Hey, everyone. I'm Stephen Glagola, Senior Equity Research Analyst covering digital assets and AI infrastructure at KBW. To kick off our inaugural KBW AI Infrastructure Summit, I'm delighted to welcome Greg Mumford, CFO of Cipher Digital, and I might add, a former KBW alum himself. Greg, thank you for joining us today.
Thanks for having me. Appreciate it.
Yeah. Greg, maybe we can start with the broader backdrop. Over the past couple of years, sentiment towards miners moving into AI infrastructure has swung pretty sharply between enthusiasm and skepticism. We've had brief periods of re-rating, but they've generally been followed by longer periods of chop and de-risking. Since June, skepticism has really taken hold. There's now a broad wall of worry. Investors are asking whether safety concerns could slow AI development, deployment, and whether token optimization and changing market shares among model providers could pressure the frontier labs, and whether political and local opposition could intensify ahead of and post the midterms. Then you layer on higher yields and concerns about circular financing across the ecosystem. Which of these concerns do you believe are overdone, which are justified, and what turns sentiment positive again here?
Yeah. Look, great question. I think it's no secret that stocks have traded down since June. I think we're at a really interesting point of the cycle for an infrastructure developer, where we've gone out and signed a bunch of leases, we've financed those projects, and now we're firmly in the physical execution stage of multiple projects, but prior to actually seeing any cash flow start being turned through the income statement. So that's a challenging time, where investors are really more focused on execution and everything between signing the lease through delivering the data center, as opposed to historically, when they were more focused on, can you sign the next deal, and when is that coming?
It is an interesting point in time. I think to your question, there are certainly a bunch of risks, some of them real and that deserve a healthy amount of respect, and some of them that are maybe a little bit overdone. As I think about the risks, one, the financing environment is more challenging than it has been historically. That is just true. The rates are up. Base rates are up. There is more supply from issuers. There is a huge amount of capital that is required to build out the AI infrastructure that we are talking about when we talk about deficits in the market. You are seeing that through various S-1s and things like that in the news. There is a lot of capital that needs to come, and investors are more focused on the execution.
Another risk, physical execution at this scale, it is huge, it is challenging. Then add into the equation that craft labor is being pulled in every different direction. The supply chain is being pulled in every direction. It is a challenging market. I think what is overdone is the thought that because of these risks and because of where we are at in this point of time, that demand is collapsing and that the fundamental underlying demand is slowing down. I think it is actually very much the opposite. We are continuing to see, and as you have seen in the market, lease terms improving despite the backdrop in equity markets. We are seeing definitely great alignment from our tenants and from conversations we are having across the board. I still think scarce power is the constraint and demand is not changing from what we can see.
Great. I want to talk more about the financing environment, funding environment in a minute, but let me turn to execution here. Barber Lake seems to validate Cipher's early bet on Texas. An AI lab committed to lease the facility for 10 years after Fluidstack's term at substantially similar economics. What enabled you to secure a commitment that far in advance, and what does it say about the durability of demand and pricing for Cipher's Texas campuses?
I am glad you brought this up. This, to me, is the best proof point that we have in the market to date of the fact that there is strong amounts of terminal value in these assets. There is a discussion since our first deal that we signed, which was 10 years, and it is at this Barber Lake site, about what is the terminal value of these sites that are being built in West Texas and other parts of the country that weren't built traditionally for cloud data centers. I think fundamentally, cloud data centers and AI campuses are two different beasts, and they require different things. The Tier 1 or primary locations for them will be different. Fundamentally, you need less latency, lower proximity, more land, and more power. There will be Tier 1 locations that are not traditional cloud 1 locations.
This is the best proof point in the market, I think, to date, of the fact that we are signing a lease 10 years from now, adding $5.2 billion of contracted revenue 10 years from now. It really validates the fact that we are building long-lived critical infrastructure assets to these tenants and counterparties that are using them.
Yeah, no, that is great. I guess on that, on Barber Lake, sticking with that, how did you get comfortable with the lab's credit that far out with no backstop disclosed for that decade? When evaluating prospective tenants, to what extent does your view of the long-term model layer landscape influence your willingness to partner with a particular AI lab, so to speak?
