Transcript 0:00 Ladies and gentlemen, there is blood in the streets. Ooh. That's right. That's right, Charlie. AI stocks are getting absolutely hammered today, and many of our favorite names are actually deeply in the red year to date. 0:16 You're listening to Blockspace Live, and if you're wondering, "Is it over?" or "Should I mortgage my house and sell my kids' tablets to buy the dip?" Well, we can't exactly answer that question for you. 0:28 But for our top story today, we will be taking a look at the market route and doing some analysis on what could be causing it. Hint, it's the credit, baby. 0:38 After that, we have Ligos Finance CEO Jay Patel on to talk about the current market route and whether or not he thinks the jitters that we are seeing among investors as to being into question the economics of bankrolling these sites is justified, or if the whole thing is rolling over. 0:54 After that, we're gonna be touching on Iren proposing a second site in Australia, an 800 megawatt site near Wagga Wagga. New South Wales, Charlie. 1:04 And then we will cap off the show with Bit Farms, excuse me, Keel, formerly Bit Farms, eyeing an 80 megawatt site in Pennsylvania. That's right. 1:15 You're listening to Blockspace Live, co-hosted by Charlie Spears and Colin Harper of Blockspace Media. We are Compute's daily live show, going live at 1:00 PM Eastern every single weekday. 1:27 We cover the latest in data centers, AI, emerging tech, and markets. 1:33 If the live stream is not enough for you, don't worry, there's more content on our website, blockspace.media, including a newsletter in your inbox every day that we can write it. Today's kind of a doozy, though. 1:46 The market's not looking too pretty. We have our tweet of the day lined up, Colin. It's just gonna blend into us commiserating about the market. I see in our chat right now- The troll box popping off today. 2:00 Really appreciate y'all- Yeah... you for all of the comments. This one's my favorite. [laughs] The punishment doesn't stop buying the dip. Man, it's getting bloody out here. It was bloody two days ago. 2:12 It's even bloodier now. It's a full-on execution in the street. [laughs] That's- We're all g- we're all gonna hold hands and sing Kumbaya together. Yeah. 2:21 And the reason we need to do that, for anyone who has not been paying attention, data center sector and the AI industry at large is just getting hammered. NVIDIA's down, NeoClouds are down, PowerShells are down. 2:34 Um, your, your parents' portfolio is now in the red. Gold's down, Bitcoin's down. Freaking things down. Everything is down. Absolutely. Is any, is anything up today? I don't think anything's up today. 2:44 Uh, brass and canned food. Okay. [laughs] Yeah. G- yeah. Bullets. [laughs] Bullets and beans. 2:52 But anyway, coming from Walter Bloomberg here, quote, "Shares of AI infrastructure firms CoreWeave and Nebius drop, steepest decliners on Nasdaq 100 index." 3:03 So Nebius and CoreWeave, the darlings of the NeoCloud sector, are currently the worst performing stocks today on the Nasdaq 100. But it's not just Nebius and CoreWeave. 3:18 This is, as we said, a market-wide route, and we're gonna hop into the details, the numbers, and what could be causing it. But first, a quick word from our sponsor. 3:29 [upbeat music] Soluna Holdings, Nasdaq listed ticker SLNH, develops data centers co-located with wind and solar plants, turning surplus renewable energy into computing power for AI and other intensive applications. 3:48 We are proud to partner with Soluna Holdings here at Blockspace. If you would like to learn more, please visit solunacomputing.com. That is solunacomputing.com. 4:04 The AI stock market is crashing, and a single company could be to blame. 4:09 Investors are starting to worry about private credit risk for the broader data center stock sector after one of CoreWeave's bonds hit an all-time high 15% yesterday. 4:20 The contagion has plagued every sector of the AI industry, so much so that a number of our favorite names are at or below the prices they fetched at the beginning of the year. 4:32 To list just a few of those names, Iren is down nine... Excuse me. Iren is down 10% today, 26% year to date, and 55% year over year. 4:43 Cipher Digital is down nearly 11% today, 36 and a half percent year to date, and 51% year over year. Applied Digital is down nearly 10% today, 27% year to date, and 41% year over year. 4:58 And CoreWeave is down a staggering 22% over the last five days, 15% today, 17% year to date, and 51% year over year. Now, the market route has bruised these names the most, but I don't want to single them out. 5:16 Once again, every single name that we cover is in the red today, and only Nebius and HUT8 are both in the green year to date and year over year. Charlie, I gotta say, this sell-off kinda makes July's look like a warm-up. 5:32 What do you think? Are we seeing the AI trade unwind for real this time? Are you a true believer? Do you believe in the AI trade? I'll put it this way. 5:41 Like, if CoreWeave goes down, if Nebius goes down, if this is a full unwind of all of these names, this has big implications for the entire AI sector, which, by the way, may I remind you, is the entire US economy right now. 5:57 So, uh, not to be too, uh, doomsayer, but if all of these unwind, if it truly is joever- Then, uh, somebody's gotta step in. 6:08 I think maybe an NVIDIA steps in, or maybe just the market shakes out some weak, uh, financing deals. But that's, that's mine. Maybe, maybe I'm blind. 6:18 I'm always, I'm always way too bullish at the top, so I'll throw it back to you. Well, this is going back to what you just said about it being the entire US economy. 6:27 This is why there are murmurings that AI will eventually get a bailout if things do turn south. So don't worry, dear listener, if you're worried about your bags, Uncle Sam will help you. 6:39 Or if we look to 2008- Or, or Father Jensen. Yeah. Father Jensen will come in and save you. Either the Uncle Sam will help you or he'll help the banks underwriting all of this. 6:48 But on that note, many investors are probably wondering why the unwind is happening now, and the simplest answer that we can find is financing risk. 6:57 For instance, CoreWeave's notes, uh, 9.6% notes due 2032 hit an all-time high yield in the secondary market yesterday, and they are climbing again today. 7:08 Today, they rose above 15% for a 1,050 basis point spread over benchmark Treasury yields. 7:16 Although CoreWeave isn't alone here, Oracle and Meta are also seeing spreads for certain bonds blow out, and the panic in the data center bond market has bled through into the stock market for these companies. 7:27 Ultimately, the market is repricing these names as borrowing costs rise, and investors question whether the next phase of the AI CapEx cycle will be in- requiring more equity from these companies to bankroll them, or if they will even get close to the equally, the similarly attractive terms we saw in the first phase of the AI CapEx cycle. 