Cipher Mining Inc

NASDAQ:CIFR

Earnings Call Transcript 2024-Q3

[Operator]

Good day and thank you for standing by. Welcome to the Cipher Mining Third Quarter 2024 Business Update Conference call. [Operator Instructions] Please be advised, today's conference is being recorded.

I would now like to hand the conference over to your speaker today, Josh Kane, Head of Investor Relations. Please go ahead.

[Joshua Kane, executive]

Good morning and thank you for joining us on this conference call to address Cipher Mining's third quarter 2024 business update. Joining me on the call today are Tyler Page, Chief Executive Officer and Edward Farrell, Chief Financial Officer.

Please note that you may also review our press release and presentation, which can be found on the Investor Relations section of the Company's website. Please note that this call will also be simultaneously webcast on the Investor Relations section of the Company's website, and this conference call is the property of Cipher Mining and any taping or other reproduction is expressly prohibited without prior consent.

Before we start, I'd like to remind you that the following discussion as well as our press release and presentation contain forward-looking statements including, but not limited to Cipher's financial outlook, business plans and objectives, and other future events and developments, including statements about market potential of our business operations, potential competition and our goals and strategies.

Forward-looking statements and risks in this conference call, including responses to your questions, are based on current expectations as of today and Cipher assumes no obligation to update or revise them, whether as a result of new developments or otherwise, except as required by law.

Additionally, the following discussion may contain non-GAAP financial measures. We may use non-GAAP measures to describe the way in which we manage and operate our business. We reconcile non-GAAP measures to the most directly comparable GAAP measures, and you are encouraged to examine those reconciliations, which were filed at the end of our earnings release issued earlier this morning.

I will now turn the call over to our CEO, Tyler Page. Tyler?

[Rodney Page, executive]

Thanks, Josh. Hello, this is Tyler Page, CEO of Cipher Mining. Thank you very much for joining our third quarter 2024 business update call. We've had an extremely busy few weeks recently at Cipher, and our business model has rapidly evolved from being just a Bitcoin miner to being a developer of HPC data centers, with a natural built-in offtake via Bitcoin mining for prospective sites.

We believe that we have found a truly unique niche by opportunistically investing in greenfield development sites knowing that we can build and operate an HPC site for a tenant, or if a high quality HPC tenant fails to materialize, we can always continue to expand our Bitcoin mining footprint, and put the sites to profitable use. We have closed 5 such deals to acquire greenfield data center development sites in the last 2 months.

Our current primary intent for these sites is to develop them as HPC data centers, but all 5 sites are located in Texas and would also be excellent sites for bitcoin mining as we can always use our proven ability to manage power curtailment and produce our own best-in-class electricity costs.

Cipher combines expertise in site origination at the front end of the development funnel, with an experienced team of construction and operations professionals that joined us from the hyperscalers. Our team has built some of the most high-tech data centers in the world, and continues to innovate in both the HPC and bitcoin mining space. Further downstream, we believe Cipher's talents in operating the technology and trading needed to manage energy prices and the curtailment process will bear fruit as the entire data center industry evolves.

According to a recent research piece from JLL, data center industry demand is forecast to grow at a 23% compound annual growth rate through 2030. The demand for large scale data centers, driven by the rise of large language models and AI, is seemingly growing even faster and those sites are increasingly hard to find. Regulators and system operators are now suggesting hyperscalers match requests for new power interconnections with new generation development in order to receive required approvals.

This trend will further extend waiting times for large interconnects, and this situation is juxtaposed against an environment where the chief executives of the hyperscalers are ramping up their CapEx spend in the race to be the winner in AI. It is against this backdrop that we launched our HPC vertical. We believe that large scale interconnects available in the next few years are exceedingly rare and valuable, and I am excited to tell you more about our progress today.

Before I talk about our new development portfolio, let me begin the call by updating some key metrics for Cipher as of the end of September 2024. While we have major growth coming in the near future with our new sites, we are also in the middle of a significant expansion of our bitcoin mining business right now.

We operated 9.3 exahash per second of self-mining capacity at quarter end, and as of this morning are in the middle of installing our Odessa upgrade and have grown to 10.5 exahash per second. By the end of the year, we expect our self-mining capacity to grow to 13.5 exahash per second with a fleet wide efficiency of 18.9 joules per terahash.

Cipher continues to manage a significant bitcoin inventory, holding 1,508 as of the end of the third quarter, and we are probably best known in the bitcoin mining industry for our very competitive all in weighted average power price of $0.027 per kilowatt hour.

Electricity represents the large majority of our operating costs, and our low price is a key driver of our outstanding unit economics. In the next 2 months, as we complete the Odessa rig upgrade, we will be pumping the industry's cheapest electricity through 1 of its most efficient fleets.

Now let's move to a review of our current operations. On Slide 6, we give a portfolio overview of our existing data centers in a near term timeline for expected scaling of our managed power capacity. Year-to-date, we paid an average all-in electricity cost of roughly $18,162 per bitcoin produced at our data centers. We are very proud of this number.

Please note that when we talk about all-in electricity costs, we mean the total cost to deliver electricity to our mining rigs. So our numbers include all taxes, transmission, and other charges, and our low numbers dramatically demonstrate our competitive advantage.

On the left side of the slide, we show an overview of our production split across Odessa and our JV data centers, along with our all-in electricity cost per bitcoin at the sites year-to-date.

