Transcript 0:00 On y'all. Welcome back to Blockspace Live, presented by CleanSpark, for another beautiful day of earnings week, phase two, Charlie. And man, do we have some news for y'all today. 0:11 Riot dropped its Q2 earnings last night, and they dropped two bombshells. 0:16 First, a new AI lease, reportedly with Anthropic, we'll get into that in a second, at its Rockdale site for 191 critical IT megawatts in a multi-billion dollar deal. 0:31 Plus, an LOI has been inked for its Corsicana site, the entire gigawatt site gross. Will they get it across the finish line? 0:44 We don't know, but we do have the deal details as well as their Q2 results, so we'll be tackling that for our first story. Following that, NVIDIA talking about $500 billion in financing. 0:58 New $500 billion financing coming from NVIDIA. Just when you think that the CapEx bubble can't keep inflating, they keep throwing dollars around like it's nothing. Second story coming up on that. 1:09 For our third story, we have requests from ERCOT to move deadlines for parts of the Batch Zero process to accommodate Greg Abbott's data center audit. 1:21 They need PUC approval, and they are trying to move the deadlines for a few things. But crucially, they're not actually moving the April 9th, 2027, deadline for final deliberations on the Batch Zero study. 1:33 We will cover that for our third story. 1:35 And then for our interview today, to cap off the stream, we have none other than CEO of AnkerWatch, Rob Hamilton, on to talk about his experience with Red Team, putting out fires following the ColdCard exploit, or rather, I should say, doing some due diligence to make sure the fires don't actually start in the first place. 1:53 He's got some great insights here, including why linear algebra might be the new cryptography, because we're talking about banning it with American AI models. 2:03 He's run into some stonewalling with their ability to actually audit these code bases for open source software. Rob's coming with receipts for what has worked and the frustrations he's encountered along the way. 2:15 Blockspace goes live every single weekday at 1:00 PM Eastern. We are compute's daily live show, live streaming every single day at 1:00 PM Eastern. 2:26 If you miss the live stream, you can catch the podcast later everywhere podcasts are found. And if you can't get enough of the audio and video, we write a lot. You can find all of our content at blockspace.media. 2:42 The website is blockspace.media. This show is brought to you by CleanSpark, NASDAQ listed ticker CLSK. More on CleanSpark later on. And Colin, it's week two, day two of earnings week. Let's kick it off. 2:59 [beeping] Yes, earnings week direct straight to your feed. 3:07 Uh, before, uh, Colin does [chuckles] a review of Riot, which was last night, uh, which if you tuned into our watch party, uh, we're, we're gonna review it, let me show you, uh, what's on the docket right now. 3:19 Earnings week two. Monday down was Keel, Bitdeer, Riot. Today, CoreWeave this afternoon. Um, a little tweak to this as we originally had, I know some people are tuning in just to see Soluna CEO John Belzer today. 3:36 He had to postpone. So we'll run it back later. Don't worry. We have the CFO of Soluna, Mike Pici, on Friday. That'll be after their earnings call, which is the Saturday. 3:47 So today, Tuesday, CoreWeave earnings call, uh, watch party later. Tomorrow, Wednesday, Sam Tabar of, uh, White Fiber, Brandon McBee of CoreWeave, Nevius watch party, White Fiber watch party. 4:00 Thursday, Ben Gagnon, CEO and captain of Keel Infrastructure, as well as the Soluna watch party. And again, Friday, Mike Pici of Soluna. So all that said, let's actually dive into the analysis. Colin, 4:18 Riot was yesterday, they [chuckles] announced a big-ass deal. A big-ass deal and some breadcrumbs for what could be a larger deal around the corner. It's the LOI for all of Corsicana. No deal yet. 4:32 Have to get that out in front. They were very clear about that on the earnings call. In fact, they didn't even put it on the press release or mention it in the 10-Q for their earnings. But that being said, uh, a huge 4:44 announcement for Riot and just a nice feather in their cap for Q2 earnings. We talked about it yesterday on the stream. 4:53 I had a hunch that they were delaying because they were getting a deal over the finish line, and in fact, that's exactly what happened. So here's the headline from Blockspace. 5:00 "Riot Platform Signs $9.1 Billion 191 Megawatt AI Lease at Rockdale, Signs LOI for 1 Gigawatt Corsicana s- Corsicana Site Q2 Earnings." 5:12 I will go ahead and go over some of the headline numbers 'cause the, uh, Riot actually broke out a data center segment for its revenue for the first time ever, I believe. 5:19 [lip smack] They might have done it last quarter, but, um, this was the first one that it was consequential. Then we'll get into the deal specifically 'cause I know y'all are probably more interested in that. 5:28 Headline numbers for Q2, total revenue, $174.2 million, up 14% from $153 in Q2 2025. Bitcoin mining revenue stood at 113.7, down 19% from Q2 2025. 5:44 Data center revenue, $23.2 million. Gonna break that out though, because it makes it sound like that's all lease revenue, and it's not all lease revenue. 5:53 The components of that are operating re- leased revenue at 4.542 million, variable lease revenue, which includes power reimbursements, at 353,000. This was the lion's share here. 6:06 Tenant fit-out reimbursement revenue, 18.318 million. 6:10 So that is Riot ordering equipment for A&D for their build-out and- And getting reimbursed for some of that equipment and installing it for them, earning a little bit of a cut on top of that. So [lip smack] 6:22 operating loss, two hundred and thirty-nine point four million. Net loss, two hundred and thirty-seven point two million. Earnings per share diluted, minus sixty-eight cents per share. 6:31 Adjusted EBITDA coming in at negative sixty-nine point seven million. Uh, H1 net loss was seven hundred and thirty-seven point six million. Seems really bad. 6:43 That being said, the loss is overwhelmingly non-cash, ninety-seven, almost a hundred million, ninety-seven point eight million in depreciation and amortization. 6:51 Seventy point six million negative mark on Bitcoin mark to market. Thirty-five point six million in stock-based comp and twenty-eight million in impairment. The impairment line is worth mentioning. 7:02 Riot wrote off twenty-eight million of long lead items sitting in construction in progress for a planned Bitcoin mining expansion at Rockdale. But that's no longer gonna be a Bitcoin mine. That's gonna be a data center. 7:14 Now, they do have a little slice left over for Bitcoin mining when you factor in gross megawatts at the site. I believe Rockdale right now is seven hundred megawatts. 7:24 When you look at gross megawatts for these deals, they have somewhere between a hundred and two hundred megawatts left over for Bitcoin mining. 7:31 Jason Les said on the call that they will continue mining Bitcoin at that portion of the site to continue to energize it and, you know, to pay their bills with regards to the power. 7:41 But obviously, they're going to be looking for ways to monetize that for AI. 7:46 Jason Les said they're going to take the best deal that they can get, and they will, at the drop of a hat, put a AI tenant in there if it makes sense. 7:53 Quote, "We are going to continue to utilize Bitcoin mining to monetize that available capacity until we have the right lease or set of leases to take the balance." 8:01 [lip smack] Looking at their balance sheet, Riot has become a big seller of Bitcoin. 8:09 It sold nine thousand six hundred and sixty-five Bitcoin for seven hundred and thirty-two point five million in the first half of twenty twenty-six. 8:17 Booked a one hundred and thirteen point two million realized loss on that Bitcoin. Holdings have fallen from eighteen thousand and five Bitcoin at year-end to eleven thousand three hundred and eighty Bitcoin at June. 8:29 Still a lot of Bitcoin on the balance sheet, but five thousand eight hundred and twenty-one of those Bitcoin are restricted as collateral for a Bitcoin-backed loan, leaving only five thousand five hundred and fifty-nine unencumbered, worth three hundred and twenty-five point four million. 8:43 Positions underwater, that's n-not a shocker. Everyone's Bitcoin is underwater. [chuckles] Total liquidity for R-Riot comes in at one point two million. 8:53 Uh, rough-- Uh, five hundred and forty-eight point nine million in cash. Some of that is restricted, seventy-seven point five million plus six hundred and sixty-six million Bitcoin. 9:02 Again, some of that is restricted as well. Debt came in at eight hundred and forty-three million. [lip smack] Two hundred and fifty-four point six million of that is current. 