7:47 Andrew Aziz lays out the logic pretty well in a recent post. Quote here, "AI infrastructure sector is getting hit hard, and debt is a big part of it. 7:57 CoreWeave has $35 billion in debt, interest expense surging, still free cash flow negative with $35 to $39 billion in CapEx plans. Oracle debt roughly doubled. Large off-balance sheet leases. 8:09 Credit default swap spreads widening. 8:12 Add rising yields, hundreds of billions in AI-related debt issuance this year, the OpenAI revenue clarification, and the market is repricing the risk that these cert- that these companies keep borrowing while returns stay uncertain. 8:24 Demand is still there. The financing math is the question." And when we take a look at that financing math, we can see this pressure building as lending tightens. 8:35 Financial Times reported that AI-related debt issuance fell to $23 billion in September, less than half the August volume, according to Morgan Stanley data. 8:46 Now, to be clear, this doesn't mean that we're facing a credit crisis. For instance, Blue Owl back in July argued that direct lending credit quality remains comparatively resilient, 9:00 reporting approximately 2% default rates in its mid-year commentary, below its cited long-term average of 2.7%. 9:08 But this could mean that AI infrastructure investors are becoming increasingly worried and pricing in the risk of a funding squeeze before there is clear evidence of a broad default cycle. 9:20 So Charlie, I think that kind of sums it up pretty well. 9:23 Obviously, we're not seeing waves of default yet, but a few project delays or snags could be a spark that sets, if not the whole thing ablaze, then a number of companies ablaze. Yeah. 9:36 Wouldn't be surprised to see, um, some, some heads roll or some... Or, you know, a few companies fall or a few strategies which had worked out very well for some companies, uh, fail to keep working out. 9:49 Maybe the debt starts repricing at a much higher rate, because that's probably more expensive to fund this build-out than a lot of people anticipated. Like, it's not just free money if you're exposed to AI. 10:01 There's gonna be some cost to it somewhere along the way. Maybe that's what this is. 10:05 But this really is kind of like an American, American economy story here, and it's surprising to me that the average person on the street doesn't really know the term Neoclouds, or may not know the company CoreWeave, because, like, that is what their, that's what their re- their retirement portfolio is basically, uh, downstream of right now. 10:26 So I, I, you know, I would say there is a lot more on the line here than just a specific, uh, market sector of just, like, no more narrowed new Neocloud da- data center companies. 10:41 Yeah, and when we look at the, the credit market specifically, going back to some of the risks for the broader economy, I think most people don't realize that pensions and insurance companies are starting to hold some of this debt. 10:54 Um, this is coming from Financial Times about CoreWeave. Financial Times always trying to... This was published yesterday. They're always trying to punch at all of the different companies in this sector. 11:04 But they say here, quote, "CoreWeave is chasing insurers," quote, "unlimited," end quote, "pools of capital, as the NVIDIA-backed AI cloud provider seeks cheaper ways to fund tens of billions of dollars it is spending on chips and other AI infrastructure." 11:18 CoreWeave is going to have a tough time tapping into some of those pools of capital because typically insurers, I mean, after all, they are insuring a number of things from homes, to life, to auto, to, uh, to healthcare. 11:32 Uh, they don't wanna hold CoreWeave debt unless it has some sort of investment, uh, an, uh, a high investment grade credit backing, right? 11:42 So we've seen insurers, um, in some of these capital pools lending to the hyperscalers because that's high investment grade credit. But the Neoclouds themselves have been boxed out from this. 11:52 CoreWeave has a pretty high cost of capital as a result, um, meaning, you know, it's, it's tapping for GPU financing and things that typically are less secured than, or, um, the sec- the security behind the loan is not as robust as, say, a data center, 'cause the GPUs do appreciate, 12:08 um, depreciate. But- 12:11 Ultimately, the, the capital that is being pushed into the AI CapEx boom is coming from a bunch of different sources, and within the private credit markets, you know, that makes its way back to some of the industries and some of the, you know, se- uh, the, some, some of the industries that end up, you know, underpinning people's pensions and also their insurers as well. 12:33 The last thing I wanna say about this is that 12:36 going back to the CoreWeave story specifically because, you know, everyone's pointing their finger at CoreWeave and also the Oracle bonds that are tied to Project Jupiter, which has been delayed. 12:46 This is not [clears throat]... This is, it, it at least the credit risk right now, I would say is not evenly distributed and it's not systemic. 12:55 Chamath has a pretty good thread on this where he does a deep dive into pri- the build, uh, uh, the deep dive into the AI build-out specifically, the, um, whether or not it kind of points to a private credit bubble, and he has an important note here 13:13 where he talks about, uh, the market is al- also already distinguishing among these deals. Apollo bought debt financing chips leased to xAI at 99 cents in December 2025. After SpaceX bought xAI, it traded as high as 106. 13:29 Project Jupiter, an $18 billion AI data center financing later traded at 89 to 91 cents as investors reassessed the project's credit. 13:38 So that's the one thing that I would look at for broader risk here is if you start seeing these bonds kind of trade more in tandem with each other. 13:45 But clearly investors and bond traders are looking at certain projects and certain companies and saying, "You have your shit together a little bit better than some of these other companies." 13:55 And one last thing that he notes here that I thought was pretty interesting that we'll kinda touch on with Jay here. The structure of some AI financing adds another layer to the picture. 14:05 CoreWeave borrowed for five years against customer contracts averaging about three, while Meta's Hyperion bonds run to 2049 against leases renewed every four years. That's the other piece of this is, 14:17 you know, you can underwrite that credit risk based on the terms of one contract, but do you really know how to evaluate changes to GPU hour pricing, to take or pay contracts five, 10, 15 years out? 14:31 You're juggling so many variables with that. GPU depreciation, advancements in AI, open weight models potentially undercutting token prices for the leading AI labs. So I think another part of this 14:44 even has, uh, not necessarily only to do with the credit worthiness of the companies, but just fundamental misunderstanding about what the next 10 years of AI will look like. 14:54 What the, what the business model will ultimately look like after the first phase of the CapEx cycle is done. How much companies can reliably expect to earn for their services versus what they are right now. 15:06 And who better to help handhold us through what's actually going on in the markets rather than us throwing darts and speculating than Jay Patel, CEO of Lagos Finance, who's in the wings. 