Cipher Digital Inc. (CIFR) Earnings Call Transcript & Summary

Okay. I think we're going to go ahead and get started. It's an exciting time to be speaking with Tyler from Cipher Digital. Just a quick background, Stephen Byrd. I lead our sustainability and thematic research at Morgan Stanley. I spend a lot of time on this theme. It's been a very exciting one over the last couple of years. One, just brief disclaimer I have to read. For important disclosures, please see morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley representative. So I'm done with my public service duty. Tyler, thank you for being with us.

Thank you.

We've talked quite a bit over the last several years about the opportunity to serve as really the infrastructure for, perhaps, the most important industry we'll ever see. Cipher is now Cipher Digital, really fitting where you guys are headed. Let's start at a high level. Would you mind just kind of giving us an overview of the company, your assets, your strategy, and then we're going to dig in from there.

Sure. So we are a developer of HPC data center campuses. The rise in the demand for large interconnects at single sites to drive AI workloads has driven a change evolution in our business. We began life as a developer of bitcoin mining sites, and that's really put us in a very strong position to be ahead of everyone in terms of building up a land bank of very attractive sites because that business requires a lot of power and the power has to be cheap. And so the power interconnects needed to be located in nontraditional locations because by very definition, if you've got to pay the minimum amount in electricity costs, you have to go to places where there's oversupply of power, which is off in places like West Texas. In the course of developing a lot of expertise for originating and developing those sites and interconnects, we watched what has been happening in the last several years with the growth of LLMs and the sort of attached growth overall and desire for large campuses or single tenants. And I think we had this advantage of being a different kind of data center developer looking for opportunities for large campuses in nontraditional places and we saw this future, and in fact, Stephen, I was thinking about it, I think it was 1 year ago at this conference, I said something that I'm sure very few people agreed with at the time was that we believe the future data center capital of the world would be West Texas. Not as many people disagree with that. Now I don't know that I've got everyone agreeing with me necessarily, but there is so much desire now for these large interconnects at a single campus. And we had the foresight to acquire a lot of them. And so that's led to this transition in the business where we are transitioning out of bitcoin mining, and we will solely be dedicated to originating and then developing these campuses for hyperscaler or potentially other tenants.

Good. That's a good start. I mean you've made a lot of good commercial decisions, and that's shown up in contracts, in the stock price. And I think if anyone saw our initiation, I think people see where we are -- we share the bullishness. We see the upside opportunity that's quite evident. I wanted to really get your view on not just the hyperscalers, but customers more broadly, the nature of your dialogue, the sense of urgency to get power. I wish I could be a fly on the wall in the room when you're talking with these folks. But would you mind just giving us a broad sense for that sense of urgency, where we see that going from here?

Sure. So I think it's best described as it's kind of a multivariable calculation on their part. So there is certainly extremely elevated demand from tenants tied to time line to availability of power. So I think we've gotten questions, we've signed some very favorable lease terms. We have one very large project we're working on for fluid stack and Google. It's a 300-megawatt gross campus we are building. We are building a separate 300-megawatt gross campus for AWS. We have some extremely favorable and certainly elevated to what those counterparties would traditionally pay in a lease for those sites. And I think that was largely driven by the fact that there is so much going on in the space where there is a race for power. That's driven by their own strategic initiatives, maybe they're working with an anthropic or an open AI and trying to bring gigawatts of availability to the table when they're making equity investments or whatever might be going on drive that frenzy and the speed. But if you can deliver a time line that is near term, and let's call it, I think they start to get very interested if it's sort of sub 18 months and it starts to get urgent, if it's 15 months. And if you have campuses like we had that were already energized with fully built substations, so you're really just talking about a construction time line for a core and shell and putting all the equipment in it. I don't want to say it was name your price, but you're closer to that than the traditional sort of framework for what might a tenant might be willing to pay. And I think that environment persists today with that same caveat that I think, if you've got 2026 power or early 2027 power, that window of 18 and sort of 15 months and 1 year just keeps rolling forward. So when we see the comments that we've seen from folks like Satya Nadella that seem to be indicative of -- I'm paraphrasing, but effectively endless demand right now. That gives us great confidence that this desire to sort of pay whatever is necessary to secure a lease continues to roll forward with this moving time line. So what it doesn't mean is like we could just name any price we want for a site that's available 2.5 years from now. But we're working pretty hard on some pretty big interconnects. So in addition to the 600 megawatts, we're constructing right now, we are marketing 370 megawatts of sites where we have firm interconnections already. That's across three sites. We are pretty far along. I mentioned on our earnings call, we've identified a preferred tenant for a 100-megawatt site that we're working on. So it's not finalized, but progressing. And we've got multiple potential tenants looking at the other 270 megawatts that I'm including there that are across two other sites. So fair to say those are entering this window of near-term power, where I think I'm optimistic we will get very favorable terms again when we come to leases for those sites. And then we've got a lot of power that is scheduled, some very large campuses we are currently anticipating available in '28. And I'm hoping that what I have heard from the -- about the demand environment from folks continues to roll forward. And hopefully, later this year, we're moving forward on more favorable leases for them.