The chart on the right side of this slide gives you a graphic illustration of the number of megawatts we currently manage as well as the projected growth for the coming year. As you can see, we currently expect to manage 927 megawatts across 6 data centers in 2025 when we bring Black Pearl and Barber Lake online. At this point, we will turn to production by site.

Slide 7 has a production summary of our Odessa facility. Odessa is the most significant part of our portfolio, as it represented approximately 83% of our bitcoin production in September. Recently Odessa became the first bitcoin mining data center to be awarded the Uptime Institute's stamp of approval for Management and operations.

Odessa is a wholly-owned facility in the middle of a 5-year fixed price power purchase agreement, and pays some of the lowest prices for power in the industry. As of quarter end we generated approximately 7.1 exahash per second at the site, using approximately 207 megawatts. Those same 207 megawatts will generate roughly 11.3 exahash per second, with the rig upgrade we are in the middle of executing now.

On this page, we also provide the observed all-in electricity cost per bitcoin at the site post halving, which was roughly $25,488. Even after the April having reduced the number of new bitcoin paid to miners, you can see how valuable it is for Cipher to have a cheap fixed price of power available on such a large portion of our portfolio.

On Slide 8, we highlight our joint venture data centers of Alborz, Bear & Chief. With the recent expansions at each of Bear & Chief, the sites now have a total power capacity of 120megawatts, and generate approximately 4.4-exahash per second. We own 49% of the JV sites and our portion recently generated roughly 17% of our overall bitcoin production.

On this page, we also provide the observed all-in electricity cost per bitcoin at the sites post halving, which was roughly $34,160. As a reminder, Bear & Chief operate as front of the meter sites, so there will be some expected seasonal fluctuations with their electricity costs, and summer months tend to be higher.

Now let's turn to an update on our development portfolio. Slides 10 and 11 show a rendering of the completed data center at Black Pearl, and photos from the current site work underway. We are scheduled to energize the site in the second quarter of 2025. Everything for Phase 1, which is the first half of our building, and the full 300 megawatt substation is progressing on schedule.

Our current design envisions 250 megawatts of air cooled and 50 megawatts of liquid cooled bitcoin mining. At full capacity, the site is anticipated to produce roughly 21.5 exahash per second of hash rate. We have continued to receive inquiries on our willingness to repurpose a portion of the data center for HPC hosting, and ultimately our final design at the site will depend on what we think will produce the best outcome for shareholders.

Slide 12 gives an overview of our new Barber Lake site that we acquired last month. We immediately recognized the potential for Barber Lake when we first saw it. The site has an approved capacity of 300 megawatts, and we purchased 250 acres of surrounding land. Perhaps most importantly, the site already has an existing energized substation, so any data center will be immediately available for use upon completion of construction.

When you also consider that it is located next to the major fiber line running along i20. This site is ideal to host a large HPC tenant. Every potential tenant who has seen it thus far has expressed interest given its optimal setup. We look forward to updating the market in more detail, as we progress in our various discussions.

Slide 13 shows an overview of the Reveille site, which is the first site in our medium-term pipeline as it is scheduled to energize in 2027. By the time we turn on Reveille, we will have already been managing our initial large HPC sites at Barber Lake and potentially Black Pearl. The site is located in Cotulla, Texas. It has been approved for 70 megawatts, but based on early discussions with the transmission service provider, we believe we can expand the site capacity to 200 megawatts by the time the site is energized.

Given the timeline to energization, we have a lot of flexibility on Cipher's strategy for the site. We may choose to build a powered shell data center for a hyperscaler, and secure a long-term lease from a high quality tenant. But we also have the potential to expand our capabilities, and will review a variety of potential business models including more of a multitenant model, or even managing our own fleet of GPUs. We have some time to watch the market develop and evolve, before we complete our strategic planning.

Slides 14 to 16, give overviews of the 3 sites covered by the purchase options that we recently acquired Mikeska, Milsing, and McLennan, or what we call the 3Ms for short. These sites are the furthest out in our development pipeline, as they are pending final approval for interconnection, and we expect the results of approval processes for the sites to be finalized in the coming year.

We hope to receive approval for up to 500 megawatts at each site. In addition to the interconnections, our purchase options also cover substantial parcels of land at each site. All 3 sites have the necessary characteristics for development of HPC data centers, but also sit in locations with demand response programs that would allow us to monetize the flexibility of curtailment used in bitcoin mining operations if necessary.

With these sites we have a lot of optionality, which is exactly where we like to be positioned in front of trends, with the potential for massive growth. As you can see, our evolution as a development company has occurred rapidly so far. We are building on our demonstrated success of originating the best sites and power deals in Bitcoin mining, and bringing that expertise to the traditional large scale data center market.

As that market is evolving and forcing large tenants to go to non-traditional areas for the scale they need now. It feels like the entire market is moving towards us. As we finalize our plans for Black Pearl and Barber Lake, and define our long-term ambitions at the sites further out in the pipeline. We are extremely confident in our positioning.

So why are we so confident in our positioning? While we don't yet have specifics to confirm today on our current HPC business negotiations, a simple review of current market conditions and the economics of operating GPUs demonstrates why there is so much interest in our sites. We have talked about the scarcity of overall capacity given the current and projected growth in the data center industry, and the dearth of large scale sites in particular.