15:18 We'll bring on up here and get another hit on this market story developing right now after a word from our sponsor, Luxor. 15:27 [fire sizzles] This episode of Block Space Live is brought to you by Luxor's Commander, Bitcoin miner management software for enterprise operations. 15:40 Luxor's Commander gives you real-time fleet monitoring, bulk remote commands across your fleet, and Intelligent Miner. 15:46 That's an automated profitability engine that runs every five minutes and adjusts your fleet's power settings to live energy and hash rate markets. 15:53 In fact, ERCOT back tests show 10% improved profitability with intelligent mining versus old-fashioned binary mining. 16:00 Commander Pro is roughly half the cost of competition, $100 per megawatt or a 25 basis point pool fee adder. But you can also try it free for 60 days. 16:08 So if you'd like to learn more, go to luxor.tech/commander to get started. Let's bring Jay up here. Jay Patel, CEO of Lagos Finance, welcome back to the show. How you guys doing? Great. 16:23 Oh, you know, just looking at the sea of red. Uh, don't wanna check my Fidelity account. Just normal things. A little fear and loathing in the AI CapEx cycle. Um, so we just spent that entire segment. 16:37 I think you, you got, you got to hear most of it, Jay. 16:39 So I'm really curious to hear your thoughts on the ripples we're seeing in the market as it relates to ripples that we're seeing in the bond market for some of these names, not all of them. 16:49 So to kick it off, what are your first takes on the route that we're seeing? I, I'm skeptical. Like, I don't think it's a full 16:58 repricing of the entire market, and I, I d- I don't think it's gonna be, like, a broader wildfire per se. 17:05 Um, but I think it links back to what we talked about last week, which is just there's a segment of the market that has backing from the hyperscalers and, you know, IG backers that I think will do well. 17:15 And there's another segment which doesn't, and I think is gonna struggle. Um, uh, you know, I, I actually think, like, you know, everyone threw in CoreWeave, Oracle, and Meta. 17:25 The Meta bonds are a little different in that this w- was, you know, the Hyperion data center was, like, some off sheet, off balance sheet financing. It wasn't direct recourse to the Meta parent balance sheet. 17:34 There's, like, the renewable four-year leases. So I wouldn't classify that in the same category as, like, debt issued by, you know, Meta, the parent, or Alphabet or the other hyperscalers. 17:45 But the CoreWeave and Oracle debt, I think it, it does get back to that point that you guys pointed out, which is just a lot of this debt, the maturity on the debt is longer than the commitments that they have from customers. 17:55 And so now you're making a bet that they'll be able to keep finding someone to offtake on the other side. 18:00 Um, and I think specifically in CoreWeave's case, you know, they've made this argument that they're gonna bring in unlimited capital from insurers. 18:09 I'm a little skeptical of that argument just because of some of the stuff we've already seen over the past couple months. 18:15 Um, but I do think this, you know, there's a chance this is just isolated to- You know, kind of NeoClouds and, and, you know, uh, really like the part of the market that's very debt-fueled that has uncertain cash flows on the other side. 18:32 So, uh, let's talk about the CoreWeave bit really quickly as they try to tap insurers. 18:37 I mean, as I understand it, insurers, like, are not going to take on credit that's not investment grade because there's real risk, and then they could not be able to... 18:45 Uh, if they lose that investment, it- they risk not being able to meet their obligations for the people holding their plans. 18:50 So you're skeptical that CoreWeave will be able to do that as the only path for them, basically an NVIDIA backstop. 18:57 Or is there a world in which, um, CoreWeave just continues to, uh, surprise to the upside, although it's not really doing that right now, and they eventually get that increscent- investment grade rating? 19:09 Like, how, how can they do this without NVIDIA? Yeah. 19:12 So I, I, I think it's actually worth stepping back and thinking about, like, what, what is the difference between a data center that is backed by hyperscalers or funded by hyperscalers and one that isn't? 19:23 Um, I think if you are... You know, if you look at, like, the 2050 Alphabet bonds and, you know, similar bonds issued by, um, uh, 19:33 Microsoft and Meta, like they haven't significantly widened in terms of spreads because I think, you know, investors have an understanding that regardless of what happens over the next five years, there's durable businesses that are gonna be able to pay them back in the long term. 19:46 The problem with lending specifically to someone like CoreWeave is the only way they can monetize these data centers is by having someone take the capacity on the other side and pay for, you know, the, you know, compute and power. 19:59 Um, there is no ability like with Meta, where even in a worst-case scenario, I don't know, if Meta struggles to get people to pay for the Muse API, they can definitely monetize this compute capacity by serving better ads through Instagram and Facebook and, you know, serving people more slop and monetizing via, you know, pretty durable businesses they already have. 20:19 Um, my, yeah, my concern with CoreWeave is really that, you know, they might be able to surprise for the upside, but the other thing that you gotta think about is there's a few IG offtakers. 20:29 There's not h- the universe of IG offtakers that's looking to invest, you know, tens of billions of dollars in compute. 20:36 Um, so I guess the other bet that they're making on this insurer angle is they either they themselves get an IG rating or OpenAI and Anthropic do, um, and then, you know, they can say that, "Hey, our, you know, customers that we're offloading to are, are IG offtakers." 20:52 I'm pretty sure OpenAI is a far ways away from that, so I don't think insurers are gonna buy that. Um, and then my- Seems like all three of them are pretty far away, right? I mean, unless I'm missing something. 21:01 They're all, they're all quite far, and even if you somehow miraculously convince the rating agencies to, you know, get you a better rating, the other question that I always have in the back of my head is if you're lending to one of these, you know, kind of NeoClouds where the primary customer they have is Anthropic or OpenAI on the other side... 21:21 I know CoreWeave is a little different in that they have Microsoft, although Microsoft didn't renew last year and they had to move that capacity to OpenAI, and they have other customers. 