Yeah. When I think about the deal that we announced at Barber Lake, it makes sense to break it down into two different decades. The first decade is a lease facing Fluidstack with a Google backstop that fully supports our debt over the first 10-year term. Nothing has changed about that first 10 years, and the way that we have structured our debt over that first 10 years is to fully amortize within that first 10 years. If you think 10 years down the road as we are starting the second binding lease for 10 years directly facing the lab, our data center will be substantially delevered at that point. That is $5.2 billion of additional contracted revenue on a data center that is fully paid off.
It is a very different underwriting decision when you think about credit support or quality of the counterparty than something that we were going to do today, and we would need to finance through construction.
Okay. I guess maybe just double-clicking on what you just said prior to this question around the terminal value at your Texas campuses. What gives you confidence that you can maintain and grow rents at renewal rather than potentially trading at a discount to the major data center hubs as supply becomes less constrained?
Yeah. Again, I think this goes back to the idea that Tier 1 locations for AI campuses are just fundamentally different than Tier 1 locations for cloud data centers. When you think about that, and I mentioned a little bit earlier, AI campuses require things that cloud campuses do not. So you require way more land, huge amounts of power, and ability to scale. Almost every tenant wants the ability to scale and grow at the site. You need skilled labor in large amounts. When you think about places like West Texas or Louisiana or Indiana, North Dakota, there is really an investment from all of the major hyperscalers into these locations to build these large campuses. So you're seeing large amount of tax dollars flow into these communities. You're seeing a huge amount of labor flow in. There's going to be permanent jobs.
We just believe that over time, these locations will be the Tier 1 locations for AI campuses. If you extend that logic to what it means for terminal value and renewal risk at the end of our base terms, we think that it's really actually de-risked. That needs to play out in the market a little bit, that thesis, but we're seeing it via discussions that we're having with our tenants. We're seeing it in the commitments that we're getting today, so another lease 10 years down the road for 10 years. We're seeing it in general lease terms across the space. Where lease terms initially started as modified gross leases for 10 years, you're now seeing 20-year leases triple net as kind of the standard.
As lease terms continue to improve and go further out in the future, I think that starts to really validate the terminal value, and the ability to re-lease these sites in the future.
Okay, great. Turning to the leasing pipeline, Tyler described demand on the Q2 call as stronger than ever. You mentioned it was still strong. Starting with the 2027 pipeline at Ulysses, Reveille, Odessa, how have the tenant discussions and the positioning of those sites evolved since then?
Yeah. So given we're in a quiet period, I can't speak specifically about demand at those sites, but I can zoom out a little bit and talk about demand in general and what I've been alluding to already on the call, which is we have a very strong and ongoing dialogue with almost every partner that we would want to be a tenant. The pool of tenants that you would want to be a tenant that is kind of the top quality has grown. There is now not just the top hyperscalers, but there's also some of the chip developers. There's other interested parties that have really great credit ratings. So we have really strong discussions with them on an ongoing basis. We've said this before on earnings calls and the like.
Demand is strong, and I think, again, if I can point to proof, it's really the fact that we were able to go out and sign that additional lease 10 years down the road at the same and increasing terms as we have today. Like I said earlier, the goalposts for what is a standard lease today, that has moved. The discussion has moved from where we initially were when we signed the Fluidstack lease, at the end of the summer in 2025, to now being 20 years, kind of the base starting point. So that is really speaking to the demand, and the fact that it continues to be strong across our portfolio.
On that point, how does your lease pricing power, and maybe terms differ across the customer cohorts, so hyperscalers, labs, neoclouds? What customer cohort do you think provides the most alpha risk-adjusted returns today for Cipher's lease portfolio, and why?
Cipher is pretty focused, and always has been focused on making sure that we underwrite each deal individually and think about the risk-adjusted return for that specific site. That changes dramatically depending on the tenant. To date, we have focused on top quality tenants, so you've seen two deals with AWS directly facing them, one deal with Fluidstack, but supported by Google. Google and AWS credit is two of the best out there. You can see that reflected in the fact that our bonds are still trading the tightest out of the peer group. We're pretty focused on that and always have been, and I think that speaks to what we talked about earlier, which is the financing market is getting more challenging. It's not getting easier. We've been thinking this for a long time before it hit Main Street.