9:09 Five hundred and eighty-eight point four million of that is long term. Only s- two point seven million interest in the quarter. 9:16 Uh, m- Riot, like Mara, took out convertible notes, I believe, at the end of twenty twenty-four, and so I believe those are at an extremely low coupon. It might even be zero percent. 9:26 And let's see, H1 CapEx was a hundred and seventy-six point two million, plus forty-one point four million in equipment deposits. [lip smack] All right. Let's get to the meat of it, Charlie. 9:39 Let's look at this new lease for Rockdale. We're also gonna provide an update on the AMD lease 'cause Riot has completed some milestones for that. 9:47 But we're gonna start with this new one at Rockdale 'cause I think people are most interested in that. 9:51 So this was signed in August, after the quarter end, and the terms are as such: a hundred and ninety-one megawatts critical IT built to tier three specifications at Rockdale. 10:02 Twenty-year initial term running through June twenty forty-eight, commence me- commencing upon full build-out, which is, uh, or full deployment, which is estimated in June twenty twenty-eight. 10:13 [lip smack] Nine point one billion base rent over the initial term with two five-year renewal options at tenant election take potential value to sixteen point-- uh, will take the total value to sixteen point one billion. 10:26 Net operating income, seven point three to eight point two billion cumulative or three hundred and sixty-five to four hundred and eleven million average annually. 10:35 They plan to deliver ninety-six IT megawatts in December twenty twenty-seven, with the full one hundred and ninety-one IT critical megawatts by June twenty twenty-eight. 10:46 They are using Rockdale's existing approval and connection. That's important for our story later on. There's no [audio glitch] zero exposure here for Riot. 10:56 And there's about a ninety, eighty to ninety percent net operating income margin. During the earnings call, Lee d-- uh, Le-- uh, excuse me, 11:05 Jason Les disclosed, uh, CapEx of roughly eleven to twelve million dollars per critical IT megawatt or two point one to two point three billion dollars total. 11:16 Now, there are some financing details here as well that they addressed on the call. 11:21 Riot has entered into a five hundred and seventy-three million dollar interim credit facility with JP Mor-- or, or Morgan Stanley Senior Funding as administrative agent to fund long lead equipment procurement and development costs. 11:35 The key items for this, it matures in October 15th, twenty twenty-six, roughly two months out. 11:40 This-- It is SOFR plus two point seven five percent floating, so I believe SOFR is somewhere in the ballpark of like three point six right now. 11:49 So [swallows] we'll call that just over, uh, j-just over six point five there and-- Or sorry, around, uh, six point five. 11:58 Secured by substantially all borrower assets related to the project, im-importantly non-recourse to Riot platforms. It is subject to customary carve-outs. 12:07 Morgan Stanley's Stephen Burd asked about the bridge financing and, 12:12 and, uh, Riot CFO, Jason Chung, confirmed the Morgan Stanley facility is, quote, "Fully secured by the tenant," and that's meaningful since it means near-term build-out costs already carry tenant credit support. 12:25 They are going to use this bridge loan Until management can button up an investment-grade credit backed off for the deal, which they said is, quote, "In the coming months," end quote, coming down the pipe. 12:38 Charlie, if you have anything to add here, I'm gonna toss it to you for a second before I go into the AMD deal and then the LOI for Corsicana. 12:49 Yeah, I mean, I don't know how much, uh, you want to reiterate just touching on Batch Zero here. 12:55 So this existing Anthropic deal is, uh, not affected by Batch Zero, as it's already the preexisting power pipeline from Riot that they're just repurposing 13:08 formerly and currently mining Bitcoin to a what looks like December 2027, uh, energization date for the first, um, some odd, I think it's like 96 megawatts in December 2027. 13:25 So, um, it-- uh, like every single deal that we, that we talk about in Texas now is just gonna have the Batch Zero cloud hanging over it. 13:34 And the big question is, um, was this deal signed, or this LOI, you know, signed before the Batch Zero news or after? And then how does the Batch Zero news or delay, indefinite delay affect these deals? 13:47 The short version is it shouldn't affect this deal as it's existing power capacity. They're already pulling. 13:56 Yeah, and that's a good thing to bring up just with regards to the power portfolio for Riot, um, with Rockdale and Corsicana. They've been building out these facilities for years. 14:04 They were at one point, you know, at least for a public miner, they were the largest Bitcoin mines of any public miner. And we were actually talking on the show yesterday. 14:13 It's like we haven't seen a deal for Rockdale yet, and then that day one comes out, right? I mean, we were literally talking about yesterday. 14:18 We're like, "They're probably gonna announce it on their earnings call," and they did. So, uh, little get our flowers, uh, put a feather in your cap. Yeah. And the Corsicana deal is, is massive too. 14:29 Just one gigawatt is huge. They were asked on the call whether or not it's, uh, with a hyperscaler, and Jason Les just ducked that. 14:37 But you have to expect, obviously-- there are only a few companies that are going to take that capacity, right? You, you could name them, you could, you could name them, and then, uh, count them on one hand, probably. 14:48 Uh, it almost like doesn't matter- Yeah... who it is. You just say they are a hyperscaler 'cause there's one of five to seven companies. Uh- Right... that's all you really need to know. It- Right... 15:01 I think it's like you, you may have a little question over is it, uh, you know, is it a specific hyperscaler? But they're all, they're all top-tier investment grade, you know, 15:12 money, and they all want infinite compute yesterday, so. 100%. And going to this deal specifically, there's this KBW, uh, outperform target that we have from Steven Gladgola. They, uh... 15:28 He's calling for a $35 price target following the Anthropic deal, and he calls it the Anthropic deal in this note. And I believe there's been some reporting on that. 15:37 It was mention-- it was described, the counterparty here was described by Riot as being an AI frontier lab. There's only two, so it's either OpenAI or Anthropic. 15:48 And according to the reporting that we've seen and from this investment note, uh, it is likely Anthropic. We don't know for sure. Riot hasn't said anything, right? So take that with a grain of salt. 15:57 But just wanted to flag that for us, for to explain where we're getting that from exactly. All right, moving on to the AMD deal 'cause there were some updates with regards to the AMD deal. 16:10 Riot completed the final 20 megawatts of AMD's initial deployment during the quarter, bringing 25 megawatts commissioned and online on time and on budget, according to management. 16:20 Construction is underway on the second 25 megawatts, 10 to be delivered in November 2026, according to projections, and 15 megawatts to be delivered in May 2027. 16:30 The option structure was restructured for this, uh, the option structure was restructured for this deal in April. 16:36 AMD exercised 25 megawatts of its expansion option and holds 50 megawatts of remaining reserve capacity, plus conditional first priority right to lease up to 100 more. 16:46 This is, uh, this is a first priority right as opposed to right of first refusal. 16:54 They weren't very explicit with this, but as my understanding, right of first refusal typically means that Riot has to go out and find a s- and someone else who would take that capacity, show AMD the specs for that deal, and then AMD can decide whether or not they wanna take or walk. 17:10 But this gives AMD the opportunity to say, "We want that now. Let's negotiate a deal for it." 17:15 Importantly, this expansion for 100 megawatts, um, which will take AMD to 200, is only applicable if AMD exercises the 50-megawatt option. 17:28 They are branding this campus on media as a me- as one of their mega labs. Uh, I think they went on CNBC saying that. I would be surprised if AMD didn't actually, didn't take all of the megawatts they have here. 17:42 In fact, if Riot continues to execute, I think it's kind of a foregone conclusion that they will end up exercising the rest of this option and get the 200 megawatts that they currently have in their potential if they wanna continue this relationship. 