start of the day. Let's check the stock chart really quickly. Down just 42 basis points on the day. Not really anything material. And the reason for this I think despite a few miss on revenue and earnings per share and there's an accounting quirk as to why they missed on earnings per share that we'll get to in a second. I think investors liked what they saw specifically with an update on the lease for Lake Mariner, most likely tied to Fluid Stack. They didn't specifically name the tenant, but given the build out that's going on there is for Fluid Stack I would assume that's what it is. Specifically they're milking more revenue from that contract because Fluid Stack or the tenant in question has upped the capacity they are requiring from Terra Wolf. We'll get to that here in a second, but here are the headline numbers in terms of total revenue. 44.8 million. The consensus estimate was 46 to 49 million. So a slight miss there. And GAAP earnings per share came in at negative $1.94 versus negative $0.19 to $0.24 range, which was a big miss. But that's largely because there was a change in the value of warrants on its balance sheet. And specifically these are warrants tied to the Fluid Stack and Google deal. I would imagine they might have other warrants outstanding. They don't break it out, but those are the bulk because Google holds like 14% stake in Terra Wolf through these warrants and so the change in those because Terra Wolf stock price went up over the quarter, those are counted as a liability based on the way that they are consolidated into the financials. But that's not the same as like a cash liability with something like debt where you know you're going to have to pay cash out of pocket. These are warrants, they're shares that will be converted to equity eventually. So it's not really, it's not fair or it's not how should we say indicative of the company's financial health to look at that and discount it. So overall revenue by segment Bitcoin mining came in at 12.8 million. Not really a factor anymore in terms of their day to day operations. HBC lease revenue 31.9 million which was zero year ago. And shows that Tera Wolf, which was I believe the first company in our coverage former bitcoin miners to actually have HBC revenue flip their mining revenue. We're, we're, we're, we can retire the bitcoin miner label. I think Charlie that for increasingly for many of these companies that obviously has no bearing on what they're actually trying to do day to day anymore.
- [20:26]
A
A, a, a tiny little violin place to play the bitcoin miners out. Even the bitcoin miners we had Eric Ellingson on Monday. They're becoming zcash miners. So everybody fleeing like rats from a ship. Who will come save us?
- [20:45]
C
If you needed a another bottom indicator you might have it or just that we're getting closer to desperation.
- [20:52]
A
Yeah, Zeta hash is both a psychological threshold and, and an actual one as the industry rotates out.
- [21:00]
C
So yeah, it's like 100k for Bitcoin's price. So going back through a few other numbers, net loss was 939.9 million. This was driven entirely by the change in fair value of warrant liabilities. That was a $755.7 million non cash change. Again that's not a cash cost but it is factored in to their earnings. And so if you're seeing skewed numbers on that, that's why there's also 83.9 million in stock based comp and a 7.1 million debt extinguishment loss. Adjusted EBITDA came in at 18.3 million negative 18.3 million versus 14.5 million positive a year ago. Management attributes the decline to higher SGA and operating costs as contracted HBC capacity ramps up. Something that we saw with Cypher as well. You know there's more. The executive compensation stock based compensation is going up as they try to as they attract personnel to lead these AI revamps. An update on operations Moving away from the financials and looking specifically at the platform level data center business side of things. Terra Wolf now has a 2.1 gigawatt controlled pipeline across five sites, 27 billion plus of total contracted revenue across the platform and a 1.5 billion targeted average annual net operating income once that rent is stabilized. They are targeting 85% net operating income margin on contracted capacity. They're targeting like a 4.5x leverage target. I believe that's for the total consideration of the values of these contracts. They have 1.9 billion in cash. Capex spend to date has been 2.3 billion with 1.7 billion in capex remaining throughout the year, backed by 3.2 billion in senior secured notes. Roughly 90% of Wolf's compute construction capacity is now under committed purchase orders. So they've only got 10% gap there for the remaining materials for these sites. A few updates on those sites individually and we will touch on the financing angle for these once Patrick joins us here Shortly. Lake Mariner, 102 megawatts energized and generating lease revenue as of July 2026. That means the CB3 data hall is now operational. CB4 at 168 megawatts and CB5, also 168 megawatts, is progressing towards delivery in either the second half of this year or the first quarter of 2027. This is the big one where I think, I think this is why Terra Wolf has been resilient even disregarding the sell off after the spike before earnings or during earnings. Actually this is probably why it spiked and then maybe sold off reverting to the mean. On July 5, 2026, ter Wolf executed lease amendments with the McMariner tenants that shifted the rent schedule and delivery for CB3, CB4 and CB5 in exchange for 500 million of incremental revenue over the lease terms. Terra wolf is contributing 150 million to fund tenant fit out costs incurred through June 30, 2026. But any fit out costs after July 5 are borne by the tenant. Interesting development here and they didn't give a lot of information as to what was going on here. But this also coincides with CB4 and CB5 picking up 6 megawatts of critical IT capacity each beyond what was contracted back in February of 2026. This seems like a design shift from the tenant, in this case Fluidstack. They found a way to juice more compute, have a denser rollout it sounds like and they want to replicate that on the future builds. And what this means for Terror Wolf. With this scope change, the tenant is getting more capacity per building which requires additional design and fit out work and this will probably push the timeline for this out a little bit. Terror Wolf is protected against that within the amendment. They they structured it as such to where there are no penalties on delays considering this is kind of like a ninth inning call, you know, from the pitcher here. But this means that they will increase the incremental revenue over this lease to $500 million for the full leasing term. A pretty nice little added bonus here for Terra Wolf and no doubt was something investors looked favorably on during the earnings call. A few more updates here Charlie, before I toss it to you. The Justified data center in Hawesville, Kentucky. This is the 480 megawatt campus in the miso market. Again 401 megawatts is least anthropic. The 19 billion 20 year term with anthropic has five year extensions that will put it at 33 billion delivery targeted for the second half of 2027, first half of 2028 and then they've also got their Muskie Data center in Grayson, Kentucky. 1000 megawatt campus and PJM they going, they're going to have an initial 500 megawatt ramp up with a target of Q4 2028. Second phase of 500 megawatts targeted in 2030. And the big one that we've covered recently on the pod, Terra Wolf having a win here with FERC clearance. The Chesapeake Data center and Energy in Morgantown, Maryland. This is a net generating campus. It has two phases. A 500 megawatt generation with 250 megawatts of battery storage and 500 megawatts of data center load totaling 1 gigawatt generation and 500 megawatts of storage. There are still some closing and regulatory approval, closing contingencies and regulatory approvals for this site. But they cleared the first big hurdle with that FER clearance as we covered on the stream recently. That site's really interesting because they are taking the, the oil plant assets that are on the site and converting them to nat gas. They have the backup power on site. That's it's going to be a big cap expert and something that I'm going to grill Patrick on here in a second once we get them on. But probably one of the most exciting sites in any of these companies portfolios if you're looking at the potential to like have your own power, have the battery back up and be able to be interconnected directly to the grid to arbitrage between actually selling power when you need to empowering your data center when it makes sense.
