Good day, and welcome to the Bitfarms third quarter 2024 conference call. At this time, all participants are in listen-only mode. After the speaker's presentation, there will be a question-and-answer session. Instructions will be given at that time. As a reminder, this call may be recorded. I would now like to turn the call over to Tracy Krumme, Senior Vice President of Investor Relations and Corporate Communications. Please go ahead.
Thank you. Good morning, everyone, and welcome to Bitfarms' third quarter 2024 conference call. With me on the call today is Ben Gagnon, Chief Executive Officer and Director, and Jeff Lucas, Chief Financial Officer. Before we begin, please note this call is being webcast with an accompanying presentation. Today's press release and our presentation can be accessed at our website, bitfarms.com, under the Investor section. Turning to slide two, I'd like to remind everyone that certain forward-looking statements will be made during the call and that future results could differ from those implied in this statement. The forward-looking information is based on certain assumptions and is subject to risks and uncertainties, and I invite you to consult Bitfarms' MD&A for a complete list. Please note that references will be made to certain measures not recognized under IFRS and therefore may not be comparable to similar measures presented by other companies.
We invite listeners to refer to today's press release and our MD&A for definitions of the aforementioned non-IFRS measures and their reconciliations to IFRS measures. Please note that all financial references are denominated in U.S. dollars unless otherwise noted. I would also like to add that we will be attending the following upcoming equity conferences: Cantor Fitzgerald's Crypto Digital Assets and AI Infrastructure Conference in Miami tomorrow, Roth MKM's Technology Conference in New York City on November 19th and 20th, B. Riley Securities' Crypto and Energy Infrastructure Conference in New York City on December 4th, Northland Capital Markets' Virtual Growth Conference on December 12th, and lastly, Needham & Company's Growth Conference in New York City on January 14th and 15th. If anyone would like to meet with us on those dates, please contact me or a sales representative from the firm.
Now, turning to slide three, it is my pleasure to turn the call over to Ben Gagnon, Chief Executive Officer and Director. Ben, please go ahead.
Thanks, Tracy, and thank you, everyone, for joining today. On today's call, I will be discussing an overview of our year-to-date accomplishments, our Q3 operational update, and an overview of the strategic moves we've made in the quarter, and our 2024 and 2025 goals and guidance. I will then hand the call over to Jeff to discuss the Q3 financials. Turning to slide four. 12 months ago, we outlined our transformational fleet upgrade. This involved upgrading nearly our entire fleet of miners across, at the time, our 11 data centers. We had always planned for 2024 to be a transformational year, but we underestimated just how transformational it would be.
Year to date, so far in 2024, we've upgraded 10 data centers and added three new data centers to our portfolio, unracked over 50,000 older generation miners, and racked over 46,000 new miners, increased our hash rate by 83% to 11.9 exahash today, improved our efficiency 40% from 35 watts per terahash to 21 watts per terahash, reaching our year-end efficiency goal three months ahead of schedule, and most importantly, significantly grew our megawatt pipeline from 324 MW today to 950 by year-end 2025, with a multi-year growth potential up to 1.6 GW. These numbers speak volumes to our accomplishments and the commitment of our global team to grow the company and drive improved performance.
In addition to these tremendous quantitative improvements, we announced our acquisition of Stronghold Digital Mining, which is the largest acquisition between two public companies in the Bitcoin mining sector yet, and we significantly enhanced our board of directors and leadership team. Turning to slide five. On our Q2 earnings call in August, I outlined several goals and initiatives as Bitfarms' new CEO. These included, first, the continued diversification of our portfolio beyond Bitcoin mining. The aim here is to maximize utility of our strongest asset, our megawatts, in the most efficient and cost-effective way possible by expanding into high-growth and high-value geographies and into synergistic business lines such as hosting, HPC and AI, heat recapture and recycling, and energy generation. Second, a strong focus on U.S. expansion, particularly in the PJM region, which we identified as the most attractive energy market in the U.S.
