Good morning. Thanks, everyone, for dialing into the session with Cipher Mining. If you had joined the start of our conference, we started at 8:00 A.M. My colleague, Brian Wheaton, hosted Blockchain.com for an interesting fireside chat there. Really excited to have Cipher here. Tyler Page, CEO, and Ed Farrell on the CFO side. Thanks everyone once again for dialing in. Thanks to management for joining us. With that, would love to get into it. For those in the audience who aren't familiar with Cipher, they're one of the bigger, larger public Bitcoin miners today. They recently announced some high-level plans to pursue HPC, which we have some questions, which we'll get into.
Core of the business today, though, is still kind of on that mining side, so a lot of our questions will focus on hash rate and all that good stuff. So I guess with that, kind of first question from my side is, you're about to embark on a big hash rate increase. I think in our model, we had you guys getting to 30 exahash by 2025. I think that's end of 2025. So walk us through a bit of, you know, are there any risks you see on the horizon that could prevent you from getting to 30 exahash?
I mean, I guess there's always risks, but why don't we talk about what we've done to limit those risks and what we're forecasting? Let's take it in stages. Our next big addition to hash rate, so we're at about 9.1 exahash right now, currently hashing at our four data centers. We have an upgrade of our largest data center in the rigs coming towards the end of this year. That will when we add those new rigs by year-end 2024, we expect to have thirteen and a half exahash, and that will also bring our efficiency down to about eighteen point six joules per terahash. So I think we're known for having among the cheapest, if not the cheapest, average power price of any of the miners.
You know, at that point, we will be among the most or closest to the most efficient fleets of any of the miners. So you'll be taking the cheapest power, running it through the most efficient machines. It will produce nice results with that thirteen and a half exahash. Only risks there are getting the rigs. That's mostly a large Bitmain order. There are also some Canaan coming. You know, you've got to ship them around the world. I guess container ships can wreck or whatever, but assuming nothing weird like that happens, that's a pretty low-risk timeline, in that, it's almost entirely paid for already, and it's just a matter of swapping out rigs at a currently operating site. So there's not a lot of execution risk to that. It's sort of unplugging one and plugging in the next.
So that's this year. Go to next year, where we've got a lot of the work going on right now to deliver this. We are building a 300-megawatt data center right now we call Black Pearl. It's basically one large building that looks like six 50-megawatt buildings connected together by a spine. When that is fully kitted out, we expect to be at 35 exahash, John, actually, not 30. And when we do that, our efficiency will be close to 15 joules per terahash. So that's new machines going into that new build. The site is scheduled to energize in the second quarter of next year, and the anticipation on our side is that we will be bringing up chunks of that throughout the year.
We will do our very best to get it sort of middle of the year, but I think the only numbers and projections we've given are by year-end, because bringing 300 megawatts online is a pretty big deal. So the construction's going on right now. We are forecasting 35 exahash there. Across our entire fleet, there's about 21 of that or so will be at Black Pearl, and that's the plan for next year. And I guess, sorry, to the risks to that. So that has more risks related to construction. The construction's going on right now. You know, we've got all the significant equipment under contract.
It's either been delivered or is scheduled to be delivered, but these are pretty complex processes with hundreds of people working at the site on some days, so there's always something that might go wrong here or there. In general, you know, we've built in time. I don't anticipate any challenges, but that is a different stage of construction than just a simple swapping of rigs at an existing site. So we've got the kind of normal risks that getting something online happens. But we've got regulatory approvals, like, there's nothing like that, that we're waiting on. And, you know, I think we're being pretty conservative, given that we expect it to energize in the second quarter by basically giving a year-end hash rate. I think that gives us plenty of room to surprise to the upside, I hope.
Yeah. That's helpful. What do you think has changed, you know, whether in your own operational capabilities and maybe peers as well? Because I remember a time when companies would come out with an exahash, and they almost seemed to never hit it.