- [27:34]
A
It's almost like you know how like Riot is not really like a mining company. They traded power. You know could see a similar thing here man. I'm looking at like you're dropping these numbers which is Wild. And I'm just thinking 2028, so many megawatts are going to come online. You'll have Vera Rubins deployed all across the world and maybe in space after the SpaceX thing. And that's when the, I believe the next Nvidia, the Feynman GPU is expected to launch. My Fable 10 queries are going to run at the speed of light by then. So very excited to have superintelligence.
- [28:11]
C
I've been thinking about this recently just as a kind of a tangent and aside, when people talk about the compute constraints and people can feel it sometimes when they're using their models and Claude or ChatGPT hallucinates or like the context is too Windows too big, there's not enough resources. You know, what does it look like in two years when we have a lot of these build outs finished? Does it get worse because there's actually just more demand or we actually see some of that congestion clear?
- [28:38]
B
Right.
- [28:38]
C
And these models kind of flourish in a way that they maybe have not been able to do so far under the capacity constrained environment.
- [28:46]
A
That is literally the 700 billion to $1 trillion CapEx question. Probably more by 2028. We'll see when we get there. I guess I'll pull up the, the scoreboard really quick this morning I've just got the companies we've covered that release their earnings. Earnings has been kind of like a. Yeah, not, not done well for the companies which have done earnings. Except for Wolf who's flat on the day, but Galaxy Hut, Ionic Cipher, all down since earnings. But to be fair, they're up like
- [29:21]
C
double digits on the week and I think that.
- [29:23]
A
Well, I think they're all up for the year too.
- [29:26]
C
So yeah, they're. Everything's still green, but we've rebounded heavily from the Leopold fallout, if you will. And I wonder how much of this, you know, I mean imagine if these earnings came in the, in the, in the depth of that sell off. I think we would be maybe telling a different story with regards to where the stocks are moving.
- [29:47]
A
But yeah, it feels like the market is just a little antsy and reactionary. We need Ken Griffin just come by the entire stock market so we can go up.
- [29:53]
C
Well, it's, it's crazy right now, man. You have like hypers. You have mag7 companies adding like 10% in a day.
- [29:59]
A
Amazon and Microsoft.
- [30:01]
C
Yeah, like 250 billion added to the market cap in a single trading session.
- [30:06]
B
Right.
- [30:07]
A
So, okay, we, we're going to stop riffing. We have our esteemed guest, Patrick Fleury in the wings. We'll bring him on up here in just a moment after a word from our sponsor, Luxor.
- [30:24]
C
This episode of Blockspace Live is brought to you by Luxor's Commander Bitcoin miner management software for enterprise operations. Luxor's Commander gives you real time fleet monitoring, bulk remote commands across your fleet and Intelligent miner, that's an automated profitability engine that runs every five minutes and tests your fleet's power settings against live energy and hash rate markets. Ercot backtest showed 10% improved profitability with intelligent mining versus binary mining. Commander Pro is roughly half the cost of competition. Either $100amegawatt or a 25 basis point pool fee adder. But you can also try it free for 60 days. So if you'd like to learn more, go to Luxor tech forward slash Commander to get started. All right, we have Patrick Fleury waiting in the wings and we will bring him up for interview on what's going on at Wolf, specifically from the finance side. Patrick, thank you for joining. Sir, it's been, it's been a minute. I think you were on first last about a year ago, so it's nice to have you back on.
- [31:21]
B
Good to be back.
- [31:23]
C
Thank you for joining and congrats on what shaped up to be a really banner quarter for y' all in terms of the operational side of things. I'll save most of the operational questions for your counterpart Nazair Khan, but for you, I've got some questions specifically regarding financing and I would like to start with Hawesville. Specifically the routes that you are weighing, taking in terms of financing. That site, you know, shaping up to be one of the most exciting sites in the portfolio with Anthropic as the counterparty on Bloomberg you said that the financing would be a mix of leveraged loans and high yield bonds. Is that structured as a ring fence project financing similar to Wolf Compute or more like a corporate level raise at this time. And if it's not, I'm curious about why that structure versus doing it entirely through a dedicated project financing route.