17:56 Per management's call, uh, per management's call figures, the 50 megawatts should generate roughly 63.6 million average annual revenue, 51 million annual net operating income on 170.2 million of CapEx. 18:10 All right, last bit here. 18:12 Corsicana, the LOI that they announced during the earnings call was not in the press release, but they have signed a non-binding letter of intent with a single prospective tenant for the entire Corsicana campus. 18:25 One gigawatt of fully approved utility power supporting roughly 75 or 756 megawatts of critical IT capacity. Management framed the full site as capable of generating more than one billion in annual rent. 18:38 It is not subject to batch interconnection process, um, totally outside batch zero, currently equipped for 400 megawatts of Bitcoin mining capacity. 18:48 There's no tenant disclosed, no rental rate, credit support, or the terms of it. KBW notes the LOI remains subject to further technical and commercial work. 18:57 Um, asked whether a single tenant for a full gigawatt implies a hyperscaler, again, Jason Less declined to characterize it, saying that it-- only that it carries uncertainty, and Riot chose to disclose this one solely for transparency and, you know, to get people excited there, right? 19:13 One interesting note though is that r- they're building it right now. They're, they're moving towards the construction phase, and they're trying to maintain flexibility. 19:23 According to the earnings call, Riot is proceeding with horizontal work and long lead procurement only for items, quote, "that will confidently support any eventual tenant's requirements, whether it is for our current LOI or otherwise." 19:36 So the first core and shell building is underway. They are not committing to a single design. 19:41 They are trying to keep their options open so that if this LOI falls through, they can move quickly and secure another tenant at this site. Some people may read that as they're not very confident in this LOI. 19:53 I think that's just hedging your bets and being smart about how you're gonna build out these massive data centers. 19:59 You don't wanna just put all your eggs in one basket, especially if the Easter Bunny is gonna make away with that basket. So that is the TLDR for Riot with that, with the LOI. 20:12 A number of banks have moved up their targets. Bernstein's from thirty to thirty-five, Citi from twenty-eight to thirty-two, Piper Sandler from twenty-three to twenty-five. KBW outperform at thirty-five dollars. 20:26 One last thing I wanted to note here, Charlie, because it shows you just how far we've come and how fast we've moved with regards to these AI builds. Riot purchased 50.9 exahashes of MicroBT orders. 20:44 I believe these were made in 2023 with delivery running through Q4 of this year. Seven hundred and ninety-five point two million dollars worth of equipment. 20:56 My biggest question, if we could get Jason Less on outside of the AI stuff, what's gonna happen to all that hash rate? I'm sure they'll deploy some of it in their Kentucky sites, 21:08 but, I mean, that's almost a billion dollars worth of Bitcoin miners that are worth nowhere near that now. Um, some of them might live in Rockdale in the capacity that is not being taken up by AI. 21:20 Maybe they'll put some in Corsicana as they continue to build that out as well. And- They can't sell it. There's no one gonna buy that much hardware. Weren't these the water-cooled ones? Yes. Yeah. Yeah. 21:31 I think some immersion as well. Like- They were the- Yeah... top-of-the-line advanced immersion and water-cooled units. 21:37 I, I think, like a lot of people, you know, I'm so sorry for those of, those who care about more about AI HPC compute. But, like, r- uh, repurposing and reselling, like even not used, but just, uh, 21:54 water-cooled rigs is different because it's not simply pull from the rack and replace it. You have to consider the entire loop, the water loop of the system. 22:06 So in a way, like you actually have to consider the dry coolers, uh, that were, uh... 22:12 So you have to consider that they, uh, you know, you have, uh, depending on how far along, uh, on the actual build of like the loop system, that kind of goes with it. 22:23 So it's not like w- in this new age of different form factors, especially water-cooled, direct to chip cooling rigs, uh, the resale game is just very different. 22:35 Maybe they, uh, you know, on the flip side, maybe these things don't appreciate as quickly because, like, uh, you have much less dust, you have a lot, uh, probably a lot less wear and tear. 22:46 Again, I haven't like really dived into the resiliency of the modern era of like direct to chip water-cooled stuff. But the resale gets more complicated 'cause there's more moving parts. 22:57 Perhaps the depreciation is less and less due to wear and tear and more just due to efficiency. So I'm kinda sp- still a billion dollars of rigs. [laughs] Where do those go? Holy smokes. Yeah. 23:10 And, uh, no questions about that on the call. No one cares about it. That's what I find most interesting. Nobody cares. 23:15 You know, uh, these companies have set billions of dollars on fire buying hardware in the years leading up to this inflection point in their industry. 23:27 And, you know, no one cares because frankly, if you like weigh that against the potential value of these leases, what's seven hundred and ninety-five bill- million when you've got a nine point two billion dollar lease? 23:40 And you might have an even larger lease across around the corner at Corsicana. So, you know- Yeah. And let, let's not default Riot. Every o- every Bitcoin miner did this. They were all buyin'... 23:52 You know, I was at, uh, you know, um, the World Digital Mining Summit in Hong Kong in 2023. 23:58 All the CEOs of all these names that are now doing AI, or at least some of them, were up there shaking hands with the, with the Bitmain C-suite and announcing massive orders of the latest Bitcoin miners, you know? 24:11 And now [blows raspberry] Bitmain has no pool. Yeah. Bitmain has no game anymore, bro. 24:21 I, I mean, a lot of people are like, "How do we, uh, how do we, uh, kill, um, how do we kill Bitmain's, you know, monopoly over the sector?" Don't worry. The sector will kill it for you. So, [laughs] 24:33 uh, that's, that's that. Uh, we're gonna keep rolling. We're gonna talk about some more AI HPC this time, how it's financed. Nvidia, half a trilly- Yeeting into the ecosystem. Will it touch some Neoclouds? 24:45 The answer is yes. But before we hit that, a word from our sponsor, CleanSpark. [gentle music] We are CleanSpark, America's Bitcoin miner, a publicly traded company with the largest operating hash rate. 24:59 Powered entirely by self-operated infrastructure across four states. This is our proof of work. We are setting the standard for what's next. 25:12 Learn more about the intersection of energy and Bitcoin at CleanSpark.com. 25:21 All right, so yesterday, NVIDIA just signed up six of the biggest names in Wall Street to raise half a trillion dollars so that everyone can buy NVIDIA chips. 25:36 This gets to the heart of the financing question and, uh, is coupled with a letter from Jensen Huang, his second tweet ever, directly addressing the circular financing story. So, 25:52 here is the letter. Let me explain what has happened. So, 25:59 the goal of NVIDIA's, uh, new fi- a- announcement is to mobilize over $500 billion in third-party capital for hyperscalers, frontier AI labs, and enterprises to build data centers, and crucially, to buy NVIDIA hardware. 26:17 The six financial firms will independently underwrite the AI infrastructure, providing capital at attractive rates to NVIDIA's customers. Who are the six? Well, they're names... 26:27 They're household names if you're in finance. Um, maybe if you're outside of finance, you'll know a couple. They are Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. 26:39 This answers a lot of the question of where's the next dollar coming from, because this is a lot of, uh, you know, we've kind of j- squeezed all that we can out of the, uh, like, uh, out of the, uh, cash from hyperscalers, from e- everybody else. 26:58 So where's, how is this, how do we get, you know, half a trillion dollars? Well, we get it from the even bigger guys, Larry Fink et al. 27:07 So let's look into Jensen's letter here, titled NVIDIA AI Factory Compute is Becoming an Investable Asset Class. His core argument here, and he, he doesn't give a lot of specifics, actually. 27:21 He's kind of br- he's, he, he creates some broad strokes here, not a lot of specifics. 