Transcript 0:00 Applied Digital just reported an absolutely smashing earnings beat and a surge in revenue year over year for their Q1 2027 financials. 0:11 That's right, revenue's up 322% year over year, and their earnings beat, still negative, but beat by 28 cents for negative 0.1, or j- uh, 1 cent per share. You're listening to Block Space Live. 0:27 We will cover Applied Digital for our lead story today after a little bit of banter on some breaking news regarding the actual ARR of OpenAI. 0:38 After that lead story and some banter on OpenAI, we have also got a story on SpaceX mulling over a $40 billion raise for its GPU deployments as the SpaceX AI cloud component of the business seeks to put 10 to 15 gigawatts online in 2027. 0:59 Bond market's not liking that one, Charlie. Near- nor, nor is the stock market for SpaceX. SpaceX down over 3% today on the heels of the news amid a larger rout in tech stocks. 1:10 We'll be covering that second in our lineup for stories. And then to cap it off, got a little bit of fun today. 1:17 It got rowdy in a Memphis City Council meeting this week as anti-data center protesters took over a city council meeting, drowned out the chairwoman for the council meeting, and generally just said, "We don't want your data centers in this town." 1:34 Charlie's got all of the notes on that for our final story. That's right, you're listening to Block Space Live, hosted by Charlie Spears and Colin Harper. Block Space Live is Compute's daily live show. 1:46 We cover AI data centers and all things emerging technology. If you like the live stream but you miss it, that's okay. 1:56 It becomes a podcast, but the live stream goes out 1:00 PM every, 1:00 PM Eastern every single weekday, and it can be found on all streaming services shortly thereafter. 2:07 If you want more Block Space content, go to blockspace.media, get our newsletter, drops in your inbox every single day, featuring things that we talk about here, but also some more stuff. 2:21 We are starting out things a little bit differently going forwards. 2:24 We're gonna kick it off with the tweet of the day, because we realize as we build the show and cover this, we prep these segments, there's almost always, like, an additional thing, something that breaks right as we go live or just, uh, you know, uh, uh, random things, uh, that we wanna show on screen. 2:42 But today, right as we were going live, this drops, um, from the Financial Times, this tweet from Mitchell Martin, "OpenAI annualized revenues $20 billion less than previously signaled." 2:58 Colin, 20 billion's a lot of dollars to be lower than what OpenAI had previously signaled. Is this bearish for AI? Is this why markets are tanking right now? Uh, yeah, what's going on here? 3:13 I don't know if this is why markets are tanking, and I have to say I'm really frustrated at that the Financial Times broke this story right before we went live, uh, because this would have made a fantastic segment. 3:23 We'll cover it in detail tomorrow. Yeah. M- my first thoughts on it are, I guess my f- initial take is, if Anthropic isn't earning that revenue, is... 3:33 Or, excuse me, if OpenAI isn't earning that revenue, is Anthropic earning that revenue instead? Or is this j- a general slowdown? 3:41 Is this the first hiccup in the demand narrative that we've been seeing for the last two years for AI compute? 3:49 Because if it is a legitimate breach in that narrative, then you do have to wonder what this means for all the other stocks in our universe if spending, i- if deal flow for Anthropic and for OpenAI are both down, that necessarily would mean that deal flow for the NeoClouds are down, and the hyperscalers, which would necessarily mean that the powered shells are not going to have as much pull when they're trying to get leases for these sites that they're building. 4:16 That being said, there's a little bit of a narrative violation here, and that's why I wanna see what the Financial Times article says- Yeah, me too... 4:21 and then do some more research, because as we'll cover with Applied Digital, Applied Digital's CEO said that the negotiations that they're having for leases now are at 15% better terms than the leases that they were signing earlier in the year and in 2025. 4:35 So if that's the case, it's hard for me to believe that this necessarily points to something systemic within the AI CapEx market, or the AI CapEx boom and the, uh, GPU market in general. 4:48 This could just be an isolated OpenAI story. And perhaps this is why they decided to push their IPO to next year, because they knew that this was in the tea leaves, and they didn't- Yeah... 4:58 wanna have to IPO with lackluster numbers compared to their number one competitor, so. 5:02 The, the private, the private markets that we see, the, the deals that are being struck right now all say up and to the right, compute demand is relentless. 5:10 I could see this as maybe OpenAI, uh, signaled a little bit too high. Think about it. They're trying to IPO. Anthropic, same deal. They're trying to IPO. 5:18 Maybe once you actually get down to the real numbers, uh, and they come out, they're like, "Well, maybe we're a little hot on those numbers." Maybe to the tune of $20 billion. 5:27 But Colin, I think we also have another tweet pulled up here that we wanna touch on really quickly. Um, I'll let you take this one. 5:36 Yeah, so, um, can you actually, uh, there's a, there's a f- an, uh, a tweet that I wanna address before that. What's the- Yeah, okay, okay... 1:00 PM one. So here we go. This is- Let's do this one, yeah... 5:46 This, this is from Jim Liu, um, AIREN, addressing the semi-analysis FUD on reliability. 5:52 Those of you who may know, semi-analysis published their, uh, cluster max rankings with AIREN getting pretty poor scores, and it caused a lot of critique of semi-analysis in general. 6:04 A lot of people basically saying pay to play, and they're de-ranking, uh, people who aren't paying them, uh, a- Colin 6:13 Yeah, so this whole thing started with the cluster-- the latest ClusterMax rankings in which Eyeron was absolutely lambasted by the SemiAnalysis team. 6:24 They more or less said, "This is like the worst GPU managed cluster that we've seen," and 6:31 it, uh, and Eyeron d- would do better to stop pretending that they offer managed clusters and instead just do bare metal offerings. 6:38 The, the problem about this whole thing is that at the end of the day, both Eyeron and SemiAnalysis kinda have a point here because following this, following this ClusterMax ranking dropping, 6:54 Eyeron's CCO went on to McNally Money, I believe. CCO Kent Draper went on and said, quote, "We don't have any live managed services today," end quote. 