- [32:19]
B
It will look a lot like Wolf Compute. So it'll be a secured asset level financing. Right. First lien security on the data center and it, it that will be, you know, likely with a compare completion guarantee. So I think it will look very similar to Wolf Compute, which again is the IP that, you know, we developed initially with Morgan Stanley and Google Fluidstack and our partners, you know, Anthropic and others at, at Lake Mariner. So I think where it could evolve is you know, the structures, I think as you point out could have some loans and maybe not be all bonds because as you know, bonds are not you typically non call to at least high yield bonds. Right. Some of my peers have done investment grade bonds which I continue to scratch my head on because you can't call those for 15, 16 years. And I'm not interested in doing those because I think my cost of capital is going to change a lot over the next few years. So I think high yield bonds and loans, loans, you know, tend to be more like, you know, 101 par call protection. So just much more flexible instrument and yeah, so I think it'll be like that. I also think the, a lot of my peers are pushing leverage, you know, they're trying to do 95, 100 like loan to cost financings. Right. Because you've got good counterparty standing behind the leases. I, I don't think we'll do that. Like I said on the call we have, I was saying internally, you know, just because you can doesn't mean you should. These projects are complicated and if you have no equity layer in them, you know, you can stub your toe and a stub toe becomes an amputated leg. And that, that's not what we're looking to do. I think we will, we will always structure them with a healthy equity layer because it gives you protection and the designs are moving around as you all know. I mean I think you saw it with our lease amendments at Lake Mariner. The reality is like our customers and Whether it's Nvidia GPUs or Google TPUs or AMD, whatever you're using, there's just not many clusters running at size today. So the customers are getting data points every single day and constantly coming to us and saying, hey, you know, what if we did this or what if we did that? And so it's a really very constant dialogue around yeah sure, we can tweak this or we can tweak that and a design change here or there, but all of those have typically a cost impact and a schedule impact. Right. And so it's a sort of a combination of us working with the customers to optimize the design on the go, but then balancing, you know, what they want to change with how does it impact cost and how does it impact schedule. And so some of that's accepted, some of it's not.
- [35:16]
C
Nothing really underscores how fast things are moving than those amendments to the, to the Lake Marin releases. It's like oh we unlocked 6 extra megawatts with this design change. And I would imagine you have to really think on your feet, not just from the op side, but from the finance side as that new information comes in. Just a kind of a side note there. I have kind of two follow up questions for this and I think I'll tackle this one first. And it's off the cuff because you just said you would rather not do high investment grade bonds specifically because you think your cost of capital will be lower in the coming years and there's less flexibility. Could you unpack that for us? Is the idea that you just expect terrible credit rating to go up so I'll have access to cheaper credit? Do you expect rates to come down over the coming years? What's, why do you want to avoid those specifically?
- [36:06]
B
Yeah, the short answer is all the above. So I think a couple of things are going to happen. One, financing a project that's under construction versus one that is operating and has had some operating history, that's a change in cost capital. Right. I was a credit debt investor for almost 20 years. Like if you have something that's operating right and has reliable cash flow and doesn't have a construction ramp period or construction risk, it's generally going to be cheaper. Right. Than something that has construction risk and execution risk. So that's number one. So yes, I do think like once all of my data centers are up and online at Wolf compute and Google's 3.2 billion of credit backstop has kicked in and we're putting that cash flow through the income statement, I think those, you know, my, the cost of financing that longer term should be lower than it was when I was, you know, raising money to build, build them and, and hadn't proven that I could do it over and over and over again. So number one, yes. And so that's why I like high yield bonds. Because high yield bonds and you can see where my Wolf Compute bonds are trading probably around 6%. Well, you know, there are some of my peers that have done investment grade bonds that are pricing 50 basis points tighter. So for 50 basis points I can tell you every single day of the week, I, I will want the ability and the option to call that bond out in two years or three years where if you have an investment grade bond that you can't touch for 16 years, you can't touch it for 16 years. Right. It's like refinancing your house. So I can, I would sit here and say in 16 years you're telling me there's not going to be a recession, Rates are going to go to zero and you can go refinance your mortgage on your house at 2 and an eighth as opposed to 6% where it is today. So again, I think it's not only having lived through a bunch of cycles, right. And seeing the great financial crisis and then Covid and other things. I mean rates will have ups and downs in the economy. Rates will go to zero versus where they are today at some point and you'll be able to finance that a lot tighter. So I think it's a combination of multiple different factors. But again, to simple analogy for me is paying an extra 50 basis points a year. I mean just do the math. 50 basis points a year on you know, $3 billion. Right. Is what 1% is? 30 million. Yeah, 30 million. So 15 basis or 50 basis points is like 15 million bucks. So basically I'm paying 15 million bucks more for two years to have the optionality to take that thing out whenever I want to take it out as opposed to being locked in a structure for 16 years. Right. That, that clearly had some pricing day one of construction risk because the thing is under construction. So that's why.
- [38:58]
C
Yeah, that makes sense to me. It's kind of playing like short term gain versus the prospect of long term gain with regards to having a lower cost of capital over the entire span of the project financing. Like you said, like those 15 year terms. One last question with regards to Hawesville recently you also mentioned that there's a that Anthropic or it was you or someone on Terrible mentioned that Anthropic is still working through vendors for the chips on this. Will that vendor end up impacting at all the route that y' all take? If there's like a backstop involved? If there's one vendor that you would rather have versus the other in terms of credit worthiness, does that change the equation at all for how y' all anticipate financing the site 100%.
- [39:40]
B
And again, that's another example of us working with Anthropic. Right. Anthropic as you know, is, is choosing hardware from Nvidia, Google, Amazon and now amd. So they want competition in that to get the best price that they can get. We want competition also to get, you know, the cheapest wrap that we can get from the, the best rated counterparty. So yeah, I think that was purposeful with us working with Anthropic, you know, to, to create a competitive environment that they want and also a competitive environment that we want to get the best terms for the project as a whole.
- [40:21]
C
I want to move to, to Morgantown really quickly if we can. You've got a bunch of projects to figure out the financing puzzle for and they're massive capex, you know, evolving massive capex. Morgantown to me seems like the one that would probably have the most out of all of them. There's this, there's this retrofit of the actual plant itself into nat gas from oil. There's the battery backup which is first phase, 250megawatts worth. When you approach financing for this site, how do the power generate generating assets change your calculus for what is most doable and also what has the lowest cost of capital for investing in this site? Does it change the calculus that much or not?