Against that backdrop, we have 2 of the largest suitable sites available that can be used as HPC data centers, before the end of 2025. Hyperscalers that want their own large site have few options in the market if they want to operate within the next 3 years. And the general view among those companies is that they are in a race to win AI supremacy and need to accelerate development as quickly as possible.

Mark Zuckerberg recently said that he would, "Rather risk building capacity before it is needed rather than too late given the long lead times for spinning up new infrastructure projects." Hyperscalers can currently generate tremendous revenue from investing in GPUs. This year NVIDIA has estimated that companies can generate $5 to $7 of revenue over 4 years for every dollar spent on their GPUs.

To our potential tenants, Cipher can offer 2 extra years of operations on up to 600 megawatts across our Black Pearl and Barber Lake sites at a critical point in the race, compared to waiting for other sites to be ready. The additional potential profit for tenants from that time acceleration, amounts to many billions of dollars. These are the exciting market dynamics that are currently driving interest in our data center sites, and ultimately giving us great confidence in the success of our HPC business.

We expect to sign a long-term lease with a high quality tenant that will generate substantial returns to Cipher shareholders, and I look forward to updating everyone when we have more specific details to share.

With that, I'll turn it over to our Chief Financial Officer Ed Farrell.

[Edward Farrell, executive]

Thank you, Tyler and hello to everyone on the call. I'd like to begin by sharing some high level thoughts on our recent site acquisitions, which are a critical part of our HPC initiative and represent significant investments for us this quarter. As Tyler has mentioned, being a leader in this space requires not only great sites, but also an experienced team and strong expertise in financing.

The ability to secure such attractive sites is a direct result of the foundational work we did when we established the company. Although we are still in the early stages of our HPC initiative, we believe the strength of our team, our balance sheet, our tech stack, are key elements that will position us as a leading developer of HPC data centers. The strategic investments we've made in these areas have enabled us to act swiftly, and capitalize on unique opportunities like Barber Lake.

Turning to earnings, it comes as no surprise that the third quarter was a challenging one, given that it was the industry's first full quarter post halving, revenues were down. However, we remain encouraged by the business' underlying performance and the company's overall growth trajectory. Our access to low-cost fixed price power and our strong balance sheet continue to be critical strengths in maintaining a solid financial position.

Slides 19 and 20 give a snapshot, which we provide every quarter of some of our financial metrics on both sequential and year-over-year basis.

Let's move on to slide 21 and dive into the numbers in more detail. Similar to last quarter, we encountered significant industry headwinds including a record low hash price, and a rising network cash rate. For the quarter we recorded a GAAP net loss of $87 million, compared to a loss of $15 million in the prior quarter and $19 million in the same quarter last year.

In the current quarter, we mined 396 bitcoin, generating revenues of $24 million at an average price per bitcoin of $61,000. This compares to 563 bitcoin mined in Q2, 2024 at an average price of $65,000, resulting in $37 million in revenue, a sequential decrease of 35%. Year-over-year, revenues decreased by 20% primarily due to the halving, the decline in bitcoin prices, and the increase in network hash rate.

As I mentioned earlier, our fixed price power remains a critical factor in maintaining attractive unit economics. In the current quarter, our cost of revenues increased by 5% sequentially. This increase was primarily driven by one-off expenses related to the fleet upgrade at Odessa.

Excluding these, our cost of revenues remained flat quarter-over-quarter, thanks to our fixed price PPA at Odessa. We'll discuss the quarterly pricing of the PPA in more detail later, but its true value is evident in the low-cost fixed price power, which is reflected in our cost of revenues.

Moving on, as you recall, we adopted the new crypto fair value accounting standard in 2023 and with a drop in bitcoin price in the quarter, we recorded an unrealized loss of $22 million on the fair value of our bitcoin holdings. However, this mark-to-market loss was offset by $20 million of realized gains from the sale of bitcoin in the period. This resulted in a net loss of $2 million, which is reported in the financials.

Our philosophy towards the growth of our bitcoin inventory, and our approach to treasury management has not changed. We maintain an opportunistic approach, continuously evaluating various funding options to support our growth initiatives. While our general aim is to grow our bitcoin inventory over time, our decisions are driven by market conditions, and our overall capital allocation strategy.

We actively assess the markets to identify the most attractive sources of capital, carefully considering the advantages and disadvantages of different funding methods to support our business and expansion plans efficiently. This may involve using our cash reserves, bitcoin holdings or issuing equity. An example of this approach during the quarter was the acquisition of Barber Lake, which we funded through the sale of part of our bitcoin inventory.

While we remain highly constructive on bitcoin, using our bitcoin holdings to fund the acquisition, was an optimal choice. As Tyler has said, we exchanged one rare and valuable asset for an even more rare and valuable one. As of September 30, we held 1,508 bitcoin in our Treasury.

As in previous quarters, I'd like to spend a minute on G&A expenses and our philosophy for managing these costs. On a quarter-over-quarter basis these expenses remain relatively flat. Compensation and benefits decreased $1 million sequentially to $15 million. This was primarily driven by a decrease in share-based compensation, and current quarter versus prior year quarter decreased 14%, also due to decrease in share-based compensation expense.

Now onto general administrative expenses, which include IT, corporate insurance, professional fees, occupancy, and other public company expenses. Sequentially, these costs remain relatively flat. On a current year quarter versus the prior year quarter, these expenses were up 31%, driven by professional fees, and public company expenses primarily related to Sarbanes-Oxley compliance.

