$iren Sunday Night 2.15.26 / The Path Forward

Hosted by @₿itcoin ₿utcher 🥩 🐑 🐷 · 2026-02-15 · Tags: IREN

TLDR

The speakers remained bullish on IREN’s long-term AI infrastructure opportunity but stressed that GPU delivery delays, revenue recognition, ERCOT clarity, and execution of the Microsoft project remain near-term risks. They expect Sweetwater, Childress, Canadian expansion, potential new customers, and improved GPU procurement to drive a future inflection point, while warning that the stock could suffer if another major deal is not announced by the next earnings period.

Speakers

Notable quotes

Transcript

₿itcoin ₿utcher: What's going on, everyone? February 15th, happy belated Valentine's Day to the Iron family. And we're 3 minutes after nine. We'll get things going. No real topics per se. I threw in the nest the posts that I had, but again, just want to meet with everyone weekly, keep everyone's spirits up. re-underwrite our investment, understand what's good about it, what's bad about it, and everything in between. I purposely wrote today's post given we're in, Franz has a subscriber group, only Friends, and that's usually a very concentrated view of sentiment, good or bad, when share prices in the 60s. Everyone's celebrating and talking about how we're going to 100. And more recently when things aren't going so well and we're tapping A3 handle, it gets a little negative and dark in there. So we have to just take a step back, be objective. And that was kind of the goal in what I wrote earlier today is just to remind everyone that there's a lot in front of us, and there's a lot behind us. And I do think better times are ahead. Specifically related to what I wrote, I think it, for those of you who don't know, I have a small business in Detroit. It's a butcher shop, family owned, had it for over 50 years in our family, since 1967. And in a parallel industry, restaurants often have, they call them soft openings. And the idea being you're trying to kind of get through the growing pains. And I think that's what we're going through right now is Canada and Horizon One are kind of our soft opening for what iron's about to become. I'd be the first to tell you that I was expecting a higher revenue number in cloud, not 17 million this past quarter, but we're here now. And I would be interested with McNally and Anthony Power if they could ask Kent what they have contracted 400 million, and I see God particle in the audience, and he properly corrected my assertion that we could assume 100 million in the first quarter now, just because that power, or I should say the power's been contracted out to third parties, which is great, but there's still the execution element where there's kind of a disconnect right now. My gut feeling is things are improving. But we don't know until we see it flow to the financials. So I appreciate God Particle for pushing back a little on my post. But the high level overview of it is this theme of patience. And we've been very patient. I don't doubt that. But I don't also think it's constructive or productive at this point to be tweeting at Dan or like one deal. And I think there's a middle ground. And I was saying to a few guys earlier that they have the tools in place now. They have $2.8 billion in cash after the direct equity offering and the retiring of the convertible notes and the subsequent issuance of new convertible notes. So the balance sheet warrants more aggressive moves going forward. but we had to secure the GPU financing for the Microsoft deal. And Franz has repeatedly said that the value of the Iron franchise will be made or broken with execution of that deal. And I agree with him. Like, if you can't deliver on your first hyperscaler deal, then all of the other signings are kind of irrelevant. So we might want the dopamine hit in the short term to say that we signed Anthropic or Google. But the key to all this is being able to deliver on our promises. And no different than that restaurant having a soft opening where they make a few mistakes in front of their friends and family. You almost don't want It's nice to have a fully booked reservation list that first night, but if your crew isn't ready to perform, you're setting yourself up for failure. So in parallel, I think Dan and Will and Anthony understand that even if they could sign the 4.5 gigawatts that we have tomorrow, which I believe they can. There's no shortage of clips on Eric Schmidt talking about shortage of power, or I put in the nest, the AWS CEO on the power constraints, or Dario from Anthropic. Everyone needs power. We've established it. It's kind of the core thesis of the investment. It's kind of nauseating at this point to repeat it. But what good is signing all those if we can't construct it? And unless we plan on selling some of it, which I don't think is the intent, given how scarce it is, and hold on to it, then yeah, we might have to just wait for them to finish 160 megawatts in Canada or the 300 megawatts in Childress. And then with Sweetwater specifically, I think this week will be nice to get some clarity from ERCOT where they finally publicly say that iron has two gigawatts secured or maybe even more potentially in batch zero, which I believe Dan and what he said in the call, but I also personally, my hunch is that we're getting priced or some of the headwinds that we're facing right now is that potential risk, because that's a lot of the promise with iron right now is this power capacity and if two of the 4.5 or even, well, when you say the two, that's Sweetwater one and two and then the new 1.6, if 3.6 out of the 900 is called into question and is discounted, then you can't help but ask yourself, it seems reasonable to me that until ERCOT finally lays out the rules of the game that there would be some uncertainty. And we see it in a parallel universe with the House and the Senate are trying to pass a crypto bill, and that's impacted the price of Bitcoin and Ethereum. So markets want clarity. And until we get some more of it, whether it's a deal signing or whatever, we just have to kind of roll with the punches for now. And if you do still have extra money, buy at 40. If you liked it at 60, I'm pretty sure you should like it at 40. And I can promise you that nothing changed from January 28th when we peaked out and had our local high to now. And if anything, we're stronger. And the 1.6 gigawatts in Oklahoma simply isn't being priced in right now because it's the same bear cases, whether it's Burry or Chanos or ever, that they don't see 1.6 gigawatts as an opportunity. They see it as a problem, like how are you going to fund all this? And I think they just assume that we have to do it tomorrow, which is not the case. And much to our dismay, They management hasn't tried doing it all tomorrow. I think other companies in the past try and grow too fast. It'd be like trying to run a marathon on day three of a training plan. Like we're probably running three or four miles a day right now and we'll scale it up to get to a half marathon pace of 13 and eventually maybe you don't train the full 26 miles, but we'll get up to 18 to 20 miles and finally run that race. But There's so much ahead of us, and I just try and come here weekly and keep things positive. And it's good to be back after last week, which was a great episode that I missed out on for a ****** Super Bowl, but I don't regret spending time with my family. But yeah, that was one of the least entertaining games I've ever had the privilege of watching, but family time was positive. No particular agenda tonight. I got a DM from Jared asking about in the Only Franz due diligence group. I believe it was in the 10-Q. I would have to double check this, but let's assume it was disclosed in the 10-Q. But the Microsoft contract, what's noticeable is there's language in there saying that the company can, Microsoft can opt for, assuming that they come to terms with Iron, can upgrade GB300s to Rubens, which explains the rack density, the flexible rack densities installed. And when we were saying they were future-proof, that looks like Microsoft was already thinking about that. So we can talk about that and how that relates to Sweetwater. I think with Sweetwater too, there's been a lot of, Franz has noted in the past that the design is modular, meaning you can replicate it and it's anticipated that that design feature, the Sweetwater data center will be featured in Oklahoma. So that's another reason too that in my post, I referenced why, even though I'm excited about monetizing the 450 megawatts in Childress following horizon one through four, it's probably more important for market perception when Dan's talking about negotiating for 24 months related to Sweetwater. Sweetwater is the unicorn. 1.4 gigawatts about to be energized. You can secure not only a client that has the ability to scale there, whether it's Google or Anthropic or wherever they end up deciding to go with, it also, Franz will tell you that it's very important if you have a construction company, if you're a subcontractor and you're trying to protect the interests of your company, it's a lot easier to sell your crew on doing a project of the scope of Sweetwater. where you can know that you're going to be in the same place for a number of months, if not a number of years, as opposed to doing a smaller 200 megawatt or 100 megawatt site. Even those are relatively large, it's still not 1.4 gigawatts. So that's kind of my spiel. Anyone who wants to come up and had something they wanted to discuss in particular, if anyone had questions over what I covered. But again, I come up here each week and just try to remind people that you're in the right place at the right time. Even if the number on your brokerage account isn't what you want it to be yet, I can't think of a better opportunity in this market. And I was hoping Nick would be in here and small cap and friends for whatever reason. I've tried inviting. I'm not sure if they're just not available right now to get their commentary on markets. But the CPI came out this week at 2.4%. The target rate's been 2%, but 2.4% is the lowest that number's been since 2021. And a lot of times on a Mike Alford space, you'll hear that in some ways, we've been in a recession already, like since they started raising rates to address all the money printing from the COVID, that from 2021, most notably 2022, what a rough year that was for risk assets and for us to start crawling back. But we really haven't had that bull market overall. It's been more separated into different sectors at different times. The irony of all this is for as bad as we feel right now, having seen 80 and riding it down and on our way back up, we're still, if you look back from January 1st of '25 to now, we would be one of the top performing equities in the world. Sometimes it takes a little of that perspective. But I see Franz is up, and we'll see if he has any opening comments. Franz, how are you this evening?

