(MARA)
Marathon Digital Holdings, Inc. misses on earnings expectations. Reported EPS is $-0.24 EPS, expectations were $-0.23.
Operator: Good day, ladies and gentlemen. Welcome to MARA’s Second Quarter 2024 Earnings Webcast and Conference Call. I would now like to turn the call over to your host, Robert Samuels, Vice President of Investor Relations. Please go ahead.
Robert Samuels: Thank you, Darryl. Good afternoon, and welcome to MARA’s second quarter 2024 earnings call. Thank you for joining us today. With me on today’s call are our Chairman and Chief Executive Officer, Fred Thiel; and our Chief Financial Officer, Salman Khan. Certain statements made during this call may be considered forward-looking statements within the meaning of the federal securities laws. In particular, any statements about our future growth plans and performance, our liquidity position, our growth opportunities and our future financial performance are forward-looking statements. These statements are often identified by the use of words such as anticipate, believe, estimate, intend, design, may, plan, project, would, and similar expressions or variations.
Investors are cautioned not to place undue reliance on these forward-looking statements. All forward-looking statements made on today’s call involve risks and uncertainties. While we may elect to update these forward-looking statements at some point in the future, we have no current intention of doing so, except to the extent required by applicable law. Our actual results and outcomes may differ materially from those included in these forward-looking statements as a result of various factors, including, but not limited to, the factors discussed under the heading Risk Factors in our most recent annual report on Form 10-K and any other periodic reports that we may file with the SEC. Finally, please note that on today’s call, we will refer to certain financial measures that were not prepared in accordance with the generally accepted accounting principles in the United States, including adjusted EBITDA and non-GAAP total margin.
MARA believes these non-GAAP financial measures are important indicators of its operating performance because they exclude certain items that we do not believe directly reflect our core operations and may not be indicative of our recurring operations. Please refer to the earnings release for a full reconciliation of these non-GAAP financial measures to the most comparable GAAP financial measures. As usual, we’ll begin today’s call with prepared remarks from Fred and Salman. After their comments, we will go through some of the more popular questions from investors before transferring to a live Q&A with our covering analysts. And with that on the way, I’m going to return the call over to Fred to kick things off. Fred?
Fred Thiel: Thank you, Rob, and thank you, everybody for joining us. Today I’m going to discuss our second quarter results and provide some insights into our newly launched business teams. In Q2, we celebrated several key achievements that highlight our innovative and strategic initiatives. We completed the successful acquisition and closing of our Garden City data center in Texas, increased our 2024 hash rate target to 50 exahash, signed a significant partnership with the Government of Kenya aimed at developing underutilized energy assets and to fund further development of the electrical grid and boost local economic growth. We began to diversify our revenues through the successful launch of MARA’s Kaspa mining operations, launched a pilot heating project in Finland, leveraging digital asset compute to offer an efficient and low cost alternative to district heating by converting clean energy into heat directly on-site, while reducing carbon emissions for the district by avoidance of wood pallet and biomass burning source of energy.
Continued to increase capacity reaching an all-time installed hash rate of 31.5 exahash through the deployment of 15,000 new miners with energy efficiency of 17 joules per terahash. Organized MARA into strategic business teams with the focus on pushing responsibility and authority as far down in the organization as possible for diversification and growth. These accomplishments demonstrate our dedication to innovation, sustainable growth, and a long term vision that leverages digital asset compute to transform the energy industry and enable a speedier energy transformation. As I discussed during our last call, we are beginning to lay the foundation for MARA to become a globally diversified company that leverages digital asset compute to build a more sustainable and inclusive future.
During the second quarter, we recognized — sorry, we reorganized the internal structure of the business to better align our growth opportunities, sharpen our strategic focus, bolster accountability, and accelerate our speed and agility as we scale. MARA is now a streamlined organization consisting of three specialized business teams, Utility Scale Mining, Energy Harvesting, and Technology, as well as support organizations. Beginning this quarter, I’m going to organize my remarks around these specialties. Let’s begin with Utility Scale Mining, the specialty that MARA is primarily known for today, large scale digital asset compute focused on providing grid stability at scale and enhancing the viability of renewable energy projects in support of the energy transition.
During the second quarter, our Bitcoin production was impacted by unexpected equipment failures and transmission line maintenance at the Ellendale site operated by Applied Digital, increased global hash rate and result in network difficulty and the April halving event. That said, transformer issues in Ellendale were mitigated and remediated post quarter end and our hash rate recovery effort is complete. Furthermore, we grew our operational fleet by over 10% to approximately 250,000 miners and reached an overall operational hash rate of 30 exahash. Our current fleet efficiency stands at 23.5 joules per terahash and we expect this will continue — this will improve through the remainder of the year as we deploy additional S21 Pros and immersion cooling.
Turning to our newly acquired sites. In Garden City, Texas, since taking over the site, we improved our hash rate by 25% and completed airflow enhancements and electrical upgrades. In Kearney, Nebraska, hash rate improved by almost 15% due to container upgrades and near completion of hosted customer exits. The site now operates almost entirely with MARA miners. Lastly, we began construction of a heat recapture agricultural site, which we will discuss further at a later date. In Granbury, when we took control from Hut8 earlier this year, we inherited a site that was poorly maintained with flawed infrastructure that led to numerous electrical outages and low employee morale. In just a few months, our team has made significant progress fixing the infrastructure, resolving electrical outages and making plans to enhance the sound wall.