There is a huge amount of supply coming, and to build out the amount of infrastructure that we're talking about and that you're seeing quoted in various news articles or on CNBC anytime you turn on the TV, there's going to be a lot of debt and a lot of equity raised. We're well aware of that, and we're focused on underwriting deals in a way that we believe we can get efficient capital to grow out and finance construction. We're focused on underwriting deals that we believe have value in the future, whether that's in some sort of refinancing or recycling of capital. High-quality tenants make that a lot easier.
Okay. Colchis received conditional base load status in Batch Zero, with final designations expected after ERCOT's audit in December. Separately, the Texas Commission on Environmental Quality has paused data center permits until the audits are complete. How are potential tenants approaching Colchis before those approvals are final? Would they sign before with conditions on final designation or wait?
First of all, we're thrilled that Colchis received the conditional approval in base load. What that means for anyone who's listening who's not aware, I'm sure it's a very small percentage of people on this call, but that means that we have been approved conditionally for 1 gigawatt in the first batch. There is an audit that is coming up in December, and after that, we are hoping and expecting to be firmly approved for that gigawatt. That's a marquee asset for us. 1 gigawatt is a huge site, and it's very interesting and exciting to the tenants that we're speaking about. As it relates to the TCEQ and the audit process, regulations and what's going on in the Texas environment and across a lot of the U.S., I believe is something that will affect timing as opposed to something that will erode fundamental demand.
There is situations, and you are seeing this across the country, where things are maybe going to be pushed out a little bit longer, maybe scrutinized a little bit harder before you figure out if you get the requisite permits. Texas continues to be a pretty business-friendly state and jurisdiction. My view is that this will potentially slow execution, and certainly in some states more than others. But ultimately, fundamental demand should remain strong. What that means is that if you have approvals and if you have sites that are outside of these processes that are slowing them down, those should be more valuable to tenants. Time to power and scarcity still drive most decision-making that we are seeing on the leasing front.
On that point, do you think we are in somewhat of a holding period in Texas for leases until post-midterms?
I think it really depends on the way that potential tenants choose to engage on sites. I think there is a framework where they could think about doing a lot of diligence on a site, potentially setting up terms for a site, and trying to secure it with some sort of optionality. I do not want to speculate on what is happening across all of the sites in Texas, but I think that there is potential for that just based on engagement and dialogue that we are having. That said, who knows if there is actually going to be a lease announced in Texas pre or after the audit. That said, we do have some sites outside of the audit, and I think a few other people too do as well. So, there still is ongoing discussions in a pretty robust way.
Great. From my talks with the Cipher team, you have described that 2.5 GW of behind-the-meter generation opportunity is additive to Cipher's grid-connected pipeline. Maybe can you just unpack that more? How much of the 2.5 GW could be available by year-end 2027, and how do the development timelines, CapEx requirements, and returns compare with a traditional grid-connected project?
Great question. Behind-the-meter generation is something that we're very excited about and focused on, and it's something that we have discussed increasingly over the past several quarters, so I'm happy to discuss it now. The way that Cipher thinks about these projects is that we would love to make it very simple for a tenant to make a decision to work with us. As a smaller company with a startup history, you have to make it easy for these large companies to work with you, and you have to connect all of the dots for them. What do I mean by that? I mean we have the land and secured the site in a community where we work closely with the community and understand what they're looking for and have a good relationship there.
We've started building the lateral gas pipelines, which is critical to get natural gas ported over to your site. We've press released that. We are working on various agreements that would actually transport the gas through the pipeline to our site. We have very strong demand from tenants to come and build a data center with behind-the-meter generation supplying power to the data center. What Cipher ultimately wants to do is connect all these dots, and make it very easy for a tenant to make the decision. That said, while that sounds easy, the actual structuring and the deal-making is quite challenging. There are a lot of different parties that are interested. It's challenging to get access to generation equipment on a timeframe that works for the tenant. Supply chain is challenging in general.