27:26 But his core argument is, well, NVIDIA hardware is broadly adopted, flexible, transferable, and lenders can now underwrite Compute, specifically NVIDIA Compute, as revenue-generating. Um, w- 27:41 we saw CoreWeave pioneer the GPU-as-a-service as collateral. Um, NVIDIA is, uh, taking the GPU as collateral and basically making the analogy that it's kind of like how Boeing did airplane financing. 27:57 You have, like, project-level financing, but with GPUs backing. Um, Jensen calls chips revenue-generating assets and said it's the first time chips have become an investable asset class. So, what is he trying to do? 28:12 Jensen is, uh, as NVIDIA has, uh, written and financed, uh, many of their own c- their own clients, he's trying to now keep further risk off NVIDIA's balance sheet. 28:27 Um, he references the late-'90s telecom disaster, for those of you who were around, I wasn't, 28:34 and, uh, is determined not to repeat it, specifically having, tapping project-level financing with Apollo, Blackstone, and KKR as the alternative. Um, every hyperscaler is stretched. 28:48 Um, Jensen says the, uh, p- uh, useful life of their now pretty legacy A100 series is extending its economic life towards a decade. 28:57 This is in contrast to a lot of folks who have accounted for the depreciation on those to be two to three years. Now we're in year six of the A100. Jensen says it's more like a decade. That's crazy. Yeah. 29:10 That blows all the depreciation models out of the water. All of these analysts that got cute citing three to five years, and all of the bears that said that at most you're going to get six out of them. 29:22 Uh, now that being said, obviously this is coming from the guy who makes the GPUs. As my grandmother used to say, there are three sides to every story. 29:30 This one's probably somewhere in the middle, but almost certainly the aggressive depreciation timelines for these GPUs are totally wrong. Yeah. 29:37 And there's a world in which that that depreciation schedule even lengthens over, you know, over the years. 29:43 And the funny thing is, like, um, you could take a really simplistic view of this and, and say, uh, this is actually, why would Jensen do this? 29:51 This is basically like Steve Jobs or Tim Cook or the new guy at Apple saying, "Well, your iPhone can actually last 10 years." Why would they do this? Because now they can't sell you the next iPhone. 30:01 And that's because this is getting to the heart of what NVIDIA is doing and, um, which is, uh, similar to they ha- similar to having the, the moat around NVIDIA chips with CUDA standard, they are building an even bigger moat with finance. 30:19 And, uh, we'll come back to that in a second. Um, actually, let's go to, let's go to it now. Here's a great take from Wayne Nelms of Ornn. We often have Kush Bavaria on. And Wayne, um, 30:36 says that, "NVIDIA's problem in 2026 is not demand. It is that the entities demanding Compute cannot fund it the old way anymore. 30:44 Hyperscaler CapEx is running at hundreds of billions of dollars a year, and their own investors have started- Asking pointed questions about free cash flow. So, 30:55 Wayne says in this piece that financing AI through derivatives as AI, as NVIDIA is doing, deepens NVIDIA's moat more than any chip could do. 31:05 Um, specifically arguing that NVIDIA is saying that their chip is the one that you can underwrite. 31:12 Not saying that compute is fungible overall, but rather NVIDIA compute is fungible within itself, so NVIDIA is the investable asset class. 31:22 A competitor's chip, as Wayne says, could be cheaper per flop, but without the financing curve which accompanies, uh, this half a trillion dollar, uh, money injection, um, you just can't, uh, you just can't, uh, find the money to turn these chips online. 31:42 Wayne also then says something that my dad [laughs] actually hit me with the other day, a, a critique. Um, uh, pushing back on why this isn't 31:55 the 2008 housing crisis, and let me pull up exactly where he says this. Um, I can't r- I can't find the section. 32:05 But anyway, uh, Wayne says that, um, 2008 was not actually a housing shortage, it was rather in- an index marking problem, saying that everyone at the time lent against appraisals of the market later refused to acknowledge. 32:18 So appraisals were wrong, people lent against those for pa- p- uh, appraisals. And so this is a structurally different financing mechanism. 32:27 And that the same risk exists today that any underwriting that model, that models a residual value above the actual compute futures curve is, uh, is, i- is doing the same thing. 32:38 But when Jens- Jensen says that if we can accurately discover the true compute fut- futures curve and lend against that, then we have a robust liquid mechanism to finance GPUs as a collateral. 32:56 That was me yammering a lot, Colin. There's a bunch of different takeaways we could have. 33:02 I think that i- that difference between the subprime mortgage crisis and what we're seeing now is really important to point out, just with regards to the fact that you still have risks of defaults if the c- money, if the, if the cash doesn't come in at the levels that the lenders think that it will, um, 33:21 then you, you, you could have a cascading effect. But it's fundamentally different in the sense that the collateral for that housing market crash were the houses themselves that were inflated. 33:32 Whereas with this, you actually have cashflow businesses that are going to be responsible for paying that debt back. 33:40 Now, the question is just like when do you do, when do you get overextended and when does, uh, when does compute compress enough and get commoditized to where actually some of the older loans don't make sense under the new regime? 33:53 And I think that gets to the heart of what Jensen's saying about if we have a reliable compute futures curve, then we can have more certainty with regards to how we finance these things. 34:05 I think that, I think that what NVIDIA's, NVIDIA's doing here i- is for AI bulls, you know, I think you can read this two different ways. 34:14 Number one, if you're a bull, like we're go- the gravy train's gonna keep chugging, brother. I read this as if you're like listening for the music to stop, the music just got turned up louder. 34:24 Yeah, the music- That's my interpretation... just got turned, cranked up to 11 and now the neighbors in the other neighborhood are asking, "Hey, is there a party going on?" 34:31 I mean, it's li- like if you wanna borrow the analogy from Jay Patel from Lagos yesterday, like, you know, um, like the police show up at your, uh, uh, show up at your door, the state, the government, what are they gonna do? 34:42 The police show up and they're like, "This is dope. We're gonna invite all our police friends and we're gonna come party with you guys." That's basically what's happening. 34:49 Um, s- s- the, the roost, the chickens will come home to roost, um, but not before they multiply into infinitely- Right... more chickens, yeah. And, and that gets to the bear thesis. 34:59 This is just, if you're a bear on all of this stuff, this just adds more powder to the keg, so to speak. 35:06 And one thing I will say, I, I do think that you have to get, you have to be a little bit leery increasingly of how important NVIDIA is to this entire thing. 35:17 Like, th- they, they, like we called them the king maker, right? I mean, they're the emperor maker at this point, right? I mean, th- they are the emperor. Yeah, the usurper [laughs] basically. 35:27 You know, it's like it's, it's, it's NVIDIA's Holy Roman Empire and we're all just prostrate at the throne. 35:34 I mean, at this point it really does seem to me like you can't ignore the fact that this entire thing hinges largely on this one company. 35:43 It always has to an extent, but it's becoming clearer and clearer every day that their influence grows. 35:49 Not only, as you said, do they have the hardware moat and the software moat for the hardware, they're starting to build out a financial moat that may make or break certain projects that can't meet CapEx requirements because they don't have financing. 36:01 And, and this appe- I think this is, I mean, my, my reading of the finance and compute people, uh, leads me to believe this is a bigger moat than, than CUDA, than this, than their, than their standards, than their, like, chip production. 36:16 Like the f- finance is the, uh, Bitcoiners know, finance and money is the final boss of all endgame markets. And this is NVIDIA, uh, basically sprinting, uh, to build that moat. 36:32 I, y- yeah, I, I just wanna double tap on that. It's like, uh, in The Hollow Men, T.S. Eliot, "Between the conception and the reality falls the shadow." It's like with, with the GPUs and with CUDA, you know- CUDA... 36:43 as long as you can get that up and running, you have it up and running. But you can't get there until you actually have the money to erect the data center. Right. 