7:05 Basically making the argument that SemiAnalysis is holding Eyeron to a standard for services that they don't offer and saying that S- uh, SemiAnalysis is using Eyeron's bare metal offering and comparing it to other cluster offerings of Slurm and Kubernetes for all these other neo clouds. 7:24 If, if that's true, it seems to be true based on what I've looked into for this, then that is really unfair to just hold Eyeron to this standard when you're basically evaluating one service and giving it a ranking based on evaluating a completely different service for all the other companies in the ranking. 7:43 That being said, in Eyeron's 10-K, they do say that they offer these managed services. 7:50 So there's a con- there's something conflicting here between what we're seeing in their most recent SEC filing for their most recent quarterly earnings or, um, yeah, their most recent quarterly earnings and what they actually, and, and what their management is saying they offer today. 8:05 They did purchase Mirantis for like 500 something million dollars recently, uh, which will, yeah, Eyeron acquired Mirantis in August 2026 for $544 million in aggregate consideration. 8:17 Mirantis does provide Kubernetes orchestration, cloud infrastructure software, and enterprise support capabilities. So perhaps the, the disconnect here is that they're saying like they offer this 8:30 in theory, um, and they will offer it soon. [lip smack] Um, that being said, uh, again, Draper saying they have no live managed services today suggests that these contracts, these services aren't operational. 8:43 There is a potential that Eyeron offers these managed GPU clusters through like dedicated single tenant contracts, so they're not offered on demand. Um, again, Draper said that they don't. 8:54 Maybe he was talking specifically in the context of on-demand services, which is what ClusterMax ranks. So there's a lot of gray area here that can't be resolved. Um, I think that 9:05 SemiAnalysis has a point by saying even the bare metal services were lackluster in the sense that they had reliability issues. 9:11 If that's true, then that is a, uh, you know, that's something that we would want to factor into analysis of Eyeron as, as a neo cloud. 9:19 Think Draper's point here though too, if you're basically taking bare metal services and then saying that they're not performing next to Slurm and Kubernetes orchestrated services, well, yeah, that's not, that's comparing apples to oranges. 9:32 That's not really fair. But again, Eyeron's kind of been flip-floppy on this. It's not really totally clear what is live today and what isn't. 9:39 The last thing I wanna say about this though, per this tweet here, um, this is coming from, uh, a poster, Jay Caines, saying, quote, "SemiAnalysis seems to have a grudge against Cerebras. 9:52 Most of what they say about Cerebras comes across as con- uh, context negative spin bias sniping. Pretty consistent now. 9:58 Makes me wonder whether Cerebras not buying their very expensive research product has something to do with this." 10:03 The reason I bring this up is this was quote tweeted by someone who did a breakdown of the Eyeron rankings in the most recent ClusterMax, and they were pointing to this as a, um, as evidence that, 10:18 uh, well, I, I wanna be careful with what I say here, but basically the insinuation from some folks here is that SemiAnalysis will target certain customers that have not paid for their paid research and their paid newsletter. 10:34 And this is what Sarah Qiang of Cerebras, one of their, their head of de- developers said outright after SemiAnalysis said that their TPUs were, uh, their clusters were underperforming compared to Google. 10:50 Sarah says, quote, "All this just to get us to pay for your newsletter." So Sarah out there throwing bows and not holding back on accusations. I have no way of knowing if this is legitimate. 11:03 I'm not going to go out and just straight slander SemiAnalysis. Yeah, we, we don't know, but the-- look, we gotta do some drama. We gotta cover the drama. 11:11 It's, we're talking this, you know, billions and billions of dollars. 11:14 This is the industry at the forefront of the AI acceleration movement, and it's about like, do we pay for this, uh, new upstart research company's, uh, product [chuckles]? So I mean, look, a call to action here. 11:29 BlockSpace newsletter is free, so you don't need to pay for that. Colin, do we have anything else or we wanna roll onto the first story after our catch-up? 11:37 Um, nothing else, but if you want to have a kind of breakdown of these, these insinuations and like why some people have a problem with how SemiAnalysis does or, or conducts its research, go ahead and check out, uh, this, 11:53 um, this, uh, uh, s- substack from John Stevens, quote, "Influence as a service: SemiAnalysis under the microscope," end quote. That's the title. 12:02 He does a breakdown of-- He basically breaks down the argument here that some folks see a disturbing pattern in the fact that companies that do not subscribe to SemiAnalysis research end up being on the short end of the stick for their analysis. 12:17 Whereas people who do pay for the newsletter and for the research tend to get glowing reviews. Make of that what you will. We're not gonna dig into the details specifically. 12:27 We're gonna move on to the news for the day After a word from our sponsor, Soluna. 12:33 [ on-hold music] Soluna Holdings, Nasdaq listed ticker SLNH, develops data centers co-located with wind and solar plants, turning surplus renewable energy into computing power for AI and other intensive applications. 12:54 We are proud to partner with Soluna Holdings here at Blockspace. If you would like to learn more, visit solunacomputing.com. That is solunacomputing.com. 13:12 Investment banks are increasing their price targets for Applied Digital after the data center stock smashed its earnings with a significant earnings per share beat and revenue up three hundred and twenty-two percent year over year. 13:25 Wells Fargo bumped its Applied share price target from fifty dollars to fifty-five dollars per share, implying a thirty-one percent upside. Additionally, [lip smack] 13:37 Citizens JMP Securities did them one better by reiterating their sixty dollar price target for a hundred and fifty-two percent upside. Now, these bullish forecasts come after Applied Digital's best quarter ever. 13:53 For the Q1 2027 financ-- Or excuse me, for the Q1 2027 fiscal year, the powered shell operator reported three hundred and forty-one point nine million revenue, up thirty-two percent quarter over quarter and three hundred and thirty-two point four percent year over year, which also beat consensus estimates of a hundred and sixteen point three million dollars. 