As we discussed last quarter, we have made significant investments in both our team and technology stack, which we believe are crucial to our long-term success. These investments have already proven to be key differentiators in early stages of our HPC initiative. We expect them to continue to drive future top line growth, ultimately having a positive impact on our bottom line.

Depreciation and amortization expense totaled $29 million, an increase of $9 million or 41% from the prior quarter and a 77% rise, compared to the third quarter of 2023. The sequential increase was driven by the recent change in our depreciation schedule, for our mining rigs.

As a reminder, we had previously accounted for the depreciation of rigs over a 5 year period. However, given our recent fleet upgrade and rapid efficiency gains with next generation rigs, we now believe that the 3-year depreciation schedule is more appropriate.

Our expectations for upgrade cycles and our ability to purchase and install much more efficient machines have evolved, and we believe this should be reflected in our accounting treatment of the entire fleet. We made this change in the second quarter and accounted for it prospectively.

Now let's turn to our non-GAAP measures slide where we reconcile our adjusted earnings. As always, I'd like to remind you that adjusted earnings exclude the impact of depreciation and amortization, the non-cash changes in the fair value of our derivative asset, deferred income tax expense, the non-cash change in the fair value of the warrant liability, share-based compensation and other non-recurring gains.

These supplemental financial measures are not measurements of financial performance in accordance with U.S. GAAP. However, we believe that these non-GAAP measures may be useful to investors for comparing, our performance across reporting periods consistently. Internally management uses these non-GAAP financial measures, to better understand, manage and evaluate our business performance and to facilitate operational decisions.

When adjusting our third quarter GAAP net loss, we add $84 million for the items I just listed. This brings us to an adjusted net loss of $3 million for the quarter, compared to an adjusted net loss of $3 million in the prior quarter, and $2 million of net income in the third quarter of last year.

Now let's turn our attention to the balance sheet at September 30. Our total current assets amounted to $152 million. Our cash position came down to $25 million, a decrease of $97 million from the close of the second quarter of 2024. Our liquidity position as of September 30 is $121 million, comprised of $25 million in cash and $95 million worth of bitcoin.

During the quarter, we made significant investments with the purchase of Barber Lake for $67.5 million, $94 million in deposits for miners and $36 million for the development of Black Pearl.

I'll quickly cover some of our balance sheet line items. At September 30, prepaid expenses amounted to $3 million. This is primarily related to corporate insurance. We reported a bitcoin balance of $95 million, reflecting the 1,508 bitcoin held in Treasury. This figure marks an increase from the 780 bitcoin held at year end 2023 valued at $33 million. In the third quarter we liquidated 1,167 bitcoin or $69 million.

Now, I'd like to turn our attention to the value of our Odessa power contract, which we record as a derivative asset. As we've discussed previously, this contract provides us with significant competitive advantage, enabling us to be a low-cost producer of Bitcoin.

To recap, we began reporting a third-party mark for this agreement in the third quarter of 2022. This mark is recorded as a derivative asset on our balance sheet and has reevaluated each reporting period. Essentially, it represents the in-the-money value of the contract based on time value and prevailing forward power prices at our Odessa facility.

As we reminded investors each quarter, seasonality and gradual expiry of the contract impact the assets pricing leading to expected fluctuations in quarterly valuation. Given the unexpectedly mild summer we experienced in Texas and the corresponding drop in forward power curve, we have seen a significant quarter-over-quarter decline in valuation. However, while this lower mark is reflected on our balance sheet, it in no way diminishes the substantial value and competitive advantage the contract provides by securing low-cost fixed price power at our Odessa site.

As of September 30th, this asset was valued at $74 million, reflecting a $49 million decrease in the third quarter and a decrease of $19 million from year-end. This change is recorded as a loss on our statement of operations. As always, fluctuations in the fair value of this contract will impact our GAAP earnings, but we exclude it from adjusted earnings. This is particularly important to note given the lower mark contributes to more than half of the net GAAP loss for the quarter.

Other significant assets include property and equipment totaling $311 million, primarily attributable to our Odessa facility. Within this category, mining rigs and related equipment account for $182 million. Leasehold improvements are valued at $138 million. Land of $25 million, infrastructure of $33 million and construction in-progress of $56 million. These figures are net of $123 million in accumulated depreciation.

Deposits on equipment of $145 million primarily consists of progress payments we've made in accordance with previously announced mining rig purchases. Additionally, we hold intangible assets totaling $26 million with $24 million attributed to the ERCOT approval at Black Pearl and Barber Lake and the remaining $2 million related to capitalized software.

At the end of the third quarter, our equity investee interest in Alborz's, Bear & Chief JVs stand at $55 million and we had operating lease obligations of $11 million. We had security deposits totaling $15 million, which represent the encore deposits related to the construction of our Black Pearl and Barber Lake data centers.

There were no significant changes to the liability side of the balance sheet from year end. And as we've reported in the past, we have no debt that hinges our capital structure. As always, we look forward to updating you in greater detail on our growth plans over the coming quarters.

I will pause now and Tyler and I are happy to answer your questions.

[Operator]

[Operator Instructions] Our first question comes from Mike Grondahl with Northland.

[Mike Grondahl, analyst]

Congratulations on the development pipeline. You guys have made just a ton of progress there recently. And related to that, Tyler, I wanted to ask, how are you -- with 2 good opportunities, how are you thinking about allocating capital going forward between Bitcoin mining projects and HPC projects? How are you thinking about that?