36:50 And then within the shadow, Nvidia basically deciding whether or not you're going to end up getting that financing, right? Yeah. I mean, not totally. 36:58 Obviously, they're not the only decision-makers in this process, but I do think it kind of underscores the point. So who gets rich [laughs] off of this? Where is this half trillion dollar- Jensen Huang. 37:09 [laughs] Well, obviously Jen, he's already rich. But, uh, let's- we've been covering neo clouds and power shells on this sh- show full time for a number of months now, um, part time for the past several years. 37:22 [laughs] Uh, and, uh, this is basically, uh, not... Like, you thought the starting gun was a couple of years ago. 37:31 I think this is more, uh, uh, just, like, a bunch of runners just joined the race and are- and a wind at your backs. 37:38 Like, this is a very, very bullish for neo clouds who are trying to figure out where the money comes from so they can buy even more GPUs and build more. This is bullish for neo clouds. Obviously not financial advice. 37:51 Obviously, like, not the, the entire playing field won't, like, go up and to the right. But, like, 37:58 yeah, I know your, I know your CoreWeaves and your Irons and your, uh, you know, what-have-yous all are already up huge over the past thirty, thirty-sixty-five, four hundred days. 38:09 Um- So you should take out a HELOC and go- [laughs]... leverage long all of these companies. No, no, no, no. What's happening is Jensen Huang is having other companies take out the HELOC to buy GPUs. 38:21 [laughs] That's what's happening. So you eventually have to pay- Don't take a HELOC out to buy anything unless- Don't... you're okay with losing your house, guys. I mean, the- Yeah. 38:29 I mean, there are so many people in the last bull market that did that with not just Bitcoin, but I knew guys who did it with Bitcoin miners- Yeah. But, yeah... like in 2021, 2020, you know. So anyway. Yeah. 38:39 Anyway, um, okay. So, uh, Colin, I'll let you make a call live here. Um- No, we should definitely just, uh, we'll do a little flipparoo here. We've got Rob Hamilton in the wings. Yeah. 38:50 We're gonna bring him up and then doercot, but first, a very important word about our sponsor, Luxor. 38:58 [applause] This episode of Blockspace Live is brought to you by Luxor's Commander, Bitcoin miner management software for enterprise operations. 39:11 Luxor's Commander gives you real-time fleet monitoring, bulk remote commands across your fleet, and Intelligent Miner. 39:17 That's an automated profitability engine that runs every five minutes and adjusts your fleet's power settings against live energy and hash rate markets. 39:25 In fact, ERCOT back tests show 10% improved profitability with intelligent mining versus old-fashioned binary mining. 39:33 Commander Pro is $100 a megawatt or a 25 basis point pool fee adder, roughly has the cof- half the cost of competition. But you can also try it for free for 60 days. 39:43 So if you'd like to learn more, go to luxor.tech/commander to get started. All right, we're gonna swaparoo, bring Robarino up to the stage. Rob Hamilton, welcome back to the show. Hey, guys. Thanks for having me. 39:58 Do you enjoy being called Robarino? [laughs] Uh, I- After, after running on two hours of sleep for the last week. [laughs] For the past two nights, I got five, so I feel- Ooh. Oh, yeah... very, very luxurious. 40:10 Uh, yeah, sorry, I'm multitasking, but I can give you guys my full attention now. Okay. We're gonna keep, we're gonna keep it quick. 40:15 Uh, you're wearing the, a red Anchor Watch hat, obviously to signify red team, which we'll get into. 40:20 But let me do a little cold open here and show this tweet from Anchor Watch, uh, engineer extraordinaire Portland HODL, says, "US-based frontier AI model. You're absolutely right. 40:32 Chinese open model, 78 critical vulnerabilities found. The implications of this are unfathomable." Rob, the ColdCart exploit happened a couple weeks ago. 40:45 You spent a while in the trenches expanding basically all your time trying to find vulnerabilities everywhere, and this brings us headfirst into the dislocation between American AI models and Chinese models. 40:59 Give me the TLDR. Yeah. Um, the epiphany moment was as the ColdCart news was breaking, coming up on two week- wow, it's been 12 days. 41:08 Um, 12 days ago, uh, that Thursday where the hack happened, uh, Wednesday night on chain, but it took people a couple hours to understand what was happening. 41:18 And the thesis that was going around was that entropy was broken. So myself and a bunch of other developers in Bitcoin thought, "Okay, if the entropy is broken, it's gonna be right here. This is where it happens." 41:30 And the first thing I did out of instinct, I've been using Codex and Claude Code a lot, so I pulled them up, I said, "Hey, look at this. Are there any issues?" And Anthropic instant downgrade, like non-real answer. 41:40 Uh, Codex was like open, but like it wouldn't be clear. And then I pulled up Kimi K3, um, using open code, and it was like, "Here's a full vulnerability report." Pop. Like, it's busted. 41:50 And that was like a, like an eye-opening moment of realizing something had changed. Now we're coming up on, uh, f- that Monday before that. So I'm pulling up a calendar so I can get my dates correct. 42:02 So this was, uh, on Thursday, June 30th, was the, what I just described happening. On thir- on Monday, July 27th, Kimi K3 went full open weight. 42:15 And so when that came out, I immediately, for my own internal infrastructure at Anchor Watch, started poking around, seeing like, seeing what's going on, right? 42:21 And I was already seeing that it was very impactful for surfacing issues or potential issues, just in general. 42:27 And it wasn't until that Thursday afternoon realizing not only had most companies probably not have been on top of this, but most just software and Bitcoin was probably not on top of this. 42:37 And a lot of Bitcoin software, if it's either handling keys or if it's like moving your money for you, it's usually open source, so you can trust it. 42:45 And I immediately, uh, that Thursday night started scanning, poking around with repos using Kimi K3, and I started finding things. And then I was like, "Okay, well, you know, these aren't my software projects." 42:58 Let me reach out to these people and ask them. And I started getting a couple, like, responses. I started getting responses back of, you know, confirm received. 43:06 A couple hours later, like, this is interesting, and then like 12 hours later being like, "Thank you, we just patched," or whatever. And that happened like several times across Friday to a point where I was like, 43:17 something's happened. Like, something just happened in just software development. And I think it, it, it's, it's not just a Bitcoin story. I think that's probably a misunderstanding of the event. 43:25 It's an all software story. 43:27 And I've said that like the bugs have always been inside the code, but we finally have the veil lifted to be able to see them all, and Kimi K3's ability to do this compared to any of the American models is shocking. 43:39 Like, it, it's just, it's, it's night and day. 43:41 And I would assume that there's just kind of a post-training tampering and filtering on the responses coming back that's kind of obfuscating the, uh, the actual insight that's probably seen. 43:52 It's not like the American models aren't able to see this. They are better. They are better- Well-... performing across all these benchmarks. Well, well, so why, why is it? Is it a liability thing for the labs? 44:02 It- They, they don't want to have their, their L- LLM's fingerprint over some sort of black hat attack? 44:08 What is your be- your most charitable read for why they're doing this, and your least charitable read for why they won't give this information up? 44:16 I think the most charitable one is something in the sense of vulner- of like liability, and they don't wanna be associated with people using AI to do very harmful things. 44:28 The, the tragedy and the paradox of it, though, is when you have a team of white hats that are trying to proactively fix things, are now unable to use those tools. 44:36 And for both Anthropic and OpenAI have cyber verification programs that allow you to actually say like, "Hey, I work with an industry, uh, I need to do this." 44:48 Uh, Anchor Watch has been signed up for those for, for months at this point. 44:51 Like, both Anthropic and OpenAI, and as of this morning I've gotten total breakdowns on both of those, even using like my approved credential seat to be able to use these models. Yep. Yeah. 45:04 So- So that's the tweet you're pulling up there. Yeah. That, this was like my big epiphany moment where I was focusing on, um, a code base pretty intensely to try and find additional issues, and I... 45:13 It broke all of my agentic pipelines because it was failing over and wasn't getting a response. 