21:28 But if the bet you're making is that OpenAI and Anthropic are going to continue scaling, they're gonna have great business, you know, structure, and they're going to need more and more compute over the next five to seven years, 21:42 why would you want a 10% fixed return or a 12% fixed re- You know? Like, why not just buy equity in SpaceX AI and Anthropic and OpenAI? 21:52 Like, it doesn't really make sense to be debt that's not even at the same, you know, that's not even lending to the frontier lab. You're, like, a step removed. 21:59 Um, and then on the, on the NVIDIA side, I think I was kind of thinking the same thing as you, you know, maybe a few weeks ago in that, okay, it kind of makes sense for NVIDIA to, you know, sweep in and, you know, provide backstops, maybe take stronger equity, take, you know, better economics, things of that sort. 22:16 But the other, I don't know, story over the past week with NVIDIA has really been, like, this rift where they're more and more in support of open models. 22:25 You know, the, um, you know, Jensen has, uh, I don't know if it's publicly, but he's at least, like, indicated that he's not as happy with the direction that OpenAI is going, building their own chips and things of that sort. 22:38 And so do you want to support the companies whose primary customers are the frontier labs if you're Jensen at this point? I mean, I think NVIDIA is probably fine either way. 22:47 Like, they'll have plenty of demand for their GPUs one way or another, and so I don't really know if they wanna even get into this space and muddy the waters. Do you think that, 22:57 that the, the bonds just, just need to yield more? Like, is, is just the low teens on up? Does it just... Th- th- this would be higher for, like, the larger banks to hold these, or, um, is it just need... 23:09 Do they just need to be more expensive? Is that, is it really that simple, uh, or am I missing something? If, if you were looking at it in isolation, like, that's what I was thinking. 23:20 You know, okay, let's say low teens isn't enough. If we get another 3, 4% out of the bonds, does that make sense for the market? 23:27 And maybe it does, but the problem with the business model for a CoreWeave or anyone else in this space is that you can't just stop building. 23:34 Like, the idea is not that, okay, we've got this, you know, portfolio of capacity, and now, you know, you're just investing in our company, and we're really just leasing this out, and that's it. 23:43 Like, the whole idea behind the equity is that they have some sort of structural advantage where they're able to stand up data centers faster than someone else. 23:50 And if they stop doing that, then I think that's, you know, kind of devastating for the equity. But the problem with that is you need debt financing to be able to scale up. 24:00 I guess, you know, Nevius and a few others are the exception, where they're going a different route. And maybe, 24:06 maybe that's some indication that CoreWeave needs to, you know, expand to converts or some other structure that's gonna let them get financing while giving the lender some upside. But yeah. 24:16 It's like if, like, if the build-out slows down or stops, everything, all of the assumptions of, like, growth don't make sense. Um, like if- CoreWeave can't deliver the build-out. 24:30 If some of the other neo-glasner sector can't continue on the, the, uh, can't build out because of financing or credit default, who steps in? Like, who does this? Is this... 24:41 You said you may have thought, like, Nvidia would backstop it, but if it's not Nvidia, who else? Is there some new financial instrument that I'm not aware of that exists out there? Does the government come do this? Yeah. 24:53 Who, who- Yeah, so- Who keeps the build-out going? Yeah. 24:56 So I think there's probably, like, you know, before the government steps in, I think you will see Nvidia at least play a stronger role as it relates to future GPU deliveries, right? 25:05 They're probably not so interested in helping you finance existing data centers or, you know, if you're not a, a, you know, a customer who's going to be paying them over the next couple years. 25:15 Like, you know, Nvidia's not gonna throw you a lifeline for no reason. 25:18 But I think if you are maybe someone, uh, you know, maybe like CoreWeave or others who are going to be taking a lot of deliveries for, you know, newer GPUs, then maybe it makes sense, um, or, or, you know, like, the, the tenants will. 25:31 The u- the other kind of, you know, piece there is maybe the hyperscalers internalize a little more of this. 25:37 Like, I know to some extent, at least with Microsoft, they had this strategy of, like, "Let's let them figure it out, and if it works, then we'll start doing it ourselves," kinda thing. 25:45 So maybe there's a fast track to that. I think this is, Colin, what we were talking about last week, which is, like, the... You know, for most of these companies, the, um, [lip smack] 25:56 the party who's most interested in seeing them succeed is probably the hyperscaler that they're closely aligned with, right? 26:00 And, like, maybe that make- means a bigger equity stake in exchange for, uh, like, a backstop and better financing terms. 26:07 I, I do think if those two don't work out or you see more widespread issues, then you probably do get into the point where, like, I don't think it's gonna be, like, a government bailout, but I'm sure you'll see some loosening of regulations on how you're allowed to package these bonds, who's allowed to lend to them, you know, what kind of transparency you need. 26:25 But at the end of the day, I think all of those things, I think they're still the issue, which is just if your investor's on the other end, I don't... And this is why I'm skeptical of what Corey was saying. 26:35 I don't think the issue is that insurers are calling CoreWeave constantly trying to figure out how they can buy the bonds, and it's just a matter of the rating. 26:42 Like, I think it is actually that, hey, you need to have an attractive enough return or a loan of risk for someone to be worth investing in your debt. 26:50 Um, and I don't think, you know, unless the government is subsidizing or providing some backstop and guarantees, I don't think that's gonna happen. 26:57 I mean, I guess that goes back to, um, what did the OpenAI CFO say last, you know, earlier this year? It feels like a lifetime ago. But, like, in some future, the government might, might need to provide backstops. 27:07 Maybe we're getting closer to that. So on that note, do you, I mean, do you think this panic is overblown, or do you think this is a little bit of a canary in the coal mine for what we might expect in 2027? 27:19 [sighs] You know, I'm... [sighs] My, my immediate reaction is that basically every few months we get a big scare, and then somehow we get over that scare. 