And third, building out an even stronger bench of operators and introducing a new organizational structure in order to evaluate, plan, and execute on the significant opportunity with HPC AI and the upcoming anticipated Bitcoin bull run. I'm pleased to say that over the past three months, we've made significant progress on each of our goals and laid a strong foundation to execute our growth strategy. I'll now dive into the Q3 operational update and talk through several strategic actions we've taken this quarter with these goals in mind. Turning to slide six. In line with the first and second initiatives regarding continued diversification of our portfolio and expansion in the U.S., in Q3, we entered into an agreement to acquire Stronghold Digital Mining. This transformative combination ensures a multi-year strategic growth path that we are incredibly excited about.
With this acquisition, we are integrating vertically by acquiring two strategically located power facilities in Pennsylvania, expanding and rebalancing our portfolio with over 300 MW of U.S. power capacity, with the potential to increase our energy portfolio to over 950 MW by year-end 2025 and 1.6 GW in the next few years, creating energy trading and demand response opportunities to minimize energy prices, and lastly, creating an opportunity to cost-effectively integrate HPC and AI into our portfolio with a merged HPC AI and Bitcoin mining data center. Turning to slide seven. Upon closing, the expected 307 MW available at these two locations would drive a 47% increase in our 2025 year-end megawatts from 648 MW to over 950 MW. Including our Sharon, Pennsylvania site, the combination of these three sites creates a PJM portfolio of approximately 427 MW in 2025, which is larger than our entire operational energy portfolio today.
Importantly, these sites rebalance our portfolio back to the U.S. and North America, representing 47% and 65% of our year-end 2025 portfolio, respectively, while LATAM exposure will scale down to 35%, significantly improving the size, quality, and distribution of our energy portfolio. We have already begun maximizing the utility of the Stronghold sites with two hosting agreements, each for 10,000 miners. These 20,000 miners, initially scheduled for deployment in our Yguazu facility, are being deployed in Pennsylvania at Stronghold's Panther Creek and Scrubgrass facilities, supporting approximately 4 exa hash. Redirecting these miners reduces our logistics, duties, setup, and energy costs, driving a more optimal deployment and providing a pathway for continued profitable growth. With these two hosting deals, we anticipate that nearly 75% of our hash rate will come from our North American data centers in the first half of next year, reaffirming our commitment to grow in the U.S.
In addition, we continue to actively pursue other site acquisitions and growth opportunities in the U.S. Turning to slide eight. In line with our third initiative to strengthen our leadership team, we made significant progress by expanding our bench and reorganizing our operational structure in order to position ourselves for continued growth in 2025. Reflecting our commitment to operational excellence, we've recruited numerous industry experts with the following leadership appointments: Liam Wilson as our Chief Operating Officer, a newly created role; Alex Brammer, who replaces me in mining operations as Senior Vice President; Benoit Gobeil, promoted from EVP of Operations and Infrastructure to Chief Infrastructure Officer; and Rachel Silverstein as General Counsel, a newly created role as part of our emphasis on growing in the U.S. In addition, we reorganized our operations into two divisions: mining operations and infrastructure, with Alex and Benoit heading up the divisions respectively.
These new divisions are already delivering greater scalability and accountability in data center construction, maintenance, and operations, and are laying the foundation for the establishment of HPC and AI operations. With these hires, we are laying an even stronger operational foundation, enabling us to execute our growth plan. Turning to slide nine. In addition, we enhanced our corporate governance and strengthened our board of directors with the appointment of two new directors, including myself, and with former lead director Brian Howlett now serving as our board chair. Our board of directors now consists of five members, four of whom are independent. With the upcoming special shareholder meeting scheduled to take place next week on the 20th, the board is expected to increase to six board members with five independents subject to shareholder approval.
On a related note, in Q3, we were able to reach a settlement agreement with Riot Platforms, which we believe is in the best interest of the company and our shareholders. Importantly, this mutually beneficial agreement allows us to move forward as an independent company and to execute our growth strategy, avoiding the costly and distracting proxy contest. Riot remains a shareholder of Bitfarms with current ownership of just under 20%. Analysts have noted that Bitfarms shares provide good value and significant upside to Bitcoin price relative to peers, and Riot remaining a top shareholder is a positive market signal underscoring the value and potential of our growth. Turning to slide 10, I will now take a few minutes to discuss our guidance. While we are now at 11.9 exa hash, we are behind schedule on delivering our mid-year 12 exa hash target for the following reasons.
