Susie: Usually, angry neighbors and zoning boards spell absolute death for big infrastructure plays. But Applied Digital's out here basically treating NIMBYism as a competitive moat.
Miro: Right, because every time a town blocks a hyperscaler from building a data center, the dirt they've already energized gets more expensive.
Susie: Exactly. Welcome to October 7th, 2026, and you're listening to Earnings Unscripted. We're unpacking Applied Digital's Q1 results, ticker APLD, and the headline's just, massive top-line acceleration. Revenue shot up three hundred twenty-two percent year-over-year to about three hundred forty-two million. But Miro, it's paired with a pretty heavy eighty-two cent GAAP loss per share, which widened a lot from last year. So what exactly did they buy with all that red ink?
Miro: Mostly? They spent a staggering amount on capital expenditures to build these massive purpose-built AI factories. The core operating performance actually looks decent. Adjusted EBITDA flipped firmly positive. But that net loss got blown out by huge non-cash hits on derivative marks, plus their interest expense jumped astronomically because of new facility financing.
Susie: So basically, the underlying business generates cash now, but the sheer cost of scaling up the footprint's just burying the headline number?
Miro: Exactly. It's real growth, just heavily disguised by the financing and paper math.
Susie: And the stock tape reflects that exact confusion, right? The stock suffered a brutal quarter, dropping almost forty-six percent going into this print. Then earnings day hits, it gaps down, they host the call, and it just, it rips hard in after-hours to basically break even against the prior day's close.
Miro: Which, frankly, feels like wild action. People were clearly panicked about the debt loads, but Wes Cummins calmed them down on the call by talking up their massive backlog and aggressive expansion pricing. And since we're digging into a highly volatile tape today, keep in mind our chat reflects purely personal opinion, certainly not financial advice for anyone's portfolio.
Susie: Right, don't risk the farm based on our theories. But speaking of farms, I loved how they handle the local community pushback we mentioned earlier. To get these gigawatt sites approved, they go into rural places like Ellendale, North Dakota, and just shower the town in tax revenue. They literally bought laptops for every, sorry, every kid in one district just to buy goodwill.
Miro: It's incredibly pragmatic. They're sidestepping the years-long moratoriums hitting the rest of the sector by fully subsidizing the local fire department. And the payoff's insane. They expect to put hundreds of megawatts into service this year while their competitors stay stuck in local planning meetings.
Susie: It makes sense. If you've got the only functioning outlet on the wall, you get to charge whatever you want. But there's a bizarre ghost in this machine, Miro. I'm looking at their non-GAAP reconciliation, and management's deliberately scrubbing out tens of millions in revenue from their cloud subsidiary like it doesn't even exist. I really want to hear why they're trying to hide that segment.Miro: It's not hiding, it's just rebranding by omission. ChronoScale's their public subsidiary handling cloud services, and it generated about forty-one and a half million this quarter. But management's completely stripping it out of their non-GAAP reporting because they claim its profile, quote, may obscure underlying trends in the performance of core operations.
Susie: Okay wait, that feels slightly backward to me. Usually tech companies desperately want to highlight cloud revenue because Wall Street loves high-margin software multiples. So you're saying they actually prefer to be valued strictly as a physical landlord? Like just concrete and power?
Miro: Right, but not just any concrete. They want the valuation of a pure-play AI factory operator. If you mix in cloud hardware sales, which drove over twenty-two million in cost of revenues for that segment, it muddies the underlying infrastructure story. They want investors focused entirely on the massive physical terminal value of their grid access. Even though they're still quietly targeting a massive one billion dollars in annual recurring revenue for that cloud business down the road.
Susie: Just casually forecasting a billion dollars. Sure. But speaking of the physical grid, I'm stuck on how you even build these structures right now. Because generative AI is moving so fast, don't the chips become totally obsolete every two or three years?
Miro: Yeah, and that's exactly the structural shift they're making with their tenant fit-outs. Or TFO. The customers literally don't know what servers they're going to need a year from now. So Applied Digital explicitly abandoned the standard five-year GPU lifecycle design. On the call Wes Cummins literally said, quote, We have taken a different approach and are building for the long term with 30-plus year horizons.
Susie: Okay, walk me through that. Because if you're building a massive data center, don't you have to know exactly what kind of hardware goes inside to plan the plumbing and the power? How do you have a thirty-year horizon for something that changes every six months?
Miro: You build fluid infrastructure. Traditionally, yeah, you'd design a facility entirely around a specific cooling loop for a specific processor. But they're creating highly flexible designs to handle varied networking typologies, no matter what chip eventually arrives at the loading dock. They basically provide the mega-shell and the raw power. Then, once the customer finally decides what hardware to rack, Applied Digital procures and installs it at the last possible minute.
Susie: Ah, got it. So they decoupled the actual building from the rapidly aging tech inside it. Which means that fit-out service is just, what, something they do on top of the lease?