- [41:09]
B
Yeah, yeah, yeah. So let me answer the first part sort of piece of that question first and then we'll come back to Morgantown. So what I tried to hammer home on the call today is we are very flush. We have a lot of cash on balance sheet and I got more cash coming in from the sale of my Abernathy jv. So I got equity to fund Hawesville, I got equity to, you know, possibly fund the first round at Muskie, which I think that's underappreciated by the market. And then I got a lot of cash sitting on balance sheet that I'm going to keep on balance sheet while we're executing these big projects so that you know, we always have enough liquidity. So again I think like I do not have to touch the equity capital markets. I think for quite some time I do have to go access the debt capital markets right to finance anthropic and I think you'll see that process evolve as we've talked about a credit backstop, you know, what the right technology is, what the full capex is. So you know, I. The nice thing is I have a lot of time because we went to the equity capital markets in April, stacked a lot of cash on the balance sheet. I have Runway so I don't have a gun to my head to go to the markets. And particularly given how choppy these markets are, feel really blessed to have that cash there so I don't have to go when I don't want to go. So that's number one two on Morgantown I think. You know, we have looked at a ton of behind the meter deals, literally somewhere between probably half a dozen and a dozen. Most of them have been in Texas or you know, in the sort of Midwest or you know, Northwest and Generally, most of them don't work for a couple reasons. Okay. The first is, you know, we come from power. In power, you know, reciprocating engines are called peakers, and they're called peakers for a reason because they typically run, you know, 2 to 5% of the year when power prices are sky high and they're inefficient, they cost a lot of. So when you're doing behind the meter deals and you know, you have three different OEMs putting in three different kinds of peakers to bridge you to a, you know, baseload ultimate power solution in three, four, five years, that gets really tricky really quickly because having enough people having mechanics having spare parts for three different OEMs, and then trying to run equipment at 99.99% of the time that typically was supposed to run like 5 to 10% of the time, there's a lot to unpack there. So that's one of the number one reasons we haven't done one of those deals yet. Number two is most of those deals have different equity owners and construction parties involved in the power side and the data center side. That creates big conflicts of interest if it's just not all the same equity ownership, because if the power falls down, there's SLAs and the data center deals where you can go with damages to the power provider. So it also gets really messy really quickly if all parties aren't fully aligned and rowing for the beach in the same direction. So our approach at Morgantown is designed to basically solve all of that, which I'll explain to you, and then also solve the political problem. And it doesn't matter if you're a Democrat or Republican. Right. Which just is data centers, you consume a lot of power and so you can't be a parasite on the grid. You have got to either, you know, contribute and ideally contribute what you're pulling. And then as the approach that we want to take at Morgantown is actually to be a net provider right through the battery during times of crisis. And so if you just. Let's try to keep it high level, but let me unpack Morgantown for you. If we can build a Data center for 10 million to 12 million per megawatt, okay. We can add a combined cycle power plant for two and a half to 3 million per megawatt, and then we can add a battery for one to one and a half, like round numbers, like midpoint of. That's probably around 15 million per megawatt. Okay. And that compares with data centers that are being built in West Texas that have a lot of elements I just talked about behind the meter peakers, but they're not providing an asset to the grid that's really for their purpose only. And those are being built at, you know, 12, 13, 14 million per megawatt. So not all that different from what we're talking about. If you are, then my lease counterparty, a hyperscaler, it's pretty simple. Instead of just provide signing a lease for the 10 to 12 million per megawatt that I'm signing right now, you're signing a lease for the whole thing for 15. But importantly, here's what you are getting power in that part of the world is generally speaking, I'm going to use round numbers here and we can drill down if you want to, at some other point, like into the specifics. Power in that part of the world is roughly 100 bucks a megawatt hour. Okay? So if you're buying from the grid, you're paying $100amegawatt hour. If you have a power plant that's selling to the grid, your marginal cost to produce that power is about a 6,000 heat rate for a combined cycle. That's buying gas at six dollars an MMB to you, six times six is 36 bucks a megawatt hour. And then you have like about 10 to 15 bucks of operations and maintenance costs. So round numbers, 50 bucks marginal cost to sell power to the grid. So if I'm buying at 100 and I'm selling a 50, I just basically netted back 50. And then if I have the battery there that's selling say from 2pm to 7pm when power is really expensive. But then at night I'm reloading that from the grid when power is really cheap and there's not much demand. And I keep doing that every time, every day, we think that's another 20 to $30 that you are netting back. So think about that for a moment. What where if you're a hyperscaler, your power used to cost you. And by the way, for a thousand megawatt data center, simple math. Thousand megawatts times 8,760 hours in a year, that's just 24 times 365, that's 8,760,000 times a hundred dollars a megawatt hour, that's $876 million a year in power cost. But what I just told you is I think I can shrink that for that hyperscaler to 20. And not only that, but you're do your, your that hyperscaler is grid connected, which is what they want 10 times out of 10. They're not only putting the amount of power that they're pulling from the grid back onto the grid, but they're being a source of incremental power to the grid during times of crisis by having the battery that accomplishes again, whether you're Republican or Democrat, what the Trump administration, what every governor in pjm, what PJM wants, what every independent system operator that runs power in the, in, in the US wants, Right? Which is you are not a data center, is not being a parasite to the grid. And not only that, but it doesn't matter. I don't care if you're Google, Amazon, Microsoft, Meta, Nvidia, saving you $600 million a year in power costs is a big deal. And not only that, but you are on the right side of, of all of those ratepayer political NIMBY discussions, right? You're, you're actually being a resource to the grid. And so that's what makes that site and I think our approach there so exciting. The other amazing thing about that site is it's basically 30 miles east of Data Center Alley. So the amount of demand we have for that site, given the location and the likely cost that we can charge for a data center there is similar to what the cost is in Data Center Alley that people are getting. I mean yields in that part of the world tend to be much higher. Right? Because you're basically, you know, that much closer to where the certainly the inference, big inference need is, particularly from government
- [49:29]
C
players that, that you just gave the whole bull case for that site I think well articulated better than I could have done it for. Why it's again one of the more unique sites in any company's portfolio for data centers for the companies we cover. Patrick, you said there that you are feel blessed to have a cash cushion right now and you don't have to tap the debt markets currently for any immediate expansion. Why is that specifically? Kind of want to get into how the capital markets have changed as we continue through the CapEx cycle order. Just a few things that kind of indicate this order. Coverage on hyperscaler bond deals have fallen from about 5x in February to under 2x in July. And Amazon also had to sweeten its July raise with a little extra yield. How is the credit cycle shifting for AI operators, if at all?