Yes. I mean it's fascinating. A few things I wanted to kind of pull on there. I mean, first, just on the returns for investors, it is wild to go back a few years and then chart where we were. So before AI was a thing, when hyperscalers have come into the power world, let's say they wanted a renewable PPA we would see terms that would suggest unlevered yields of 7% to 8% would be sort of where the hyperscalers. And you're lucky if you don't have to get them warrants, it was very difficult. Then last summer, really, the dam kind of burst last summer as it were, and we started to see deals happen starting at unlevered yields of 12%-ish, going into the fall into the 15% level. Then we saw a deal in December that was above 18%, which is mind-blowing. And technically, I guess the deal we saw in January is 27%, though I don't expect to see that again. That was fairly unique for lots of reasons. But nonetheless, the trend is phenomenal. I share your view. It does feel like we're going to progress and that urgency is going to be there. If anything, the math we're doing suggests that the shortage of compute is going to get worse, not better. So many examples of that, but let me give you one. In December, a Google exec laid out their 5-year demand for compute. And we looked at that growth rate of compute implied compared to what our semiconductor analyst numbers are, the Google number is about 3x higher than our semiconductor numbers. And we can't even find enough power for our semiconductor analyst numbers, which is where you come into the equation. Given that though, I do want to dive in a little bit into securing additional power. I do -- would you mind just kind of also talking us through ERCOT and the situation there, where you all stand? I think we're going to have a good outcome there for you, but it's caused a lot of agita.

Yes. I think this is one that -- I think in general, the folks in the audience know this, it feels like the stock markets and equity investors are just nervous around this whole space right now. It's a nervous macro backdrop.And I think people are quick to jump to negative conclusions around this space pretty quick, and we just have a volatile market. And so I think this feels like one of those situations where we've been meeting with a lot of investors this week and I've explained to them that we think what's going on in ERCOT is very good for our business and very supportive. So for those that don't know, ERCOT is inundated with interconnection requests, largely driven by data centers. I don't know that I have the exact up-to-the-minute numbers, but I believe there's about 240 gigawatts in the load interconnection queue in ERCOT. And historically, the way they've decided on granting interconnects is to do a series of studies around the stability of the grid and allowing that particular point of interconnection to have those megawatts and assuming the operator feels comfortable granting that and moving forward, and they've sort of done it on a one-off basis. The other thing is that there has not historically been a requirement that folks requesting an interconnect have to put a lot of skin in the game. So you've ended up with a lot of sort of grid speculators on -- as they see this growth of power that they can have duplicative requests, and they may just be someone that's undercapitalized, speculating. And so it's kind of anyone's guess how much of that 240 number is real. We have a little bit of an in-house sort of wager going on it. I think with the requirements they're bringing, you may see 2/3 of that disappear. That's still a crazy amount of request. I mean, like the entire, I think, peak operation of the grid in Texas is like 85. So you're just going to add a whole another peak grid from the real request. So it's still a lot. So that gets back to kind of what's evolving in ERCOT. So as a way to deal with several issues around this, not the least of which have some better clarity and think about running a more stable grid as they interconnect these massive loads, they are putting in a series of requirements where they are going to approve interconnect requests in batches and look at what the sort of batch effect will be on the grid. And the requirements they've laid out to try to get the serious players to come forward and remain in the grid and get that interconnect is there a requirement that you demonstrate site control, so you own the land related to the point of interconnection. You have paid all the relevant deposits that you need to pay for the transmission distribution service providers. So you are a well-capitalized person that is serious and going to actually move. It's a good proxy for that because you're talking about millions of dollars that you'd have to pay. And then third, that you've submitted all those relevant studies, and they've been approved that yes, this is not going to crash the grid by adding this load in this particular location. As it relates to our sites, we highlighted on our earnings call, we have 3 sites in particular, where by the time they get to this batch process in June, we will have satisfied all the requirements they have laid out at this point. That's specifically for our sites. We have a site we call McLennan, one we call Mikeska, and one we call Colchis. We have requested 500 megawatts, 500 megawatts and 1 gigawatt at those sites. So it's not finalized yet. But given where things are tracking, we have done all the things we understand to get in that first batch, which has changed names. It's been batch [ 00A ] and batch alpha. And I'm just calling it the first batch. So our expected time line for interconnects at those site remains, our best forecast is 2028. And I think this will accelerate the finalization of those interconnections, which then puts us in a position to get a lease at those sites. People are not going to want to sign a lease if you don't have a clear interconnection for a particular site. So that's what's going on at ERCOT. There's also an element of even more deposits about socializing some of the costs around data centers coming online and avoiding your transmission upgrades being socialized by the rate payers. This is also a good thing. I mean, I think there's still some work on finalizing what that number exactly will be. Potentially, they're talking about a pretty big number. I think I've seen $100,000 per megawatt discussed. If that's the case, again, that's great for Cipher. We'll pay that. Happy to do it, get the interconnect and then get a lease and that will be recouped anyway as part of the budget to get the site ready. So investors, I think, should be looking at what's going on in ERCOT is a very thoughtful forward-looking way of dealing with this crush of data center interconnection, and it benefits particularly well-capitalized serious developers that have a demonstrated track record. Don't forget, we've built from scratch 5 greenfield data centers in Texas, all delivered on time. So like we are known as someone that is serious and connects our data centers and pays what they need to pay and delivers on time line.