[Rodney Page, executive]

Mike, thanks for the question. So -- and it's a good one because it's one that's occupying a lot of our head space. I think the easiest way to answer that is to say that we're focused on what's going to drive the greatest shareholder returns. We're driven on getting the stock value higher and the decision making goes around that. I think they're both very good operating verticals to have. They're just very different.

As you know well, the Bitcoin mining, process is really subject to kind of a cyclical market. We go through 4 year halvings, we get chip upgrade cycles, we get booms and busts, and bull markets and bear markets. I think of the HPC hosting business is a very different kind of profile. That is a business that you're going to have at least the way we are playing it, looking to have a really high quality counterparty on a long-term lease. So very reliable cash flows that can be heavily debt financed from a CapEx perspective. It's just a completely different profile than Bitcoin mining.

I think one of the challenges of Bitcoin mining, at least where the market is today is that lenders and a lot of investors view it as obviously volatile, but also having a lot of correlated risk. And a lot of it is wrong way risk if you look at some of these miners in the sense that their business relies on hash-price the value of the equipment they own is related to that. And if they have a Bitcoin treasury, that's correlated as well. And so, what's interesting about the HPC business is it really is a completely different profile. If you've got a high-quality counterparty, you can be looking at loan-to-cost percentages north of 80% even.

And so I think when we look out, we're really driven on what drives the most value, that revenue stream will be valued differently by investors, but I think it's very complementary to Bitcoin mining. I think what's been interesting about the way the market has developed over the last couple of months is, we have developed great strengths at finding large scale sites that may be overlooked. And again, Bitcoin mining traditionally was in different locations than HPC tenants were looking. That is changing because they need such larger sites now. And so that market has really kind of come to us.

So as we go forward, I think of Cipher really as a data center development company that has this built-in kind of hedge that, let's say HPC turns and it's not as hot as it clearly is right now, we've got these sites that work great for Bitcoin mining if you can manage curtailment, which we do. So I expect to have both businesses. It's really just a question of site specific, market specific, and where can we be the most opportunistic.

[Mike Grondahl, analyst]

And then in terms of some of your initial discussions with HPC customers, hyperscalers, and really financing partners. I mean, are you getting past the initial stages of those discussions kind of where are those and how are they developing?

[Rodney Page, executive]

So let me speak really broadly and generally because I'm obligated to not get too specific on some of those discussions. But look, if you look at that whole marketplace, I will oversimplify it. It is more complicated than this. But roughly speaking, you're going to have a client that wants a long-term lease, say, 15 plus years and they are going to pay some percentage of the total cost to build at the site, including the value of the site.

And the percentage they're going to pay in that lease is going to be somewhere from the high single-digits to the mid double-digits per annum. And there are elements to that, it's a little bit more complicated in that, you're contributing these sites probably at a different value than we paid for them, for example, because we got such a great deal.

I mean, at Barber Lake, we've had multiple offers to buy it for multiples of what we paid for it already. So, that's basically how that business is going to work depending on the credit quality of the counterparty, they would pay a lower percent on the lease and if they're a little bit further out on the credit quality spectrum, they'd be more in the double-digits.

And then your ability to finance that, that market is pretty deep with an executed lease with a known counterparty. And again, the amount that can be financed is somewhat driven by that credit quality and the specifics. But I would say, we have a lot of interest. We've got -- we've had many discussions. I would say we're pretty advanced. So I can't give too many more specifics than that. But like I said and you heard in the comments at the beginning of the call, we're very confident in this marketplace and the level of interest we've received that we will get it all the way to the finish line.

[Operator]

Your next question comes from Paul Golding with Macquarie.

[Paul Golding, analyst]

Thanks so much and congrats on all the development pipeline. I had a quick question on Black Pearl. Given all the conversations you're having with a potential HPC tenants, is there any change to how you're thinking about Black Pearl on a -- at a higher level in terms of whether there's optionality to slot in some HPC earlier if demand is there? And then I have a follow-up. Thanks.

[Rodney Page, executive]

Sure. So the way we are operating right now is that Black Pearl construction is going full speed ahead with the envisioned 300 megawatt Bitcoin mining data center that we have planned and everything is on-track.

I think that large scale sites that are energized in 2025 are so rare that we have received a lot of unsolicited reverse inquiry on the site for hosting HPC. And so what I'd say is, we're flexible. Again, we're kind of driven by what we think will produce the best returns for the company. If someone wants to offer the Moon and the stars for Black Pearl and we think it will produce a better investment return than building the Bitcoin facility there, we'll certainly entertain those offers.

So it's hard to say what may happen all the way at the end when it's up and running, but nothing has changed about our process and we've got hundreds of folks working out there every day making the progress that we showed in the pictures of the presentation.

[Paul Golding, analyst]

Thanks, Tyler. And then when I think about Black Pearl, it has 250 air-cooled 50 megawatt split for liquid cooled. Is there anything to glean from that in terms of how you see data center development bifurcating? Is there a potential for that site to pivot maybe more over time towards liquid-cooled if you're looking to be positioned well for GPUs, direct to chip, et cetera? How should we think about, maybe more macro for your whole fleet in terms of how you're approaching liquid since historically you've been air-cooled? Thanks.

[Rodney Page, executive]

Great question. So we view it as really important to have hydro as something that we do. Candidly, when we have modeled out investment returns and you look at the difference in CapEx and the spend for hydro, I'm not sure you're going to make it back-in Bitcoin mining. And so that's why we have generally favored doing air-cooled. We just think the ROI works better. We're pretty good at managing a fleet of air-cooled machines.