27:31 Um, so I don't think this will be, like, the beginning of the end, but I do think you might start to see kind of a bifurcation in the market where, you know, maybe the equities for a lot of these companies don't, 27:46 you know, severely outperform like they did last year. Um, and y- you know, I, I think you might have some more, like, tampered expectations for where things go. 27:56 I don't think it's gonna be like, um, like, uh, uh, the bubble popping or anything of that sort. 28:03 Like, I do think we have a sense in this space that there's, like, overblown hysteria, everything, everyone thinks everything's gonna come crashing and burning down, and then somehow just time resolves all things. 28:13 And so I think we will be fine. 28:15 Um, but I do think that some of the expectations for what people had in terms of where both, you know, like, the common stock for a lot of these companies is going, as well as, like, their ability to continue raising debt, I think those will have to be kind of recalibrated. 28:28 Um, and you know, the other side of things is also it seems, you know, on the model side, like a four-horse race again, you know? I think it wasn't just OpenAI and Anthropic. 28:39 You know, I think Meta is definitely back in the game. Google is back in the game. Uh, Microsoft is training their own models. 28:46 So you might actually see more of this shift back towards the hyperscalers where, you know, I think there was this sense that, okay, OpenAI and Anthropic are the two companies that are best able to monetize Compute, and so they will increasingly become a larger and larger portion of the offtakers. 29:01 I, you know, I don't wanna call it now, but I think maybe we're back in a world where they're actually a minority of offtakers over the next five years, and it's actually, you know, the, the labs within the hyperscalers. 29:14 One last question on this, Jay, before we toss it, uh, a final one to you about what all this means for Bitcoin. It, it definitely makes sense to me that you'll probably see a bifurcation going forward. 29:25 We covered Firmus' IPO last week, and they scrapped it because they were asking for an absurd valuation for basically having, like, not even 100 megawatts online. It was something absurd like that. So I, I, I... 29:38 The optimistic side of me wants to believe that, yeah, you're just starting to see the market get a little pickier, and as we covered in the segment, right, like not, uh, we're not seeing contagion across the entire bond market for these things. 29:50 It's very specific operators that investors are starting to ask questions about. What would you look for 29:56 if you were trying to see whether or not, um, you know, this has graduated from, like, a isolated panic to fears of outright credit risk or default across the sector? 30:10 Yeah, I think what I'd probably be looking for is if this, you know, widening of spreads actually makes its way towards issuances from the hyperscalers, because then there's some concern that... 30:25 You know, I think particularly Meta. Meta's probably most exposed but, you know, um, I think, uh- You know, Microsoft and Alphabet to some extent as well. 30:34 Um, because if there's concern that, hey, not even the hyperscalers are going to be able to properly monetize this CapEx to get us paid back, I think that is a major concern. 30:44 The other is obviously, you know, let's say that we are in this kind of new regime where rates are just higher and, you know, if you're building out infrastructure, you're going to need to pay a higher rate. 30:55 At what point do, you know, planned and announced build-outs get canceled purely because the financing couldn't make sense? You know, there, there's, there... Obviously, CoreWeave would love to borrow at a very low rate. 31:07 But I do think that they make enough money from, you know, the, at, at least the existing build-outs that they've had, have, you know, been quite successful and they're able to sell the capacity. 31:16 Um, so if, if, if they start canceling announced build-outs because, you know, they just can't get a positive ROI, I think that becomes concerning. I think right now we're... 31:26 And, and you guys are probably more familiar with this space than me. I think we're still a ways away from that. 31:31 Like, there's a good enough ROI on building out data centers and compute that, uh, I don't think the financing cost alone is gonna kill the incentive to build things. 31:39 But I do think at some point it becomes, you know, like, like, that's the math they have to do and decide, is it worth borrowing at 14%, 15%? May- maybe, maybe not. 31:50 Maybe people will just have to start paying for ChatGPT finally. Most people don't pay for it. Maybe people are gonna actually have to realize intelligence ain't free, doesn't grow on trees. 32:00 So, so Jay, what does it mean for Bitcoin as the majority of my holdings? What's gonna happen? 32:08 Yeah, I think, um, on the BTC side, as we've al- as I've, you know, kind of said over the past couple of months, I do think regardless of what happens, we probably find ourselves in some scenario over the next couple of years, and maybe sooner rather than later, where the federal government has to come in if there's an air pocket, if there's issues with financing for this build-out. 32:27 As you've mentioned, you know, it's not that this is a part of the economy. Like, this is the economy. If there is no AI CapEx, we are in a recession. Like, this is the majority of the economic growth in the US. 32:38 And so I think for Bitcoin, it's optimistic in the sense that, look, A, there's probably a fairly high chance that sometime over the next couple of years, the government is going to, you know, increase liquidity in the market to keep the, the train going. 32:53 Um, and then B, what I'm starting to think about is basically that for, for equities, right? There's been this competition for capital. 33:00 Um, but, you know, I think equity investors are starting to realize that even though we're in this new world of AI where there's amazing businesses and revenue growth, it's not like software, you know, 2012 to 2023, where the marginal dollar of revenue was basically earnings, right? 33:18 Like, for, specifically for CoreWeave and some of these neo clouds, but also even for the hyperscalers. The incremental dollar of revenue has a lot of cost associated with it. And so... 33:27 A- and, you know, there's no free cash flow for buybacks, right? Like, they're issuing new equity, they're taking on debt. It's like the opposite of what was going on over the past five years. 33:36 And so, you know, maybe in that world, you know, there, the Bitcoin does become more attractive, right? 33:42 Because the equity market's not just returning a bunch of capital to you constantly, it's actually kind of swallowing capital on the other side. 33:50 Yeah, I've been keeping an eye on NASDAQ correlations, at least for near term swings, and it's down like 1.7% today. Bitcoin kind of correlating with the NASDAQ again. 