That's a fantastic overview of the situation. And let that sink in for a minute what Tyler said. Even if we eliminated in the state of Texas, 2/3 of the proposed interconnect projects, the new demand would be equal to the state of Texas, which that's just how big this dynamic is. I did want to build on what you mentioned about sort of being a good citizen to the grid, managing that well. Obviously, we were talking beforehand that I think AI needs a new PR firm out there. It's just day-to-day, week-to-week, it's not great. One of the things I'm thinking about is being a good citizen, which -- and you come from a background as a bitcoin miner where you were used to being a good citizen to the grid. How will that evolve for data center design and operation? How do you see that?

Yes. I mean, look, it's an evolving story in the PR around AI. I mean, I think for is overwhelmingly impressed as I am from the raw intelligence of the innovators in that space, I'm not overwhelmingly impressed with the raw emotional intelligence quotient of the people in that space. And sometimes wish they would take a hiatus from podcasts for a few months because it may put us in a better spot. I mean, I think, yes, you highlighted something that working in bitcoin mining is a great training ground for the challenges that are coming for AI data centers. And so that's an easy political target that is misunderstood and is actually -- and thank you for highlighting Stephen, traditionally. Bitcoin miners are great assets to a grid because they're a very flexible load, and they can effectively shut down their data center in 1 minute. So at times where the grid is feeling stressed, that power gets returned and bring stability to the entire grid. Now I can tell you from experience, no one wants to have a nuance to understanding of that, and it's very easy if you're a politician to score points by picking on bitcoin miners as something that's bad. And if you have to get into a very technical discussion about how grids work and how you're actually good, it's very hard to win that argument. I think we see a similar backdrop with AI data centers that -- you've seen not -- in our case, I mean, again, this is yet another strength of areas like West Texas. We're building data centers that are in somewhat remote locations. It's not like we're putting a data center down in the middle of a neighborhood or something. We generally have sites that have hundreds of acres of space and that are surrounded by not much. Furthermore, from this kind of PR good citizen perspective, Texas has a setup where the tax system is related to public school payments. And so like, for example, on the PR side of just being a good corporate citizen, not just a good grid citizen, like our Barber Lake site, from our contract there with Fluidstack and Google, we will be the largest taxpayer in that county. Just from that contract, we will pay over $100 million to the Texas Public School system. So like we have a lot of -- in addition to having job fairs and getting to meet the locals and doing all the things about being a good neighbor, there are some very good objective facts about the growth of this industry, particularly in Texas. As far as being grid citizens, there will be elements that probably evolve in AI. A lot of the lessons we've learned about -- we are having discussions with some of our tenants around peak shaving and thinking about ways to be more flexible customers going forward. That all has a long way to develop. But suffice it to say, there will be a lot of scrutiny on the industry. It will be an easy punching bag for politicians. Unfortunately, we're used to that from a bitcoin miner. The good news is here is in a similar fashion, we have a very good story to tell about being good citizens. So nuanced understanding certainly defends our position and hopefully, AI in general gets a little bit better PR.