That said, the industry more broadly, the data center industry is certainly moving there over time. And I think given that we've got such a large scale site at Black Pearl, it was our view that we should do something meaningful there in hydro so that we have credible experience managing that.

And exactly as you suggested, Paul, we want to be prepared for a world where everything is direct-to-chip cooled and we've got operational experience that we can show off doing that. And it will be interesting to see how that progresses and to check-in on where the ROIs are over time on those megawatts versus the other megawatts to check our assumptions and our modeling.

[Operator]

Our next question comes from Mike Colonnese with H.C. Wainwright.

[Michael Colonnese, analyst]

First one for me is a bit of a follow-up to the previous question there. So it sounds like you're open to utilizing a portion of capacity at Black Pearl for HPC. So I'm curious how that would impact your 2025 hash-rate outlook, which currently calls for 35 exahash per second, assuming you find a great deal on the HPC side?

[Rodney Page, executive]

Thanks, Mike. So the 35 projection -- the 35 exahash projection by the end of next year envisions the full build-out for Bitcoin mining at Black Pearl. And Black Pearl represents an estimated 21.5ish of that exahash calculation. So to answer your question, it would depend on what portion would end up being HPC and it would proportionately ramp that down depending on how much of the site we would potentially repurpose for HPC.

Again, I think from our perspective, we are not changing anything about our scheduling, the progress we're making, et cetera. It's just if someone -- I think the market is still kind of defining the pricing for large scale sites that can be energized in 2025 and it only keeps moving in one direction and there's just an astounding amount of enthusiasm for sites like that, which is why we want everyone to be knowledgeable that we're open to if we see something much better giving up a portion of that hashrate for something even better. But as is projected, 21.5 exahash coming out of Black Pearl by the end of next year.

[Michael Colonnese, analyst]

And if you could just remind us of the remaining CapEx needed to complete the full build out at Black Pearl and really how assuming a portion of that capacity is allocated HPC, how that figure could be impacted as we progress through the coming quarters?

[Rodney Page, executive]

Sure. So it's hard to give exact specifics on what could end up there, but let me talk about Black Pearl as we're thinking about it from the Bitcoin mining perspective because that's a little bit more well defined.

Let me remind you that it's a front of the meter site, so we don't have any kind of take or pay obligations there beyond hitting some minimum megawattage for certain deposits we've put down. The first phase, which is the first 150 megawatts of the site has about $77 million of infrastructure spending left to go. And you could roughly double that if you were assuming the build out for the full 300 megawatts of Bitcoin mining. Now that's just -- that's sort of everything but the mining rigs.

The Reason why I'm giving some flexibility on the rigs is, if you do the full build-out of the 21.5 exahash as envisioned, recall that we have deposits down on options to buy rigs at very attractive prices, both the S21 XPs from Bitmain and the next generation Canaan machines. If we were to buy all of those next-gen machines to get to that 21.5 exahash at the site, we'd be paying about $340 million for those rigs.

That said, those are options, it's not an obligation. And I'll remind you that we are currently upgrading Odessa and unplugging about 4.5 exahash of machines that we could go repurpose somewhere else. And at a front of the meter site, we could, of course, just run curtailment and operate them profitably.

All by way of saying that our plan is to do the full spend, get the really cutting-edge machines to get that full strength going into what we expect to be a Bitcoin and Bitcoin mining, bull market next year, but if conditions didn't materialize or something like that, we do not have excessive obligations and so can kind of manage that opportunistically.

[Operator]

Our next question comes from Bill Papanastasiou with Stifel.

[Bill Papanastasiou, analyst]

Congrats on the recent developments with your hyperscale discussions and thanks again in advance for answering my questions. For the first one, just hoping you could share some more color Tyler, on the amount of capacity that these hyperscalers are looking for demanding. Do you have a potential customer in mind that is considering multiple sites? I'm just curious to hear how advantage it has been to have one of the largest pipelines in the bitcoin mining space?

[Rodney Page, executive]

Let me give some color. So at least in some of the discussions we have had, there seems to be a very outsized focus on 2025 capacity that effectively like near term quotas have not been met at some of those very large users of compute. And so, the real focus has been on what's available sooner in the pipeline. That lines up very well for us as I mentioned, that we can sort of potentially expand to other pieces of that business over time.

But our first attempts will be, let's call it on a little bit more just the vanilla hosting kind of version of the business. Most of them screen for sites that are at least 100 megawatts. I think that's a general screen. Sometimes it's even bigger, 150 megawatts. I don't know how much more color I can give than that. Other than, you know, we've got, I don't know, a half dozen folks have expressed interest. Some of that has been reverse inquiry and they've found us.

And certainly, as I've mentioned with Black Pearl, we really went into those discussions knowing how fantastic and rare Barber Lake is. It's only after some of those discussions progressed and they got to know us that, people have been pretty aggressively asking also about Black Pearl, but that's generally kind of the state of the world.

[Bill Papanastasiou, analyst]

Appreciate the color there. And then just sticking with the power portfolio or power strategy, can you share some data points in terms of how the level of difficulty has changed to secure greenfield sites and Texas over the recent quarters, and how important will it be to secure more capacity at this point now that you have roughly 2 gigawatts of unallocated power, yet you know you're seeing a number of hyperscalers getting anxious due to the long lead times?