Frans Bakker: Hey, Butcher. Yeah, I'm good. I'm joining in from my phone now because I got kicked off the desktop version. There's something wrong with X that every now and then, There is this bug that when you are listening into an X space on a desktop, you get like, you lose a sound. It's really weird. And then when you rejoin it, it gets back, but it's just, I can't work with that. So I'm on mobile now and I'm just using nothing else, no microphone, just only my smartphone. But yeah, so. I listened to your introduction for the most part, and I was kind of making a comment. I was preparing to reply to the God particle on his pushback on your post. So I will try to form this reply, but basically what I'm going to argue is that from the $0.5 billion of annualized revenue that they contracted by earnings, I think Iron is basically saying that most of that, or at least a very large part of that is also already operating. So if we can, you know, construe from everything they say during the call, there is still a decent chance that the revenue for the quarter will you know, near triple digits. I won't say it will absolutely be $100 million. Obviously, we don't know when those last GPUs will arrive. But if you dissect the call, I think you can see that, you know, they're typically iron sandbagging, under promising and over delivering. So I I'm confident that there will be a very decent number reported in it during the next earnings. But all things considered, I also agree with you that if the company doesn't have anything by next earnings in terms of a deal, then the stock will be sold off regardless of achieved revenue. So, so far I'm on par with you for the most part and I will share the post in the nest after I post it and then we can continue to talk about that. And other than that, yeah, there are some developments we are following at Childress. I will speak to that in a little bit. And we also got a new satellite image from Sweetwater that just arrived like 10 minutes before the space. So I haven't looked at it yet, but I think we have a lot of material to work with today and in the coming week. So yeah, pretty exciting.

₿itcoin ₿utcher: Great. I got a comment from-- I want to get the name correct on here. Jman711, I did not speak to the index inclusion yet, MSCI index inclusion. So thanks for bringing that up. For those, I had a post. I can throw that in the nest that briefly describes what that is. I'll do that later while someone's speaking. But the index references on February 27th, IRIN shares will be bought by various institutions that will include them in their indexing where It's a net positive for the stock, is the short answer to it. And there's going to be more institutional ownership as a result of it. This index is widely held, and the expected impact of it-- well, the index itself, I should say the MS-- keep screwing the acronym up-- MSCI, multiple indexes will use those participants on the list, which Iron's now included in. And as a result, I said those indexes will buy shares so that it mimics that list. And that's a net positive that was announced on the 11th or the 10th, I believe. And then by February 27th, that should be a net inflow. So I'm looking at that to be a positive-- or not a headwind, it should be a tailwind for the stock, as well as more clarity this week. As a reminder, things are closed tomorrow, Monday. But later in the week on the 20th, we'll get the batch zero announcement from ERCOT. And even though Dan and Kent spoke to their expectation that Sweetwater is included in that batch zero process, which would essentially mean that there's nothing material that will change, but it's just more clarity. You don't know until they formally announce it. So we had some members on-- there was a call, I believe, on the 12th this week, which was-- just looking at my calendar, that call was on Thursday. Some people attended, and that was establishing the criteria, because at a very high level, ERCOT, is the power system that Iron operates in Texas. So that impacts Sweetwater and Childress. And they've had so many power requests. I think it's something like over 200 gigawatts requested, while the current system's only 60. So well over three times the existing capacity Of the grid is being requested by multiple parties, so there's new criteria that's being added, and they're trying to come up with a framework on how to process all of these requests effectively, but a lot of them... It's also been speculated are almost dummy requests in the sense that there's people with land that are speculating that they can just build a data center on it without more than owning the land. So there's going to be criteria, most notably deposits. I believe it'll be some form of not pay to play in the sense that it's corrupt, but pay to play in that showing that you have skin in the game. and to see who's real and who's not. And I look forward to kind of putting that behind us, because if anyone's been at Iron Bowl long enough, I can't tell you how many times the question of whether Sweetwater has power or not, despite every single photo that Franz shows that they're building upon it. I don't know why you would rationally commit millions of dollars on long lead items, on Transformers, all this. power infrastructure just to be told that you can't connect, but that's a whole other story. So I'm looking forward to, if anything this week, hearing batch zero, hearing that it's secure, there's potential that there are other sites that will be included for iron. It's just whether I have a hard time believing they'll disclose them. It's whether ERCOT would disclose them or not. And maybe Franz, you can speak to whether you think they would do that or not. But I think in the past that there's been an apprehension from management, rightfully so, for strategic reasons to announce sites because there are usually, like with Sweetwater, noticeably, there's competition right down the road where from a real estate standpoint, if you're telling everyone what you own or where you're at, potentially, there's an incentive for people with male intent to screw with your plans and buy parcels right next to your existing site only to try and sell it back to you or even possibly develop right next to you. And then there's a risk of losing contractors and poaching and all those things. So it kind of does create quite a bit of noise, but I do think it's there's a middle ground and they've released the Oklahoma site during earnings. So I there's no reason to believe that we're at 4.6 right now. Why that wouldn't grow to an excessive five and probably an excessive six in the next 18 months, if I had to guess. But Franz, do you have or does anyone in the crowd have any knowledge as to whether that batch zero, if that's disclosed privately to the companies or if that's a published list to the public?

Frans Bakker: Well, I don't know for sure, but going by what ERCOD usually does is they protect their customers. commercial, you know, they usually don't talk about names from stakeholders. I know that the stakeholders get a chance to form a reply and those are public, but I can't imagine they will specifically mention projects or sites or substations or something, or even capacity numbers. I think that's, you know, Information that is commercially sensitive, so I don't think they will talk about that. I I I can't imagine that.