14:17 Adjusted EBITDA came in at sixty-four point four million, up fifty-two percent quarter over quarter and twelve thousand seven hundred and eighty percent year over year. 14:27 Earnings per share came in at negative one cent per share, which still smashed consensus estimates of negative twenty-nine cents per share. [lip smack] Net loss was two hundred and twenty-one million. 14:41 Applied Digital carried six point three eight billion in long-term debt, up twenty-eight point two percent quarter over quarter and a hundred and twelve point six million in short-term debt, up five hundred and eighty-six percent quarter over quarter after a quarter's end. 14:58 If you're wondering why Applied Digital had such a tremendous revenue beat, it's because analysts underestimated Applied Digital's largest revenue segment, and that segment isn't data center rent. 15:10 And it also generates very little profit for the company. Digging into that further, a hundred and fifty-seven point two million of Applied Digital's revenue last quarter came from what it calls tenant fit out services. 15:26 [lip smack] This is essentially construction work or additional construction work to customize and equip the facility according to a tenant's requirements. 15:38 The cost of this revenue was a hundred and fifty-one point one million, implying a gross margin of three point nine percent at six point one million. 15:48 For the rest of the revenue, if we look at data center rental and other revenue specifically, by contrast, this segment includes its base lease revenue and tenant cost recoveries. 16:01 It earned Applied Digital seventy-nine point one million against forty-three point nine million in cost for a gross profit of thirty-five point three million and gross margin of forty-four point six million. 16:13 The rest of the revenue came from twenty-three million in GPU hardware sales and eighteen point five million in GPU services from a subsidiary, Chrono Scale, and thirty-seven point eight million in Bitcoin mining hosting. 16:25 Taking this in hand with Applied Digital's loss, we can see that the company's underlying data center hosting operations are generally positive segments generating income, but Applied Digital is not yet profitable as a consolidated company. 16:38 Its biggest operating drag is compensation and corporate overhead. 16:42 It paid two hundred and forty-five point seven million in services and cost of overhead for the quarter and another a hundred and ele- a hundred and fourteen point seven million in SG&A. 16:52 While its net loss is exacerbated by seventy-seven point four million in financing cost and sixty-seven point four million in non-cash mark-to-market adjustments in Babcock, Bag- uh, Babcock and Wilcox investments. 17:06 So part of that net loss is driven in some instances by that sixty-seven point four million non-cash mark-to-market loss, so that's a paper loss. 17:15 Also, there is a significant amount of non-cash SG&A over the quarter as well. So a lot of that loss is coming from specifically non-cash charges. 17:25 [lip smack] Looking at operational matters for the company, Applied Digital has one point four, uh, one point four one gigawatts of contracted AI capacity, and it said that it's marketing another one point three gigawatts. 17:38 Additionally, [lip smack] management has said that it, it expects approximately two hundred and fifty megawatts of expansion leases to be signed by the end of calendar twenty twenty-six. 17:51 And CEO Wes Cummings, speaking to the improving leasing economics for these deals, indicated that the company is targeting a lease rate increase of more than fifteen percent compared with earlier agreements, potentially combined also with longer lease durations. 18:09 Looking beyond the near term, Cummings, uh, Cummins outlined a target of approximately two gigawatts of operating capacity by the end of twenty twenty-eight, rising to three point five to four gigawatts by the end of twenty thirty. 18:24 All right, Charlie, a few nuances to unpack in those earnings, specifically from that tenant fit out revenue line, which was new to me. I have not seen that from any of the other powered shell operators. 18:34 But Applied Digital clearly emerging as one of the premier providers with that massive contracted, uh, capacity and also looking forward to admittedly pretty Herculean task to get three to point five to four gigawatts by the end of twenty thirty Yeah, I mean, $341 million in revenue, which is the consensus estimates are 116.3. 18:59 Uh, so 300%, uh, or 200% increase in, uh, revenue. Uh, granted, it's not high margin. It's what you said, 3.9%. 19:10 So but still, like, it's huge gross revenue, uh, change, uh, real big outperformance. 19:18 Like, maybe this is something we should have our eye on as other builders actually start building these data centers out, and it could really come down to the specific terms they have with the, uh, the tenant fit outs. 19:31 Yeah. I, I was really surprised by this 'cause, again, I haven't seen this specific tenant fit out or anything similar from some of the other leading powered shell providers. 19:41 As I understand it, it's specifically for, like, additional asks in the construction process for the build. So if they have something that they want kind of custom fit or, um... 19:50 I mean, mo- all of these are bespoke though, so I, I really do actually- If we consider, I mean-... kinda wanna know- Yeah... what, what exactly-- what, what are the terms of that agreement? 19:59 Why is that baked into it specifically? And going back to the beat, it seems like analysts just completely underestimated how much they would actually net from that. Um, and they did that last quarter too. 20:10 So Applied Digital beat last quarter basically from the same, uh, underestimation of how much they would earn from that. But again, only 3.9% gross margin from that business segment specifically. 20:21 Compare that to the actual lease revenue and those services, 44.6% the margin on the actual AI leases. 20:29 So, I mean, I think this is part of the reason why Applied Digital really didn't do too hot today after the earnings, even despite the beat. Last I checked, it was down. Yeah, it's down 3%. 20:41 It did, it did surge in, uh, early morning trading hours, actually, following the earnings or after the market had digest the earnings that hit yesterday, but it has basically totally retraced all of those gains. So, 20:57 I don't know if that was people digging into the numbers more and recognizing that most of the revenue bump was not from services that will be recurring. 21:05 Like, once the data center's built, that's not gonna exist, and then you just have the lease revenue, but. Yeah, but a lot of companies do get valued off of, uh, multiple of revenue. So, uh, we could see. 21:18 Uh, little bit of more context. You know, yesterday we reported that, uh, Applied Digital announced they had a one gigawatt power pipeline in Finland, the eastern most of the Nordic countries. 