While we're very excited about the opportunity, it still is one of those situations where until you have it signed up, you don't have a deal yet. We are continuing to do what we do, which is, I think, try to connect the dots and make it very easy to make a decision. From a return profile, like I said earlier, we try to underwrite every deal based on the risk-adjusted returns. These deals are slightly different than what a front-of-the-meter deal looks like, and we're structuring the deal to make sure that we're appropriately compensated for that risk profile.
Okay, great. Turning to the interest rate backdrop. The 10 year, the 30 year, whatever you're looking at, but the 10 year has increased, I think roughly 80 basis points in Q3 from Q2 end. The leaseback debt for these projects has moved with it. An environment of increasing rates or higher for longer, can you pass the higher cost of capital through to customers in terms of higher yields on cost on these leases? If so, what are you seeing in current negotiations around customer willingness to accept higher lease rates?
This is very topical for us and for, I am sure, most of our peers in the space as well. Yes, base rates have increased, but I think in addition to that, you have actually seen the spreads of the data center debt widen more than just the base rates have increased. I think that goes back to something I said at the top of the call, which was investors are now more focused and probably pricing a little more risk into what happens in between signing the lease and delivering the data center. There is a lot that happens in there. The physical execution of an infrastructure project of this size and scale is very large, and it takes an extremely well-coordinated effort to deliver on the timeframe. You have to deal with bumps and hiccups along the way.
I think that that is one of the big reasons for the actual spreads widening on top of the fact that base rates are going up and we are in an inflationary environment. What does that mean for us and other data center developers? For Cipher, we certainly are focused on protecting our returns through appropriately underwriting each deal. If that means taking into account the fact that we think financing will be more expensive, then that is something that we will do. If it means taking into account the fact that we are in an inflationary environment and the price of transformers is probably higher than it was two years ago, then that is what we will do as well. We are still in a situation where the ultimate tenants or end users that are contracting with us are really focused on credible megawatts.
What I mean by credible megawatts is us telling them that we will deliver a megawatt of data center capacity by a certain time. That is critical to them as they want to get their compute up and running, and the return profile to them is so important that they are willing to support the development. You are still seeing a lot of credit support come out from various parties, increasing lease terms. To loop that all back to what the question was, is can we pass that along to customers? I think you can. I think you can find ways to structure deals to make sure that you are compensated for it where it is not specifically or only yield on cost that is moving.
There are other levers in the structure of a lease that you can pull on as well to make sure that you are appropriately protected in an environment that is changing very quickly.
Well, can you just maybe elaborate a little bit what those levers potentially are?
Yeah. For sure, there is SLAs in leases t hat have penalties. If you have to deliver certain phases by certain dates, those penalties could be more aggressive or less aggressive. You could have termination clauses where the tenant has the ability to actually cancel the lease if you are delayed by a certain amount. Those terms are becoming more favorable to developers as well, to account for supply chain and the fact that labor markets are challenging. There is also the ability to bake into the lease some risk transference in some situations. Some of our leases have a risk transference where it is a pre-agreed split of costs to Cipher and costs to one of our tenants. There is things that you can do to kind of be creative, aside from just purely the economics.
Okay. Thank you. We have seen Cipher's three project bonds have been priced progressively tighter. I think Cipher Compute was at 7.13%, Black Pearl Compute was at 6.13%, Stingray, 6%. Given the movement rates, how has pricing changed on new, I guess lease back debt or issuance today?
Yeah. Spreads have certainly widened. I think our bonds are still all trading relatively close to par. I think our Barber Lake financing, which is Cipher Compute, is trading maybe a point above par, and the others are maybe just a point below par. So the bonds have held in very well and are yielding on average the lowest and most efficient cost of capital of our peer group. But that said, yes, pricing has changed and I expect that it will be a little bit higher if and when we go to the market for another financing in the future.
That is a function like we talked about earlier of higher rates, but it's also a function again of the fact that investors are asking now the harder and the right questions, which is, what's your track record of delivering on an infrastructure project of this size and scale? What's your ability to go out and execute and deliver credible megawatts? What are the out clauses in your lease? So I think certainly investors are getting smarter about the right questions to ask and probably pricing in a little bit more risk. Like I said, we're okay with that. We will make sure that we are underwriting each deal based on our understanding of a slightly higher cost of capital. But generally, we're feeling pretty good about where our bonds are trading and our access to capital.