45:18 So I've largely, at this point, removed all American models from my ability to do this research because they actively just hinder it, and it's just difficult to trying to, at scale, run, run these scans when you have constant just like triaging because, uh, endpoints are failing. 45:36 And this has like, like, like, like, patriotic... I, I mean, it has patriotic implications. 45:43 You, you say, quote, "It absolutely guts me, as a patriotic American, to have to do this, but I will be going back to using Chinese open source models to conduct my research to protect Bitcoin infrastructure." 45:54 Like, this obviously, we narrowly in, in our world is, is focused on Bitcoin, but, like, this has huge implications for software, especially open source and security critical software. Yeah. Um, 46:08 like, if the US does nothing, if this doesn't change in the next couple months, uh, what happens? Well, um, two things. One, I think it's a Sputnik moment. 46:21 This is a digital but Sputnik moment of you, you genuinely have the... 46:25 And for those, for history, that, you know, the Russians were the first to get the satellite in space, but America got their act together and got on the moon first, right? And this is kind of that shot across the bow. 46:34 And, uh, listen, it's at a point now with the amount of compute... I, I probably personally on, as part of the red team efforts, have spent somewhere between 30 and $40,000 of AI spend alone, just me. 46:46 Because I have set up infrastructure to do mass scanning across many, like hundreds of code bases. Um, whereas other people on the team are doing like domain specialization work and finding... 46:59 I could go on for hours about what I've learned, but, um, I think it's a, it's a, i- it's almost like a national security problem. 47:05 Because now, like here's the thing, too, is just 'cause you're using an open source model, like those may be hosted at Chinese companies. 47:12 So now I'm just, now, now I'm just trying to fu- fix things because the black hats are gonna do it anyway, and now I'm trying to fix things. And there's a great, um... 47:20 Uh, I was talking to someone who referred me to Fireworks AI, which is like an American-hosted, American-owned company that runs these models. 47:25 If you're trying to be compliant, I would say if you're working in IT, like this is really, like relevant as a path to look into. I've been using them so far for our own internal stuff, and it's been going great. Um, 47:36 there's so many implications here. Uh, my head's, my head's kind of spinning. Um, the ultimate... [laughs] Sorry, yeah, just so much has been happening. 47:44 You, you've had, you've been having less than five hours of sleep for, for 12 days straight, so- Yeah... uh, I'm just glad- Yeah... we got you for 10 minutes. Of course, of course. 47:53 Uh, yeah, no, th- th- and I think if you're taking the cynical take, I, I'm under the impression that they in- they want to lock these down 'cause they can charge a higher markup for people who pay tens of thousands of dollars for cybersecurity scans. 48:04 Um, you could get a better quality report than what you would do for a traditional third-party pen testing or security audit, and you would get more value for a lesser price, and that would massively scale. 48:12 And so there's gonna be way better margins than trying to sell metered intelligence through, um, buying tokens, and when you just buy a service, that could be marked up even higher. 48:21 I will say from an AI spend perspective, um, it's gonna accelerate. Like, like, people are gonna want access to these models. People are gonna wanna be able to like at scale. 48:28 I'm actually hitting throughputs right now where I'm going through three if not four different providers through OpenRouter, through, um, OpenCode's Zen product, through, um, Wafer, uh, through Fireworks, and I am... 48:42 'Cause I'm, if I'm running 50 agents at once, even if I'm willing to pay for it, they throttle me and they don't let me do it, right? And I'm just a guy trying to help open source Bitcoin software. 48:53 If I was any major IT company, I w- I, I would, I [laughs] would be spending- Tens, if not hundreds of thousands of dollars a day. 49:00 Because the, what I've learned and what I'm seeing of the ability for these things as you kind of learn how to use them, how it accelerates your understandings and the ability to find additional issues, there, there, there have been moments where I would scan something and it would come up kind of middling, and then a domain expert with that type of software would look at it, prod it a little bit, and you'd get this massive cascading, because they know exactly where to poke, of more issues, right? 49:23 So th- like this is not like... This is the thing, too, with LLMs, they're d- they're, they're not deterministic, they're probabilistic, so every time you query them you get something different. 49:32 So you are literally pulling a slot machine and instead of getting coins, you're getting vulnerabilities coming out. And it's, it's, it- this has been insane. 49:40 Like I, I, this has been the wildest two weeks of my life, trying to do everything I can to help many, many maintainers. 49:46 And at this point, like I just got, uh, Portland just this morning, uh, for a confirmed vulnerability, tried doing additional triage in Fable, and got downgraded, like, like, pushed out. 49:55 And it's just like, what's the point? 49:57 I'm just gonna use a bunch of KimiK3 instant endpoints across five different, like, open s- like five different vendors, so I can- don't hit rate limits, and I just can keep on going. 50:06 To me, this problem seems all the more relevant and salient right now given the ch- chatter about regulation for open-weight models. 50:16 And, and open-weight has got, is kind of being used as a synecdoche for Chinese models, so I'll just say that. Like, for the Chinese open-weight models. 50:25 But you know, if you've been on X recently, you've seen OpenAI's Dean Ball going back and back with blows with David Sacks. 50:33 And I, I just can't help but see all of this, Rob, as a kind of sort of Damocles moment for these frontier models in the sense that there's a real existential risk here that, yeah, maybe they are gonna try to charge these companies for more, uh, for, for, for a higher tier to have access to these capabilities for cybersecurity. 50:52 But how many people are just gonna go for the open-weight models? 50:55 Unless there is some sort of, it, th- there's some sort of, you know, uh, regulatory fencing and moat-building on, on- I, I hate to say it, but it's even worse than that. 51:04 It's not an industry thing, it's a question of the American experiment. To be able to make math illegal is the death of the American experiment. And I say, Molon Labe, like, come and take it. 51:12 Like, you're gonna take my linear algebra? Like, not a chance. Not in a free country like this. Linear algebra and matrices are not illegal. You cannot make them illegal. 51:22 And to live in a society which makes them illegal is, i- is the death of the American experiment. You cannot have a free country where you're not allowed to do math anymore. Like, like, w- where do, where do we go from? 51:31 'Cause it just starts with math, and if the math is a proxy for intelligence, you're gonna make thoughts illegal next? Like, this is, it is en- entirely un-American. 51:39 And the, and the, and what I have seen over the past two weeks, it, it, there are serious gray policy concerns if there's... 51:46 Like, the American experiment, as I said, is cooked if we're gonna actually talk about banning math. Like, you can't do that. We, we had a run-through of this in the '90s with cryptography. 51:54 We tried making prime numbers illegal. It lost. This is the new frontier, and it needs to be defended to, for the continuation and the prosperity of the American experiment. Like, bar none. 52:03 No, like, I, I try to be pretty reserved in my opinions at time, in most of the time, but this is not one of those things. This is a die on the Hill situation. You cannot stop math. You cannot make math illegal. 52:16 Not in a free country. Mic drop moment. Okay, to wrap this to like, to like, p- try to tie a bow on this. Um, you know, the average person has no idea how their chatbot works. 