We have also seen a notable uptick in employee satisfaction. Work continues to upgrade the site with the installation of single phase immersion containers, which should make up 50% of the infrastructure by year end. We have also made good progress on transitioning hosted customers with an expectation of completing all transitions in Q3. Moving to our third-party hosted sites, we deployed software upgrades resulting in 20% increased efficiency of installed miners from 29 joules per terahash to 23 joules per terahash. By deploying software upgrades and purposely underclocking machines, we have been able to mitigate issues resulting from faulty transformers, while increasing efficiency and maintaining high uptime. We currently own and operate approximately 54% of the 1.1 gigawatts of power in our diversified portfolio of digital asset compute.
We will continue making owned and operated sites a greater percentage of our fleet over time and expect to see material cost savings as this occurs. Longer term, our intention is to be amongst the lower cost operators in the industry. We remain focused on delivering on our 50 exahash goal and continue to install and energize miners according to plan. As we look to diversify our portfolio of digital asset compute, we announced that MARA is mining Kaspa, a proof of work digital asset. We currently mine roughly 9% of the global hash rate and once fully deployed, we expect that Kaspa will consume only 1% of our energy capacity, or 11 megawatts, further highlighting the benefits of diversification. Furthermore, we continue — we currently mine Kaspa at 80% plus margins, which is highly accretive as those margin dollars can be used for everything from covering G&A to acquiring Bitcoin in the open market.
This brings me to the recent change in our HODL strategy. We have long been advocating that Bitcoin should be a strategic treasury asset for corporations and sovereigns. In early 2021, MARA purchased $150 million of Bitcoin, which formed the basis of our current HODL. As market conditions changed in late ’22, MARA began selling Bitcoin from operations to cover operating expense. Market conditions have since improved and regulatory pressures have eased, as exemplified by the successful launch of Bitcoin ETFs and increased institutional and sovereign investments in Bitcoin. Our recent announcement was opportune given several statements from presidential candidates, and Senator Lummis signaling support for Bitcoin as a strategic reserve asset at the recent National Conference.
We continue to strongly advocate that corporations and sovereigns should hold Bitcoin as a strategic reserve asset. And we will make open market purchases of Bitcoin opportunistically from time-to-time as our in-house digital asset trading desk takes advantage of dips in the market, such as our recent purchase of $100 million in Bitcoin. Shifting to Energy Harvesting. Its mission is to support the energy transformation by converting underutilized sources of energy, stranded clean energy and environmentally harmful waste energy into economic value. Additionally, we can eliminate redundant expense and waste associated with generating heat, leading to energy and cost savings and a more sustainable approach to energy use. In June, we announced the launch of a 2 megawatt pilot project in Finland to warm a town of one — sorry, a town of 11,000 residents using digital asset computing.
I’m pleased to report that we’ve maintained over 98% uptime over the past 30 days and mined approximately 4.3 Bitcoin in June and July, all while generating valuable heat for the community at a total energy use of less than what each home would have individually required. We plan to increase our heating capacity before the winter season by an additional 1.5 megawatts to the current pilot and start a new 1 megawatt pilot at another location. By year end, we expect our total capacity in Finland to reach over 4 megawatts or an equivalent of approximately 9,000 more finished homes. One distinctive aspect of our Energy Harvesting strategy is that unlike Utility Scale Mining where electricity is typically purchased from utilities at market prices, our Energy Harvesting projects typically benefit from subsidized power costs by utilizing recycled energy or unlocking stranded energy resources that lack grid interconnect.
This approach aims to substantially reduce our cost to mine over time, which is part of our goal towards achieving zero cost energy in the future. As our Energy Harvesting business continues to grow, we expect to seek projects that have energy generation components with longer lives than traditional Bitcoin mining sites, resulting in higher potential returns. Energy Harvesting provides appealing opportunities for potential joint venture partners and infrastructure investors to collaborate with MARA and make a positive impact on the global energy landscape. Turning to our Technology business. The focus is on creating advanced technologies that transform digital infrastructure and enable AI and digital asset compute data center builders and operators to benefit from best in class cooling and power optimization strategies.
Since its launch earlier this year, our 2PIC technology has received lots of positive attention, and we are actively working with OEMs, reseller partners and end user customers on evaluations and potential orders for delivery later this year. We have already begun using and deploying our own technology to improve the operations of our site. The development of our immersion technology will increase our uptime and stability and allow us to operate without curtailing due to heat or other weather related issues that in the past may have impacted our air cooled sites. Over time, we expect the majority of our sites to be converted to immersion and our new build sites will be predominantly based on immersion. Together with our firmware, immersion will also help us to overclock and underclock with greater ease and effect, as well as provide the perfect complement to balance inference AI loads.
We recently introduced a version of the tank that is designed specifically to be flexible to use for either Bitcoin mining or AI. As a data center operator, this means that you can run Bitcoin mining and if or when you want to move to HPC, the infrastructure does not need to change. We are optimistic about the conversations we’re having with OEMs on the AI side as they like this concept since it allows their customers to deploy infrastructure and have it ready for AI, while utilizing it to mine Bitcoin in the near term. Providing infrastructure to data centers is a long term project, and we expect to see revenues begin to scale in the next 18 months to 24 months. In the mining space, we expect technology sales to grow incrementally as companies continue to build out and update their facilities with immersion technologies.