21:32 And that so they, uh, purchased a site after value, a bunch of-- after evaluating a bunch of locations in Europe. And the, uh, first hundred megawatts of that power is expected to become available in 2028. 21:46 This is what we learned from the earnings. Uh, capacity could expand progressively to about one gigawatt by 2031, which again, significant. 21:56 And the, the framing we gave yesterday is that this is probably because we can interpret this as, uh, indicating that getting a gigawatt in the United States is marginally more difficult. 22:09 So if you can get a gigawatt in Finland, you go for it. I think it's more of a story of how difficult it is to get these large power allotments domestically here in the US. 22:20 Yeah, and I, I found the earnings really helpful here with regards to contextualizing this because as we said on the segment yesterday, I think it's like there are more questions than answers from this. 22:29 And you just outlined them there, but I do wanna highlight first hundred megawatts of power is expected to become available in 2028. 22:35 So pretty far out, honestly, compare, uh, w- or, I mean, it's only a ca- uh, a year and some months out. 22:43 But per one of the notes that we covered, I believe the Wells Fargo note that we covered yesterday, they're basically making the point that investors want to see 2027 power because that's what, uh, that's what tenants wanna see. 22:54 That's what's most valuable right now. People are no longer ascribing multiples to these companies for pipeline power further out than 2027. 23:03 Uh, obviously, as 2027 enters the picture, they'll start valuing that 2028 differently. But the other thing is, too, they said up to one gigawatt in the press release. 23:12 Uh, approximately one gigawatt by 2031, so that's even further out. 23:16 That being said, still very, uh, still very good for Applied Digital to lock in this power, no doubt, or, um, to at least get their foot in the door for it. 23:26 Speaking of their foot in the door, they also discussed in their earnings call that they have entered into a long-term power purchasing agreement with Base Electron for approximately 1,200 megawatts of natural gas-fired power near Polaris Forge III in central North Dakota. 23:43 Applied Digital also owns approximately 10% of Base Electron. They expect power deliveries to commence in 2030, and the project is specifically front of meter and grid connected. It's not isolated and behind the meter. 23:56 Uh, they said that they're also exploring significant generation development elsewhere in North Dakota with Base Electron, including around Harwood and potentially Jamestown. 24:05 Cummins described roughly 2.4 to 2.5 gigawatts of generation equipment under procurement across the broader initiative, but only that 1,200 megawatts right now is the long-term purchase agreement that they have agreed to. 24:19 This is really interesting. 24:20 This is basically Applied Digital trying to get ahead of not just today's power bottleneck, but [chuckles] you know, the power bottleneck at the end of the decade, assuming things continue as they're going. 24:31 And, uh, the strategic implication here is that they are seeking m- ultimately greater control over future power supply instead of relying exclusively on grid interconnection queues. 24:41 So I think that was one of the more interesting developmental aspects of this, of, of this, uh, earnings release. 24:50 Specifically apply digital thinking strategically about how it can have more control over power flow as it looks to get to that four gigawatt target by the end of 2030. This will certainly help. 25:02 All right, we're gonna keep on going. We, uh, have a SpaceX story about inking a chip deal with NVIDIA, and then we have a story about how the people hate SpaceX's data centers in Memphis. 25:16 But before we go to the first story, a word from our sponsor, Luxor. [static sound] This episode of Blockspace Live is brought to you by Luxor's Commander, Bitcoin miner management software for enterprise operations. 25:34 Luxor's Commander gives you real-time fleet monitoring, bulk remote commands across your fleet, and Intelligent Miner. 25:40 That's an automated profitability engine that runs every five minutes and adjusts your fleet's power settings to live energy and hash rate markets. 25:47 In fact, Ercot backtests show 10% improved profitability with intelligent mining versus binary mining. 25:53 Commander Pro is roughly half the cost of competition, $100 per megawatt, or a 25 basis point pool fee adder, but you can also try it free for 60 days. 26:02 So if you'd like to learn more, go to luxor.tech/commander to get started. 26:10 SpaceX stock's price tumbled today after news surfaced that the company is seeking $40 billion in debt to finance NVIDIA GPUs for its rapidly expanding AI data center business. 26:23 According to Reuters, which broke the story, SpaceX is seeking $10 billion in bank loans and another $30 billion in investment-grade debt, with Apollo Global Management leading the arrangement and Pimco emerging as a potential lender and investor. 26:38 The deal is expected to close in 2030. The raise would be the first major debt raise for SpaceX since it issued $25 billion in senior secured notes in June 2026 across five tranches. 26:52 And SpaceX will need every drop of liquidity it can get to roll out the millions of GPUs it will need to hit its end of 2027 10 to 15 gigawatt target. 27:06 SpaceX operated approximately 780,000 GPUs at the end of September, and it expects deliver-- to deliver three tranches of more than 220,000 GPUs each month from October through December to bring this total to roughly 1,440,000 GPUs. 27:25 These GPUs will power the AI cloud services behind SpaceX's $121 billion in total contractual commi- commitments, including its $30 billion Google deal and its $85 billion deal with Anthropic. 27:40 These contracts equate to roughly $1.1 billion in monthly revenue starting in December of this year, corresponding to $13.3 billion in ARR. 27:50 SpaceX CFO Brez- Brett Johnson recently reiterated the company's ambition to reach $100 billion ARR in the future. 27:58 Achieving such a starry-eyed ARR, though, clearly comes with significant costs, as SpaceX's $40 billion planned debt raise and $25 billion cash raise in June can attest. [sniffs] And the bond market is taking notice. 28:16 SpaceX's five-year credit default swap spreads climbed to approximately 194 basis points, up from 110 basis points in June, which puts it next to Oracle, which still has the worst spreads on its credit default swaps, but puts it next to Oracle in terms of one of the highest spreads for credit default swaps among mega tech companies. 28:37 The widening spread suggests bond investors are demanding greater compensation for SpaceX's glowing-- growing leverage and AI investment risks. 