52:28 Um, they're just like their boyfriend, if you will. So like, the average person, probably difficult time, like, relating these two. Um, if I'm kind of a dumb... I'm not gonna say dumb. 52:39 Inexperienced AI user, and I just interface with the internet as normal, and I'm out here for all, you know, I'm just blissfully unaware, um, what is, like, something you could, a practical piece of advice you could give to someone who uses the internet in a re- in a regular fashion to psychologically prepare them or give them, like, a guiding light of, uh, like, to help, like, help them navigate the next several months? 53:11 I'm an optimist at the end of the day. And just because there's great instability and fast changing of things does not take away from the prosperity and the abundance and the growth that will come from these tools. 53:28 So while things may be chaotic and hectic, they were when the internet was starting, they were with the introduction of electricity. 53:35 This just may be one of those things of on that tier, if not starting to become larger than that. 53:41 And I would say that, um, being lo- local, connected to your local community, spending time with real people, um, having those ties to kind of see these things through together, uh, are the things that I, that I would call out to for things. 53:55 You know, your family, your friends, and, um, staying in touch. Um, and, and having that network to be able to, to, you know, have a sense of, uh, comfort. 54:05 And the inevitability of what's going on at the moment, um, things change. And that's, you know, it could, you know... I, I'm an optimist. I think it'll all be for the better. 54:15 But we have to carefully navigate the next couple of months, if not maybe a year, to make sure that's being all handled responsibly. Uh, it's like the meme, uh, "Hey, I know it's bad now, but it's gonna get way better." 54:29 [laughs] So- It's gonna get way better... uh, Rob, thank you so much for your time. Go back to red teaming and solving the world's- Happy to... problems. We're all a huge champion. 54:37 Also, uh, if you wish to donate or support Rob- Mm-hmm... Rob's efforts, give me a CTA, Rob. Yes. Um, what can people do? OpenSats. Yes. So OpenSats started a red team fund. 54:47 Um, you can go to opensats.org/red, and y- there is a link there. You can donate in Bitcoin. Um, it's a 501(c)(3). 54:56 Um, if you wanna get any tax advantage status, you could do dollars, you could do Bitcoin, and you can help support the efforts at, uh, opensats.org. All right. Thank you, Rob. Man, I hope he gets some sleep someday. 55:10 He went gloves off with that second alert. Yeah. Rob, a very balanced neutral party. Getting fired up. Yeah. I mean, it reminds me of the way-- the, the impassioned 55:20 speeches that Bitcoiners used to make about the separation of money and state and all of these things. 55:25 Uh, it to me, it makes total sense that you would have, especially as, as someone who works in software engineering like he does, of course, you would have a strong stance on this. 55:36 And to me, it, it, it, it is, uh, as someone who doesn't-- who's not running code reviews for this stuff, it's pretty, pretty bleak, man, thinking about the ramifications that we can't use these things for cyber defense. 55:53 Like, I mean, I'm assum- I assume on Wa- in Washington, they will be able to do that like at the drop of a hat. Yo, but like, you know, there's the NSA and the CIA and the, and the other three-letter agencies. 56:05 But what about, you know, your- Well, you're leaving your entire domestic industry and population kind of railroad here. Yeah. What about your community-- Yeah, what about your community bank? What about your- Yeah... 56:15 uh, what about all your baby photos and your personal information? Um, what about being able to know that it's your mom calling you and not some, uh, North Korean? Like, you know, that, that kind of stuff. 56:25 So, um, this is probably gonna be a fun and terrifying story to cover over the next, uh, several months, uh, uh, perhaps longer. All right. We're gonna keep on going. We're gonna-- We saved the best for last. 56:41 Batch zero, uh, equally impactful as existential AI threat, which is, uh, the power pipeline in Texas. We will go and talk about the latest batch zero updates right after a word from our sponsor, Ligos. 56:57 [fire swooshing] You know what sucks? Losing your Bitcoin to the cold card exploit. But you know what also sucks, but a little bit less 'cause let's be honest, this is kinda your fault if it happens. 57:11 [laughs] Is losing your Bitcoin in a Bitcoin-backed lending scheme that was playing funny games with your hard-earned Satoshis. Well, I got good news for y'all. 57:21 If you want a Bitcoin-backed lender, but you don't want to give up control of your keys, our sponsor, Ligos Finance, is the company for you. 57:30 They are our preferred Bitcoin-backed lender here at Blockspace because they use Bitcoin native smart contracts to make sure that you always control your Bitcoin. You heard that right. 57:40 When you're lending with Ligos Finance, over the course of the loan, you always have control of your keys. 57:46 With Ligos, there's no rehypothecation, no bridging, no wrapping, no funny business, just competitive rates as low as eight point five to ten percent APR. Go to ligos.finance to learn more. 57:59 All right, Charlie, we are gonna bring it on home by going to the Lone Star State and looking at some of the hiccups that are surfacing as a result of Greg Abbott's data center audit. 58:14 Headline here from Blockspace reads, "ERCOT seeks to suspend Batch Zero deadlines for Texas data center audit." 58:21 There's a lot going on here, Charlie, and I'm gonna do my best to untangle the intricacies of this because there are a number of deadlines that are being-- that are potentially going to be pushed back as a result of this. 58:32 Now, specifically with regards to this headline, ERCOT is making this request to the PUCT, that is the Public Utilities Commission of Texas. 58:40 ERCOT filed with PUCT on Monday requesting good cause exceptions to suspend several Batch Zero deadlines while it conducts the data center audit that Abbott has mandated. 58:51 That audit, Charlie, is projected to take a few months. 58:56 Now, a few months could mean two to three, it could mean four to five, but no doubt it's going to take some time, and they need the PUCT to approve them to push back on certain deadlines. What deadlines are those? 59:11 Number one, the August 7th classification deadline, they already missed it. ERCOT was required to notify, um, interconnecting large load entities of dynamic data deficiencies by that date. 59:22 Couldn't get through all the two hundred submissions. They cited internal processing constraints and also variations on how these interconnecting large load entities formatted their data. 59:32 Now, specifically, the dynamic data deficiencies angle is a data-- i- is basically data reporting on how your data center or your large load would be impacted or impact certain, uh, interruptions within the grid. 59:49 So this is front of mind with regards to curtailment events, whether that be four CP, which is forecasted, or whether that be things like the winter Uri storm that snapped the grid. So 1:00:01 the August 7th deadline was specifically for ERCOT to understand how these large loads were approaching that problem of what would happen in the event of a shortfall or some sort of interruption with regular transmission. 1:00:15 Again, that was missed. The August 1st and November 1st quarterly stability assessments, um, it, uh... Oh, sorry, hang on. I've gotta move on to that in a second. I just skipped over the other, um, the other deadlines. 1:00:31 Sorry, Charlie, let me get here back on my notes. Okay. Here we are. 1:00:44 Um, yes, the August 1st and November 1st quarterly stability assessments, these are the other deadlines ERCOT wants to, um, include unclassified loads in both. We'll get more to that in a second. 1:00:55 And the September system-wide study start. That can't proceed on schedule because they have to do the audit. 1:01:00 So those are the three key deadlines for Batch Zero, which again, is the new inter- large load interconnection batch study process as a result of all these data centers that are coming to the grid under these new rules. 1:01:13 So those were the deadlines that are being pushed back. The audit will begin shortly after August 20th and quote, take several months, end quote. Going back to the quarterly 1:01:27 assessment, the quarterly stability assessment, ERCOT is requesting that six projected Q1 2027 loads and 17 Q2 2027 loads be potentially eligible for these quarterly assessments. 1:01:44 They're quote, requesting authorization to include them. The PUCT hasn't granted ERCOT this yet. 1:01:51 This is important because this may include some of the companies that we cover, namely, I believe HUT 8 and Galaxy, but I know definitely HUT 8 with regards to its Beacon Point data center. 