- [50:24]
B
Yeah, look, it's a huge capex build. I've seen this twice in my career. Once was in telecom, late 90s, early 2000s. The other was in oil and gas in the 2010s, you know, when you have these industries that right, explode and need a lot of capital, you know there's, there's going to be winners, losers, there's going to be, you know, a couple blow ups along the way, right? So I think we're early days and everyone's still trying to figure out where do I want to put my money and who do I want to give it to and what's the right return on that capital. And so you have all these companies, right? I mean take like Google for instance, you know, who I think is one of the best companies in the world and are super privileged, you know, to be partners with them. But when, when we were negotiating the first wrap last summer, Google was a 2 trillion dollar public company. Right now they're 4. They at that point in time had 32 billion I think of balance sheet debt. You know, I haven't looked today but I bet it's closer to like 100, right? So still on a $4 trillion company like 100 billion of debt isn't a ton. But it, if you think about each one of the hyperscalers, right, and their free cash flow profile, their debt profile, their massive market caps, like there's change happening. And, and to be honest, this is one of the things that we think about all the time when we're signing long term customers. It's not what is the credit quality today. It's like how durable is your business model. Right? I mean, I mean that's a huge factor and reason why we have anthropic and perhaps not somebody else because they're targeting enterprise customers long term and we think that's really sticky. But back to just one thing to correct from earlier. So I don't need to tap the equity capital markets that I don't need to touch. I do need, I'm going to go to the debt capital markets for Hawesville, you know, and then likely Muskie, you know, in the next 12 months, you know, I'll be, you know, looking probably to finance both of those.
- [52:36]
C
Thank you for the correction there. I wanted to just two more questions here. Patrick, I know we're running up on time here, but I wanted to get a audience question here real quickly from JZS Squared. Patrick, on the call you mentioned that the loan to cost of Mariner was now 70%. Is that a function of rising cost of debt? Will Future deals be 70 cent? 70% LTC also?
- [52:59]
A
No.
- [53:00]
B
So what I was trying to point out there is when we initially financed Wolf Compute in October of last year. It was roughly 75, 25 debt equity with the agreement, lease amendments and the agreement for us to fund incremental capital of about 150 million which again is recovered in rent. If you take, you know, that amount of capital, you know, plus some additional capital that we're putting in to basically solve the labor problem that I think is popping up everywhere, particularly with regard to electricians, that takes the capitalization to about 68% debt, 32% equity. So we're effectively putting more equity into the project. We're getting a return on that equity because it's 500 million more of rent over the lease period. But I think what I'm trying to point out is I'm going the other way. Most of my peers are taking leverage wherever they can and levering up. That's not what I want to do long term. I want to have a healthy equity layer. I want to have robust cash flows, I want to have robust deleveraging and I ultimately want to be investment grade at my project companies and at the parent. And so that's where I'm driving long term, which I think is in stark contrast to a lot of the private data center developers in particular, but then also a lot of my peers that are doing things that are 95, you know, 100% loan, loan to cost. So I'm actually again just going the other way where we're over equitizing on purpose.
- [54:38]
C
Patrick, last question. We'll get you out of here. I want to just briefly touch on convertible notes. I believe that you all have notes, Note tranches maturing 2030, 2031 and 2032 equity. Wolf's equity has re rated this year. So how are you thinking about addressing or those convertible notes repurchasing, refinancing industry debt or just letting the converts run? Like how do you look at those when you approach thinking about managing the capital stack?
- [55:06]
B
Sure, that's a great, great question. So generally speaking, I'm trying to get the company to be four to five times levered on a consolidated basis. Okay, so if I. And I'll come back to the converts in a second. If you take Wolf Compute, wolf compute has 3.2 billion of debt, 600 million of NOI. So like right out of the gates, it's five and a quarter times levered, but it amortizes down by a little over 300 million a year. So it's effectively half a turn of amortization each year. Those bonds are callable next October and then they step down to like a 101 call, I think in October of 28. So my intent ultimately would be to, you know, refinance those bonds into a more, you know, longer term, lower cost of capital bond. If I'm investment grade, then great. If not, I'll stick with high yield. But if you take wolf compute of 600 million of NOI and then what, you know, we just signed at Hawesville with anthropic. If you take that, it's, you know, another, I think on average, I think about 800 million of NOI over the life of that, that of that lease total parent NOI, you know, call it billion four. And then you take out some SGNA, right, 100 million of SGNA. And I'm at, you know, again, round numbers just make math easy. Let's say, you know, it's a little over a billion dollars of EBITDA. More like a billion, 2 billion, 3. But let's make it easy and say it's a billion. The way I think about the converts at the parent, again, if I'm trying to get to four to five times, I have two and a half billion principal amount of converts up there, my view is those will I will always look to refinance the underlying principal amount of those. But the in the money portion, that's the portion that I may or may not let convert into equity. And so my first convert, the 2000 and 30s, I think actually becomes callable end of next year. So I've got, you know, a bit more flexibility around what I want to do with that. But I'm not in a rush to take care of those. I think, you know, it's a cheap cost of financing for us. I do again expect I will have two and a half billion. So again, when I think of four to five times leverage, if it's a billion of ebitda, two and a half turns of leverage is just through the principal amount of those converts. And then I've got, you know, another, call it two, $3 billion, you know, of, of debt capacity up there where I can ultimately move like once wolf compute is up and has delevered over a year or two, then I can move that capacity, if I wanted to up to the parent. And that's where I'm kind of targeting that like four to five times consolidated leverage on a stabilized basis. So again, I think the converts, I view it as two and a half billion. That'll kind of always be there as principal. And then the in the money portion is the piece that I'LL try to figure out like okay, am I going to repay that with cash, am I going to flush them, am I going to let them convert into equity? Like what am I going to do? But not in a rush to figure that out. That's a high class problem.