34:01 But hard not to be happy about it being above 80K when we were at 60, you know, for most of the summer. So Jay, thank you so much for your thoughts. As always, man, really appreciate it. Love talking to you. 34:13 And, you know, we'll, we'll see what's going on next week. Maybe we'll have, uh, some better tidings for our listeners to talk about than all of this doom and gloom. Yeah. Thank you, Jay. 34:22 If, if the past is any, any lesson, you know, the, the doom and gloom usually subsides in a few weeks. We'll forget all about it. It will have never happened. We'll brain hole it entirely. Jay, CEO of Legos Finance. 34:33 Go to legos.finance to learn more. Thank you so much for coming on the show. Take care, guys. [clears throat] We are gonna keep on going. We are gonna go talk about Iren, this time striking a deal. 34:49 Colin's gonna lead this one. Take it away, Colin. Iren just announced its second data center on its home turf in Australia. 34:58 The Bitcoin miner turned Neocloud has proposed Redbank, a 900-megawatt data center near Wagga Wagga, New South Wales, potentially expanding its Australia development pipeline substantially. 35:11 But the project is still in the preliminary planning stage, and it is not included in Iren's existing five gigawatts of secured power. But it's estimated... 35:22 And it's estimated to cost $10 billion over six to eight years of construction. Charlie, last week we covered Firmus scrapping its IPO. Looks like Iren smells blood in the water. 35:37 It's trying to eat its lunch with a second data center down under. Yeah, in a place which is totally real called Wagga Wagga, Australia, they're going to build a, a data center, or at least it's proposed. 35:49 Look, um, Firmus obviously, uh, had people who wanted to buy the IPO, but it couldn't cross the threshold. It's dead in the water. Iren's there. 35:59 They're the longstanding Australian genesis-ed, uh, data center company, and they're really, really popular. 36:05 So why don't they come in, take up, uh, that, those spare dollars, the spare Australian dollars, which might be a little more, uh, available down under. Back to you. Yeah. 36:19 So Redbank would be its second data center, just to reiterate that. It already has an 800-megawatt campus called Bundy in South Australia under development. 36:31 It's got grid interconnection secured, and it's targeted to be energized in 2028. So still a little bit of time off. Obviously, that 800 megawatts will not happen overnight. 36:43 But if Iron secures Red Bank, the project would increase Iron's total portfolio from five gigawatts to approximately 5.9 gigawatts, a 19% increase. 36:53 The biggest uncertainties as ever for a project like this though, are grid interconnection. Grid interconnection, again, this is just a proposed site. They don't have the grid in-interconnection or permit secured yet. 37:04 Other permitting for zoning and environmental matters, financing, and customers. 37:10 And also, New South Wales is dealing with substantial data center connection demand, including 1.4 gigawatts from CDC data centers near Wagga Wagga. CDC is... 37:21 was the data center partner for Firmus on their flagship campus. That, that deal or that partnership disintegrated though in September. 37:30 And given the 1.4 gigawatts of hunting that CDC is doing in the area, power availability is particularly important for this site. Obviously, power is always key, Charlie, as we talk about. 37:43 But we just spent the entire show talking about financing concerns, so it seems to me like that will be another huge hurdle for this project as rates soar and financing tightens a little bit for the data center build-out. 37:57 Two months ago, this would've been a power as a bottleneck story. Right now, with the market absolutely obliterated, it's, uh, financing is the main, uh, bottleneck story. It's probably both. 38:09 Um, just right now, as everyone sells off the entire NeoCloud sector because of potential CoreWeave credit default swap contagion, uh, where is Iron gonna get the dollarydoos to fund this Wagga Wagga [laughs] build-out? 38:23 I'm sorry that I keep, like, [laughs] joking about this. It's just that, look, I'm from Oklahoma. We have our fair share of funnily named cities, so I gotta, I gotta throw, uh, I gotta throw some, uh, jokes over to them. 38:36 Uh, Wagga Wagga and dollarydoos. You're really, uh, you're- [laughs]... you're coming into your own as- Yeah... an Anglophile there, Charlie. I know. The, the Brits and their, 38:46 uh, the, the Brits and their descendants in their- Yeah... colonies across the world always have... They just have this penchant for just putting the most absurd names on things. 38:54 [laughs] Um, somehow we escaped that in the US. I think we have the thing- Escaped... Well, we just normalized it when you think about it. We, we call, um, you know, we... 39:03 The, the stuff that we wear on our legs, pants, that those are underwear for them. So think about it from their perspective. They, you know, we're kinda weird. 39:10 Yeah, but they also, you know, call, like, a lollipop, like, a licky licky lolly stick, and things like that. They do? Are you sure about that? [laughs] No. That's me, that's me joking. Okay. 39:20 [laughs] But they do call everything pudding. Pudding can be a dessert, and it can- Yeah... also be Yorkshire pudding. Yeah. Um, these people gave us the English language. They have no idea how to yield it. 39:29 Sorry for- That's true... any British listeners we have. Yeah. 39:31 Um, one more thing to note on in terms of hurdles, Iron says here, quote, "We're speaking about it now to provide local residents with information on the proposal before we lodge a scoping report. 39:43 Red Bank has the potential to create jo- local jobs and economic stimulus as well as community benefits for the Wagga Wagga area," end quote. So, uh, again, local pushback potentially being a hurdle for this site. 39:54 Uh, we're not poo-pooing Iron's, uh, their effort here. They've obviously proven themselves able to execute across a number of sites internationally. 40:04 Just m- m- merely adding context to the fact that 900 m- megawatts on paper, it's a lot different than 900 megawatts in practice. There's still a lot of unanswereds for this site as they seek to get it off the ground. 40:17 And six to eight years of construction time, I guess that's from proposal stage to actually getting the full thing built, long time off. But they have an incentive to expand in their, uh, home country. 40:30 You know, they wanna see themselves succeed there, so I have no doubt they'll put forth their best foot as they try to get this data center off the ground in Wagga Wagga. 40:41 I'm excited to hear Dan Roberts say that on their Q3 earnings, coming up in two or three weeks. So, uh, thank you Iron for our one non, uh, s- uh, blood in the street stock market story today. 