Yeah
... the environment was a little bit different. Part of that was, you know, debt markets kind of dried up, and they a lot of times they just couldn't get the funding to do it. But is there any operational, maybe know-how, that miners and in particular you guys have learned over the years that now hitting those stated exahash goals, it's a bit easier?
I think part of it is you get a survivorship bias that, like, the companies are still around are the ones that could actually deliver on what they said. I mean, let's recall that, access to capital is a big part of being successful as a Bitcoin miner, because, you know, we are an emerging industry. Even though we, in a lot of ways, look like a traditional industrial company, we build these CapEx-intensive projects that then throw off cash flow, we do not have a mature debt market that we have access to.
Like, I'm reading a book right now called The Big Rich, about the original big Texas oil fortunes, and I'm reading about these wildcatters from the early 1900s, and it's a very similar business, in that they couldn't get loans and, you know, they sort of had to scrape it together, and then there'd be a boom market and a bust market. So capital markets access is a big part of things in this space. We had really excitable capital markets a few years ago, with zero interest rates and, you know, lots of companies going public, et cetera. In addition to some good operators, there were probably some less solid operators that got through there and raised money. And so now I think what you're seeing is probably a higher likelihood of firms being able to execute and deliver.
But there's still a lot of challenges, and there's still... You know, a big part of this whole industry is still how much confidence do you have in execution risk? I tell people that, we invest a lot of time, effort, and money in trying to minimize those execution risks, so you can believe what we say we're gonna deliver. We've got a great team that has delivered in the past. They built our previous data centers. We have a good record of delivering on time. So I, you know, I don't think everyone will deliver on time, 'cause it's just very complicated. It's a very complicated process to deliver a brand-new data center. If they're building one, things get delayed, things go wrong. Sometimes the rigs have issues, whatever, there's any number of things that can go wrong.
I think it's all about being able to have a repeatable process that contains those risks, and having really good people. So I'd like to think that we trade with less of a, let's call it, lack of execution discount than some of our peers. But you know, I think you'll probably get a higher rate of delivery in the industry than you did before, because some people have washed out, but you still won't have everyone deliver what they promise. I would imagine.
That's, that's a helpful way of characterizing it. Okay, that's, that's great. I wanna turn a little bit to, to the HPC piece. It's, it's honestly one of the, the biggest kind of questions we get around Bitcoin miners. And in Cipher, I think you guys were kind of, you know, wanted to, I think, really dig into it and see if it made sense from an operational standpoint to go after HPC. So understand that, you guys are coming at it, I don't necessarily want to say later than the other folks, but 'cause you- you were doing your due diligence, which I think is fair and appropriate. And, quite frankly, I think a lot of miners aren't gonna be successful at going down the HPC route. So it is something that should be well thought out.
But within that, talk to us a little bit about high-level HPC plans. You recently acquired that new 300-megawatt site, I believe it's kind of more West Texas. The press release almost framed it like you'd be going towards HPC. So just kind of talk about how many megawatts you feel comfortable, kind of targeting. Nothing necessarily to hold you to, but just high-level targeting.
Let's talk a little bit, because you're absolutely right. We've been pretty deliberate and been circling this business for a long time. And I think we were pretty slow compared to some to jump in and say we were gonna do it, because I think we wanted to understand, you know, just from an ROI perspective and a risk perspective, what does this business look like versus Bitcoin mining, which has lots of risks, but we're also very familiar with, so we understand it pretty well. I think, first of all, to contextualize, you know, the AI or HPC opportunity, it's helpful to talk about. You know, I'll say, at least among the Bitcoin miners that are talking about it, not more broadly, you know, traditional data center companies, but among the Bitcoin miners, there are several different business models people are lumping together.
So I think the first thing investors need to understand is, like, it's a very different business to buy your own NVIDIA GPUs and sell the compute output to enterprise clients. There's, like, an enterprise sales element to that. It may be a fantastic business. We don't have a negative opinion of that business. I think that business is not necessarily immediately adjacent to what Cipher already does and has an edge in. We don't have any salespeople at this company, like, we're a self miner. So, you know, I don't think we really... In addition to the fact that it's expensive, you're speculating on kind of GPU upgrade cycles, which are analogous to the rigs in miners, but you're betting a much larger amount of money on it.