Yes. It is a good point about the school systems. In states like Texas and Louisiana, this can be life-changing to a school system that has very little money, the amount of money you're talking about. In my home state of Virginia in Loudoun, the schools went from very poor to very good over a period of years. I do want to talk about execution risk, which comes up a lot. It is interesting to me. So I talked to lots of investors from tech, generalists, industrial. I come from power. So I'm kind of used to delays to be very honest, like in the world of power, if you're delayed a month or 2, not a big deal, but there was a neocloud, I won't mention the name where the earnings call had one positive data point, one negative. The positive was they released some old ships at 95% of original pricing, which I thought was amazing, but they were a month or 2 behind on a project and the market went to that and the stock performed very poorly, and I was totally wrong. Can you speak to measures to minimize risk of overruns? And also maybe we can talk about just contractual approaches just kind of allocating the risk, if you don't mind?

Absolutely. So we get this question a lot, and I will put this back in the bucket of what we discussed earlier that it feels like investors are looking for something to get scared about. And so again, we have such an excited environment right now for new data centers. It's funny to watch even if something goes wrong and it gets delayed a month. I don't really have concerns that like desire for AI data centers will be gone in the extra month or something. But that's neither here nor there. Let's talk about our sites and how -- just how well I sleep at night because I'm not worried about this. So the things that would make me worried about delivering on an aggressive time line, I think we have reasonably aggressive time lines for our data center builds. I'll work backwards from the things that would really scare me. The long lead time items are things like substation transformers, right? Again, in the case of the two sites we are building right now on aggressive time lines, not only are the substations completely built already, they are already energized. So there would be this extra risk about the transmission distribution service provider have to do work on their side of the wires and what if they're late because they're busy on other things, and I can't control that. And gosh, that's scary. Completely derisked at our sites. So now we're working backwards to like, okay, what's the next pragmatic risk from our seat, I would be worried about would be what if war breaks out and something goes wrong and no one wants to finance anything and we don't have the money. In the case of the two sites we're building, we are completely funded. We have done two project-related bond offerings. They are incidentally both trading up in the market. So that does tell you that bond investors who are very focused on things like delay risks, think not only do we deserve money at a lower rate than pretty much all of our competitors, they think even the rate that we're paying is generous and our bonds are trading up in the aftermarket. So we're completely derisked from a sort of we don't have the money or someone's not releasing our next draw because we missed some milestone or something. So now you're talking about like okay, what I usually term like more garden variety type supply chain risks. And these happen on every single construction project. If you've ever remodeled the bathroom, I say this always happens, the tiles aren't ready on time or something. So like the kinds of things I'm talking about are, let's say we need a key piece of equipment that's supposed to arrive at the first week of April, and it gets delayed to the fourth week of April. And then we have a work crew that's coming in the second week of April that needed to work on the piece that was going to be there. Now they're not available the fourth week of April and they're pushed back. And so you suddenly have to rearrange this puzzle of work streams that you need to do to make sure the data centers delivered on time. Again, I expect things like that actually have not happened touch wood on either Barber Lake or Black Pearl thus far. I fully expect they will because they always happen. We have an expert team that has built many Tier 3 hyperscaler data centers, as well as the five data centers we have built for bitcoin mining. Incidentally, all five of those data centers were built in Texas. They were all delivered on budget and on time. And that was actually in a much tougher supply chain world. That was like post-COVID. There's no copper cabling in the United States, oops, what are we going to do? How do we solve this? So my point being back to your original question. we are derisked on the major things that would cause bad delays that would scare me. Things can happen. Could we have a pile up of those garden variety rigs that pushes it and it ends up a couple of weeks. It's possible to never happen with our team before. Possible. the penalties and the setup for late delivery under our contracts are, I think, pretty market standard. They run out 5 or 6 months, and there's typically increasing penalties. If you look at them, the penalties are none for the early day -- first couple of weeks or months. They might ratchet up to like 10% of a day's rent for another period. They might ratchet up then at the end of like a 5- or 6-month period might be a full day or 2 days rent. But basically, if you think about a 6-month delay, that's 50% extension to the time line we're talking about for a site that's already energized with a substation. -- that we already are fully funded. So like I'm not saying it's impossible, but it rounds to that. These are not concerning for our sites.

Very helpful. I mean even in the context of the value creation, like we -- the simple math, when we look at these projects, we calculate value creation depends on the project, call it, $10 to $18 a watt of value creation in a really nasty outcome where you're 10% over budget and you're eating all of that, that's $1 a watt off the base of $10 to $18 for a stock to trade way below that value creation. So I don't want to say it's 0, but that's helpful. I'm going to pause -- I have a couple more questions, but I'm going to pause to see if anyone here has a question that they'd like to ask. I'm going to keep going. Looks like none. Very good. The financing, I've been really impressed by the thoughtfulness in terms of the approach you've taken. This is not your first rodeo, but I'm really interested in what you see in terms of the evolution of financing approaches. One concept in particular, I'm thinking a lot about is the degree of credit support from some highly creditworthy entity. I see a potential trend in that direction where the market wants to see that and also where those parties are increasingly willing to do that to get projects financed and done. But I could be wrong in that. What do you see in terms of trends there?