₿itcoin ₿utcher: See, I'm wondering. They've already announced the site secured. I do think if Mike Powers in here. And incognito. I I still hope that IR. I hope they don't take for granted that. They're included in batch zero. I have almost every reason to believe that they will be based off of what Dan said. But I still think that the share price, it's a reflection of a risk-off environment with rotation into lower-risk sectors that got overlooked last year as AI got outperformed the rest of the market. But at the same time, I do think they're still-- we're only showing $17 million of quarterly revenue in AI right now. And everyone's looking for a reason to ding the stock. And this is one of them. And I do think it's an opportunity for a cheap post later in the week just to say, hey, we secured this, and things are great. It's in their nature to ever-- they're not going to be like Sue from-- I forgot her last name-- from Hutt. But I think there's value to what she does. A lot of times, we're watching it, and she's educating their shareholder base and us by extension on the process and how the company's going through it. And while someone like Franz or there's other members of our group are in the weeds and attending those meetings, it's a good middle ground for them to advise their shareholders. And I've benefited from it, and I enjoy some of the work that she does. And it's not going to be Mike on camera saying it or Dan, but a cheap post on Thursday saying, good news, everything's according to plan. anchors away, let's keep moving forward. That feels like an easy middle ground where no different than-- and to be fair to them, they included a headline on the index inclusion this week, which was great. So I think going forward, I know their priority is institutional ownership, but I hope they just remember that retail is been a big part of this story. And I think part of the reason why we're still at 40 right now and not a lower share price is because of a lot of the people in this room that refuse to sell, and if anything, keep buying at these prices. So that's all I have to say regarding that. But I also got a question from W. Peel, who I've met in person, fine Canadian gentleman. He was asking me if I thought they were going to have an ATM or financing in the coming weeks. I don't see that happening, because as of January 31, they had $2.8 billion in cash. They secured 95% financing for the GPUs, including the Microsoft prepayment for Horizon 1 through 4, and the retrofits for-- what sites? Canal Flats and McKenzie. So that's 30 and 80 megawatts. It's 110 megawatts. It's probably roughly, we'll say 80 critical IT. It might be a little higher, maybe it's 90, but it's about 250 to 300 million. estimating figure 3 million per megawatt on that. So if it were 80 to 90 on that, that's where I'm getting the 250 number approximately. That's not, it's a material amount of money, but it's not going to break the bank. Like that's less than 10% of their existing cash balance. And that's part of the reason I think what you'll see going forward is that emphasis on contracting the rest of Canada. Franz already alluded to the 500 being almost done at Prince George, and then numerous people have sent us the related job postings for Ops and Canal Flats and Mackenzie. And I think there's an opportunity there, whether My gut feeling is that even though they left an opening for Childress to be used for air cool in the future, I think that the next contract that's coming up, if I had to guess, will likely be Sweetwater, given they've been negotiating it for 24 months. or in and out with certain parties, or McKenzie or Canal Flats is a whole site for an enterprise. But that contract specifically for the software, which last week was a good episode if you guys missed it, but related to the Kubernetes installation that an enterprise could layer their software on top of, as opposed to a hyperscaler that wouldn't need Irons Kubernetes install. That leads me to believe that Canal Flats or Mackenzie will be either a sovereign, potentially the government of Canada, and/or there's been enough speculation on Anthropic or whether it's another company of that size that can dominate a whole site as an enterprise. And the reason you would do that primarily is those companies, instead of using Hyperscalers in the past have used primarily Oracle just because Oracle's charging a cheaper rate and the hyperscalers tend to charge a higher premium for their services. So if you can find a way to team up with a Palantir, an Anthropic, or a TikTok, or one of those companies that demands all that AI compute, you can not only have a sexy brand on your next presentation, but you also have someone that can grow into the space more appropriately. Franz, with respect to, can you speak more to, earlier Jared had asked about Rubin, and can you kind of refresh everyone, your understanding of it, the contract, the ability to upgrade to Rubin's at Horizon, but even if they don't do that specifically at Horizon One, do you anticipate that Sweetwater is going to from the get-go be a site dedicated to Rubin.

Frans Bakker: Yeah, I mean, it makes a lot of sense for Iron to be working on a data center design that is lowering CapEx, but, you know, suitable for the Ferra Rubin. In particular, because, you know, the site is still being prepared for actual data center construction. I was just looking at the PNG from the satellite image, but it's overexposed as typically delivered as a PNG. So I will need to wait for someone to work on the TIFF file. But there is a lot of new stuff to be seen at the Sweetwater site. So maybe this could give us some hints as to what they're planning. But Yeah, it's my it's my expectation that they will make this site ready for data centers that are going to be running Vera Rubens. I'm not entirely sure if there's going to be like if they're going with the old architecture. I think that is the VR200 is the last model. that will run in this architecture. And then I think it's going to move to, what was it, 800, 400 or 800 fault DC architecture. But Yeah, the thing is that we don't know, and we don't know which customers they're talking to, and we don't know the time to data center that the customers are looking at. I mean, right now, all you can really work on is if you want to have GPUs running in the coming, let's say, 12 to 18 months, then it's probably going to be the VR200, so the the normal Vera Rubin, so to say, not the ultra version or anything beyond that. So yeah, I mean, I guess we can also deduce from the fact that they have said that they will build the Oklahoma side based on the Sweetwater design. So based on that, I still conclude that there's going to be a new design for Sweetwater that is different from what they're doing in Childress. So, you know, if you see the token output per megawatt that's estimated for the Vera Rubin, it just doesn't make any sense to go with the GB300 data center only, you know. So we already know that Iron started to build Horizon for 200 kilowatt racks. with the VR200 in mind. One side further into the future is Sweetwater One. It just wouldn't make any sense for them to make a data center design here, which would include lower rack densities again. I think it will be a future-proof design, a modular design, a lower CapEx design, in that way, they could have like operating leverage and economies of scale on a large site like Sweetwater. So it would make a lot of sense for it to be VR200 and beyond, if that's even possible, of course. I would have to look into some of the Details around the architecture, but it's probably going to be VR200 or the ultra version, but that would mean that the whole design of the power step down would have to be different. So I'm sure there are people that can speak to this better than me, but. Let's just say very simplified. I think a portion of the site is going to be dedicated to the VR200, so the Ferro Rubin. And I would not be surprised if they would dedicate a portion of the site to a future proof model with a whole different power architecture where they could potentially expand direct densities up to 600 or a megawatt, 600 kilowatts or a megawatt. So that's just my expectation based on the fact that Iron was already front-running the Vera Rubin with Horizon. So I can totally imagine they will be front-running Vera Rubin Ultra with the Sweetwater design, you know, so it's just a bit of deduction and following their logic. But yeah, I will have more info about Sweetwater once I have the updated picture. The one I'm looking at now is-- yeah, there's not much to see on this one because it's edited by Planet, and they're not really as good as the people we have. So I'll have to get back to you about that.