As previously stated, our long term objective is for 50% of our revenues to originate from non-Utility Scale Mining. Achieving this goal will take time, but these new ventures are already showing progress. I’m now going to turn the call over to Salman, who will discuss our results in more detail. Salman?
Salman Khan: Thank you, Fred. As Fred mentioned, our hash rate started to climb back up during the quarter as transformer issues at Ellendale started to improve and were subsequently resolved post quarter end. The Ellendale facility where we are hosted by Applied Digital is now fully back on track. During the quarter, despite increased deployment of miners and capacity, a higher global hash rate resulting in increased network difficulty and lower Bitcoin production due to April’s halving event impacted our results. Let’s dig into details. Due to the unfavorable fair market value of digital assets and a decrease in Bitcoin production due to the halving event in April, we reported a net loss of $200 million, or $0.72 loss per diluted share in the quarter.
This compares to a net loss of $9 million, or $0.07 loss per diluted share in the second quarter of last year. The average price of Bitcoin was higher this quarter versus the same quarter last year, but the Bitcoin price being lower than the price on the last day of the quarter of 2024 resulted in a significant loss on fair market value of our digital assets. The price of Bitcoin deteriorated on June 30, 2024 versus March 31, 2024, resulting in a loss on digital assets of $148 million during the second quarter of 2024. The price of Bitcoin after quarter end has recovered by roughly 6% as of July 30, and we expect this volatility to result in a temporary gain or — book gain or loss due to one of the largest Bitcoin holdings in corporate America on our balance sheet.
I want to point out that if we combine our almost $490 million paper gain in the first quarter with our roughly $150 million paper loss in the second quarter, we are still up almost $340 million year-to-date. As we continue to hold a larger number of Bitcoin on our balance sheet with the recently announced HODL strategy, we expect the volatility in Bitcoin price to impact our earnings to a larger extent. For example, a $10,000 change in Bitcoin price will result in over a $200 million impact in our earnings purely due to our large HODL position. Revenue increased 78% to $145 million from $82 million in the second quarter of 2023. With the average price of Bitcoin mined 136% higher this quarter than the prior year period, the increase in revenue was primarily driven by a $79 million increase in the average price of Bitcoin, partially offset by a $24 million impact due to lower Bitcoin production.
We produced an average of 22.9 Bitcoin each day during the quarter compared to 32.2 Bitcoin each day in the prior year period. The company produced 868 less Bitcoin in the second quarter as compared to the prior year period, primarily due to the halving, increased global hash rate and the continued impact of unexpected equipment failures at third-party operated sites and transmission line maintenance, partially offset by an improvement in average operational hash rate. Subsequent to June 30, 2024, these were completely resolved. We successfully integrated 590 megawatt of three operational sites in the last six months. As a result of these acquisitions, we are temporarily providing hosting services to existing hosted customers. The revenue we generated from hosting revenues was $9 million, which was not present a year ago.
We expect this revenue to taper off in the coming quarters as we work with current tenants of the sites on their transition plans and use that space for our own growth. Our hosting and energy costs were $86 million compared to $55 million last year. The $31 million or approximately 55% increase was primarily driven by the growth in the company’s hash rate from the deployment and energization of mining rigs in existing and new facilities, which increased hosting and energy costs compared to the prior year period. Partially offsetting the increase was the impact of unexpected equipment failure and transmission line maintenance, which resulted in downtime that reduced hosting and energy costs. Our non-GAAP cost of revenue without depreciation has improved 19% from $50.4 per petahash per day in the second quarter of 2023 to $41 per petahash per day in the second quarter of 2024.
Sequentially, we have improved this cost from $45.2 per petahash per day in Q1 of 2024 to $41 per petahash per day in Q2 of 2024, which reflects a 9% improvement in sequential quarters. As we realized synergies from our recently announced acquisitions, reenergized production at our third-party hosted sites and spread the fixed costs over a larger capacity as we gear towards our 50 exahash goal, we expect our cost per petahash per day to improve at times. Depreciation and amortization was $88 million, a $51 million increase from the same quarter in the prior year. The change was predominantly the result of deploying additional mining rigs since last year. Our energized hash rate grew from 17.7 exahash to 31.5 exahash from the same period last year.
Our non-GAAP total margin, excluding depreciation and amortization, was $51 million this quarter compared to $27 million in the same quarter last year. The change was predominantly related to higher average Bitcoin prices and increased operational efficiency. General and administrative expenses, excluding stock based compensation were $29 million compared with $15 million in the prior year period. This increase in expenses was primarily due to the increasing scale of the business and acquisitions, including payroll and benefits, professional fees, facility and equipment expenses, and other third-party costs associated with growth in the business. Our headcount grew from 40 employees at end of Q2 last year to 109 employees at the end of Q2 of this year.
As Fred mentioned earlier, we have structured the company in three distinct strategic business teams and we expect to continue funding diversified growth initiatives as we scale. Primarily due to an unfavorable fair value of digital assets and lower Bitcoin production from the halving, we reported an adjusted EBITDA loss of $85 million compared to a gain of $36 million in the prior year period. As a reminder, these numbers include the fair value gain or loss on digital assets. Part of our treasury management strategy has involved partially selling Bitcoin to cover operating expenses and then HODLing the remainder on our balance sheet. Last week, we announced a significant shift in our treasury policy and adopted a full HODL approach to retain all Bitcoin going forward.