First, we have some minor construction delays at several sites that have been overcome but delayed energization by weeks or a few months. Second, over the past few months, we've experienced shipping delays and extended minor warranty servicing that goes beyond the 2,700 miners replaced a few months ago. Despite improvements in recent miner shipments, continued warranty servicing has impeded the achievement of our hash rate target. We've been working closely with Bitmain to address these issues and are implementing an efficient solution. We will be upgrading the remaining 18,853 Bitmain T21 miners to be delivered by Bitmain as part of our fleet refresh announced last year. Bitmain will now be sending 18,853 more powerful and efficient S21 Pro miners operating at 234 terahash and 15 watts per terahash, representing more than a 20% improvement from the T21 miners in both energy efficiency and hash rate.
This capital-efficient upgrade allows us to expeditiously upgrade the remaining miners to be delivered for better performance with minimal impact on our cash while preserving the full upside of the Bitcoin. While all of the miners are scheduled to ship before the end of the year, this significant upgrade and continued RMA servicing will push our year-end target of 21 exahash into the first half of 2025. It will also improve our energy efficiency by an additional 10% from 21 watts per terahash to 19 watts per terahash. The majority of the upgraded miners will be deployed in Pennsylvania and Quebec. Amidst the backdrop of rapidly improving macro conditions, we believe this is a rare opportunity to upgrade miners for better performance before miner prices increase. Turning to slide 11.
Looking ahead to 2025, demand for immediate capacity for both HPC AI and Bitcoin data centers is surging, and turnkey assets are highly sought after. Data center development continues to accelerate, driving more scarcity and making access to power very valuable. KKR estimates that $250 billion will be spent annually on the data center boom. The U.S. is the biggest developer of data centers, consuming about 16-18 GW of power compared to about 6 GW each in Europe and Asia. KKR's global head of digital infrastructure believes over the next three to four years that the 18 GW could double, if not triple. We have just over 500 MW of secured power and infrastructure for 2025, which represents one of the largest portfolios of flexible megawatts among any publicly traded miners. This gives us unparalleled flexibility to approach this bull market and take advantage of strategic opportunities.
With this in mind, we've taken important steps to ensure we're getting the best value from our assets. First, we've changed our deployment plans by reallocating the miners from Yguazu to the U.S. with the Stronghold hosting agreements. In doing so, we have freed up the 200 MW at the Yguazu site entirely and have applied for a delay in energization of 60 days to March 1st, 2025, in line with the contract. With 200 MW of cost-effective, high-uptime, renewable energy, and new infrastructure, Yguazu is a marquee asset going into a Bitcoin bull market that is capable of producing approximately 15 exahash if filled with the latest S21 XP miners. Management is currently evaluating all strategic options to best monetize the site. Second, in Q3, we assumed control of our site in Sharon, Pennsylvania.
This is a 120 MW site strategically located in the PJM Interconnection market, providing access to low-cost, flexible power. The site is very well suited for both Bitcoin mining and HPC and AI, and we are currently evaluating the best way to monetize these megawatts. Third, and as discussed a few minutes ago, we acquired high-value megawatts in Pennsylvania through Stronghold. Of the 307 MW available in 2025, these sites are currently contracted to utilize approximately 100 MW for hosting, and we are evaluating the best way to monetize the remaining 200 MW. We now have sufficient megawatt capacity to go well beyond the previously guided 35 exa hash in year-end 2025. However, as we continue to refine our 2025 plans and actively rework around Bitcoin mining and HPC and AI, we believe it would be premature to maintain the 35 exa hash guidance.