I think the cost may change, but I don't think our access to capital is diminishing.
What do you make of the argument that the higher rates or the increasing financing cost could increase sort of the replacement cost or constrain new supply, which is sort of positive for existing lease value?
Question basically being, do you think cap rates are going higher?
Well, I'm going into cap rates in a minute with another question, but I just wanted to kind of play off what you just said a little bit of, is there a spin that you could think you can make here that, or are you seeing where the higher financing costs actually could be somewhat of a positive to the existing lease value?
Potentially. I think at the end of the day, I'd love capital to be as efficient as possible. What I do think is important is the ability to go out and secure financing. I don't think that every developer will have the same ability to sign leases, execute on the physical construction, finance it efficiently, refinance, and start from the beginning with their front end of the funnel on the origination side. I think that's really what sets Cipher apart, is we've built the platform to be able to do all of that and have started to demonstrate that we can do all that and build a track record. What I mean by that is we have a fantastic origination team.
We've greenfield sourced all of our sites to date, and that's in the double digit of energy assets, and we've demonstrated an ability to mature those through the interconnection process and get them connected. We've demonstrated an ability to deliver on the physical construction side, specifically with delivering a data hall at Black Pearl multiple months early. And we've demonstrated ability to go and get financing. Like I said, our bonds are trading tightly. We have a bunch of investors on the credit side that follow the story and I think would be supportive of us in the future. So that's really what we're trying to do, is build a platform that has the ability to continue to do that and cycle assets through each of those different phases.
Okay. I want to touch on that too again, but before that, on the cap rate side, I think investors, at least my conversations that, I think there's often this implicit assumption that higher rates translate directly to sort of higher cap rates for AI data centers. I would say maybe funding costs and stabilized asset values don't necessarily move one for one. I'd be curious to get your take on that. We haven't really seen many private market benchmarks yet in this space. How do you think higher rates affect the value and implied cap rates of Cipher's AWS and Fluidstack/Google leases?
Yeah. This is a good question about terminal value and proving out the fact that there is terminal value here. I think the data points that we have, and I've talked about these a little bit before, are that lease terms are getting better, credit enhancements are getting better, and you're seeing Cipher sign a lease 10 years down the road. What we don't have yet, like you said, is a lot of precedents of people recycling sites or capital with a cap rate to some of these yieldcos, BXDC, for example, or others. I think that will come, and I think really the reason for that is right now we're still at the kind of beginning of that first vintage of data centers being delivered. You're seeing us start to deliver data halls. Some of the others who signed leases around when we did start to deliver.
It'll still be a little bit of time before you have a bunch of stabilized assets, which would be prime targets to be recycled or taken out at a cap rate. That's where you'll start to see the cap rate come in and kind of put a line in the sand of what the value is. What I would say specifically is, as soon as you deliver the data center, execution risk comes down materially upon delivery, and you should see that reflected in your cap rate. When you have a look-through to someone like AWS, there is only so high that the cap rate can be on top of that as soon as the data center is de-risked.
In addition, these are kind of unusual leases when you look at data centers historically, in the fact that they are directly facing an extremely high quality counterparty, and they're very long. There is an argument too that these leases should have premium cap rates associated with them because of that look-through and the length and the infrastructure-like nature of them. It remains to be seen where asset values stabilize and where cap rates are in the first few transactions. We're feeling pretty optimistic that the terminal value will be reflected in good cap rates.
Is there anything on what you are seeing on the private side there that maybe the public markets would not see in terms of transactions that sort of justify what you are saying on that?
Well, I have not seen anything on the private side that you would not have seen either. What we are seeing is potential opportunities to buy powered land that needs to be matured through interconnection processes and things of that nature. We have not really started looking at buying stabilized assets in any real way.
Okay. Once these projects, Greg , you have mentioned this a couple of times on this call, but once these projects are stabilized, what is the long-term plan for Cipher? Do you expect to hold Cipher's interest for the cash flow or monetize the assets over time and recycle that capital into new projects?