I don't think we have any edge, so we are not doing that business. Some of our competitors are, God bless them, they may make a lot of money. It is not what we are focused on. I think within that space, what we are most focused on is effectively originating sites, which we've proven we are very good at. Originating sites that are large scale, with attractive power features, and then building a data center there. You know, the truth is, on the one hand, an HPC data center is very different, 'cause it's more expensive, and it's got backup power supplies and generators, and things like that. On the other hand, from a project level perspective, it's not that different, when you're taking a greenfield property and turning it into a data center.
What we see as an opportunity here is leveraging our strengths and originating sites that are large scale, which are in very high demand right now from hyperscalers and others, and then leveraging our best-in-class construction and operations team. So we've got, you know, close to a dozen folks on our construction and ops team that joined us from hyperscalers, so Google, most of them are Google alums, Meta, you know, places like Vantage that build traditional data centers. So we've got a very experienced team that knows how to deliver a tier three or tier four data center, and shows very well.
And if we look at those advantages, I think what's interesting to us is the business of building a powered shell for a hyperscaler-like tenant, or for a, depending on where you draw the line, hyperscaler, a very high quality from a credit perspective tenant for a long-term lease. That they would bring their own GPUs and operate at the site, but we're delivering a powered shell for a long-term lease. The kind of third leg of that stool, so we've got a great portfolio of sites. We've got a great team that any technical team, at least in our early discussions, at any of those big clients has been impressed with, thus far. You know, the last leg then is capital. These are more capital-intensive builds.
We will only do this business if we debt finance it, kind of independent of the cost of capital for Bitcoin mining. Bitcoin mining is typically funded by equity. Most companies in the space use an ATM shelf, typically, and the truth is, that's really because it's the lowest cost of capital. If you look at the limited debt offerings, most traditional lenders don't wanna take exposure to the kind of booms and bust in Bitcoin price and hash price. Contrast that with HPC data centers, there's a pretty robust lending market at different stages of the build-out. Everything from bank financing, if you've got a really good credit counterparty in a long-term lease, you can debt finance the build largely with, you know, loan-to-cost ratios of 70%-80%.
So, it's a fundamentally, let's call it, more stable business. It does not have the kind of boom and bust necessarily of Bitcoin mining. Frankly, we're really bullish on Bitcoin mining over the next couple years. I think that, you know, that may produce a better cash return over the next couple years, but, having a very stable long-term lease that would produce tremendous margins has a lot of value independent of that. And it's a unique way to monetize our site. So I'd say it's still early. When we've acquired the sites, every site we've acquired, has the flexibility that we think they would make fantastic Bitcoin mining sites as well. So what's interesting about these sites is that we have a huge advantage in that we have a natural offtake.
If for some reason the whole HPC data center craze just fizzled out, and those phone calls stopped happening, we would be very happy to build Bitcoin mining data centers at these sites. They are all sites that are in locations that have demand response programs, similar to the ones we work with in Texas, to monetize curtailment, and that's really the secret to building a sustainable Bitcoin miner. So, you know, what we loved on these sites we've acquired is the flexibility. Now, one I'll talk about in particular, because you highlighted it, we did put out a press release last week. We are acquiring a 300-megawatt site in West Texas. What's unique about that site? It's the interconnect and approvals have already been done, so it's approved for 300 megawatts.
Substation is already constructed on site and has been reviewed, and we think it's a very high quality one. So what I have found in our discussions with hyperscaler-like tenants is that there's a bit of a sliding scale. Yes, they want access and capacity at large-scale sites, but that the enthusiasm for those sites dramatically increases if they can be ready sooner. And so this site is ready effectively immediately, in the sense that there's no data center built there. It is a dirt field, but we own a large acreage there when we close, and the substation's already built, which is one of those big execution risks, is getting the substation. That's often a long lead time item. So it's early. We haven't closed on the site yet. We're gonna close in the coming days.