40:57 Um, well, actually we got two. Keel's coming up. Keel signing, uh, getting some permits in Pennsylvania. I'll take the story after a word from our sponsor, Ligos. 41:07 [static sound] If you're using a Bitcoin collateralized lender, you need to ask yourself one question: Do you know where your Bitcoin is? 41:19 If you can't answer that question with confidence, you need to do yourself a favor and check out Ligos Finance. 41:25 Ligos Finance is our preferred Bitcoin-backed lender here at Blockspace, because they use Bitcoin native smart contracts that ensure you never relinquish control of your keys. That's right. 41:35 With Ligos, you always have self-custody of your Bitcoin during the lifetime of your Bitcoin collateralized loan with their services. 41:43 With Ligos Finance, there's no rehypothecation, no wrapping, and no bridging, just competitive rates as low as 8.5 to 10% APR. Go to ligos.finance to learn more. 41:59 Keel, the company formerly known as Bitfarms, picked up another permit in Allentown, Pennsylvania. 42:07 Last night, the Allentown Planning Commission approved a 75 to 80 megawatt data center in an empty warehouse with Keel lined up to operate it and rent it to an unnamed tech company. That's good news, right? 42:24 Well, today the stock is down about 8%, which is about $2.65, roughly 64% off its 52-week high. So what's going on? 42:37 Well, back in May, CEO Ben Gagnon said Keel would sign three major leases by the end of this year. It's mid-October, and the count is still zero Keel does have 2.2 gigawatts of power pipeline. 42:53 It had about $819 million in liquidity as of August, and it has a giant stack of permits. What Keel does not have is a signature from a tenant, as Ben Gagnon, like Babe Ruth, had called that shot earlier this year. 43:11 However, the market is tanking right now. NeoCloud's power shells are selling off left and right because of the CoreWeave c- credit default swap contagion. 43:21 We're in a market regime, though, that pays for signed contracts on real delivered megawatts, not just the megawatts alone. So Colin, we had Ben Gagnon on this show back in August. What's going on? 43:37 Does this sound like this is a company holding out for a better price, like they're negotiating? Are they waiting for a better market environment and the market's just not playing ball? 43:47 Or is there some level of they just can't close a deal? I don't think they're waiting for a better price. I think they'll take... Um, I think they'll get, you know, favorable economics for the site now. 44:01 We've seen economics for these leases trend upward over the last year, and demand is no doubt insatiable across the United States. I also don't think that we can necessarily read this as Keel incompetence. 44:13 There's nothing that I've seen that would indicate that. I think a lot of people are quick to assume that if you haven't signed a deal yet, that means that your sites aren't good and you're not good at negotiating. 44:25 I think people underestimate how long these deals actually take to get through the door. I mean, you have to first, you have to first gauge the interest of your potential tenants. 44:36 You have to scope what the specs will be for the site. 44:39 You have to go through round after round of discussions on what the innards of the data center will look like, go through all of their specific requirements for their GPU fleets, uh, securing vendor lists, securing lead times, things like that for all the equipment. 44:55 There's so much that goes on behind one of these deals that I don't think the average person realizes how many man-hours go into actually getting a deal like this inked. 45:05 All that said, um, Ben said that they would have three leases before the end of the year. To me, that indicates that he was very confident that that would be the case when he said it, um, in, in the middle of the year. 45:16 Um, my question is, i- is like, why would you say that if it might take a little bit longer or if there had been unforeseen snags? 45:25 Also, maybe, like, Ben should just be a little bit more careful going forward, and I'm sure he's learned that from doing, from saying that. 45:32 Last thing I will say is there's a, there's a chance that the, all of these talks were in motion, and then Josh Shapiro slapped down that executive order. 45:40 [laughs] Now, Keel is grandfathered in for their flagship sites in Pennsylvania 'cause they had the permitting already, um, secured when they, wh- wh- when that order hit the ground. 45:54 Um, that being said, I do wonder how much that makes some of these potential tenants a little shaky about signing a deal in the region, just because you d- might not know how to price in the political risk of operating there. 46:06 Yeah, think about how much has changed in the past five months. 46:10 I think it's been five or six months since, uh, Ben said something to the effect of, "My number one priority is going to get, is three tenants or three leases this signed this year." So much has happened. 46:22 People just forget how quickly things have moved. The big one is the Pennsylvania soft moratorium. Um, [clears throat] let me give some stats about this particular site and permit, 'cause it's a little bit different. 46:38 So it's not one of Keel's three priority sites, Panther Creek, Sharon, or Moses Lake, and there is a third party developing this site. 46:45 This is a smaller fourth site, so I'm kind of curious where, what the reasoning and thought process behind that is. 46:53 Obviously everyone wants more power, permitted power, but this is not, you know, one of the three of its flagship sites. 47:00 Um, then it does have a little bit of potential local friction because if I'm understanding this correctly, the site could have more than 30 backup generators. 47:12 Backup generators with noise and diesel storage can present new challenges, especially today, as everybody, uh, hates the actual behind-the-meter generation, um, at, at, at the local level. And, um, 47:30 then, uh, I'll also point out there's, you know, there's only 11 weeks left in the year to sign these, these three leases. 47:41 11 weeks can either be a really long or a really short time, depending on how you frame it, 'cause the past 11 weeks has felt like a lifetime. 47:50 So if by that measure, we could be looking at a very, very, like, a different landscape 11 weeks from now at the end of the year. 47:58 Um, co- you know, however, in the normal business cycle at the scale of these large industrial deals, 11 weeks is not that long. There- Yeah, and I also wonder if it's even shorter than 11 weeks. 48:09 Like, is it really more like four or five? Because after Thanksgiving hits, bro, people check out, you know? I mean- You say that, but think about it. We're, I mean, this is... 48:18 Look, oil and gas, the, the business that produces the world's most important commodity, those guys go out to the well site on Christmas and work. This is why I left that business. Yeah, but those guys- But-... 