[Rodney Page, executive]

Sure. So, I mean, obviously for context, I'm sure, as everyone probably knows, having a 2.5 gigawatts is a very large portfolio. We have found a lot of opportunities just, because historically bitcoin miners have used these big sites. So we've got some expertise in looking for them. I do think it's getting more challenging to find them. The sites we have found, they tend to be kind of sticky situations.

I don't think, though we have been very active in participating in brokered deals. We always get a call and a look, and we typically will go through and provide bids. I'm not sure we've even ever made a final round of bidding for a heavily marketed site. That's not really our sweet spot. Our sweet spot has tended to be sites that, again, for whatever reason, they need to close quickly.

There's been some history at the sites that's complicated. Maybe there's various constituents that are having challenges getting to agreement. Whatever, we've been able to kind of swoop in, and those have been sites where we've been able to really extract a lot of value. If you look at what we have paid for sites, we've done a really, really good job on it.

That means we don't get as many sites as we bid on, but we do tend to source these. I think if you look at our progress over the last quarter, in how we look at different timelines, it sort of is indicative that it's getting more challenging to find sites. The options on the 3M sites, we mentioned are earlier in the development process than we have ever gone.

So typically we have bought sites that are greenfield, but they have an approved interconnection. In the case of the 3Ms, because we see this market getting tighter and tighter and tighter, and every element of the marketplace seems to be making them harder to get. We've gone earlier in the development cycle to participate and we've been able to again, lock in extraordinarily favorable rates to acquire those data centers with some risk of the overall sizing of what gets approved.

Again, we're confident that those 3sites will be very successful in securing large interconnects. They are bidding or they are applying for 500 megawatts each. And our pricing framework is fixed based on the number of megawatts that get approved. So I do think sites are going to be harder and harder to get. As I mentioned earlier, if you speak to some of the system operators, and maybe you've seen some of the quotes in the press.

But hyperscalers now want these very large-scale sites and the operators of the grid are saying, you know, that is not a curtailable load typically the way they are run, and that's a lot of power that's being drawn off of our system. Please think about providing generation to match. And that's why you've seen things like Microsoft buying Three Mile Island or restarting Three Mile Island.

It's to have these discussions where they can match that generation with the load that they're putting on the grid. If that continues to go the way it seems to be going, it's going to be harder and harder, to get large scale interconnects and we think that puts even more value on the sites we're securing. As we look out to the 2027 sites.

Again, there's not as much market pressure from the business development folks at the hyperscalers on 2027. They're so focused on the near term, but we see that changing over time and that's why we've moved earlier and earlier.

[Operator]

Our next question comes from John Todaro with Needham & Company.

[John Todaro, analyst]

Two for you, guys. So first off in the prepared remarks, Tyler, it sounds like there's a few different avenues you guys could go down with HPC. Obviously it's a big CapEx list though. So wondering if are you going to put any CapEx dollars towards building a dedicated site, without a lease in place from a customer or a major lease at that. And then I have a follow-up question?

[Rodney Page, executive]

So I think there's some basic CapEx we would spend, you know, preparing a site, grading it, et cetera and for future sites doing things like arranging to build a substation that can be used. A lot of the sort of high voltage to mid-voltage infrastructure looks the same at both kinds of sites. And so that's the kind of things. Those are the kinds of things we would do.

We probably would not build a data center on spec without a tenant in mind, at least at this in our evolution, because based on our discussions, there are very particular design and build requirements depending on the prospective tenant. And so to sort of build something on spec and hope that, they like it is again at this point in our evolution is probably not what we're going to do.

We would look to, you know, I think what you should expect would be we would get a term sheet, letter of intent kind of thing in place with a tenant that would have the design and build requirements envisioned. We would make progress on that as we come towards an executed final lease. And then on the lease we would look to debt finance as much as possible that build costs.

So I do think we would have some spending in the months between letter of intent, and executed lease. But fair to say there are a lot of avenues to provide that. There are many folks have been calling us up offering all different ways to sort of finance that CapEx. So I think it's all easily within reach.

[John Todaro, analyst]

And then just to follow-up, there's been some talk of 2025 HPC, but unless you kind of -- it almost seems like would be cracking ground pretty quickly here. Is it more so that something would be signed in '25 and be more so generating revenue for the customer, or generating revenue in the lease in more so like a '26? Because it just seems like a very short timeframe to get a site up and running to capture that '25 demand?

[Rodney Page, executive]

Yes, no doubt that if you're starting from scratch, that's going to be challenging. But the thing about Barber Lake of course is that the substation is sitting there and humming. So it's really the downstream construction from there. And that is typically a long lead time gating item, after that to figure out a timeline. It's a little bit build and customer specific.

If it's a customer that wants 5, 9s of uptime from day 1, the long lead time item is going to be generators. And you're probably not going to be getting generators in 2025 until maybe you'd get some at the very end of 2025. But that would be the other question. If they're willing to run with say 3, 9s of uptime forever, or in advance of receiving generators, you could be up and running by late next year at least.

You could be getting started theoretically, but you would have to move quickly. I think the one other thing to mention, is there's a lot of different models of this business. There are some clients that have their own design and all their own equipment, and you're not procuring it. You are providing basically the land and constructing the shell for them, on a build to suit basis. But they may be sitting on an inventory of generators.

So in that case it's really just a construction timeline, which admittedly is many months, it's going to be very late '25, but that is doable. So it depends a little bit on the tenants.