₿itcoin ₿utcher: Great. There was a question from Laz about the-- I'm trying to pull it up, but essentially how Rubin will impact margin, and I can answer that with we don't know. Now we know that the token output's higher, but we don't know how NVIDIA plans to price it. So if it's three to five times more efficient, I think it was for inference or training, I can pull up the stat. Point is, directionally, we know it's more efficient. It's just whether Jensen is willing to give some of that margin back to his customers, which I'm assuming they're self-interested, that they're going to take as much as they can. But what gives me hope that running Rubin will be better is the alternatives that are now appearing in the marketplace, whether it's a TPU or Trainium, they don't necessarily perform the same specs for training as the GB300 or the VR Rubin. Those are still high-end, but there's still an element of competition that's coming into the market where I think that is keeping NVIDIA a little more honest going forward. But that's yet to be seen. And until they announce a price on it, it's hard to say. They can promise how great the performance is. But ultimately, guys who are modeling this out want to know how much per GPU it will cost to figure out what the token cost is. So I don't have a great answer for that beyond we don't know yet. But I also know that for Iron to be future-proofing Horizon and most likely Sweetwater, that they're prepared either way. And if the customer wants it, odds are they have to kind of roll with it too. And I think any good partnership, no one's going to be 100% happy, but both parties will get what they need. But to Laz's point about margin, I do want to see on this next deal specifically for Sweetwater, you have a unicorn site. It's your flagship site. It's not your soft opening anymore. You're on track. You haven't necessarily handed the keys to Microsoft yet. They're on track to do so and have made good progress from the acceleration that's been priced in. I think you get one mulligan with the market. And I think even though, whether it's Google or Anthropic, whoever decides to inhabit Sweetwater, I can't think of many other one plus gigawatt sites available that people can scale into and run larger clusters like that. It should command a market premium. And Iron has done enough in the background. They might not be able to walk someone through Microsoft's site, but I think there's enough respect for Microsoft to know that if Iron got through that probing from Microsoft, that they're doing high-quality work that warrants the attention of everyone else, and it will command the appropriate premium. We won't know it until we see that sign, though, for now. trying to think what else we're kind of post-earnings and still in that waiting period. And the obvious question is when deal. And I've said earlier that I believe it's going to be McKenzie, Canal Flats, or Sweetwater that are next are kind of the three sites that make sense to me. But until, if anything else is coming up off the top of my head, Otherwise, it's really just macro-related. I did watch a good speech today from the Secretary of State, Marco Rubio, in the Munich Defense Conference. But Franz had his hands up. So Franz, what do you have?

Frans Bakker: Yeah, so I would like to talk a little bit about my expectations for Childress. In recent days, we have seen on satellite imagery that Iron's been making some new laydown yards at the north end of the site. So if you look at Childress, you know, from the overview that we usually share, on the top of the mining buildings, Iron left out like a large plot of land, like empty land. It's like on the north of the data centers that they built last year and the year before. So block one to five. And they are now making these laydown yards. One of them is a very large one just above block four and five. Those are the most new mining buildings they finished last year for the jump from 30 to 50 exahash. And there's a smaller laydown yard just north of the block three, that is the 150 megawatt that they use to get to 30x hash. So I managed to obtain some data from, well, actually some images from what's actually going on there. And it appears that Iron's been hoarding long lead time items and materials that are currently also used in Horizon one to four. So it's very striking that they are making these lay down yards very far away from Horizon one to four, but with the same materials that are being used in Horizon one to four. So we've been kind of like guessing and anticipating something happening with the mining buildings Are they going to demolish them? Are they going to retrofit them? We've seen the renderings from Iron about converting the remaining 450 megawatt to liquid cooled capacity. And I think what we're seeing right now is a preparation of the site to start this conversion. So I've been a very, I've been holding back with talking about a deal for Horizon 5 to 10 in the last couple of weeks. But after seeing this new development, I am starting to feel like maybe Microsoft is going to take the whole site. I know that I've been kind of adamant that it wouldn't be Microsoft because As far as I know, Microsoft didn't get any options. They didn't get any dips on Horizon 5 to 10 or anything like that. I know they wanted it, but I'm pretty sure they didn't get it. And so, you know, now to see that Iron's going ahead and ordering all these equipments like I'm literally looking at these like these cylinder kind of tanks. where that they were also using for Horizon 1 to 4, so liquid cooling, you know, those liquid cooling repackaged plants. So it's it's starting to look like we are really going to see the conversion of Childers to Horizon 1 to 10, like they're they're going to do a full liquid cooled conversion. which is kind of odd because I know that during the earnings call, there was a lot of talk about air-cooled capacity and how air cooling was now all of a sudden interesting again. But I am still convinced that the 810 megawatt of operational data center capacity was just a number they threw around. I think that they are in advanced negotiations with Microsoft, and I think that what we're going to see at Childress is just going to be a copy paste of Horizon, you know, all the way to Horizon 10. That's my personal expectation right now. It wouldn't make any sense to me to see all those long lead time items be piled up there if it was not for the same customer, because this is a very capital intensive project and I would assume that only the Microsoft would be, you know, willing to have these same kind of materials like stainless steel piping inside the data center and, you know, an overkill of rack density and cooling capacity. It just makes a lot of sense that Microsoft is the one who is now, you know, has done all the due diligence for this design. who is okay on the shelves as they are so far at Horizon One. And I think that there is like some kind of a, you know, not a deal, but like an understanding or sort of agreement that if nothing changes, we just keep going, something like that, if that makes sense. So it appears to me that that's what's going on right now. I mean, it just, it really looks like that. So, I mean, I could be wrong. It could still be that all these long lead time items and these components are just for horizon three and four, maybe, but it doesn't really make sense to, you know, to put them down all the way on the other side of the site, very close to the mining buildings. It looks more like it's... being prepared in anticipation of converting the rest of the site to liquid cooled capacity. But that's just something I observed. And of course it's not doesn't mean that the deal is imminent or anything. But I would I would still say that if Iron is able to deliver Horizon One operationally with ships and everything and it's working the way it was intended, I think it would make a lot of sense for Microsoft to, you know, stick with Iron for this site. It just, you know, I was very much considering a different option earlier on, but now I'm starting to see that maybe it just makes a lot more sense for Microsoft to take the entire capacity. And from what I'm hearing, Microsoft is very content with the progress and the pace. Another thing to mention about Childress and Horizon is that it appears that Horizon one had a slightly different design than Horizon two to four. So we've observed some difference in the liquid cooled plant that has the fans now on top instead of on the bottom. So it's kind of weird. But it appears that Horizon 1 was kind of like the test tube baby. And now they've kind of figured out how to do it really well. And I think this also means that they're going to increase the pace of the construction. So I think maybe a lot of people were like oh but they said they would uh deliver Horizon one by December blah blah blah and now it's still not ready and uh you know the revenue is expected to flow in the next quarter so um I I kind of what I'm hearing is it's going to be May or June but I think that after that you're going to see like a it's going to go hockey stick up so you're going to see Horizon two three four going to uh you know, being built out staggered. And I think it's going to go faster. And I think that's going to keep on going for Horizon 5 to 10 as well. So I think they kind of like have a slow startup phase like with Horizon 1, but I think after that it's going to go a lot quicker. And I, you know, I think that also you know, what we're hearing about Microsoft's reactions to the project, how fast it's really going already. I think it all bodes very well for a new deal for the remainder of the capacity. And yeah, so it would be great if we know more, if we hear more about the upgrade for Horizon 3 to 4 to Fira Rubin, because I think that could also mean that maybe Horizon 5 to 10 wouldn't have to be flexible rack densities anymore. So that could also potentially save some CapEx and maybe there won't need to be any super clusters for Horizon 5 to 10. So it could still be a cheaper build with a better chip, with better margins. So all in all, I am kind of hopeful that There will be something released around this in the coming weeks or months. At the end of the day, I do believe they're going to tear down these buildings, the mining buildings, because I don't think they're suitable for a full retrofit. But that's just my personal belief. I don't have any proof for that. But let's see how it develops. But I think we should keep a close eye on Childress because After being a bit dismissive about the next deal is going to be Horizon 5 to 10, no, I was saying it's going to be Sweetwater. Now I'm sort of flip-flopping back to the idea that it could be children's after all, because of what we're seeing on the ground. So that's just something I wanted to share.