In addition, we plan to periodically make strategic open market purchases of Bitcoin. The adoption of this strategy reflects our confidence in the long term value of Bitcoin and our belief that it is the world’s best treasury reserve asset. In Q2 of this year, we increased our Bitcoin holdings 7% from 17,320 Bitcoin to 18,488 Bitcoin. Subsequent to quarter end, we purchased $100 million worth of Bitcoin and currently hold over 20,000 Bitcoin on our balance sheet. Digging more into our Bitcoin holdings and cash position, unrestricted cash and cash equivalents totaled $256 million, up from $114 million a year ago. Also, at June 30, we held approximately 18,488 Bitcoin with a fair value of $1.2 billion on the balance sheet. Combined, our balance of cash and Bitcoin was approximately $1.4 billion as of June 30, 2024.
We sold 1,048 Bitcoin during Q2, realizing cash proceeds of $69 million. These proceeds were utilized to fund operating expenses, including cost of revenues for energy hosting and other cash operating expenses and G&A. During the six months ended June 30, 2024, we purchased $16 million worth of Bitcoin and subsequently announced adding $100 million worth of Bitcoin purchases to our HODL position. During the quarter, we raised $345 million from at the market equity sales, which we primarily intend to use for Bitcoin purchases, miners, operating costs, acquisition of infrastructure and other general corporate purposes. We expect our future Bitcoin holdings will generally increase due to our HODL strategy. We intend to add to our Bitcoin holdings primarily through our production activities and periodic open market purchases.
As Fred mentioned earlier, the company recently introduced Kaspa to its digital asset compute portfolio. As of June 30, 2024, we held approximately 89 million Kaspa coins on our balance sheet. The company intends to add to its Kaspa holdings primarily through its production activities. As of now, we incur significantly less cost to produce Kaspa in U.S. dollar terms, which helps pay for our expenses and allows us to hold a larger amount of Bitcoin on our balance sheet. And that completes my update. I’ll now turn it back over to Fred, who will talk more about our future plans. Fred?
Fred Thiel: Thanks, Salman. For MARA, Bitcoin mining is a vehicle rather than the end destination. Unlike our Bitcoin mining peers, MARA is very focused on executing a long term vision that leverages digital asset compute to transform the energy industry and enable a speedier energy transformation. Over the past few years, the intersection of Bitcoin mining and AI data centers has been at the forefront of our industry. While several of our competitors have announced their intentions to pivot parts of their business model to hosting and operating AI HPC, we have chosen to stay the course. However, don’t take that to mean we’re idly sitting by. We’re constantly evaluating opportunities and looking for the right approach and entry point.
While LLMs form the bulk of AI data center investments so far in the industry, AI at the edge will be the predominant application use case for AI. The majority of traffic in inference will not be human to machine, such as when you query your phone about where to find the best widget you’re going to buy or how to write the best speech. The predominant volume of inference AI will be machine to machine operated at the edge in oilfields, on jet planes, by telecoms at the base of telecom cell phone base stations, vehicle-to-vehicle communications networks, etc. The challenge with inference AI is that it’s an uneven load, not dissimilar to demand patterns on the grid, which vary throughout the day. The energy load on the compute varies as demand varies.
Typically, these large sites that are operating predominantly inference AI have power purchase agreements and they need to consume a consistent amount of energy. Bitcoin mining can act as a load balancer for inference AI, similar to how it does with the electrical grid, due to its flexible and scalable use of curtailable compute. It can be used to smooth out the energy load of the compute as demand varies throughout the day. During periods of low AI activity, excess energy can be redirected to digital asset compute, helping to maintain a more consistent energy consumption pattern, which is more cost effective and efficient and preferred by utilities. We believe that co-locating AI with Bitcoin mining is highly synergistic, especially when utilizing our technology where we can significantly overclock and underclock our miners in perfect synchronicity with the energy demand at the AI site.
This is just one example of the types of opportunities we’re currently exploring. You won’t see us purchasing GPUs and compute directly for a customer. As we develop greenfield sites, we may offer AI data center operators the opportunity to co-develop a site with us, where we do the mining and they do the AI, or we might build a data center and do the hosting, whichever path we choose to go down, we believe it is still too early in the cycle to be placing big bets. There are already concerns with overbuilding and overinvesting. M&A remains a high priority for us. We have a dedicated focus to go long capacity, which means a pipeline of greenfield power, partially developed power, acquiring additional power or even a competitor. Our corporate development team of seasoned former energy M&A professionals is building a great pipeline of opportunities for us to execute on.
We are not prepared to do a transaction for the sake of grabbing headlines or just to do a transaction. If you look at the transactions that we have done historically, we’ve set the bar rather low on our cost per megawatt basis. As these types of opportunities are presented to us, we will do our work, but we will only pull the trigger on the right one. In closing, I’m proud of the way our team performed this quarter. We are past our operational challenges, and we believe our 50 exahash goal by year end is in sight. I’m a fan of Professor Scott Galloway, and a recent podcast of his really struck me. He pointed out that in 1980, the most valuable companies in the world were energy companies, and today the most valuable firms are tech companies.