Yeah. I think this is a really important question, but I think the answer is that there are multiple different paths we could go down. I think that is honestly the truth. A critical piece of this business model to make it work is recycling capital. Whether that means through refinancings and you are able to take out some additional capital on higher leverage and lower cost of capital, better credit rating at the project level, that is one way. There is also potentially monetizing a portion of an asset, monetizing a whole asset. All of those are available to us, but finding ways to recycle capital in a non-dilutive way is critical to making sure that we generate a strong return for equity holders.
We are open to any and all options as we think about what the best way to capitalize the company and do that is. Certainly, there is an element of we like the cash flows that we are creating and want to free up those cash flows over time to have access to them, and ultimately eventually use those to recycle back into development as well. But we will evaluate each situation on a situation-by-situation basis, and we will do the right thing to make sure that we are capitalizing the company in the most efficient way possible.
I wanted to ask you maybe a higher-level question on just sort of industry structure evolution. I mean, you are not competing just with other miners pivoting to AI infrastructure. You have established private data center developers, and then you have a growing number of new entrants. I guess, how do you see the industry evolving over time? Do you think it kind of consolidates around a few scaled platforms, or do you think it remains fragmented based on site scarcity and so forth?
Yeah. Based on looking back at other industries where there have been massive infrastructure build-outs, it probably makes sense to think that over time there will be some sort of consolidation. I think the consolidation and the value, excuse me, accrues to the platforms that have developed a track record of being able to go out and deliver credibly on what they say they are going to do, being able to go out and finance efficiently, being able to recycle capital. So in the short term, I think you will continue to see new players enter the market. I think you will continue to see credit enhancements support the development of those sites. I think general contractors will be very busy in the short term. But I think companies like Cipher, and the reason we have set ourselves up the way we have, should accrue value over time.
We have built a company where we in-house our EPCM function. We have procurement and engineering and construction management in-house. We have the ability to originate greenfield sites. We have a great ability to go out and finance them. I think that is what ultimately will lead to consolidation over time, is access to capital and track record of execution.
Okay, great. I think we're coming down to the final questions here, but I wanted to touch on something you mentioned at the start of the call around the risk around labor and cost and so forth, inflation. I know that's still a very topical thing on investors' minds. How should we view your average build cost or total build cost, all-in build cost going forward, in terms of CapEx per megawatt? Maybe you could touch on, because I think there's some differences in how companies report this in the market. What does that include? Does it include maybe all the soft costs, like design, engineering, permitting, and so forth? That'd be helpful. Thank you.
Yeah. Spec drives dollars per megawatt. We don't go and build anything based on just a Cipher design at this point. What we have done to date is we've worked very closely with our tenants to understand exactly what they're looking for us to build, and we go and we execute on that. That can give a wide range of dollars per megawatt in the CapEx forecast, depending on how much redundancy do you have. Do you have backup generators? Do you want it to be evaporative cooling, or do you want liquid cooling? All of these things can completely change the dollar per megawatt. So it's hard to give an exact answer of that because we're really focused on delivering for what the tenants want, and then underwriting and pricing the deals accordingly based on that.
As it relates to Sorry, what was the second half of the question there after pricing?
I mean, it just was kind of-
Oh, what is included in the budget?
Yeah. You have 11 million a megawatt.
Sure. If that is where you are. Imagine we would include everything that you need to get the site from acquiring the site all the way through handing over the keys if it is a triple net site and delivering down to the racks. Yes, cost of the land will be in there, permitting, soft costs, engineering, development fee to us. All of the things that you would expect to complete the construction project from front to back are included in that budget.
Awesome. Greg, I think we're going to leave it there. We're about at the time. I want to thank you and Cipher for kicking us off here at the inaugural KBW AI Infrastructure Summit and for sharing your perspective with us today. Thank you. Our next fireside chat for everyone will begin in 15 minutes, with Sean Glennan, CFO of Hut 8, and we hope you'll join us then. Thank you, everyone.
Thanks a lot, Stephen. Thanks, everybody.