₿itcoin ₿utcher: Franz, what was your understanding? Can we speak more to The so part of the conversation, the discourse with God particle has to do with we're not seeing it in the financials yet and that primarily related to if I had to guess the delivery of GPUs and you had spoken previously to potentially seeking a third party outside of Dell to contract GPUs. Can you speak a little to that dynamic on installs, what your understanding is, if you've had any contact with any sources on if that is being, if those delays are being worked on or if there's been improvement? And then the second piece as a follow up related to Rubin, if we're going to provide infrastructure as a service, by definition, we're buying it, the chip itself. But at some point, it feels like Microsoft has to come to bat for iron. And it feels like, and again, I'm speculating, but Dan on the call had spoken to how people assume that parties like Nvidia haven't sought out iron when the opposite's most likely true with respect to his quote, that Nvidia probably has sought a piece of equity out of iron, or at least it has entertained it and presented an option to Dan and Will, but they haven't followed through with it for, I'm sure, good reason. Having said that, if you're not under Nvidia's tent, and Vera Rubin is the most efficient GPU in the world, how are we able to deliver that chip to the marketplace if we're not on their good side? And I think devil's advocate on that is how many people have the racks ready to fit Vera Rubin. So that might be the answer is the technical prowess, but I just want to hear your thoughts on that. rambling and brainstorming out loud. But I think where my impatience kicks in is not so much new deal, but the existing infrastructure that we have. In some ways, they're on track. But if we're being honest, I don't think they met guidance at the end of the year. I think they've shown enough progress to suggest they'll meet the 500 million ARR by the end of March 31st. But if you have any commentary related to those two thoughts, I'm sure everyone would appreciate it.

Frans Bakker: Yeah, so I think the guidance is a little bit distorted because first they said 225 million of ARR by the end of December and then, but they also said they're going to deliver Horizon by the end of December. Then they said 500 million by the end of Q1. I think the the contracted revenue and generated revenue is there is a little delay between that. And I think it's the word delay is striking because that's exactly what's happening. I think there they are, you know, the GPUs are just arriving too late. And I think that's what's causing this. That's also what we heard in the previous earnings call in November when they said that the GPUs that aren't contracted yet are the ones that are the furthest out on the delivery scope. So what we're, and then remember in the filing they said that they would expect the first part, like the first, what was it, 12,000 or 13,000 GPUs. to arrive, oh no, sorry, the first 9,000 to arrive in the in before the end of December and then the remainder 12,000 something by the end of Q1. So the end of Q1 is March 31st. So we're still not there. But judging on the wording they used in the last earnings call, I would say they were talking about the coming weeks. I think it would make a lot of sense for them to receive all the GPUs that they have purchased so far under the Dell contract, or I don't know if they actually said it was coming from Dell, but I think so. For Prince George, I think it would make a lot of sense that they are installed in the coming week. So let's say by the middle of March or something. I personally, I'm going with an expected quarterly revenue of somewhere between 60 and $80 million for this quarter, which is, you know, not $100 million as it should be if $400 million of ARR was actually operating the entire quarter. But I think they're going to, you know, install a bit more than $500 million of ARR generating GPUs in this quarter. So it should be somewhere between 60 and $80 million in my opinion. If they are going to absolutely fumble the football on that and it's going to be below $50 million of generated AI revenue in this quarter, I think they really have to explain the market. Like just be upfront and say, yeah, look, we were expecting the GPUs to arrive in February and March, but they all arrived in the end of March and early April. So give us a break. We have them now, you know, but they're not very clear in their language. I made this post about that in November last year. where I said, look, all the GPUs are going to arrive a lot later. And I got a lot of backlash from people that said I had to delete that post because it was too bearish, but it's just turning out to be true. There's a severe issue with the delivery of these GPUs, and they are all arriving late, and it's causing us to have a delay in actual generated revenue. So to your second point, I'm happy to inform you guys that Iron is actually procuring GPUs from a different source now. I'm not going to tell you what source it is, but it is actually another Neo Cloud. So that is, that shows that Iron is aware of the issue and that they're fixing it. They're stepping up and they're buying GPUs elsewhere because, you know, there is a lot of, there are a lot of GPUs on the market. There are a lot of companies that have, you know, inventory that's gathering dust, if you remember correctly, even Microsoft said it. They have GPUs in their inventory that they can't put anywhere. Well, Iron has a different problem. They are a preferred supplier at Nvidia and they work with Dell and a bunch of other vendors, but those guys are all servicing their biggest clients first. And that means Iron is, you know, they are preferred, but they're not prioritized, so to say. So I guess that's why they are now buying their GPUs elsewhere. And I think that's going to be a positive surprise to the market because that's another thing that you're going to see. It's not just going to be the known GPUs that are going to be installed in this quarter, but I think during the next earnings call, we're also going to find out that Iron has procured other GPUs that are going to arrive a lot quicker because they're already you know, in inventory somewhere nearby. So, you know, I think we just have to sit out this year. Yeah, it's kind of like, I know that's very hard for us, impatient and we are very hungry for information every day, but You know, there's a reason why Iron is not updating the market monthly anymore, because there's just not that much to inform because these things just take time. But I think we're building towards an inflection point where all these things are going to, you know, come together at the same time, where Prince George is going to be generating 600 plus million dollars of annualized revenue. and the new GPUs are going to be installed in McKenzie, and you're going to get a deal for Kennel Flats, and Microsoft's going to get delivered. You know, the market is still thinking that it's all going to fail and Iron won't be able to execute. Even now when they de-risk the financing, the whole funding is de-risked, basically. The market's still not interested because we haven't shown the revenue. So I think someone in only France made a good comparison between Nibius and Iron. The market thinks Nibius glass is half full and it thinks Iron's glass is half empty. And I think that is basically The problem here is our Bitcoin history. I think that is the reason why the market thinks we're a bunch of criminals laundering money. And that's why I hope that we can get rid of this stigma and drop the whole Bitcoin mining label in this calendar year. And then, yeah, I think the stock will get rerated once we're going to show that we're for real with AI. Unfortunately, having a Microsoft as a partner and, you know, getting $400 million of annualized revenue in Canada, it's not enough to prove to the market that we're no longer a Bitcoin miner. But I think it may not be the next earnings where the actual revenue will show, but it's certainly going to be one of the earnings this year where it's all going to show that everything we've known since last year is all happening. And yeah, as investors, we just have to be patient and let the team execute. And this is all outside of any deal. So yeah, that's my two cents on this whole topic.