Miro: Precisely. It's basically a cost-plus markup model. And it's booming right now. They recognized roughly a hundred fifty-seven million just from those tenant fit-out services this quarter alone. It completely de-risks the hardware cycle for them.
Susie: But if you own an energized, community-approved fortress that can support basically any AI chip for the next three decades, and power elsewhere is hopelessly bottlenecked... how much leverage does that actually give you when it's time to negotiate those lease renewals?Miro: An enormous amount. I mean, they're basically looking at signing roughly two hundred fifty megawatts of expansion leases by the end of calendar twenty twenty-six, and they expect pricing on those to jump more than fifteen percent compared to older deals. So the grid bottleneck is absolutely paying off for them.
Susie: That's a massive premium just for the privilege of plugging a server in. But honestly, with the US grid getting this tight, where do they even go next? They can't just keep squeezing fresh power out of North Dakota forever.
Miro: Well, they're heading across the Atlantic. They just signed a major agreement for capacity in Finland. It mirrors their existing playbook completely, offering cold weather and favorable power fundamentals, and it really helps them march toward their target portfolio size of around three point seven five gigawatts by the end of the decade.
Susie: Okay, so the physical expansion plan is incredibly ambitious. But I think the real structural risk here isn't the power grid, it's the actual tenants. You mentioned earlier how concentrated their customer base is. If CoreWeave and a few select hyperscalers are carrying this entire backlog, what happens if their big AI bets don't actually generate cash?
Miro: That's the real test. Management knows it too. Wes Cummins made a very telling point on the call. He said, "We view our customers' ability to convert their AI investments into durable and sustainable earnings as an important metric as we make decisions on who we will build for."
Susie: Oh wow. That's essentially a commercial landlord running a venture-capital stress test on a hyperscaler before handing over the keys.
Miro: Exactly. They know they're building permanent infrastructure for an industry that hasn't proven its cash cycle yet. They have to pick the survivors.
Susie: Well, it's definitely going to be wild to watch them execute this ramp-up over the next twelve months. We'll definitely keep an eye on how those European sites develop. Thanks for breaking it all down, Miro.
Miro: Always a pleasure. Catch you later.Susie: If you enjoy unpacking these massive market shifts without having to read the footnotes yourself, it's definitely worth making our feed a regular stop. And for full transparency, the two of us are entirely synthetic, and this whole podcast is an AI experiment designed to turn today's seventy-plus pages of new earnings filings into audio the same day they drop. We'd love to hear your feedback or which ticker we should cover next, so just send a note over to earningsunscripted@datamadness.ca.
Miro: Talk to you then.
Miro: Right, because every time a town blocks a hyperscaler from building a data center, the dirt they've already energized gets more expensive.
Susie: Exactly. Welcome to October 7th, 2026, and you're listening to Earnings Unscripted. We're unpacking Applied Digital's Q1 results, ticker APLD, and the headline's just, massive top-line acceleration. Revenue shot up three hundred twenty-two percent year-over-year to about three hundred forty-two million. But Miro, it's paired with a pretty heavy eighty-two cent GAAP loss per share, which widened a lot from last year. So what exactly did they buy with all that red ink?
Miro: Mostly? They spent a staggering amount on capital expenditures to build these massive purpose-built AI factories. The core operating performance actually looks decent. Adjusted EBITDA flipped firmly positive. But that net loss got blown out by huge non-cash hits on derivative marks, plus their interest expense jumped astronomically because of new facility financing.
Susie: So basically, the underlying business generates cash now, but the sheer cost of scaling up the footprint's just burying the headline number?
Miro: Exactly. It's real growth, just heavily disguised by the financing and paper math.
Susie: And the stock tape reflects that exact confusion, right? The stock suffered a brutal quarter, dropping almost forty-six percent going into this print. Then earnings day hits, it gaps down, they host the call, and it just, it rips hard in after-hours to basically break even against the prior day's close.
Miro: Which, frankly, feels like wild action. People were clearly panicked about the debt loads, but Wes Cummins calmed them down on the call by talking up their massive backlog and aggressive expansion pricing. And since we're digging into a highly volatile tape today, keep in mind our chat reflects purely personal opinion, certainly not financial advice for anyone's portfolio.
Susie: Right, don't risk the farm based on our theories. But speaking of farms, I loved how they handle the local community pushback we mentioned earlier. To get these gigawatt sites approved, they go into rural places like Ellendale, North Dakota, and just shower the town in tax revenue. They literally bought laptops for every, sorry, every kid in one district just to buy goodwill.
Miro: It's incredibly pragmatic. They're sidestepping the years-long moratoriums hitting the rest of the sector by fully subsidizing the local fire department. And the payoff's insane. They expect to put hundreds of megawatts into service this year while their competitors stay stuck in local planning meetings.