28:45 And even worse, if we look at the 2056 bonds, they traded at a yield premium of approximately 236 basis points above comparable Treasuries. 28:55 Still, investment banks are remaining bullish on SpaceX's stock, with Morgan Stanley maintaining an overweight rating and a $300 million-- or [chuckles] excuse me, with Morgan Stanley maintaining an overweight rating and a $300 price target, and Dan Ives of Yorkville and Ives maintaining an outperform with a $225 price target. 29:19 All that said, SpaceX is still down 3.7% today amid a general rout in tech stocks. Well, Charlie, we talk about SpaceX a lot on the show, becoming a favorite here and a fan favorite at Blockspace. 29:35 Some amazing projects to be sure, but when you look at the 10 to f- 15 gigawatts, you know, target, and then the 100 gigawatt target later on, um, or sorry. Yeah, actually s- Elon Musk did say that. 29:50 And then you look at Terafab, which is going to be the largest factory in the world if Elon Musk gets his wishes, and I'm just stuck here thinking, man, like, how are they gonna finance all of this? 30:01 I mean, like, unless you have King Midas in there, you know, turning all the robotaxis into gold, like, where is this money gonna come from? Uh, it comes from the banks and financial instruments. 30:12 The-- I mean, Colin, this has been the story. Zoom out. Look at history. Whenever you have these major, like, uh, societal-level 30:23 industry shifts, like the railroads, pipelines, like power plants, you lever up and you sign contracts with strong customers. 30:33 So yeah, SpaceX, uh, has, uh, these large, long-year, multi-year contracts with r- a few different counterparties, which all are the big AI labs. Um, yeah, there is a problem because the chips bill is due today. 30:50 You gotta pay for the chips now, and you have revenue coming in down the road. I don't think that the widening spread, uh, because of the bond market is that much for concern yet. 31:01 Again, zoom out on the, the yields for these bonds. Uh, uh, it's not, it's not totally unprecedented. I think the real concern is not just the yields because people are... The lenders are still lining up. 31:16 They just want to be paid more for their risk. I think the real question is, are these AI labs around in 5 to 10 years to pay the bills? Like, that's the big question to me. 31:27 Um, certainly there will probably d- be demand much higher than it is today for compute. Will the specific counterparties and customers be around? That's a really big question. 31:39 OpenAI was not generating hardly any revenue three to five years ago. Same with Anthropic, who didn't even exist five years ago. 31:46 So, you know, if they are going to be, uh, core counterparties and tenants for the success of right now one of the largest companies in the world, 31:58 that's the big question, not like what the yields on the, the credit default swaps are to me, so that's my take on this. I'll throw it to you. Yeah, I think that's... Uh, I think you're right to have that question. 32:10 But I, I, I also have... 32:12 My, my question still stands with regards to like if you look at Elon Musk's portfolio of companies or, or, or, or specifically the vision for Tesla and SpaceX kind of operating in concert as this mega corporation, right? 32:26 It- it's going to require hundreds of billion dollars in investment. I mean, Terafab alone is gonna be $119 billion. That seems low to me. 32:34 Uh, there are independent industry analysts that th- that say that full global scale, um, for, uh, let's see, investment research firm Bernstein estimates that achieving Musk's goal of one terawatt of annual AI chip computing power would require between 142 and 358 modern fabs carrying a proto- total price tag of 5 trillion to 13 trillion. 32:56 So- Weirdly enough, I... Weirdly enough, that seems more reasonable to me than a pithy 100 some odd billion. Like, to me, I hear Terafab, and I think the number of its total cost has got to start with a T as well. Yeah. 33:07 So put it in trillions. Right. I mean, at the end of the day, the one terawatt, like who knows what timeline that's on. I mean, this is one thing that Elon Musk i- is like he's, he's an incredible salesman. 33:17 He has this gleaming picture of the city on the hill where all of the techno future op- uh, you know, optimists will live, and he's going to build it, and maybe it'll be on Mars, you know? 33:29 And then, you know, questions how we get there be damned. 33:32 So, like, I don't think that at least in our, you know, in any time soon, you're gonna get to that terawatt of, uh, of, um, an- annual chip computing power at this place. 33:43 I mean, that just like outstrips what we're capable of right now. 33:47 Um, but even if you don't get to that scale, ev- you know, $119 billion in the grand scheme of things is someone's got to eat that up, and right now, the insurers and the pension funds are, um, buying a lot of these corporate bonds. 34:03 Uh, and as we talked about with Jay yesterday, the appetite for these corporate bonds is part of the reason, uh, some people think that the Treasury market can't calm down. 34:12 Because if you're gonna make, you know, 100 to 200 basis points or 300 basis points more in the private debt market, why not just take that if they're from, you know, some of the most, uh, some of the companies with the highest investment grades in the world? 34:24 So all that being said, my question just continually becomes how long until the lenders get exhausted? 34:35 And with every new raise, you know, another 40 billion, and I think the credit default swap spreads and the reaction today to SpaceX e- even if it's caught in a larger market route, to this, uh, to this reported raise that they're trying to do, to me shows that, like, we're probably closer to that than a lot of people think. 34:53 Unless these labs and the NeoClouds, uh, which we can include SpaceX in this at this point, or AI cloud providers, uh, maybe, maybe some people would argue they're a hyperscaler. 35:03 Um, unless they can just continue to just absolutely smash it with earnings and show that their revenue growth for these cloud segments is just beyond, uh, uh, beyond and above what most people would expect, um, 2027 could be interesting. 35:18 Like, if AI CapEx spending starts slowing down in 2027 and investors and, and, uh, financiers show that they're getting a little leery about throwing cash at this thing, you, you could be looking at, um, some pretty tur- uh, tumultuous market conditions. 