[Operator]

Our next question comes from Reggie Smith with JPMorgan.

[Reginald Smith, analyst]

Tyler, question for you. Obviously there's a lot of interest in HPC. We've seen some longer range deals from nuclear plants announced. We've seen it's one big deal of core scientific analysis, core weave. But I guess my question for you Tyler is are you surprised? And it kind of builds on the last question. Are you surprised at the pace of development, and kind of the near term, because again, like if things aren't signed soon, it becomes very hard for something to be up by 2025? What do you think is, is causing HPC application users to hesitate? Or why hasn't a deal been signed yet? Is it just nailing down the pricing or like what are some of the things that are kind of going through their heads as you kind of see it?

[Rodney Page, executive]

Well, I mean, I'd say based on our experience, there is a fair amount of diligence. So there are site visits, there's a fair amount of study of, geo-tech studies, things like that, understanding the sites, the risks that the companies that are spending billions of dollars, on this also tend to be very large companies with hundreds of thousands of employees.

And frankly while there's a lot of enthusiasm to go quickly and again, if you listen to the earnings calls of the hyperscalers, you can hear about the revenue, the CapEx investments, et cetera, related to their cloud services. So I think that's kind of the big picture driver, of why they're in a hurry. I think between diligence and just working with large organizations these things take time, these leases are complex.

It's not, we're a very nimble company that does self-mining. So it's very easy for us to execute quickly here. Other things are a little bit more bureaucratic. Also not for nothing, there is a bit of a dating process. They've got to get to know us. I would say look, broadly speaking, the bitcoin mining industry in general, has by design built things on the cheap.

We are at the low end of the revenue spectrum for data centers, often very high margin, but that margin is driven by squeezing costs out of builds. And that is very different than the high end of the revenue spectrum where building [ 5, 9s ] of uptime requires a level of attention to detail and craftsmanship that I would say, not every company in this space has.

We happen to be blessed with a very talented and experienced construction and operations team that, have worked at places like Google and Meta on their data centers. And so, I think in the diligence process we show very well. It's part of the reason for my confidence in addition to having these great sites, is we have a team that impresses. And so, I think overall, since your question is kind of about the industry.

I think those are some of the reasons I would imagine and I would imagine, if we're not leading the charge, I know there are other folks that are involved in these types of discussions as well. And given just the setup in the marketplace, I'm confident you'll probably see multiple folks with these deals. But I'm extremely confident in our prospects.

[Operator]

Our final question comes from Greg Lewis with BTIG.

[Gregory Lewis, analyst]

I guess we're on the hour, so I'll just keep it to 1. So Tyler and, I guess, Ed, as you kind of called out the write-down in power, can you just give us a little bit of thoughts around how you're thinking about the potential power book? Just as we look at, if we were to look at like '26 electricity pricing in Texas, it looks like it's kind of like down and it's like 2-year low. If we were looking at a range, just kind of how are we thinking about that? Is this something that we're going to continue to build or just as we think about, we've been hearing from some miners that prices are low and really we can just kind of be in the spot market, and maybe not look to hedge as much as maybe, we might have thought about doing in the past?

[Rodney Page, executive]

It's a good question. I think that, listen, we're always going to look for the most favorable way to lock in margins. Sometimes that's floating in the spot market. Sometimes the market presents you with an opportunity to do a fixed price contract. Let me speak to the contract and give some context on ERCOT. So of course, I'm sure everyone knows this.

But the power contract is mark-to-market each quarter and that's going to depend on how much time is left and what the forward pricing curve is over those coming years. And so the contract has always been in the money. We have hedged at a very cheap price over the summer. I believe in the second quarter on a forward basis, those prices hit the highest they have, I think, maybe ever been, but certainly at least on a local high, on a forward basis.

And so the contract marked up quite a bit. Of course we have always stripped out that mark-to-market in our adjusted earnings, because it is a non-cash item. It moves around a lot. I think the bigger thing is nothing changes about the fact that we're paying $0.027 per kilowatt hour for power. And that's really the value of that contract is in the ongoing operations. So it came down, those prices on a forward basis came down from highs over the summer, and of course we expired another quarter of time value remaining.

And that's the mark-to-market on the contract on that specific question. More broadly speaking, look, the market continues to develop and it has implications for both the bitcoin mining industry and potentially where HPC goes over time. You know, again, if grid operators are not going to want these huge loads coming onto their grids without generation and the users of the HPC are not going to want to wait for the time it would take to build generation.

You could imagine a world where perhaps if you look at Texas dynamics, you get a solar ramp in the afternoon in the summer, and it's hot and prices spike. You could potentially manage going between generators, backup power, and not drawing from the grid for a couple of hours, and bringing kind of curtailment to an HPC load. We're very excited about things like that, because we've been doing that business for a long time in bitcoin mining. And think we can bring a lot of expertise. It depends. This summer showed a lot of growth in batteries in Texas and so Power Dynamics were really different and will continue to evolve. But I think that's always been one of our great strengths and we expect it to be going forward.

[Operator]

I'll turn it back to Tyler for any closing remarks.

[Rodney Page, executive]

Yes, thank you, everyone, again for joining. We're very excited about the future and look to providing more updates as soon as we can. Thank you. Happy Halloween.

[Operator]

Ladies and gentlemen, that concludes today's presentation. You may now disconnect and have a wonderful day.