₿itcoin ₿utcher: That's encouraging with the procurement of GPUs. If I was speculating, if it is an outside party, I think about the companies that are struggling with, I would guess it's Oracle or Coreweave personally because of their balance sheet problems, but I obviously don't want you to comment on that. But both companies have struggled with debt perception and have had to raise funds to operate more recently via, I think, Core, we've had the additional equity capital from Nvidia. And then on top of that, you got Oracle that is going through a multitude of different debt offerings and equity offerings, again, to share up their balance sheet. They both have these huge RPOs, remaining performance obligations, but they also have ****** balance sheets. And I think that kind of to circle back to my original point, I have a lot of faith. When I hear Anthony Lewis talk on the earnings call, I feel really good about where we stand, where they're going to scale at the right pace. And if you're driving your car down the street, I could give you 100 analogies, but not going over 3,000 RPMs or maxing out the engine, having it fall out. It's kind of the same thing where it feels like we're slowly even if it's not fast enough for Bitcoin AI guy and the rest of us asking one deal, we want to do something that's sustainable and slowly pick up speed. And before you know it, think about too, like if you ran it, I think I could look back. I wish Comma was on. Franz would be able to speak to it well, but they've been here since. the company was below, even though we hate mining right now because Bitcoin is in the doghouse, it's a very good use case. to show how they scaled their business. Like for the longest time, they had one or two exahash, and then they built up to 10, and then from 10 to 20 to 20 to 30, and then 30 became 50. And if you look at the growth in exahash on a chart, it would reflect that kind of parabolic growth that Franz has spoken to with the potential of Horizon being built out and being solely for Microsoft, like think how long Horizon One took. And he's saying if it's handed over in May, fine, it's there, but they got rid of all the imperfections and now they have the right processes in place to almost no different than having a 3D printer and just start printing data centers. And I think too, Sweetwater. There's a lot of thought being made into how do we if Brian Fry will often say that iron is the AI factory of the future. Well, the only way to produce at scale is to have a process that you can replicate, and that's where the modular design comes in. But Bitcoin AI guys usually are pretty. Instead of saying one deal and all the comments, if you got something to say, you're free to come up right now. Small cap, I think, was still out to dinner. I don't think you had much else to add right now. Franz, you're obviously free to talk. I see marbles in the crowd.

Frans Bakker: What about these backlogs? We keep talking about backlogs. But with the current development of the hopper prices, don't we think it's better to, you know, strategically sign your deals rather than locking everything in at current economics? I mean, it just, it's kind of absurd to know that Iron has sold the Microsoft contract for $2.91 per GPU hour for the GB300 that's not arrived yet. but hoppers are going for $2.40 right now. So, you know, the spread between these two generations is pretty small. I know there is a lot more to it, but there are some people that want Iron to contract all their capacity and, you know, be just like Nibius that says, yeah, we're sold out for the quarter, we're sold out for next quarter. But that means that you're signing for periods of time under current economics. And Iron lives by the philosophy that they know how scarce the power is and they know why they own it. And they take their sweet time to make a deal because they know it's only going to get worse. Well, if we follow the pricing of the hoppers, you know, there was a At first, when it got released, it went up to like $8 per GPU hour. Then the market got flooded and the prices got absolutely destroyed. And now it turns out that there's actually a shortage of data centers. And now these older models are going up in price because the the new models that are being trained need so much more tokens. So on AgroGrad, you need a lot more older models to be able to fulfill, you know, the amount of tokens that are needed to build these new models. So basically what you're seeing here is the power crunch in effect through the shortage of data centers. which is having its effects on the prices of the GPU. So I think what we're seeing with Iron, they are taking their time, but I think it's also going to mean that the mix of GPU hour across the complete portfolio is going to go up. And this first 10% that we had with Microsoft, looking back, you can see, well, the deal was pretty bad maybe compared to if you would sign it today. But, you know, if they wouldn't have signed it four months ago, you wouldn't be having, you know, all the prepayments, all the financing tools, and you wouldn't have the GPUs as well, by the way. That's a bit of the other side of the coin is, okay, we also got the GB300s for a lower price. So there is a thing to say about signing earlier to get the chips cheaper. But let's say Iron is able to procure these GPUs from a different source. for, you know, for market prices or below market price, but they are able to sign the deal for economics that are better than the market right now. You know, I would hope that Iron, by building data centers before they actually sign the deals, is able to have some arbitrage on the market where some of these some of these other AI cloud providers, they hoard all these GPUs, and then they are going to look for data center capacity co-located. And some of these other data center infrastructure operators are building data centers, but they're delivering the data centers in 2027, 2028, and '29. Iron is a bit in the sweet spot where they have the data centers in '25, '26. And they can just buy the, you know, scoop the GPUs up from the players that bought too many, but don't have anywhere to put them. But they can put them in the data centers in 26 and get the elevated economics under the shortage of data center capacity and power. So I think we will be surprised by the next set of deals that Iron will sign. I think it's going to show that this first anchor deal was not a template of their economics for deals. I think it's going to look very different. And I think waiting and not building up a backlog right now is a lot better than selling out your entire capacity and start showing revenue to the market. I think I think it does help stock in short term to do it that way. And I think there's nothing wrong with what Nibius is doing. I think this is one way to do it. And I think that what Nibius said during the earnings call that some of their contracts just expired and they're going to upgrade them to new pricing tiers, that's all very good and well. But I really hope that You know, for example, the Finland deal for Nibius, I think that was all of their liquid cooled capacity in the United States for 26 and maybe 27 outside of maybe one small data center they have somewhere in Kansas, I believe. But that was only 40 megawatt, I think. But so Iron has only signed away a portion of their data center capacity of 2026 and 2027 in the US. So I think, you know, I think we're going to have to wait until the next deal, but I think it's, it's going to improve a lot. And I think the elevated prices of the older GPUs are going to help with that. It's just really a matter of finding out what the prices are going to be for the Ferro Rubin from from Nvidia. I I'm not really so skeptical that I think Jensen is just going to increase the price to match the token output. That would mean that Umbisom is right and there is never going to be any margin for AI cloud providers. I think it's going to be a bit different than that, but I guess we'll have to find out. But I'm I'm optimistic about the next deal's economics. I think everything is showing that the Robots Brothers were right about the constraint of power, the time to power, and now the time to data center. So yeah, that's just what I wanted to add. And I think there are exciting times ahead, but we're just gonna have to be patient for a bit longer.

₿itcoin ₿utcher: Yeah, I think the only, to be clear, Agree with you. But I would say the counter argument to that, while everyone's keeping themselves honest, is when you look at the graphic from the earnings call and assuming even though it's not all fully contracted yet, but the 3.4 billion of ARR guidance is again for 300 megawatts. at Childress and 160 megawatts in Canada, which is 460 megawatts total. And then after the inclusion of the 1.6 megawatts added to the legacy 2.91, that's 4.51 gigawatts of power. So one 460 over that 45, you know, roughly 10%, a little over that. At the same time, you could argue there's an element of waiting and not and getting a better price given the direction of the market. But there's also, if you have this monopoly of power, if you're still making money on it, I do think there's something to be said about creating the reputation in the market and making multiple parties happy and creating more bidding power in the future. I know they could sign the full amount tomorrow, but at what price? And I think they would command an even higher price if they get Google at Sweetwater or Anthropic or whoever these parties are, even if they don't get the best deal and yes, the directional price is going up over the long term and they could sign it for more, they still have a perception as a Bitcoin miner. And I think the only way to turn that is by doing a deal now, even if the terms aren't optimal. We've seen that with Microsoft, how many new phone calls inbound came in and the upgraded pricing that flowed to Canada. I still think one more deal, they had their zero to one moment, but one to two is still doubling that. And I think by having that next big hyperscaler, big enterprise deal, that the market perception will flip, and we'll have that inflection point that you're speaking to. We want to see the revenue flow to the income statement for sure. But I do think that signing a party of the caliber that I've mentioned with some of those names would really help catalyze the share price right now, which It's not everything, but if they plan on using convertible notes in the future to fund this business, which I'm pretty sure they still plan on doing, at some point they have to play to the market and have something, not something that compromises their future. I'm not saying that, but you acknowledged earlier, Franz, and I feel this very strongly, that failure to execute a deal, if not two deals by may with what the tools that they have in place now, I do think just the share price would get clobbered, and it will create a lot of mass fallout. It's not something that-- you have to entertain it. We have to say it out loud. I have no reason to believe that they're not ready to. But at the same time, the reason he mentioned Oklahoma on that last call and the reason that he mentioned the deal that they're negotiating right now for the software piece is because he knew that they need to provide something to the market. I think they're pretty close and when we're the most frustrated is probably the day after that they're going to release something in the morning. So those were just some additional thoughts. It's really encouraging to see hopper pricing, as you said, at 230 or 240 an hour, whatever the metric is compared to what we signed Microsoft for. But we've already been down the Microsoft road. I thought it was necessary to get our foot in the door, make a hyperscaler happy. and make another hyperscaler jealous and wanna do business with us. So that's all I wanted to add.