- [58:20]
C
Patrick Fleury, CFO Terwolf, thank you so much for joining man. Congrats on the quarter and we'll be keeping ourselves a prize of updates throughout the year, specifically looking at Morgantown. You really sold that one on your answer there. So thanks for joining man.
- [58:33]
B
Appreciate you guys having me, thanks.
- [58:34]
A
Thank you, Patrick. All right, we're going to keep on going again. It's earnings week so it's data centers. Data centers. Data centers. Which means that SpaceX, which is now a data center company, is on our radar. So that's what we'll close out with. But before we move move on to SpaceX, a word from our sponsor Lygos.
- [59:00]
C
If you're using a Bitcoin backed lender, there should be one question you're always asking yourself and if you're not satisfied with the answer, you should look elsewhere. Is your Bitcoin safe? Unfortunately, Bitcoin's annals are rife with mismanagement of other people's funds and this is not except not an exception with the Bitcoin backed lenders. Don't become the next Celsius or Blockfi victim. If you're using a Bitcoin backed lending service or looking into one, look no further than Lygos Finance. They are our preferred Bitcoin backed lender here at Blockspace because they use Bitcoin native smart contracts to make sure that you self custody your keys when using their Bitcoin collateralized lending services. That's right, with Lygos Finance you never relinquish control of your keys. With Lygos there's no rehypothecation, no wrapping and no bridging. Just competitive rates as low as 8.5 to 10% APR. Go to Lygos Finance to learn more.
- [59:55]
A
All right. SpaceX. SpaceX earnings yesterday. Their first earnings as they went public and their earnings were decent but the stock was decent. Stock is down so initially rose intraday but it fell about 8 to 10% after hours I believe. Currently it's sitting at down 9% on the day. But some very interesting announcements specifically regarding a Nvidia exclusivity and their satellite we reported on our website. You can see here SpaceX Max maps path forward towards 10 gigawatts of compute after signing 14.1 billion of cloud contracts. So let me go over their earnings first and then I'll get into the data center side, specifically the fun stuff, which is compute in space. So their revenue hit 7.81 billion, which is up 92% year on year from 4.1 billion last year, beating estimates which were at 6.8 to 6.93 billion. That's quite, that's quite a beat of estimates. Net loss narrowed to 541 million from 1 to 1.01 billion year over year, which is about negative 9% per share loss versus estimates around 0.23 and 0.$26 rather 23 to 26 cents per share loss, also beating estimates considerably. Where, where's the revenue? Well come from three sectors. One, the connectivity. Starlink, of which it was 4.29 billion, up 66 year on year. Operating income of 1.66 billion or plus 79 from last year with about 38 to 39 margins. And here's the kicker. I think people aren't like looking at this chart enough. The subscribers doubled year on year to 12 million. And they're adding 1.7 million Starlink subscribers net each quarter actually, because I have this pulled up. Here's the chart. Look at this chart, Colin. I mean down from, you know, thousands to 2 million in August 23rd to now 12 million. This is a parabolic advance and with massive growth happening in just the pasture alone. So Starlink.
- [62:28]
C
Yeah, that's an impressive chart. I can remember in 2020 when the, when Starlink was being hyped up.
- [62:35]
A
Yeah. And I mean, have you been on a United plane and used this? Oh, it's.
- [62:39]
C
Yeah, it's incredible. It's a game changer.
- [62:42]
A
It's really difficult to like not want, you know, I'd take delays as I get if I could just get that start like. Okay, so back, let's get back to the numbers. So the connectivity was the first sector, then the, the big one, which is the one we focus on, which is AI, which includes the Xai Grok Cloud Compute Services. Revenues 2.65 2.56 billion, which is up 247% year on year. Operating losses narrowed to down 1.6 1.26 billion. And it's driven by the deals we've covered, mostly new cloud services agreements totaling 14.1 billion in contracted sales of their third line, which is the smallest, which is kind of funny to me, Colin. The smallest part of SpaceX is their space line, which is their Starship line. Revenue not Quite a billion 962 million up 29 year over year, still at an operating loss of 542 million. A lot of that apparently is spent on heavy research and development on starship.
- [63:52]
C
Yeah, it's. It's spent on sending in test, test rockets up into space and then maybe blowing up. Or not. And that's not a dig.
- [63:59]
A
Right.
- [64:00]
C
You always see these videos of like a starship blowing up and then all the Elon Musk haters are like, see bro?
- [64:05]
A
Oh, the tech doesn't even work. Yeah. But yeah, this is really where I did. I have, I have trouble going. Get your head out of the gutter, guys. Are you all of a sudden ruthless capitalists who are like, doesn't make sense because you lost money shooting literal space, you know, stuff into space. Anyway, okay, so. And then we'll, we'll do this before we get. And I'll wrap up this earnings before we get into the fun stuff. So cash that ended with 100 bucks. Billion.
- [64:32]
C
That's crazy.
- [64:33]
A
Billion with a B. Which is mainly the IPO proceeds plus the bond issuance. They have a backlog of about 47.5 billion. And they have the cursor acquisition agreement for 60 billion. Still crazy to see that for AI
- [64:50]
C
that was all stock, wasn't it?