1:02:05 HUT 8 has a PPA in place with AEP Texas for the expansion at that data center. That will likely classify it as baseload as a result of the Batch 0 process. 1:02:18 And so if the PUCT actually allows ERCOT to include some of those expected Q1 and Q2 2027 loads to be included in these quarterly stability assessments, that should help usher them along. 1:02:32 Basically, it's ERCOT saying, look, we have these deadlines that we are going to miss, but we'd still like to keep some of the more serious projects in the queue. They didn't name any by name. 1:02:42 We don't know for certain whether or not HUT 8 will be included in that inclusion. But it's worth noting that it will certainly affect some of the companies that are in our universe of coverage. 1:02:55 Just following up on that note, absent a good clause exception, these large loads would be excluded from the August 1st and November 2nd or November 1st QSAs and ineligible to energize on their projected date solely because ERCOT hasn't finished that classification. 1:03:11 So if these companies are not included in those quarterly stability assessments, that will delay their energization timeline. 1:03:20 That's why ERCOT wants to get those six from Q1 and 17 from Q2 in terms of their projected energization date or approval date to be ushered through so that they can keep those timelines intact. 1:03:33 There is a, and just to note, that doesn't authorize energization, the stability assessments. That's just one other milestone to clear in this whole process. 1:03:44 There's also an August 31st deficiency cliff here to take into account. Buried in the same section of the filing is this contingency under 1:03:58 a planning guide that interconnecting large load entities must resolve those dynamic data deficiencies by August 31st, 2026, or be removed from Batch 0 entirely. 1:04:09 This is part of the reason I think that Abbott is pushing this audit. Certain folks have not had their paperwork in order. This is Abbott's attempt and ERCOT's attempt to chase off phantom load, i.e. 1:04:24 projects that are submitting multiple requests at the same time for different interconnection agreements from different power providers. 1:04:32 And specifically, if these entities do not resolve the deficiencies in their dynamic data reporting, they're going to be removed from Batch 0 entirely and said you have to get in line for another batch down the road. 1:04:45 Again, ERCOT has reviewed like 200 plus submissions and notified those developers that are not been up to muster. They haven't finished all of the rest. So that's like a hard removal date, roughly three weeks out. 1:05:01 So if applying to projects that may not have known they have a deficiency. 1:05:07 So that is a clear cutoff for some of the projects in this queue right now and will be, I think, one of the first like separating the wheat from the chaff moments here. Now, importantly with this, 1:05:22 the ERCOT is proposing with the PUCT to keep its April deadline for the study. So this is stage three of the Batch 0 process, a screening study with an April 9th, 2027 deadline. And ERCOT basically, 1:05:39 this is them saying this study will determine two things, mostly the maximum peak demand that can be reliably served in each year according to these new inter-large load entities interconnected to the grid 1:05:53 and also the transmission upgrades required to support these new allocations. So this is a system-wide steady state and stability screening study for 2028 through 2032. And this is the 1:06:10 big one with regards to whether or not these companies will make it through the Batch 0 process and actually get approval. 1:06:20 April 9th, 2027 is when ERCOT must deliver the study results and is the date developers learn how many megawatts they will actually get and what transmission upgrades they are on the hook for. So 1:06:31 ERCOT's got a lot to plow through, man. If so, the again, the 1:06:40 audit is pushing back this September system-wide study start and it can't proceed until that it can't proceed until the audits are done. 1:06:53 So the study that is due in April 9th, 2027 was set to start next month. We don't really know when it's going to start now and it's going to take for it's going to have to be until the audit is completed. 1:07:05 And just to be clear one more time, this is ERCOT requesting the PUCT to accept these amendments to the Batch 0 timeline. And I would imagine that most of these contingencies will be pushed through. 1:07:18 But there are some legitimate questions with, can you get the studies done in time? Let's say the audit takes four months, and we're bumping up on the end of the year here. 1:07:27 Can you get that study done before the April deadline? I think that's a question that's up in the air. But needless to say, it just shows that, you know, on the one hand, 1:07:38 there are a lot of hiccups here with regards to how the ERCOT and how the Texas grid is processing this influx of demand. On the other hand, I don't really know what you expect them to do. 1:07:50 If you have data center developers coming in and just, like, not really filing their paperwork accordingly, you can't get an idea for which projects are serious. 1:07:58 You have to figure out how much demand you're actually going to be bringing onto the grid to make sure things don't break. So I think they're probably moving, moving along as quickly as they can here. 1:08:09 Yeah, like what do you expect them to do? Like, they have to do something. 1:08:13 They're, like, they're l- progressing linearly through the, the batch, the interconnection request, as is just, is not gonna work, so they have to do something. 1:08:23 Uh, maybe there's multiple things driving this, but one of which is gotta reform the process. 1:08:30 And, um, uh, time will tell 'cause i- it's, it, you know, it'll be a little clearer probably six months from now whether or not this was primarily to try to reform the process or whether it was, like, 1:08:45 a, a community pushback ha- like, hampering, like, political move. Um, given the fact that ERCOT has 1:08:54 s- uh, in some ways clumsily, but overall somewhat effectively, um, added data center capacity over the past four to five years, uh, I'm a little optimistic. 1:09:06 I just think there's, it's just too big of an opportunity for them to totally fail and fumble it. Like, um, with the amount of capital with, uh... 1:09:15 I mean, I saw or read a story that Dallas, um, overtook New York in, like, finance jobs, uh, over the past year. You see Elon's bringing a, his trillion, multiple trillion dollar companies down to Texas. 1:09:28 You see tech in Austin. You see energy in Houston. Like, I, there, you have too many smart people and too much capital having formed in Texas for them to totally fumble the bag here and with the data center plan. 1:09:42 Yeah, as, and if they don't, Te- Texas will be the most important state in the United- Yeah... States in 10 or 20 years. Yeah. And the- And Oklahoma. 1:09:52 [laughs] I was gonna say, and Oklahoma- And Oklahoma's the sec- look, I like to say, you know, [laughs] we're gonna ride on those coattails. 1:09:59 Uh, we'll, we'll play our little real, uh, our, you know, Red River River, Red River rivalry football games, but, you know, if they seceded, we'd go with them, I like to say, so... 1:10:10 [laughs] You can't though, because you're too- No... you're too important to- Ah... to the union. Yeah, we are. We're important. That's, that's true. We're very important. That's, that's Oklahoma's motto now. Yeah. 1:10:20 [laughs] Yeah, Oklahoma, the, the sooner you realize we're important state. Ooh. [laughs] Okay. All right. Yeah. [laughs] Okay. I don't know. I think we're, uh, we're a little off track. 1:10:29 I think that's prob- do you have anything else you wanna wrap up with? [laughs] No, that, that's it. Uh, just, uh, yeah, deadlines are potentially being moved. Nothing confirmed yet. Yep. All right. 1:10:40 Thank you so much for watching Block Space Live. I- if you haven't, if you're stick, if you're still around, I will call your attention to earnings week. This is day two of week two of earnings week. 1:10:52 We have a big rest of the earnings week. Tomorrow, Wednesday, kind of the heavy hitter for the day, Sam Tabar of, of Whitefiber, Brandon McBee of CoreWeave, Nevius and Whitefiber watch parties on our channels. 1:11:04 Thursday, Ben Gagnon of Keel, Soluna watch party, and on Friday, Mike Capicci, CFO of Soluna. Make sure to like and subscribe, follow Block Space everywhere. Go to our website, blockspace.media. 1:11:17 This show is brought to you by CleanSpark, NASDAQ listed, ticker CLSK. I'm Charlie. I'm Colin. We'll see you tomorrow. 1:11:28 [outro music]