Moving forward, he argues, it’s going to be the marriage of power and compute. This is the future and we want to sit right in the middle of it. And with that, I’ll turn it back to Rob for Q&A.
Robert Samuels: Thanks, Fred. At this time, we are going to commence the Q&A section of today’s call. We’ll start by answering some of the most popular questions submitted by investors through our Q&A platform. The first question comes from John P. who asks, can you talk a little bit about your investment in Auradine, and are you currently using any of their miners? Fred, do you want to take that one?
Fred Thiel: Sure. So we’ve disclosed in the 10-Q orders and payments to Auradine for orders of machines. And we believe the Auradine 3 nanometer miners, which will begin shipments towards the back half of this year and into next year, are amongst the best in the industry. In Nashville, they were demonstrating the 3 nanometer machines and getting a lot of very positive feedback from a number of miners who are evaluating it currently. And we believe that the Auradine miner potentially is a better choice and option due to its performance than the best MicroBT machines and a very strong competitor to Bitmain going forward.
Robert Samuels: Great. Our next question comes from Ian M, who asks, with the announcement of your 100% HODL strategy of mined Bitcoin, how do you plan to fund continuing operations? Salman, you want to take that?
Salman Khan: Sure. Thank you, Ian, for asking that question. Look, this is not new to Marathon. MARA has bought $150 million worth of Bitcoin in 2021 and has had a HODL — full HODL strategy until we changed that last year. So circling back to the strategy, this is not new to MARA. Now in terms of how we plan to fund it, we have cash on the balance sheet and we have access to capital markets. On top of that, we — in our prepared remarks, we’ve talked about many other business initiatives that the MARA team has been diligently working on during this year. And some of the examples that come to mind, for example, Kaspa, that Fred mentioned earlier, is very high margin compared to our current Bitcoin operations. And that — those kind of initiatives, in addition to that, the Technology businesses that we’ve launched earlier this year, will result in additional cash flows for us as we move from here into the next few quarters.
Robert Samuels: Thanks, Salman. Our last question comes from David A. who asks, you have a number of really great initiatives ongoing. When can we expect to get some more color on when they will begin to contribute to revenue and earnings? Fred, Salman, you guys want to take that?
Fred Thiel: Sure. I can take that. So, listen, we’re very focused on developing businesses that are resilient and businesses that have long term growth potential. And if you look at the synergies of Energy Harvesting together with our Technology business, what you have is effectively a business that can begin to mine Bitcoin at very low cost, where you’re not buying electricity off the grid as the majority of utility scale miners do. The other thing we believe is we think that over time, what is going to happen is you’re going to see a movement of Bitcoin mining into a lot of technologies related to energy storage, a lot of things related to energy generation, such as combination with residential solar, commercial solar, etc., and we’re going to be playing a very active role in those businesses.
So we feel the key to success here longer term is a diversified business where you’re not just diversified geographically, which is why we’re diversifying internationally, but also from a perspective of the use case of digital asset compute. So it’s not just to mine Bitcoin for balancing a grid, but it’s also for heat offtake and other uses where Bitcoin mining is simply a means to an end. And we’re very excited about what we’re doing there. And, as we said earlier, our Technology business, we think is definitely going to give very good benefits, not just to us, but it will also be able to provide technology to other players in the mining business, but also around AI and around telecoms and other industries, where we think a lot of that technology is going to shine.
So we’re very excited about that.
Robert Samuels: Thanks, Fred. In the interest of time, I think we’re going to wrap up this section of the Q&A. Again, we really appreciate the questions and the interest. I’m now going to turn the call back to our operator to open the line to questions from our covering analysts. Darryl, back to you.
Operator: Thank you. We’re now going to open the call to questions from our covering analysts. [Operator Instructions] Our first questions come from the line of Reggie Smith with J.P. Morgan. Please proceed with your questions.
Reggie Smith: Hey, thanks for taking the question. I guess you’ve talked about, operational issues the last two quarters, and I was hoping — I don’t want you to get too technical, but just can you explain, maybe in plain English, what exactly went wrong, how it was repaired? And I guess what I’m really trying to get at is, I guess the likelihood that, that could happen again either at that site or another site. Just trying to understand a little bit better about what drove those issues? Thank you.
Fred Thiel: Sure. So if you look at the Ellendale site operated by Applied Digital, what you predominantly had were transformer issues where the transformers that had been supplied by the vendor did not operate to spec when they were operated at — over the long term. And so you started to see failures. And so the folks at Applied Digital did a heroic job of not just sourcing replacements, but also modifying the transformers such that they could be put back into service so that they wouldn’t have this issue again. There were also some transmission line maintenance issues, which again is outside of their ability to control as — or hosting partners’ ability to control as that’s done by the electrical utility. So, could it happen again?
Once we became aware of these issues, we started looking at all our other sites to see what we could do to make sure that these types of issues wouldn’t happen. That’s one of the reasons why in our prepared remarks, we talked about the fact that we were purposely underclocking miners at some of the third-party hosted sites so that we wouldn’t strain the transformers and run a risk of this. So again, these are sites which we don’t own and operate, and so we have little control over the infrastructure that’s there. But we’ve been working very closely with the hosting partners to make sure we put in place strategies to mitigate the potential for this to happen again.