We have already had interest in the site. I don't know what we'll do with it. It will also be a fantastic Bitcoin mining site, if that's where we go. But given that it could conceivably deliver three nines of uptime as soon as late next year, there's just almost... There are very few sites available with those characteristics, and they are in extremely high demand. You know, that said, we're still early, so I'm very enthusiastic about what I've seen from potential tenants and potential capital providers to debt finance a powered shell there. But it's early days, so we'll see, and if we get that to the finish line, I think we could have a transformative-type transaction at that site.
If not, if we built it for Bitcoin mining, it will also be available next year. You know, again, we'll have to walk through the ways we get to finance it, but if we financed it, we could be the biggest Bitcoin miner in the world in 2025 if we put Bitcoin mining there. Again, you know, no promises, because we haven't gotten into how we would finance that on the Bitcoin side. We're starting out with the HPC discussions.
Got it. You did mention, you know, all the sites you could participate in demand response programs. You know, so are the hyperscalers, if they're doing HPC, are they comfortable with that? That's a program that you're not required to opt into, you just have the right to do so, correct?
Correct. Typically, they don't wanna curtail any time. The discussions there are typically, you know, they want five nines of uptime, so 99.999% uptime. The delivering those last two nines instead of three nines, which is 99.9% uptime, those last two nines create a lot of expense and a lot of timeline, because the key for that is the availability of backup generators. So that if you don't have access to regular way power, those kick in. The lead time on those is very long. It's typically, you know, 18-24 months. And so to get five nines, I think what we're seeing is there's such a push based on the conversations we've had. So listen, it's a somewhat limited sample set.
I'm sure there's investors on this call that know more about the space, frankly, than we do. But based on what we've seen, there is such a push for large-scale sites that could be available in the near future, that we have had a lot of discussions about, like, a lease that contemplates taking it with three nines of uptime, until generators are available, and then delivering five nines of uptime. So a lot of that is around timeline, but that's generally where the conversation is. Now, listen, I don't think a lot. There weren't tons of folks looking for large sites in West Texas, necessarily, two years ago. So the space, I do think, is evolving.
Could it evolve in the future if you see a stratification of the various compute needs to move to, like, a split of how they arrange their compute needs, such that they do take advantage of curtailment opportunities? You know, maybe, but that is not where the dialogue is right now. They are envisioning buying backup generators, drawing power all the time.
Got it. Okay. And within that, as we think about West Texas, so just some numbers for the folks in the audience. Applied, which is doing an HPC data center, their first 100-megawatt site is running at $10 million per megawatt, and we would expect this, they want to bring another 300 megawatts online if they get financing. That would be, those builds would be maybe a little bit higher, $10-$12 million per megawatt. It seems to vary, right? Core Scientific is out there saying it's gonna be $5-$8 million to retrofit some of their sites. We've heard from other miners who wanna get into West Texas, that West Texas might be a little bit more costly. In a lot of locations, you gotta get the fiber built out and access that.
So, you know, are you guys thinking, would it fall into that $10-$12 million range? Would it be higher than that for West Texas? Any kind of guardrails?
So yeah, I'll try to give you some color based on what we've seen, but let me also, again, set the framework. Depends a little bit on which sort of piece of the industry or business you're targeting. So, for example, building out fiber, it depends on the site. Again, our sites have good access to fiber, so it's not inconceivable that someone could have something way out in the sticks, not along a fiber line that is more difficult to run fiber to. But I just in general, I have been kind of pleasantly surprised that in our conversations with hyperscaler-type tenants, fiber's not as much of a concern. I think they do so much work with the ISPs, that they'll get the fiber out to the site if they like the site.
Now, that's very different if you're buying your own GPUs and want to sell the compute or something, and like, a typical Bitcoin miner is not necessarily gonna have that kind of pull with the people they need to run the fiber that's necessary to the site. So that, in other words, it's not inconsistent with what you're saying, John, but that is not necessarily what we've seen. On the cost side, again, I can show a bunch of analogies to Bitcoin mining.