- [64:51]
A
Yeah, this is all stock deal. So not. Not. I don't think. Well if it was majority stock at the very least, if not all. And of course they do have some bitcoin that's down. Just you know what the proportional amount of bitcoin. So fair value hit. So the big one, this is probably why. This is one of the reasons why the stock is probably down is their capex jumped six times to 18.737 billion in Q2, of which 15.83 billion was AI not surprised there as every single large company with cash on hand is just. Is yeeting it into capex. So what's the lookout, what's the, what's the look for Data centers? Well, SpaceX expects 2 gigawatts of compute online by the end of 2026. Let me say that again. 2 gigawatts of compute near term, end of this year. What other companies have 2 gigawatts of computer in total by the end of the year? Very few.
- [65:55]
C
Yeah, I mean what other company can scale up a gigawatt campus in half a year like they did with Colossus? Now I don't think the first phase wasn't a full gigawatt, but they got that plant online.
- [66:06]
A
Yeah.
- [66:06]
C
It was like so fast.
- [66:08]
A
Yeah. And so there, that's Colossus 1. They're rapidly expanding Colossus 2. And Elon and gang say they are scaling towards 10 gigawatts by the end of 2027. 7. So just so out of. I mean that will put them more. It's just, it's hard to imagine, it's like hard to like say that they're a rocket ship company that will be the size of multiple of the largest data center companies.
- [66:36]
C
I would love to see that mapped out. Like if you had the 10 gigawatt, just like take the 10 gigawatt target for 2027. Where does that stack up against Amazon, Google, Meta, Microsoft.
- [66:49]
A
Right.
- [66:49]
C
In terms of compute under. In terms of power and compute under. Under management.
- [66:56]
A
So yeah, yeah. And then, so then they've contracted an additional 6.7 billion of cloud services revenue ramping from October. And that's just the Google and Anthropic deals alone could throw off $26 billion per year. Okay, so that's the earnings.
- [67:21]
C
Let's get you just glossed over that though. 26 billion a year from two companies.
- [67:25]
A
Yeah, I glossed over because.
- [67:27]
C
Crazy.
- [67:28]
A
There's a lot in here. It's like really difficult.
- [67:31]
C
This is a lot to unpack. This is a Leviathan for sure.
- [67:34]
A
Yeah, it's really difficult to, to like not be kind of a cheerleader for Elon here. I see why I feel like I'm catching the bug because he's like, this is just one of his companies and he's executing these wild things, one of which. So we're going to ease into it. Big takeaway is that Elon SpaceX is committed to, quote, using Nvidia GPUs exclusively because, as Elon tweeted, quote, because they are the best. He's not wrong. So that's the first part. And they'll be shooting these Nvidia GPUs up into space. They released the new renderings of the StarMind AI1 satellites. These are the star the satellites which they anticipate putting up in constellation form as orbital data center platforms. These will include the Nvidia Rubin GPUs and Vera CPUs in a kind of condensed Vera Rubin NVR 72 rack stack. So the Vera Rubin, which comes as a 72 GPU rack, which we've talked about previously on with Luxor, they're basically making a more compact space version of one of these racks to fit on one of these satellites. And it apparently supposedly is a little bit custom built to accommodate heat dissipation. And more being a little more robust, et cetera. So Elon, of course doesn't stop there because this same type of Starmind V1 satellite design, the same kind of compact NVR 72 like rack stack won't just be deployed in space. That'll be deployed on their terrestrial data centers because apparently they come with increased data efficiency, data center efficiency. At Satira, They are planning launches for these StarMind A1 satellites in 2027. Their FCC filings referenced up to 1 million satellites for orbital compute. That would be in 2027.
- [70:08]
C
Wait, 1 million in 2027? There's no way.
- [70:12]
A
Yeah, perhaps, perhaps that is a little bit.
- [70:15]
C
That has to be like a multi
- [70:17]
A
target to 1 million. So maybe they're kind of covering.
- [70:20]
C
Okay, I see.
- [70:20]
A
Yeah, yeah.
- [70:21]
C
I mean that would be an incredible output, right? Especially for something that's not proven yet.
- [70:28]
A
Yeah, well, you say that uh.
- [70:31]
C
Oh, is there already one of these floating around?
- [70:34]
A
So here's the Nvidia site and they say the final frontier of space computing has arrived. And they basically give a little insight into, you know, this particular class of Vera Rubin like terrestrial CPU or terrestrial compute stack. But a lot of people may have not been paying attention. Nvidia has actually launched earlier versions of compute satellites into space. Last year they did little test runs with some other smaller companies. They did an H100 build back in 2025, launched on a SpaceX Falcon 9. And apparently that test worked, worked pretty well
- [71:33]
C
then.
- [71:36]
A
These are, those are slightly different modules, the Jetson Orin modules. And apparently those are starting to implement the optical data relay satellites that we kind of covered a few weeks ago during the SpaceX heyday. And this is not just SpaceX. There are other partners who Nvidia is working with, but Nvidia has a few of these, I don't say prototypes, they're just proof of concept like in space. And they do in fact work. So at scale with what I think the Star Mine satellite says it will be some 30 to 35 times more powerful per, you know, in agri compute over a stack of H1 hundreds. This would be pretty significant. The. Yeah, I just want to maybe like review like the stats of the satellite for the average person because this is. You have these, what, gigawatt scale terrestrial data centers that Elon Musk is supposedly building. But these satellites are roughly 120 kilowatts each and they're arranged in the constellation almost like a grid in the sky in low earth orbit. And they just launch teeming hordes of these up and they just sit in kind of static orbit around, around the Earth and perform compute workloads up there with optical communication between each other and then. Optical communication, oh, don't quote me on that. Not some kind of communication between satellite and the Earth.
- [73:27]
C
That's my question is what will these be used for? Will these be for training or would they be for inference or both? Is there not a latency problem? Because we were told for years that you have to have an inference data center close to a Metro hub or else the latency is too great and it's not the performance degrades, you know.