Reggie Smith: Understood. And I appreciate the color on your HPC kind of strategy and thinking. I was curious, and I know it’s early days, you talked about some, I guess, different use cases at the edge. I guess, how do you evaluate away the economics of mining versus maybe what you can earn doing some of that edge type work? Do you have a sense of the value of that work today maybe on a megawatt basis or how — you follow my question?
Fred Thiel: Right. Yeah. So again, as we’ve said, we’re not going to be the operator of the AI inference operating at the edge. If you just look at NVIDIA’s roadmap for equipment, you’ll see that they’re quadrupling the power and throughput, if you would, of their technology over the next three generations in 24 months, which makes the technology transitions in the Bitcoin industry look fairly benign. So we definitely don’t want to be in what is a hugely more CapEx intensive business. That being said, we do want to be a provider of infrastructure for people in this business. And if you look at our immersion technology and you look at our cooling technology, it is specifically designed for use in these areas. And if you look at the sheer amount of infrastructure spending that is going to go into all of these inference data centers, especially inference at the edge, it is an ideal market for us to sell immersion and cooling technology to.
And there are very interesting opportunities to partner with people who are building bundled solutions that include the GPU, etc., and we’re just going to be a vendor to them. We think that is the best way for us to play in that marketplace. I don’t believe that Bitcoin miners should really be in the business of doing — operating AI. They can be in the business of hosting AI if they feel they want to build that infrastructure. But the real future is around containerized immersion based infrastructure at the edge, and that’s really where we’re going to play in that area. Now when you look at running inference AI data centers of larger scale, we believe there’s a great synergy between Bitcoin mining and the AI data center because of the electrical load variances in inference and the perfect synergy with Bitcoin mining there.
And so we will partner with data center operators on that. But again, we don’t want to be the operator of the AI. We don’t want to own the compute load because we are so early in the technology cycles here that the CapEx spend in that industry is going to make the CapEx spend in the Bitcoin mining industry look fairly small, a data center for AI costs over $12 million a megawatt today. And in the Bitcoin mining space, you’re looking at $1 million a megawatt. So we believe it’s an industry where it’s better to be a fast follower and to be an infrastructure provider. And by being an infrastructure provider, you could partner with the people who are the leaders in the industry and you can learn and you can see exactly where you can benefit from it.
Reggie Smith: Understood. That makes a lot of sense. Thank you.
Operator: Thank you. Our next questions come from the line of Tyler DiMatteo with BTIG. Please proceed with your questions.
Tyler DiMatteo: Yeah. Hi, everyone. Good afternoon, and thanks for taking the question. Really appreciate it. Fred, I’m curious. I’m following the rationale for the three business units and streamlining and breaking it out. I guess what I’m wondering, though, is how do you think about deploying capital amongst the various units and I guess what goes into that decision in your mind? And then my follow-up question here is that you put some notes in the prepared remarks on the speed and the agility. I’m also curious, beyond the capital part of the equation for the three units, how do you think about being flexible among deploying other general resources, whether that be personnel, what have you, among the various units? Any color there?
Fred Thiel: Sure. So you’re familiar with the matrix organization, because that’s the way most investment banks operate today. You have teams that do different things. And you’ll notice in our remarks, we’re talking about business teams. So the groups that are actually effectuating the business, that are out there winning the business, that are out there delivering on the business, are divided up into teams of specialists who are able to go in and execute these transactions, execute the projects. The support organizations, finance, communications, HR, IT, etc., are centralized functions that support the various business teams and their purposes. So it’s a very traditional matrix organization. The goal is to not have redundant infrastructure or redundant support people, but rather be able to benefit from being able to have specialized teams.
It’s like a football team. You have special units, special teams that do special things. That’s how we look at the business. Utility Scale Mining is a very well understood, very easy business to operate. You have swap teams doing O&M, you have business development doing acquisition, you have operations folks running NOCs and running the actual sites. In the case of Energy Harvesting, a lot of them are one-off projects and they start as pilots and then they grow. And so you have to have these specialized project teams that go in and do those. If you tried to do that in a traditional siloed organization, it would be very difficult, you wouldn’t be able to do it quickly. And more importantly, you would lose the benefit of the specialized knowledge.
And so we’re able to, for example, move very quickly in deploying these sites. We do a test, we do a pilot, we figure out how we’re going to do it and then we can scale it. From a CapEx perspective, where it becomes interesting is on the Energy Harvesting side, you have an opportunity to bring in co-investors on the energy generation asset. And that lowers your CapEx need for the overall project from a specific Marathon perspective. The other benefit you have is that a lot of the Energy Harvesting projects, the actual cost of electricity is subsidized. So we are constantly evaluating projects based on an IRR model or a ROIC model, return on investment capital — invested capital. And so Utility Scale Mining is a business which is all about operational efficiency, right?
How do you get the lowest cost power? How do you operate the miners with the maximum amount of uptime with the lowest energy consumption per miner, so you can mine the maximum amount of Bitcoin? In the Energy Harvesting world, it’s all about how do you get zero cost energy, which is very different. And when you look at the ability to finance the CapEx on the energy side using traditional means and bringing in third-party partners, and you look at essentially operating a site where the cost of electricity starts becoming insignificant because it’s being paid for by the value you’re creating for the recipient of either the heat reuse or how you’re dealing with the waste energy, those projects tend to provide a more attractive ROIC over time than some of the traditional utility scale mining sites.