Lots of people like to quote, "You know, we build our data centers at X per megawatt." And often they'll have a really cheap number, and they'll say that, "You know, we're better than this other miner that spends more than that." But of course, like, quite often, you make design decisions where you save CapEx, and it leads to spending more OpEx in the future. So it just 'cause the number is more, doesn't necessarily mean that there aren't other knock-on implications for that. I say that because a lot of, at least the piece of the business we're talking about, is more building to spec. So if you're talking to a big hyperscaler, they'll give you a list of requirements that's got several hundred line items on it. And, like, they're not all yes, no.
Like, some of them are like, "What can we get?" And, "Oh, we don't normally compromise on this, but if you're saying we could have a data center ready next year, we would." And then, a lot of the cost stems from that. Like, what are you delivering? So again, I can give you a kind of rough picture. If you're delivering three nines of uptime, it might be about $4 million a megawatt. If you're delivering five nines of uptime, it might be closer to, like, $8 million a megawatt. Ten's not unreasonable, like you said, depending on what you're building.
If you're building a more sophisticated build that perhaps has a better PUE, and so is delivering more of the megawatts to be critical IT load as opposed to like, you know, using for cooling and things like that, it could change the cost significantly. So I think what we've seen is that kind of $4-$8 million, depending on how many nines we're delivering. But that's really early, and that's based on kind of rough, you know, work we've done. That's not based on an actual build. And there's a lot of variables, like land and so forth. So like, you know, again, you said West Texas, it depends on how remote, I guess, but obviously, like, buying land really close to Dallas-Fort Worth is gonna be very different than in Wink, Texas, you know, where Black Pearl is-
Yeah
... for example.
... In those conversations with hyperscalers, so you said there's hundreds of line items they wanna go through. Obviously, you know, timing is a big one, the backup generators, all that. Is there anything in there that would be kind of a differentiator between Texas and MISO or some of the other regions?
I think some of the factors make it easier. Look, I know, like, if you're in a cooler environment, right, obviously, these buildings have to be kept cool. You know, if you're in Upstate New York, it's gonna be cooler temperatures a lot of the time than West Texas. So there are things like that. I don't think any of that prevents interest in having data centers in places like Texas. It might mean, again, that your efficiency's a little worse or your build cost is a little more expensive to get to a targeted PUE, but it's not that they're not interested. I think the upside of places like Texas is there are opportunities to find data centers that have 300 megawatts of capacity. There's just not that many of those in some of those other locations.
Got it. Okay, that's great. I have maybe one more on HPC, then I want to get to your power dynamics and some of the operational leverage that should come with all this new hash that's coming online. You know, I think from... You know, I guess just the last point. So you talked to some of these potential hyperscalers on a lot of the various line items. How far along would you say those processes are? Are you talking to, you know, like, 20 different private and public hyperscalers, or is it kind of a earlier date?
I'll say, I'll say fewer than 20, but several. I think the hard thing is that, you know, some of those companies are very, very large companies, and just because they are very large companies, processes are a little bit more deliberate. They take a little bit longer. And I don't think I'm saying anything that people don't already know. Every lender in the space has opinions about different potential tenants. But you know, we move. We're a 40-person organization working in Bitcoin mining. We, you know, it's a 24 by 7 market. Like, we don't sleep a lot here. We move pretty quick. So I'd say all conversations are proceeding, and I've been very pleased with the direction of those conversations, but it's hard to comment on timing. I wish I could.
I know everyone's excited to hear more about that, and I have a feeling my experience is similar to other miners that have said they're in those types of discussions. I think it's just the nature of these discussions.
Yes. That's from what we're seeing, it definitely is. It seems to be a long period where you're having those conversations with hyperscalers before you get to an LOI, and then months before you get to an actual agreement from there. Okay, so wanted to turn to, and just a reminder to the audience, we will take Q&A. We're gonna reserve the last five minutes. I think we're scheduled, the full session, to 9:25 A.M., so at about 9:20 A.M. Eastern, I will take a few from the audience, but next one or two from me. I do wanna talk about the operational leverage piece, but first, I want to talk about the power dynamics.