- [73:45]
A
That's a good question. I believe I was looking at low Earth orbit latency numbers and they were like 30 to 60 milliseconds, which is fine if you're gonna play, you know, I'll take that, that ping if I'm gonna play a tactical shooter. So that doesn't really worry. I'm not too worried about that.
- [74:05]
C
Kind of sounds like a question like, you know, like a midwit on Reddit would ask. It's like these data centers in space can't even work. So you often see these takes. It's like as though the company investing billions of dollars into this endeavor has not thought about all the problems through.
- [74:24]
A
Yeah, so for me it's unclear. I wouldn't be surprised if it's some kind of hybrid. But I mean let's just do a little bit of math here. Up to a million satellites in a mega constellation as they call it, at 120 kilowatts per satellite. How much is that?
- [74:43]
C
It's a gigawatt
- [74:46]
A
actually. I got to do some math here.
- [74:49]
C
Yeah, it should be like 1.2, right?
- [74:51]
A
Yeah, like 1.2 gigawatts of, of satellites up in space. Solar powered. Exclusively solar powered.
- [74:59]
C
But if you think about that, that's not that much. Do you know what I mean? Like, if you actually think about like, okay, you need a million of these to equate to 1.2 gigawatts. What change in form factor do we need to have to where these are like a megawatt in size or something like that.
- [75:17]
B
Right.
- [75:17]
C
In terms of like actually because 1.2 gigawatts is, is nothing at this point. I mean obviously that's a lot of power. Obviously it's a lot of power. It's enough to, to power a mid sized city in the US but in the grand scheme of things that's not going to be that much for how much material is needed to get these off the Ground I'd be really curious about the total capex cost for getting one of these up and running because your OPEX cost is basically the repairs for them.
- [75:46]
B
Right.
- [75:46]
C
Like there's not, you're not paying for power up there, so that's going to be a huge boon. But like how much does it cost to actually build one of these satellites and then throw it into space? Also a million. How many rocket launches we talking about here? Yeah, obviously that's like the extreme end, probably over multiple years, if not a decade plus. But even so, look, if you look
- [76:10]
A
at SpaceX's like cost per unit of payload that it takes them to push stuff up into the stars, into low earth orbit, it's plummeting. So it gets lower every year. And with their new rockets they get, you know, incredibly much more efficient. They already send satellites up into space. They're the primary satellite, you know, launch company in the world. And so I mean, you know, eventually the cost and benefit curves do, will
- [76:45]
C
probably intersect at some point.
- [76:47]
A
Yeah, and my thought is like Colin, if, if they are actually able to get to a million satellites in space on whatever timeline, whether that's next year or sometime next several years, that to me is a very successful, not just proof of concept, but gigawatt in space of compute, I feel like that quickly becomes 10 to more than 10 to dozens of gigawatts of compute because Elon's looking very, very sharp. And I brought this up months ago when we first started covering this topic. Given the rise in anti data center sentiment in the United States and having power, water, you know, noise, whatever, all the issues, the, this idea that it's always sunny in space may give way to the fact that you have no NIMBYS space short of maybe geopolitics.
- [77:48]
C
Yeah, I was thinking about this. Like what is the regulatory and approval process for this?
- [77:54]
A
Right.
- [77:54]
C
I mean SpaceX is kind of, I mean grandfather in is not the right clause but like they are one of the US government's principal contractors. Like they're the most important contractor for the final frontier. And so how, what do you have to go through in terms of rigmarole to get actual approval to do something like this is like you said, there's no NIMBY ism in space. But I will say I would actually take the other side of this from the sense of like, I think that Elon Musk has become so politically radioactive for so many people that anything that he did, people would find an angle to scrutinize it and they may, they won't be talking about, you know, okay, rolling out data centers and sucking up resources on earth, but they will clutch pearls over the sanctity of space and how we can't have all of this space junk in there if people are not already freaking out about this already. I think most people don't even know that this is happening. And the ones that do, they either are gigaboolish SpaceX because of it, or they think it's a pipe dream that will never come to fruition.
- [78:56]
A
Yeah, there's nobody who knows about this side of Space X. Pretty much every single person thinks that SpaceX is a rocket company. There are rocket, there are. They're a rocket company insofar as they transport their primary business into space, which is phase, which is either just heavy load or just other fee paying customers.
- [79:16]
C
They're a data center company. The rockets are just the cherry on top. Yeah, the rockets are the seed that will eventually grow into something that is more fruitful than anything else if, if SpaceX can actually lick interplanetary travel.
- [79:29]
A
The rockets are the trucks that hold haul all the GPUs and the conduit and the containers and the building materials out to the data center site. That's like, really, it's like calling, you know, the truck company that services a data center build out the primary business model.
- [79:46]
B
So
- [79:48]
A
that is all I've got. Again, disclaimer. We're not telling you to buy any specific stock. We're talking enthusiastically about some fun trends happening in the space. This is earnings week, so if you haven't been paying attention, this is day three of two weeks of earnings here on Block Space Live. Tomorrow we've got Clean Spark and Mara on Friday we have Core Scientific. Asher Ganute from Hud8 joins the show on Friday to talk through this. Make sure to like and subscribe. We go live every day at 1pm Eastern. We are COMPUTE's daily live show featuring quick hits on the latest in AI data centers, markets and emerging technology. If you like what you hear, you'll love all the content on our website at Blockspace Media website Blockspace Media. This show is brought to you by CleanSpark. CleanSpark is NASDAQ listed ticker CLSK. I'm Charlie.
- [80:50]
C
I'm Colin. And one quick correction. Corsi was on July 28th. So we will not be covering Corsi on Friday. Yes, we will be covering Mara and Queen Sparks earnings on Friday morning because they come out Thursday evening. So tomorrow we won't have any specific earnings to cover but we will be back at it on Friday with Queen Spark and Mera. And with that, I hope you all have a beautiful Wednesday.
- [81:16]
A
See you. See you tomorrow.