And that’s why as we look at the two businesses sort of side by side, we say we’re going to be diversifying and investing more in these non-traditional utility scale mining businesses. On the Technology side, you’re looking at a business where you’re essentially working with OEMs, channel partners and large end users for large projects where you’re providing technology, which will then eventually also have ongoing O&M fees and recurring revenues tied to maintenance and managed services projects. So we view that as a very separate type of business. But again, we invest in it just like one would a startup company. In the same way that we invested in Auradine, for example, where we leveraged our intellectual property together with great talented teams and venture capital from Silicon Valley, you’ll see us continue to kind of build and invest in things like that as we go, as we continue to diversify the business.
Tyler DiMatteo: Okay. Great. Thank you, Fred. Really helpful overview there. Appreciate it. I’ll turn it back to the queue.
Operator: Thank you. Our next questions come from the line of Lucas Pipes with B. Riley Securities. Please proceed with your questions.
Lucas Pipes: Thank you so much, operator. Good afternoon, everyone. My first question is on the cost side. When I back into kind of cash cost, direct cost, whatever you may want to call it, it — I shake out around $59,000 per coin. So first, I wondered if you could comment on that. I could see how there might be some pluses and minuses around that number. Obviously, this excludes non-cash items such as D&A and some other things. And then more broadly, in an environment macro wise where there’s maybe a bit more volatility, do you have specific cost targets that you think about where could you squeeze some additional unit cost reductions out of the business? Thank you very much.
Salman Khan: Yeah. Let me try to answer that question. So first of all, Lucas, that’s right. The cost per coin, the way you’re viewing it, it includes a certain cash and non-cash component to it. So G&A is included in that cost and so is the mining cost of revenue. The way we view this, we — a better way of viewing the way we run our business, and we believe this is a better matrix from a KPI perspective, is the hash cost per petahash per day. And that is a more better representation of our business structure and also the hash that we’re deploying which is effective and the costs associated with mining. So from that standpoint, if you view it from a going-forward perspective as we deploy more hash and we energize as we head towards the 50 exahash target that we’ve set for ourselves, we expect the fixed cost to be spread across larger exahash capacity available.
And we expect the cost to decline over a period of time systematically as a result of that. Secondly, what we have to understand a little bit is that we’ve got some one-off costs in our cost structure in the Q1 and Q2 both. We closed two major acquisitions, and there are some legal costs and repairs and maintenance costs. And certain things that Fred alluded to in the prepared remarks about the quality of assets that the way they were managed previously versus how we are managing the stand — bringing them back to our standard, it requires some additional money and that is reflected in our operating costs as of now. The other thing is we also have some one-off legal costs in there that we expect over a period of time to disappear. So there — in the last two quarters, we’ve experienced some additional one-off cost that is driving our G&A at least a little bit on the higher side.
Now when you think about our business, the current business is Bitcoin mining, a majority of the business, from materiality standpoint. But then we also have introduced Kaspa, which is more than 80% margin to produce Kaspa with less than 1% of our energy capacity. And those kinds of businesses, as we bring them mainstream from pilot phase, as Fred alluded to previously, in our energy harvest business, we expect to see our — not only increasing our revenues, but also driving our costs down and squeezing that margin with the passage of time. So hopefully that answers the question.
Lucas Pipes: I appreciate that very much. Thank you. My second question is a bit higher level for both of you. With the HODL strategy, obviously, you have some U.S. dollar expenses, should we think about you liquidating some Bitcoin to pay those U.S. dollar bills, electricity, salaries, what have you? And then more broadly, is the goal here really to just maximize the absolute number of Bitcoin in treasury? Is it maximizing Bitcoin produced per share, held per share, or is it ultimately an ROIC measure that’s still the most important? Thank you so much for your perspective.
Salman Khan: Yeah. So Bitcoin purchases, as we announced fairly recently, which is again not a new strategy for us, we used to have the strategy two years ago, we intend to HODL — exclusively HODL. So we don’t intend to sell Bitcoin at this stage for the foreseeable future. The macroeconomic conditions and the geopolitical situation and what’s with the presidential election this year, we expect lots of tailwinds in this space and the Bitcoin price, which was already projected to be going up based on the demand and supply and based on the projections that we see publicly discussed. So from a HODL strategy standpoint, yes, we are a Bitcoin miner and we produce Bitcoin, and we were HODLing a big portion of our Bitcoin historically and we were selling some portion of that Bitcoin to pay for our operating expenses.
With this change in strategy or going back to our original strategy, we will stop selling Bitcoin for paying for our operating expenses. So what does that mean? That means that we will continue to HODL more because we see, from an ROIC perspective, from a shareholder value perspective, any capital that we deploy in building our Bitcoin reserve asset, that is highly accretive to our stockholders.
Lucas Pipes: All right. Well…
Fred Thiel: I think also…
Lucas Pipes: Yeah. Sorry. Please, Fred.
Fred Thiel: Sorry. I was just going to say that I think the — we are just at the beginning of the institutional wave of investing and holding Bitcoin. ETFs have now accumulated a substantial position in Bitcoin, and we’re starting to see businesses like Cantor Fitzgerald start offering financing businesses — financing offerings around people who have Bitcoin holdings. We’re starting to see pension funds invest. We’re starting to see many long term holders invest. And at the end of the day, it is a finite resource. And the only place to get it is what’s available in the market or what people are willing to sell. And when we’re able to go into the marketplace, as we did, in the past 30 days, and buy at very attractive prices, it just makes sense for us to do that.