35:33 But, you know, that's kinda just my, my final take on all of this is at a certain point, 40 billion to 100 billion stops being, "Yeah, sure," and starts being, "Mm, I don't know. 35:46 Like, are you gonna be able to pay this back?" And the default swap spreads are kind of telling that story for us. 35:51 And just one note on that, um, 'cause I had said that it was close to Oracle, and I just wanna give some context for that. So five-year spreads on credit default swaps in late August for Oracle were 203 basis points. 36:03 SpaceX were 100 and, uh, 194 to 195 basis points as of yesterday. To put that in context, Meta 93 basis points in late August, Amazon 66 basis points, Alphabet best of the bunch, 58 basis points in late August as well. 36:20 So, you know, ultimately, SpaceX up there with Oracle in terms of investors saying, "You seem to have a higher likelihood of default than some of your peers." We are not done talking about SpaceX. 36:36 We're gonna talk about SpaceX by way of Memphis, Tennessee data centers. As Roger McAskill- Yeah... in the, in the audience says, "SpaceX out here ruining data centers." For everyone, you are not wrong. 36:47 You don't know how true that is because this next segment is a direct consequence of Colossus, actually. Oh, yeah. It's a doozy. 36:54 We've been critical of Colossus at its, uh, local level and also pretty admirable of SpaceX's ability to execute from a- Company level. But before we go to Memphis, Tennessee, a word from our sponsor, Ligos. 37:09 [firework explodes] If you're using a Bitcoin-backed lender, you need to ask yourself one question, and if you can't answer it, then you need to look elsewhere. Do you know where your Bitcoin is? 37:24 If you can't answer that question, like I said, you need to do yourself a favor and look for an alternative, and our preferred alternative for Bitcoin collateralized lending is Ligos Finance. 37:34 That's right, Ligos Finance is Blockspace's pre- preferred Bitcoin-backed lending provider because they use Bitcoin native smart contracts to make sure that you always control your keys. You heard that right. 37:45 With Ligos, you never relinquish self-custody of your Bitcoin during the lifetime of your Bit- Bitcoin collateralized loan using their services. 37:53 With Ligos Finance, there's no rehypothecation, no wrapping, no bridging, just competitive rates as low as 8.5 to 10% APR. Go to ligos.finance to learn more. 38:09 Fistfights break out at a city council meeting in Memphis, Tennessee this week as citizens clash with the city council about data center permitting. 38:20 More than 200 people showed up at a city council meeting, where on the agenda was a one-year freeze on new data center permits inside city limits. 38:31 This one-year freeze on data center permits had already passed twice during the first and second readings earlier this summer. 38:39 This past Tuesday was supposed to be the final vote, but the city council voted 7 to 5 to delay the final vote until October 20th. As I said, more than 200 people showed up, and they let the city council 38:55 hear their opinions. Chants of "Sellouts" filled the air. Council members were walking out of the meeting. Uh, reports say that you actually couldn't even hear what the city council was trying to say during the meeting. 39:10 Police ended up clearing parts of the chamber. There was even a fistfight, as shown here, in the lobby. Nobody was hurt, and it's unclear what started it, so we don't have those details, but yeah, punches were thrown. 39:23 For some context, Memphis has been the city where Elon Musk builds AI data centers fast, standing up xAI, now SpaceX's Colossus I and II data centers in record time starting back in '24 and '25. So how did we get here? 39:39 The bill's sponsor, one JB Smiley, says he wrote the bill after learning xAI was building a fourth Memphis data center. He found this out from Elon's post on X. 39:52 Smiley says, "When the community shows up, we just have to listen." 39:57 The opposition to the bill, to this bill says, if you freeze Memphis, the data center builders, the data center builders will just go across the county line. 40:05 Alternatively, opponents say that to skip the freeze and focus on writing real rules, like stricter review on anything over 20 megawatts, is preferable. 40:17 Again, for context, each SpaceX Colossus site is over 300 megawatts. Here is the catch for this particular permit freeze. 40:25 Colossus I and II would already be grandfathered in, but only if they stay the size that they currently are. 40:33 Meanwhile, the TVA, who is the power supplier, recently approved, as of earlier this year, another 150 megawatt expansion for SpaceX, possibly even higher. 40:44 This is not really just a SpaceX story, Colossus data center story. The day before this vote on Tuesday, a developer called 5C filed a $575 million permit 40:58 to turn an abandoned corporate headquarters building into another data center. 41:02 The city council is scheduled to go back in session, I kid you not, in 30 minutes from the airing of this broadcast, so if you're not listening to this live right now, I bet you that by the time you listen to this, uh, on podcast or on YouTube later, another round of videos or events are probably hitting the timeline, so don't pause this, go check Twitter. 41:25 So Colin, we're the data center defenders here. We've been critical of, of this particular site, Colossus I and II, but there were 100, 200 people in the room yesterday chanting "Sellouts." 41:40 At what point do we admit that data center proponents, data center builders are losing the public opinion game bad? I, I think we admit it right now. Yeah. 41:51 I think that ultimately this is clearly a battle that the defenders are on the wrong side of history according to the majority of Americans, right? 42:03 That being said, I would like to understand what their complaints are, because so far the best complaint I have found against Colossus, and I have a video queued up here that I'll queue in a minute, is that it actually is incredibly noisy because of the turbines on site. 42:20 It was slapped together in 122 days. They took a lot of shortcuts. In fact, those shortcuts ended up kind of, uh, biting them in the ass because I believe they had downtime, they had, like, flooding problems. 42:32 They ended up firing, or Musk ended up firing his engineers and designers for the site and then tapped a, a, uh, Falcon rocket engineer actually for the, for the next phase of the build. 42:45 But I would like to know specifically, like, what, what are... Like, when you're saying sellout, okay, well what, what's your problem with the site? 42:52 Because I think that there is, uh, d- I mean, do you have more info on that? I would- I have a little bit of info. So one, the noise is particular egregious. 43:01 However, my cursory look of, like, who was organizing, like, the protests at this particular city council meeting, the noise was not really, like, the main... uh, pushback. 