I believe one of the sites that has fixed power at $0.027 per kilowatt hour. What is the timing of that contract again? I, the PPA. I believe it's-
So our largest site is in Odessa and has a PPA, right around there, our portfolio average of $0.027 per kilowatt hour. That has nearly three years left. I think it's July of 2027.
Okay, and so as you look forward past the 2027 date, kind of what is your confidence in having power costs around the $0.03 range, or-
Yeah
... or what can you get?
I think it's pretty high. I think it's just that we will get there. It depends on the market. That fixed price contract has been a wonderful thing for us to have. It's been a very valuable asset. But the truth is, the way that, the nuance of that contract is, the reason it's so cheap, is that we've traded a curtailment option to the power provider. And so 5% of the time, they can tell us to turn off because market prices for power are high, and they can make a bunch of money selling power at much higher levels on a hot summer day in Texas. For trading that 5% option, they give us that cheap fixed price the other 95% of the time, and this is because of the way power prices work in Texas, in ERCOT.
In the locations where we are, prices are generally very low almost all of the time, except when they're not, like a hot summer day, when you maybe don't have a lot of wind or the sun's going down, and everyone's running their air conditioner, and prices can go up 100X for a couple hours. If you look at from the power provider's perspective, they make the lion's share of their revenue in those couple percentages of hours per year, where they get to make 100X the revenue. We've let them keep that. We've traded them that. Bitcoin mining sort of looks the same all the time, and we get the cheap price 95% of the time. That's the way that contract works.
Now, what's happened is, basically, we've built a whole tech stack and operation stack that pulls live market prices and makes automated decisions to have our machines running or turned off. At a front-of-the-meter site, so a site where we are not co-located with the power provider, but we are buying from the grid, effectively, at that floating price, we manually recreate the same process, which is to say, we are looking at what live market power prices are, and in those couple percent of time where they spike to very high levels, our machines shut off. So let me give you a sense with some numbers, and these move around, so don't hold me to this exactly. I haven't done this math like yesterday, it's been a couple of weeks.
But if you look at Texas, and the sort of front-of-the-meter pricing in West Texas, broadly speaking, over a huge sample size, you might expect to pay, like, $0.065 a kilowatt hour or $65 a megawatt hour. If you just cut off that 5% of time, like we do typically at our sites for the most expensive hours, your average price would go to about $0.035 or $35 a megawatt hour. Now, that's a really attractive price just right there. That's cheaper than almost everyone's price to mine Bitcoin among our competitors.
And then furthermore, if you participate in the various demand response and ancillary services programs, where you can offer up your capacity to the grid, you can also be paid for giving them the right to curtail and get paid for that. And sort of net of that, you know, we think you'll be able to drive our overall pricing below $0.03 or $30 a megawatt hour in large sample size. And that's for front-of-the-meter sites in West Texas. That's what Black Pearl - that's how Black Pearl will operate when it's up and running. And at Odessa, it depends. We've got, again, almost three years left on that contract. It depends on what we decide to do going forward there. There's a chance we just switch to, like, floating prices there and do the same math.
And so, you know, we would end up being a few dollars higher per megawatt hour, at least in current markets. On the other hand, power prices have been coming down recently. The curve, you know, was not as hot or, you know, power prices were not as high this summer in Texas. There's been more capacity built or more generation built, and also batteries have been very influential this summer. So you know, what we've seen is that the curve overall is coming down. If we get back to a place where prices are really low, maybe we strike a fixed price contract. And by the way, we could do that financially at our front-of-the-meter sites too. So we're always either gonna, I mean, part of the reason we have built out...
I think if you compare Cipher to some of our competitors, we've got far more resources invested into trading and hedging and the, and the technology that drives that, not necessarily, we also do lots of technology around running rigs, but, a lot of, like, markets and trading-based technology. And so that's to be able to create our own, power price effectively. That's very attractive.