I would agree. I think -- look, we were a part of being at the forefront of that and that we're doing a deal with Fluidstack that has credit support from Google. They've done several of those deals, not just with us, but with others, that has certainly put a marker down that is on many large investment-grade players in the space. It's within their head space about what they might have to do to get things done in the space. So we have had a broad range of discussions around credit support from investment-grade entities to noninvestment-grade entities, neoclouds, folks that might be a test for us where they know cost of financing is a huge part of our success and frankly, the money we're going to make. And also, that's going to translate into the lease rates that they're going to pay. The good news for people in our seat is that, given that there are several interested parties in the success of these various channels, they might be distribution channels for their chips, there may be other strategic investments or rationales for success, it's sort of squarely obvious that they need to do things to enhance credit if they want these limited assets, which are soon to power interconnect availabilities. I think we've seen it move in a couple of different ways. Obviously, there's the explicit credit wrapper I think over time, we'll see how much appetite there is for that because at some point, ratings agencies may start to apply that to the company doing the wrapping and that may not be as attractive. We have seen given the desire for compute, and we've seen -- and I know I was watching a neocloud speak the other day talking about they were getting some fully prepaid contracts for compute. That obviously creates a situation where if you're in our seat, a fully prepaid contract provides availability to fully or partially prepay us to now make us not have to do as much debt financing. That's the own flavor of credit support. So I think you will see creative solutions like that. As it pertains to our portfolio, we are very much focused on the largest, highest quality credit counterparty hyperscalers doing leases at our sites. That said, we have some places in the portfolio where it might make sense. We have a 70-megawatt site, a little bit small for the typical hyperscaler. It's exactly rightsized for a lot of neocloud. So we'll be working on opportunities like that, that are, for us, a little bit smaller, but still a very big data. I mean you're still talking about many, many hundreds of millions of dollars to build that and lots of rent coming our way. But I do think you'll continue to see an evolution in those space.

It's a great overview. I mean your last point, I mean, you are in a negotiating position as well where you can, to some extent, sort of dictate or push in the direction that you want just as an equity investor, as a shareholder, the highest quality counterparty as well, right? So you can do that.

Yes. I mean, look, that's our goal -- I think there's an opportunity for us to create really a very unique company that doesn't have a comp we're aiming at. And what I mean by that, and I know you've done some already on this, Stephen, about the potential for like conversion or treatment of a company like ours, if you flash forward to a future state where we've got several cash flowing leases. I think the interesting thing that we're still trying to get our head wrapped around is traditional REIT investor and a traditional REIT structure probably has a leverage ratio that's closer to like 25% or something like that. If we have the types of leases we have right now, so really bulletproof counterparties on long-term leases at single campuses, it would be hard for me to not think you'd want to run the leverage a lot higher on that. If we have a stream of cash flows backed by AWS for 15 years, I would feel very comfortably heavily levering that up. And so we may fall in a slightly unique bucket where I would argue we should trade at a premium. If you say -- imagine a future state where we have a half dozen of those types of leases, I think we would sort of try to maximally lever that and then trade at a much larger premium. But it remains to be seen. Like you had talked about overall scope of financing earlier. In my mind, if we do all the things and we've got gigawatts in the pipeline we referred to, we have a very big land bank coming to power in the next couple of years. We have great tenant relationships. We're working on behind the meter solutions that people are very interested in. So if everything goes our way, I mean, we may need to raise $50 billion of capital over the next 4 years to build that out. So we're very cognizant of all the different pools of capital we can raise that from and starting with really high-quality tenants or wrapped tenants is really key to advancing that scenario.

No, that's well said. I mean the -- I don't know what structure will be, whether REIT or otherwise, but there's this universe of investors that are the logical habitat. And I just see you moving in that direction over time. I'm not sure exactly how we get there, but that feels...

Well, and there's no question, again, if you imagine that state of the world, we look so different than our profile like our equity trades today. We are a stream of massive cash flows coming from the most trustworthy counterparties in the world. There are going to be 1 million interesting ways to do financial engineering around those reams of payments to create the most valuable equity we can.

Yes. I don't worry about whether we can find investors who will want to own that stream of cash flow. So I think that's a good place to end it. Tyler, thank you so much. That was great discussion.

Thank you very much, Stephen. I appreciate it.

Thank you.