Susie: It makes sense. If you've got the only functioning outlet on the wall, you get to charge whatever you want. But there's a bizarre ghost in this machine, Miro. I'm looking at their non-GAAP reconciliation, and management's deliberately scrubbing out tens of millions in revenue from their cloud subsidiary like it doesn't even exist. I really want to hear why they're trying to hide that segment.Miro: It's not hiding, it's just rebranding by omission. ChronoScale's their public subsidiary handling cloud services, and it generated about forty-one and a half million this quarter. But management's completely stripping it out of their non-GAAP reporting because they claim its profile, quote, may obscure underlying trends in the performance of core operations.
Susie: Okay wait, that feels slightly backward to me. Usually tech companies desperately want to highlight cloud revenue because Wall Street loves high-margin software multiples. So you're saying they actually prefer to be valued strictly as a physical landlord? Like just concrete and power?
Miro: Right, but not just any concrete. They want the valuation of a pure-play AI factory operator. If you mix in cloud hardware sales, which drove over twenty-two million in cost of revenues for that segment, it muddies the underlying infrastructure story. They want investors focused entirely on the massive physical terminal value of their grid access. Even though they're still quietly targeting a massive one billion dollars in annual recurring revenue for that cloud business down the road.
Susie: Just casually forecasting a billion dollars. Sure. But speaking of the physical grid, I'm stuck on how you even build these structures right now. Because generative AI is moving so fast, don't the chips become totally obsolete every two or three years?
Miro: Yeah, and that's exactly the structural shift they're making with their tenant fit-outs. Or TFO. The customers literally don't know what servers they're going to need a year from now. So Applied Digital explicitly abandoned the standard five-year GPU lifecycle design. On the call Wes Cummins literally said, quote, We have taken a different approach and are building for the long term with 30-plus year horizons.
Susie: Okay, walk me through that. Because if you're building a massive data center, don't you have to know exactly what kind of hardware goes inside to plan the plumbing and the power? How do you have a thirty-year horizon for something that changes every six months?
Miro: You build fluid infrastructure. Traditionally, yeah, you'd design a facility entirely around a specific cooling loop for a specific processor. But they're creating highly flexible designs to handle varied networking typologies, no matter what chip eventually arrives at the loading dock. They basically provide the mega-shell and the raw power. Then, once the customer finally decides what hardware to rack, Applied Digital procures and installs it at the last possible minute.
Susie: Ah, got it. So they decoupled the actual building from the rapidly aging tech inside it. Which means that fit-out service is just, what, something they do on top of the lease?
Miro: Precisely. It's basically a cost-plus markup model. And it's booming right now. They recognized roughly a hundred fifty-seven million just from those tenant fit-out services this quarter alone. It completely de-risks the hardware cycle for them.
Susie: But if you own an energized, community-approved fortress that can support basically any AI chip for the next three decades, and power elsewhere is hopelessly bottlenecked... how much leverage does that actually give you when it's time to negotiate those lease renewals?Miro: An enormous amount. I mean, they're basically looking at signing roughly two hundred fifty megawatts of expansion leases by the end of calendar twenty twenty-six, and they expect pricing on those to jump more than fifteen percent compared to older deals. So the grid bottleneck is absolutely paying off for them.
Susie: That's a massive premium just for the privilege of plugging a server in. But honestly, with the US grid getting this tight, where do they even go next? They can't just keep squeezing fresh power out of North Dakota forever.
Miro: Well, they're heading across the Atlantic. They just signed a major agreement for capacity in Finland. It mirrors their existing playbook completely, offering cold weather and favorable power fundamentals, and it really helps them march toward their target portfolio size of around three point seven five gigawatts by the end of the decade.
Susie: Okay, so the physical expansion plan is incredibly ambitious. But I think the real structural risk here isn't the power grid, it's the actual tenants. You mentioned earlier how concentrated their customer base is. If CoreWeave and a few select hyperscalers are carrying this entire backlog, what happens if their big AI bets don't actually generate cash?
Miro: That's the real test. Management knows it too. Wes Cummins made a very telling point on the call. He said, "We view our customers' ability to convert their AI investments into durable and sustainable earnings as an important metric as we make decisions on who we will build for."
Susie: Oh wow. That's essentially a commercial landlord running a venture-capital stress test on a hyperscaler before handing over the keys.
Miro: Exactly. They know they're building permanent infrastructure for an industry that hasn't proven its cash cycle yet. They have to pick the survivors.
Susie: Well, it's definitely going to be wild to watch them execute this ramp-up over the next twelve months. We'll definitely keep an eye on how those European sites develop. Thanks for breaking it all down, Miro.
Miro: Always a pleasure. Catch you later.Susie: If you enjoy unpacking these massive market shifts without having to read the footnotes yourself, it's definitely worth making our feed a regular stop. And for full transparency, the two of us are entirely synthetic, and this whole podcast is an AI experiment designed to turn today's seventy-plus pages of new earnings filings into audio the same day they drop. We'd love to hear your feedback or which ticker we should cover next, so just send a note over to earningsunscripted@datamadness.ca.
Miro: Talk to you then.