43:13 It was actually about air quality and the air permitting required for more turbines, and that is actually downstream of one, 43:23 uh, MCAP is the organization, which is the Memphis Community Against Pollution, which is a grassroots environmental justice group based in the South and Southwest Memphis neighborhoods around the XAI sites. 43:35 KeShawn Pearson is the president and has been one of the loudest voices against XAI's gas turbines. Um, main- And a lot of, and, uh, to quote, it's, uh, "You are, uh, you're making us breathe poison." 43:50 So it's interesting that it's actually that the, the main, like, flare-ups were about air quality, um, when that certainly is an issue because Elon stood up, I say Elon, the company stood up these, these turbines without proper, uh, air permitting, environmental permitting. 44:07 How much does that actually really do? Like, what is the gap between, like, actual real harm and then EPA standards and local air permitting standards? Think it's a little ambiguous. 44:19 Um- But wasn't that part of this site specifically? It's like, didn't they set up the turbines across state lines so that they wouldn't have to deal with the permit? 44:26 Or was it in, did they put the turbines in Mississippi, or did they put them in Tennessee? They put the turbines across the border just south in Mississippi. Yeah. So, yeah. 44:34 To me, this seems kind of like, okay, um, yeah, you can say that that's not great, but is it illegal? Right. Well, I mean, uh, you know, I- It, it's not- There, there's plenty of law- Okay. 44:46 Maybe not, maybe it's not illegal, but the thing is, like, if this is a letter of the law versus spirit- It's not gentlemanly. 44:52 Well, look, I mean, air doesn't, like, go up to the, you know, border of Mississippi and Tennessee and go, "Oh, can't cross state lines." Uh, but- Uh, yeah, no, I understand that conceptually. Yeah. 45:02 My point is, though, for an actual legal challenge of this, if it, it complies with Mississippi regulations, it's good. 45:10 I think, I think that there is a little bit of, like, ambiguity here, which probably gets into really obscure- Yeah... things about generator and turbine permitting. 45:18 But the thing is, like, you know, those are not within Memphis city limits necessarily, or the turbines aren't. 45:24 The data centers are, but the turbines across the state border which power the data centers, uh, don't follow the same rules. 45:31 Um, I'm gonna point out that MCAP, this particular community grassroots group, uh, actually started against Memphis Community Against the Pipeline, and it beat the, the One Bi-Halia oil pipeline, uh, which, uh, was gonna pass through the same part of town in 2021 and then broadened its name and mission to be generally anti data centers. 45:55 So, uh, a state representative, Justin Pearson, is one of the co-founders, and KeShawn is his brother. So, uh, there's a few thi- yeah, there's, there's a lot going on here. 46:08 So that was actually, like, some, like, a lot of people think, like, that it's just organic community pushback who shows up at these meetings. 46:17 Usually, there is some kind of, like, ringleader or group of people who, under a, the banner of some organization or even just, like, a Facebook page or something, rally people, and then that was what brought the bulk of the 200 people to this meeting. 46:33 Yeah, this- So they're kind of de facto, like, under... 46:36 And so it's like a lot of people think, a lot of people will try to map their own understandings or biases of what specifically, uh, communities are pushing back against onto the specific people of these city councils. 46:47 And I can say this from experience, I'm, like, kind of familiar with how these things go, and, like, the actual people who show up and get angry at a city council meeting, uh, often have, like, uh- No stake in the community... 46:59 unrelated... Yeah. [laughs] Well, they're just like, you know, like, the national story is, is very different than, like, what it actually looks like on the ground. So, um, the thing is- Yeah, and I, I-... 47:10 the people hate, the people don't want these, hate data centers, and they're calling the city council, like, uh, they're, they're accusing the city council of being in the pocket and selling the, the community out. 47:21 I don't know how much this is the community, the community council can do. 47:23 Sorry to keep going, again, but here, another thing is that, uh, the data centers, Elon data centers have generated significant tax revenue for the county. 47:33 In fact, the city council is currently trying to decide how to spend about $3.3 million, 47:39 which has been generated from the SpaceX AI property, not in aggregate, that's just the latest, like, amount of money that they've got that they're trying to figure out how to spend. Yeah. 47:47 So, so this, this is my problem with all of this. Like, you're saying selling out the community. Y'all, this is Memphis, Tennessee. There are not very many people champing at the bit to move there. Hey. 47:59 Businesses are not moving there. Like- It's got Beale Street. Beale Street, home of blues. Beale Street's great. Yeah. The barbecue in Memphis is some of the best you'll get in Tennessee as well. 48:07 Look, I'm not saying that, um, Memphis doesn't have things that are redeeming about it, but if you look at the growth of N- Nashville, Murfreesboro, Chattanooga- Yeah... 48:18 if you look at these places, they are crushing Memphis- Yeah... in terms of benefiting from the influx of, uh, of intrastate immigrants from, like, California, elsewhere, and stuff. 48:31 A lot of other problems come with that. I'm not gonna mention those here. But, like, Nashville has had a golden age in the last decade, and Memphis- Memphis has not... Memphis has not. Memphis has- Yeah... 48:41 continued to deteriorate. So, like, that's what bothers me about this, is you have these, like, high-minded activists who probably don't even live there. 48:48 They might have come in from Arkansas or Mississippi or somewhere else down south- Which, granted, those are nearby. That's, again, you know, you're breathing the air of those dirty, dirty turbines. 48:56 Yeah, but, but, o- okay, no, no. Come on, now. Look, here's the thing. I know, I know. I'm gonna push back on that, too, because- Yeah... in Memphis, 29 to 45% of the regional mix is coal and nat gas. Yeah. 49:06 So, like- I don't- these turbines are not going to, uh, th- they're, they're not going to probably meaningfully add to any of the pollution. 49:16 Um, and y- you're, you're, you're accusing these city council peoples of, of selling out your town. Guess what, buddy? No one's moving there anyway. And you know what? 49:25 I don't even know if you're from there, because these non-profits will bus people in to these protests. They did it for... I'm not trying to get political, but it's just true. You can look it up. 49:34 They did it for the No Kings protests. They did it for protests, uh, leading up to the 2026 election. 