And then it makes sense for us to hold it. The $150 million of Bitcoin we bought in January of 2021, we never sold. And so we think right now, we are in the beginning of a very attractive time for Bitcoin. We believe it’s an asset that’s going to continue to perform better than other areas. And if we look at kind of the historical returns that Bitcoin has offered, you’re looking at something north of 30%, 35%. And as Salman said, holding Bitcoin can be very accretive for our shareholders. So that’s one of the reasons that drove this.
Lucas Pipes: Fred, Salman, thank you, both very much for your perspective and best of luck.
Salman Khan: Thank you, Lucas.
Operator: Thank you. Our next questions come from the line of Kevin Dede with H.C. Wainwright. Please proceed with your questions.
Kevin Dede: Hi, Fred, Salman. Thanks for taking my question. I was wondering if you might add a little more detail to your 50 exahash target, at 31 now — 31.5 now, congrats.
Fred Thiel: Thank you.
Kevin Dede: Just kind of curious about power source, where — is that increase going to come on the change? I know you talked about the S21 Pro introduction. Is it efficiency or how much more power do you think you’re going to need and will you be able to address to hit that target?
Fred Thiel: So Kevin, we talked about, when we did the acquisitions and we announced the target, that essentially the vast majority of this is all going to come from the capacity that will be made available to us by owning these assets. And that remains the case today. So we’ll continue to add other sources of power along the way. But at this point, we’ve essentially — we have the machines ready and we’re plugging them in as we vacate customers from the sites and as we continue to do the refits of the sites that help us expand the capacity at the sites. So it’s not like we have to go out and buy 500 megawatts of power to do this. We’re well along that way with what we already have.
Kevin Dede: Fair enough. How would you recommend we judge your perspective on moving the company closer to power generation versus like the Generate deal and the Applied Deal in Garden City, right, where you own the facility, but it doesn’t really get you closer to that zero power cost target.
Fred Thiel: Right. So Energy Harvesting — sorry, go ahead. Sorry, I didn’t mean to…
Kevin Dede: Yeah. No, no, you go ahead, Fred. Yeah.
Fred Thiel: Yeah. I was just going to say that the — owning our own infrastructure removes the middleman in the infrastructure, allows us to own control of the power, generate the revenues from trading of the power and control our destiny. The next step is going all the way back to looking at the actual power source. Now that being said, I’m not going to go out and ask our team to go buy a bunch of power generating – large scale power generating assets. But there are opportunities out there for us to partner with people who have power generating assets, which to date aren’t able to monetize those assets effectively, where we can share in the benefits of very, very low cost power. And then on the Energy Harvesting side is where you really see this benefit of zero cost energy.
And there are locations in the world where today they are shutting down power generation because they can’t generate revenues from it. And those are locations where the marginal cost to operate those sites is de minimis. And so partnering with those types of partners allow us to get very low cost energy as well. So you’ll start seeing the Energy Harvesting business over time become involved in projects where the actual energy cost is low and then the additional service the Energy Harvesting business provides, whether that’s heat recapture, whether that’s waste mitigation, whether it’s consuming methane flare gas from oil fields or landfills, essentially provides a subsidized cost, which will help lower our overall mining cost per Bitcoin across the board.
Kevin Dede: When you look at your pipeline, though, on the M&A side, Fred, is it fair to assume that more of your emphasis will be on that Energy Harvesting initiative versus then maybe getting involved in a hostile takeover that seems to be going back and forth in the business?
Fred Thiel: We don’t comment on M&A activities, though sometimes it can be fun to watch. But as I said in our prepared remarks, we have kind of multiple buckets we look at. We look at full true greenfields where, for example, somebody has — owns an interconnect, wants to monetize that, or they have potentially a renewable energy project that they can’t monetize effectively, or somebody has a substation and at really attractive energy prices. Then you have kind of the next step down, which is somebody who has started and built a Bitcoin mining site with maybe, let’s say, 100 megawatts of power, but the substation and the interconnect they have can provide power up to 300 or 400 megawatts, which is kind of what you see a lot of people doing in the M&A world around today.
And then you have, buying an operational miner who maybe has a pipeline of stuff. We’re looking at all these things. The market pricing is kind of where it is today, which is very different than where we were buying sites at the beginning of this year where we were paying under $500,000 a megawatt. So we’ll continue to be opportunistic, but we’ll look at where we think we can generate the most long term value and as opposed to just trying to do nameplate deals.
Kevin Dede: Perfect. Thanks for the color, Fred. I’ll turn the floor over. Appreciate it.
Fred Thiel: Yeah. Thanks.
Operator: Thank you. At this point, there are no further questions. I’m going to turn the call back to Rob Samuels for closing remarks.
Robert Samuels: Thanks, operator. And thank you all for your time today. If you have questions that we were — that were not answered during today’s call, please feel free to contact our Investor Relations team at ir@mara.com. Thanks, and enjoy the rest of the day.
Operator: Thank you. This does conclude today’s teleconference. We appreciate your participation. You may disconnect your lines at this time. Enjoy the rest of your day.