Got it. That's helpful. Last one, I do wanna touch on the operational leverage aspect, because I do think it's a big piece. So you have a lot of exahash coming online, as we talked about as the first question. G&A, and some of your other costs, we should, you know, I think for the most part, we have kind of fixed moving forward, at least in the mining side. Walk us through the dynamics a little bit there.
Yeah, so the thing I tell people is, so we have a great team. I think it's the best in class at every position. Please go to our website and read the bios of our people. They are extraordinarily qualified, and they're experts in their space. Expertise is a little bit more expensive, it costs money, and they work in advance of the sites coming online. We are a company that our edge has been finding greenfield sites, negotiating our own power contracts, taking them from a patch of dirt all the way through to a functioning data center, where then we're trading, hedging power, and making decisions around that every day. We do that, I think, better than anyone else in the industry, and that scales very well.
So that same team, and basically, that exact same cost structure, is building the 300-megawatt site at Black Pearl right now, and that cost structure will not change very much, even though we will quadruple our exahash output. And so all these costs scale quite a bit, and so that's the best way, in our opinion, to grow. If you look at the economics around mining right now are challenged. You know, hash price is at a historical lows. So hash price is the amount of dollars we get paid every day for the compute we produce, and you can think of it as taking into account both Bitcoin price and network hash rate and transaction fees, sort of all the things that we get paid for on a dollars basis.
And so the key for any Bitcoin miner, if you're evaluating them, is they wanna have a good spread between their hash cost, what they pay to produce that hash, and what they get paid for it, the hash price. We think the best way to drive the hash cost down is to build more capacity, grow, and scale. And so to give you a sense, you know, recently, our hash cost was about $43. And we've seen hash price recently float in the low $40s, down below that even a couple of days recently with the recent sell-off. But as soon as we plug in those machines later this year, so in the next couple of weeks, basically, at Odessa, just having that extra top-line output drives that hash cost to $29.
When you bring on Black Pearl, I frankly, I haven't done the math. It will go way lower than that because you'll be multiplying that times, you know, nearly three X in terms of the top line, with very little growth in costs. You know, there'll be some growth, maybe around the edges, some extra bodies that have to do work at the new physical facility, et cetera, but that all scales really well.
... That's great. Thanks for that. Let's do. I think we have time for about one question from the audience. This one, I think kind of wraps everything together nicely, but it asks on the, you know, and it kind of brings us back to HPC. It asks what expected margins would be on HPC, and if you think you're gonna have to do additional hires that would drive G&A costs up because of that, or if your current team could just kind of add HPC capabilities.
I think it's a little bit early to be precise about that because it depends on the scope of the opportunity. I mean, we now have five large-scale sites that are unbuilt in the pipeline, and if it just turns out every hyperscaler wants to give us a huge lease there, we'll hire more just to be able to get five sites simultaneously running and online. And it depends on the scope of the opportunity. Currently, we don't anticipate much additional SG&A or additional hiring. We are using our current team, which has a lot of expertise, and can build, you know, one other data center while they're building our first giant data center. I'll remind people, when we started the company, we were building four data centers simultaneously. So I know we're capable of doing that.
The process gets a little more complex with HPC, but it's a very similar process. It's just got more component pieces, really, that you're operating. So I don't anticipate much. If there is hiring, it would be a good thing because that means the opportunity is even bigger than we're sort of first investigating.
Got it. Understood. That was great. Tyler, I want to thank-
Oh, and hey, John, wait, I forgot-
Yeah
... I guess one other thing. When it comes to the leases, I think what we've seen from hyperscalers is typically, they, they look to provide, like, a 10%-11% unlevered yield. So it depends on, you know, how your financing ties out to see what your return on equity is there, but it's significant. Sorry, I know, but I had left that out on the answer.
Got it. Okay. That's good. That's helpful. Okay, great. With that, I wanna thank Tyler, Ed, both you guys for joining us. Appreciate the time and insights as always today.
Thank you so much for having us today, John.
Okay, great. Thank you. Take care.