49:40 Like, this just, this stuff happens in these non-brella organizations, and under, you know, this kind of shadowy web of, uh, of, uh, funding that gets passed around to all of these non-profits and all of these non-governmental organizations. 49:55 Um, all that said, I do wanna play one clip here really quickly- Play the clip... 50:00 'cause I think that this is the strongest argument against having, uh, this data center in your backyard, and for maybe the, uh, SpaceX cleaning things up for the rest of the community. It's September 19th at 12:30 AM. 50:17 That's half a mile. Turn, turn it up. Turn it up. Turn it up. Can people not hear this? I- yeah, I can hear it a bit. I don't know, could you, could you hear that all right? 50:34 I think- Yeah, I could hear it, but-... you had audio... I mean, like, think about it. So that, that's like a low roar. That's half a mile away. Look, I'm not gonna go to bat for... 50:42 I will say that you're dumb for environmental concerns. I'm gonna say that you're dumb for water concerns. 50:47 If you think that these, these data centers aren't doing anything to help your locality, I'm gonna say you're dumb because the revenue, the tax revenues and some of the additional money that they put into community engagement is good. 50:58 But I'm not gonna say you're dumb for not wanting to live next to that, because that would drive me insane. That would drive me crazy. Yeah, the... Uh, yeah, I'll probably use softer language. 51:07 I think that the air permitting concerns is, like, not really that significant. 51:13 It's more of a procedural thing that is, that is a challenge 'cause there's the EPA and the various permitting organizations wanna have a process to, like, register and track emissions from various industrial actions, and I don't really think it would materially impact your community. 51:33 The noise, though, 51:35 uh, you know, if you are a pers- if you're an American citizen who owns property, and you've built and owned a house and everything, and that goes in next door, um, like, that could, that not just significantly improved your life, that could, like, ruin your life. 51:52 Um- Yeah... sleep, imagine trying to have a family. You can't leave, 'cause then you can't sell your house, 'cause who's gonna buy the house? 51:58 Like [laughs] Then who's gonna buy your house with, like, mortgage rates at s- 7.6%? Like [laughs] Who's gonna buy your house? Yeah, and, like, they'd go and they're like, "Oh, wow, it's noisy. That wasn't on Zillow." 52:08 So, like, yeah. Um- I think that's a legitimate problem. I think this is maybe where something like Bloom could offer a good bridge here. They're not gonna be able to do it for everything. 52:21 Um, you know, obviously, like, SpaceX, like, they're probably not gonna do that. Yeah. This is a legitimate concern. 52:27 Last thing I wanted to show here, Charlie, was the ac- some footage from the city council meeting, and, um, we just gotta play this clip in full because there's, there's just some absolute gems here. 52:38 This is from TMZ of all places. Thank you, TMZ. Real news. Uh-uh, uh-uh. Uh-uh. Those items will be held. Those items will be held. She can't stay here for those items. We're gonna have to have a recess. 52:52 We're gonna recess the meeting. [gavel pounding] Right now, this is- Item is in recess. Mu- meeting is in recess. [laughs] That's not true. That's misleading. Can we have order? 53:08 [laughs] That's not- Clear it out. Officers, we need to clear it out. Get them out. Where our officers at? They in the back. Yeah, so they cleared the meeting out. 53:26 [gavel pounding] They cleared the whole city council meeting. Clear out the auditorium. Fight back. Ain't no power like the power of the people when the power of the people fight back. Did you, did you hear her though? 53:41 She's like, "Where are our officers at? They in the back?" It's like the officers, at first, they don't do anything 'cause there's just this swarm of people, but they just keep chanting this, um, the, yeah. 53:55 Ain't no power like the power of the people when the power of the people fight back. Yeah, ain't no power like the power of the people when the power of the people fight back. 53:58 I mean, this is just performative outrage by a bunch of young, uh, you know, y- a bunch of youngsters who don't actually know how the world works. 54:07 I, I, I will go, I will be brash and say that much, because it reminds me of a lot of the performative outrage when I was in college. 54:15 And when you're young and hip, you don't wanna be the guy who s- says that actually Elon Musk isn't Hitler, because then your friends will hate you. 54:24 [laughs] And then you graduate, and then you realize that, well, he might not be the best guy in the world. He's certainly not Hitler. 54:31 And I hate to do this, but I would refer to you, refer you to- Gonna get us demonetized on YouTube... 54:36 I'm gonna refer you to Peter Thiel's podcast where he talks about the difference between good and bad, as good not being a, uh, a, a, good's antonym not being evil, but being bad, and his kind of Faustian breakdown between Marlowe and Goethe's Faust. 54:55 Well, yeah, I mean, I also think that, um, God, I'm gonna appeal to, like, civil procedure, but, like, uh, causing a ruckus at a city council meeting, uh, is, uh, very ineffective. 55:13 200 people just showing up, warm bodies in a room, uh, does do a lot. Uh, I think you do that a few city council meetings, and then you show that you're, the constituents of each city council member don't want this. 55:27 They listen to that, and then they vote. Um, I think that should be the strategy because the re- it's really like, what do your constituents do? Do you vote in favor of what your constituents, uh, want? 55:38 And I think that's where the whole Republic process, oh, God, I'm becoming an institutionalist, Colin. I can't- Becoming a monarchist. Oh, no, I'm not. [laughs] Not a monarchist. 55:49 Thank you for listening to Block Space Live, hosted by Charlie Spears and Colin Harper of Block Space Media. We are a Computes daily live show. 55:55 We go live at 1:00 PM Eastern every single week day, featuring stories on AI data centers, and now Memphis City Council meetings. If you want more Block Space content, you can go to our website, blockspace.media. 56:09 Get our newsletter in your inbox every single day, covering the stuff we talk about on the show and more. We will be, uh, here again tomorrow. 56:20 We got Brett Harrison of Architect AI, and we will be talking about what else? Compute futures and liquid compute. Thank you for watching Block Space Live. I'm Charlie. I'm Colin. We'll see you tomorrow. 56:47 [dramatic music]