Frans Bakker: Well, if I look at this Sweetwater picture right now that we just got from, it was captured on Sunday. Yeah, you guys are still in Sunday here. It's a Monday already. So I guess it was captured somewhere in the PM of Sunday. It just looks, it looks like they are moving towards data center construction. The site is just advancing at such a rapid pace in multiple directions on two sides of the road that they are working on. It would be a surprise to me if they would not be in, you know, final negotiations around a deal. for a particular capacity of this site because, you know, maybe they are waiting for the final confirmation of this batch zero thing in order to actually, you know, sign the escrow. But I think it's something in very final stages because there's just not much more to do. once you have a primary substation energized and, you know, the site is, you need to start building, you know, earthworks and cabling and all these things in order to actually start building data center shells. And there's just not much left to do after you energize and you already have like, there is a primary substation that's already there and they can just fire it up and then what do you do? You know, so depending on the kind of deal they're making and when it's going to be energized and delivered, that also will define what kind of chip they're going with and that will define the architecture of, you know, the electrical infrastructure for these shells. So I think there's not much to do. If Dan is so adamant about time to data center, they're going to have to start building data centers. And in order to do that, they need to sign a deal. So I think this can't drag on for months. I will agree with you that if by next earnings they don't have anything for Sweetwater, I think that they are, yeah, that they're making a mistake. But I guess that's all I can say for now. I will have to analyze the picture better to see what's actually going on, but I don't have the enhanced version yet. But it looks primed and ready to go, in my opinion.

₿itcoin ₿utcher: Great. That's what... That helps. That helps probably a lot of people have a little more confidence. I just don't... I, seeing the work that was already done there, that's more than just prep work. That felt more analogous to last year when you showed photos of the Barbara Lake site in parallel for Cipher. It was a little tougher for Horizon 'cause they were kind of building it and build it and they'll come, whereas like Barbara Lake specifically, once those things were in motion. It felt like they were close to signing. And then a few weeks later, you figure out that it was Google and FluidStack. So yeah, I would personally be really ****** if we got to May and we're a month into energization of the largest unicorn site, arguably in the world, and they couldn't figure out the right deal after his own words, 24 months of negotiating and or discussions like something's got to give and the commentary again is I'm just reiterating points that you made Frans. But if time to DC is so important to a hyperscaler or a larger enterprise, I have a really hard time believing that they can't agree on economics. So that's good to hear and I look forward to seeing that enhanced photo with the help of our friends and only friends. I don't think we had any other topics per se, and I'm not going to keep this open for two hours for the sake of doing it. So consider this a last call to anyone else who had questions for Franz or myself, or if they wanted to discuss anything in particular. I saw Marbles was in the crowd. If I had a question for Marbles, because I always enjoy his view, if he's able to join and come up, I would be, I think what I would find for everyone else, I'm going to try and invite him up right now and see if he can answer the question that I have. As it relates, though, I don't have the statistics in front of me, but if I had to guess, it will turn into an 80/20, where the vast majority of use cases for AI in the future will be inferenced. And if that's true, I would like to ask Marbles, as it relates to the legacy GPUs, I think Iron I'll frame the question this way. I think Iron benefits from, we talked about flexible rack densities at Horizon, giving them the opportunity to feature Vera Rubin. Dolce has famously said Iron will succeed because they try and do the hardest things. And I think there's truth to that. as well, that they need to be at the forefront, no different than Nvidia's on the forefront of chips, and keep pressing so that they maintain a lead over their competition. All that being said, we see hopper prices where they're at right now. And Franz has alluded to multiple parties sitting on GPUs. If you were running the company, Marvels, I would want to hear your take on, do you think there's an advantage for Dan and Ko to just take advantage of the upswing in inference that is being driven by use of OpenAI and Claude by just leaning into those legacy chips? Or do you want to see them hold out their space for Vera Rubin? But if you have any thoughts on that, I'm sure you've thought about this before. Everyone would like to hear from you, given your technical background. Hey, Marbles, I can't hear you that well. Maybe it's just me, but it might be your mic.

Marbles: Is this better? Hello? Wait, yeah, I had my earphones in and I think the mic is broken, but. Anyways, yeah, so to clarify, Butcher, when you say lean into the older generation chips, are you referring to H100s and purchasing more of those, plugging them into DCs?

₿itcoin ₿utcher: Could be anything that's not GB300 or later. So GB300 and Vera Rubin being the most recent top of the line on the market, my understanding. If there's something else, feel free to mention it. But those tend to be used for training would be my understanding. But if the market is leaning heavier towards inference and there is value to time to DC right now, is there something to be said about buying these lower cost GPUs? And since the spot rate is so high, someone mentioned B300 or B200, H100 or H200, is there an advantage to buying those given the spot rates right now just to generate cash flow? And is there Also, given the high bandwidth memory shortage right now, it appears the tech obsolescence, if they're able to de-risk it and lease these out and pay them back in two or three years, is there something to be said for that? Are you happy with what they're doing right now?