48:29 aren't in the ones in the office writing deals, right? Like, they're the ones who- Yeah, but I mean, if you're, think about it. 48:34 If you are negotiating the, the deals which are at the forefront of the Amer- American economy, I w- you know, you're probably- Yeah, you're probably not... still in the office. 48:42 Maybe it's different if you're one of the, the Wall- the, the lush Wall Street, like, banker folks who you actually do take off three weeks for the holidays. 48:52 But- Um, the, the operators, the people actually trying to get deals done probably gonna full steam. They're, they, they're not gonna do more than a quick Thanksgiving dinner and a presence on a Christmas morning. 49:04 They're probably gonna steamroll right through the, uh, holiday season just 'cause of the, the timeliness of these deals. So, you know, it's, it's America, baby. We don't take o- we don't take two months off for summer. 49:17 [laughs] Uh, yeah, I, I'll also make, um, I'll drive a, a another... 49:24 I'll, I'll drop a comp here, which is the last major deal we- we've seen in kind of the power shell data center category is the CleanSpark one, where the CleanSpark Sandersville deal, um, they put their project cost at 10 to 20 mi- 10 to $12 million per megawatt of critical IT load. 49:42 Um, so that's kind of the la- the last comp we have in this category for, uh, any imminent tenants or deals to be signed. I feel- And I, I think that's going up. Yeah. 49:54 I mean, as we saw from Barber Lake for the Fluid Stack deal with Cipher. 49:58 Now granted, that is due to, like, a design rehaul that Fluid Stack is requesting for that site, so that's part of the reason why it's now over budget. But, um, we're starting to see, uh, within 50:13 the data center, the powered shell industry, that average cost per megawatt ticking up just a little bit, especially as, uh, inflation continues to press on the, uh, raw materials for a lot of these data center build-outs, and lead times not really getting any better. 50:30 In fact, Cipher kind of mentioned in that, 50:33 uh, in, in that, uh, press release for the convertible note related to Barber Lake that they said that, um, their, their inputs are going up and that, um, it's just getting a little bit more expensive to, to build these things. 50:50 So, um, we'll have to see. I know Keel's a hot, uh, hot stock for a lot of our audience. Um, kind of divides people. Some people love it, some people hate it. 51:02 Um, [clears throat] seems to me that people are very reactionary as to whether the stock's going up or down, whether they love or hate it. Uh, but that's just markets, baby. 51:12 Colin, that's kind of all I've got for this sec- this segment. Anything you wanna wrap up here with? 51:17 No, I guess just to put a bow on a lot of the market commentary, obviously there are operators that are executing at a higher level than others. 51:26 Um, for example, HUD8's one of the only ones that's up year over year and year to date. 51:31 Um, their credit is pretty cheap compared to some of their competitors, and overall they've just done a really good job of turning RiverBend and Beacon Point into some, uh, premier sites. 51:41 So there, you know, there are definitely... And then there are companies that have just like languished and haven't been able to do much at all within AI, um, the AI space, like Mara. 51:51 So, like, within individual operators, there's clearly some key differences. 51:55 But when there are, like, absolute bloodbaths like today, you'll often see in our comments on Twitter and elsewhere people saying, like, "Oh, the management for this stock sucks. It just keeps going down." 52:07 Um, and this is after the stock was up, like, at its peak, like 400% over a year. So, um, don't blame management if you bought the top or if you're buying falling knives. 52:18 Some of these companies do have management that are not operating at a higher, at, at, as, as, as high of a level as some of their peers. But, like, every single company is getting absolutely bloodlet today. 52:29 So I would just urge people to zoom out when there are moments of extreme weakness in the market and when everything is getting destroyed and not blame their individual stocks for the fall, uh, for, for, for the downtrend that they're seeing. 52:44 I mean, here's the five-year for Keel. It switched to Bit Farms, I'm sor- or it switched from Bit Farms to Keel sometime around here. Um, yeah, down. It's actually... Is it up on the year? It's- No. 52:59 Okay, it's not up on the year. I don't think so. Yeah, it's down on the year. Yeah. It, it's been pretty brutal. I mean, Keel was ripping to, like, six bucks, above six bucks back in July. Yeah. But everybody was. 53:11 I mean- Yeah... Ju- late June was gangbusters for the entire sector. Yeah. 53:17 And, and you do have to wonder, given, like, the current environment, if they do get these deals signed, like, how much do we actually think that's gonna pop the stock price? 53:25 Because, you know, we covered, uh, was it the Wells Fargo note a while back? They were basically making the argument that signing deals right now isn't enough for a lot of investors. 53:34 Like, they, they basically wanna see that you actually are generating revenue on that lease and that you've delivered the lease itself. 53:40 I do think that for better or worse, barring, like, a couple of names maybe, um, y- you're not gonna see the same premiums given to a stock value, uh, a stock's valuation just for signing a deal that we saw at the end of 2025 and the beginning of 2026. 53:58 I think that was a Goldilocks period, where you get a deal signed for billions of dollars, you immediately get re-rated. Everything else was re-rated after that. 54:07 Like, any company that had megawatts got a soft re-rating, and now we're at the point where the market's demanding actual execution if you're going to get that valuation that we were seeing earlier in the year. 54:21 Yeah, it sounds to me like, God forbid, um, you may actually have to, uh, sign a deal, deliver on the deal, start generating revenue, and then the market rewards you. 54:31 Uh, as opposed to saying, "We've got all this stuff which can happen. Now we just have to do it." 54:38 Um, if the worst is that the, that the sector reprices more towards real delivered revenue, uh, and, uh, real build-outs, then I would say that's good. That's good for the long-term health of the data center business. 54:53 You are listening to Blockspace Live. Thank you for watching. Charlie and Colin, we do this every weekday at 1:00 PM Eastern. 55:03 We are Compute's daily live show, featuring quick hits on AI, data centers, emerging technology, and markets. 55:10 If the live stream's not enough, head to our website, blockspace.media, and get our newsletter and see the expanded universe of content that we produce here at Blockspace on the data center and AI sector. 55:21 Whenever Bitcoin pops, we'll touch on a Bitcoin story, but it's gotta go up first. Thank you for listening to Blockspace Live. We'll see you tomorrow. I'm Charlie. I'm Colin. 55:31 Peace. [outro music]