Marbles: So yeah, thanks for the question. First of all, I'm no expert. Obviously, I trust management. But in my view, I think it really depends on where your view on demand is, right? If you think demand is so insatiable, and no matter what people do in the real world, no matter how much they try to do in terms of increasing the supply of power and data centers, if you believe that there's no way they're going to catch up in the next few years, then yeah, I think for sure you should just, to bridge that gap, I think for sure you should invest more in B300s or B200s, which I think Ireland is already doing, right? They're already doing that in their Canadian data centers. So I think in the next few weeks, some months, we're probably going to see higher prices for their GPU hours, I suspect, based on what Franz is saying. So in a way, I think they're doing that. And I think, look, I'm not an expert. So again, I trust management. I defer to them on exactly how they think about this from a strategic standpoint. But I think that the important thing is everything that we're seeing in the market is that demand is increasing. And for example, the H100 price going up and the spread between the H100 GPU hour and the GB300 GPR is closing. And it just goes to show you how hard it is to get the GP300s online and into a data center because of the enormous amount of rack density. So I think that's incredibly bullish because a year ago, we were seeing the H100 saturate the market and the GPU hours drop substantially. I think what's happening right now is that there has been an exponential improvement in all these models, like Cloud Opus 4.5, GPU, sorry, OpenAI Codex 5.3. I don't think anyone saw this coming. It's just been something that's been happening in the last few months. And you can see that token consumption is increasing at an incredible rate. So I just anecdotally from my job, I've mentioned this before, I don't really need to code. I just give it to Opus and it does everything for me. And all my friends, they don't really code anymore either, right? And if you just look at all these coding interviews for software engineers, People don't, you see a lot of companies transition away from something called LeetCode, which is basically just a pattern recognition test. Because really, if you just have AI now, there's no point, there's no point in testing someone that way because, you know, the best barometer of whether someone's gonna be successful now is how well they can leverage AI or these models. So I think you're just seeing incredible, incredible innovation happen so quickly. And I think what I'm observing is, and what I'm seeing in OnlyFrans is, There's a small cohort or maybe a large cohort of people constantly complaining about, look, if we don't get a deal now, we're screwed. And I think that's not really the right viewpoint because I think that means you believe that supply is going to catch up the demand pretty quick, maybe in the next two years. And my view is that every single time one of these models improves, there's no way Anthropic or OpenAI can service any of this demand. There's just no way. I mean, you just see it in the H100 price. You see it in how they throttle the models. Right? And I think how, if you think about how existential it is, it's like if, because Anthrop actually throttles opens a lot, which means they kill the model, they make it less capable. So they can, A, decrease demand in the short term or they can service more demand because it requires less tokens. But they can't do that too much. Because if you do that too much and OpenAI never throttles their models, then it's really, there's not really a high amount of switching costs. You can just switch to another model or you can just download codecs and pay for that and never be throttled. So it's pretty existential for all these companies. So I think, look, that's kind of a roundabout way of answering your question, but I think maybe we lucked into it. Maybe Dan and Will, they didn't plan this way, but I'm pretty sure they did. I think I agree with Franz here. I think that by kind of playing it slow, and we're into a space where demand is so insatiable, I think this is going to benefit the company in long-term with higher GPU per hour price. And the great thing is we have so much power, we can monetize it in increments, unlike Nebius, where all of it has been monetized at end of 2025, beginning of 2026 rates, prior to Opus, prior to Codex, prior to these amazing models we now have. And these things will Actually, these things are-- these models will only improve because we still haven't got a black well-trained model. Once that comes out, imagine the new amount of demand that's going to surface because of that. So I tend to agree with Franz here. And I think we should all be incredibly bullish and just be patient. I trust management. I think that one year from now, we're all going to be very happy shareholders.

₿itcoin ₿utcher: Thanks for coming up, Marbles. You always You're a man in the field. Tomorrow, I'm going to be playing with a bandsaw and a breaking knife. And as much as I read about this stuff, it's a little different than someone who's actually-- even if the model is doing it for you now, you're still interacting more with AI than I can. So I always like to talk to someone in the field as opposed to making new avatars on my phone. whatnot. So thank you for coming up tonight. Franz, do you got any other comments? Otherwise, I'm probably going to move to close this unless someone else wants to come up last minute. I'll take that as a no from Franz. Marbles, do you have any closing thoughts? I think you hit it on the head that one year now we'll be from HAPI. happier than we are today. But is there a middle ground to what I was saying earlier that-- do you think it's unreasonable for people to have some expectation by earnings in May to show a little more progress as opposed to-- personally, I think it would be a disappointment to only have the Oklahoma announcement and have $500 million ARR by the end of the next quarter. I don't think that's what's going to happen, but especially after the evidence submitted by Franz tonight with the progress of not only Sweetwater, but Horizon. But you have any thoughts on that the next few weeks or months?

Marbles: Yeah, no, I don't. I don't think that's unreasonable, but I do think that Iron can only build so many data centers, right? Franz mentioned in Childress. It's really hard to build Horizon 5 to 10 and Horizon 1 through 4 in parallel. And I think at Sweetwater, you can only really build 200 megawatt increments with so many people. So I think lastly, you can also only build so many data centers so quickly in Canada. So they're constrained by the physical world. So I get why they're taking it slow. They want to assign the best deal possible because Yeah, I mean, even though you have all this power, it sort of doesn't matter at this point because you can only build so fast. So I think that's point A. I think point B, yes, I think if they dilly-dally and take too long, that might be a red flag because in this crazy demand environment, if you can't sign a deal either, that means management is being too aggressive and with negotiating, maybe they should just be a little bit more or chill, so to speak. I don't know, I don't have full visibility, but I don't think it's unreasonable, Butcher. But on the same note, I think that they're aware they can only build so many data centers. And if that's the case, if you get a deal, you want the best deal possible. Because once you're locked in for 200 megawatts, those 200 megawatts aren't going to be monetized for another year. And this is kind of what's happened for Saffron Wolf. Like Cipher signed away all their megawatts. Imagine if they waited like another few months, they could probably get better better economics on a new deal if they signed it now. They signed AWS and Barbara Lake. Barbara Lake and Black Pearl pretty close to each other. Imagine if they waited a bit longer for Black Pearl and then signed it now. So I think I get why they're being patient. And I also think that, yes, you can't take forever. But I think the other point is you said that, look, they need to sign a deal so they can legitimize themselves in the market. I think that's fair, but my question is, why can't delivering on the Microsoft deal, generating revenue, delivering on Horizon 1 through 4, why can't that be enough for the market? That shows the market that they're not just a Bitcoin miner, they can deliver at scale. They'll be one of the first companies in the world to deliver 200 megawatt data centers that are full of black holes.

₿itcoin ₿utcher: I'll tell you why. The reason why that's not good enough, it addresses the execution risk. They've addressed the financial risk on their balance sheet, but it won't appease the people that got scared away by the commentary of the Chainoses of the world. And otherwise, it's just anecdotal. Until we see it flow through to the, we have no visibility. We can trust that management is saying that they're securing higher rates. But if you have that mega deal with higher than $2.91 an hour, which I have every reason to believe that's coming, I think that right there gets the bears to, quite frankly, shut the **** **. You cannot tell me that handing it over for $2.91 an hour should make everyone happy because the conservative operating assumption at that point is that they're going to do a similar deal, even if the evidence suggest otherwise, and we have every reason to believe it. There's something to be said about that headline and being able to plug it in and show higher than a high single digit or low double digit unlevered IRR. There will be power and saying to the market that we were right, that first deal was just getting our foot in the door. Now we actually have pricing power and can operate as we see fit, and we can be more selective because we've earned the right to that. So that would be my personal pushback on it, Marbles, just to play devil's advocate. Franz, I appreciate you coming up this week. I know you have family obligations and make yourself available to everyone. Again, it's better than the meat cutter talking about AI and satellite imagery from your subscription group. So thanks for making time for us this week as always. Marbles, thanks for pinch hitting here and speaking to your industry experience. I'm going to close things up for the week, guys, and I appreciate everyone's time. I really do believe, despite some of my comments tonight, especially that last one, voicing a little of my objection to taking too much time, there's a middle ground, and I write a post about being patient, yet I just told everyone that I want to tell the Bears to **** ***. So I think we're going to get what we want here in the next few weeks. and definitely by Q2 earnings. So I look forward to that. I appreciate everyone's time tonight and we'll talk soon. Have a great night.