We will issue updated guidance once these strategies are finalized. Turning to slide 12, we own and operate a portfolio of high-quality energy assets that are monetized through Bitcoin mining. At Bitfarms, we are not growing hash rate at any cost. We are focused on getting the best utilization and driving the maximum value of our energy assets for our shareholders. When we take a step back and look at how we do this, we believe that a combination of self-mining, hosting, energy trading, and HPC and AI creates a more powerful and resilient portfolio. We're conducting a thorough evaluation of all of our energy assets, looking at a host of different metrics, both qualitative and quantitative, to determine which sites would be best suited for which activities. That said, we're pleased to announce that we have two different U.S. sites for a 1-2 MW HPC and AI pilot.
While still in early stages, we have the land and power secured at both sites and are in active conversations with potential partners and suppliers discussing potential accelerated deployment in 2025 and are currently in the process of finalizing terms, equipment, plans, and budgets. We believe both of these sites are well suited for development for a number of reasons. First, they are well located within a few hours from major metropolitan centers on major fiber trunk lines and are in or near data center clusters. Second, the power is secured and cost-effective with high uptime and no curtailment requirements. Third, the land is cleared and ready for accelerated construction.
We believe that initiating work at a pilot site will allow us to prudently and cost-effectively evaluate the significant opportunity and to get a firm handle on the space, technology, and infrastructure before committing a significant amount of capital on a larger HPC and AI site. This pilot reinforces our commitment to diversifying beyond Bitcoin mining and is an important first step into an exciting new business line. We expect to provide an update on our progress in 2025. Turning to slide 13. Before I turn the call over to Jeff, I want to reiterate the significance of what we've accomplished over the past 10 months.
We've refreshed nearly our entire fleet of miners, significantly improved our mining economics, built from the ground up a 70 MW state-of-the-art data center, acquired three new sites in the U.S., embarked on an HPC and AI strategy, and completely revamped our operational structure and strengthened our leadership team. We are now uniquely positioned with a strategic pipeline of over 950 MW in 2025, with nearly 0.5 GW of power infrastructure that is highly flexible, representing a massive, secured, and cost-effective growth opportunity as we enter into the anticipated 2025 Bitcoin bull cycle. We are now a stronger company with a significantly expanded energy portfolio with up to 1.6 GW available over the coming years, and we are well equipped to take advantage of the significant opportunities in front of us.
It's an exciting time in the Bitcoin industry on the heels of the election and with Bitcoin hitting new record highs, and I've never been more excited or confident in our growth prospects and our ability to execute. Turning to slide 14, I'll now turn the call over to Jeff.
Thank you, Ben, and thanks, everybody, for joining us today. I will provide an overview of our third quarter financial performance and position. Turning now to slide 15, let's start with Q3 results. During the quarter, we earned 703 Bitcoin, 14% more than the 614 Bitcoin mined during the second quarter.
The increase is driven by the 54% growth in our average hash rate from 6.7 exahash in the second quarter to 10.3 exahash in the third quarter, which more than offsets the impact of the April 2024 halving and a 3% increase in the average network hash rate quarter- over- quarter. Total revenue was $45 million in the third quarter, a $3 million, or 8% increase from the second quarter, and a 30% increase year over year. Revenue from mining activities was $43 million, compared to $40 million in the prior quarter, an increase of over 7%. The balance of the revenue was earned by our electrical subsidiary, Volta, which provides commercial services in addition to cost-effectively supporting our data center operating activities.
General and administrative expense, or G&A, excluding non-cash stock-based compensation and the second quarter Canadian sales tax refund, was $22 million, up from $11 million in the prior quarter. The $11 million increase is due largely to one-time non-recurring costs, which totaled $12 million in the quarter and accounted for $9 million of the quarter-over-quarter increase. These one-time non-recurring charges include the Riot requisition, expenses for the settlement of the employment dispute with the former CEO, and expenses associated with the acquisition of Stronghold and the Sharon, Pennsylvania site. The balance of the increase was due to higher compensation expense as the company made several key management hires to build out our leadership team, as well as higher insurance, duties, and other expenses associated with the company's growth.
For Q3, our operating loss was $44 million, up from $24 million in the second quarter, driven by the factors I just referenced. This operating loss reflects non-cash depreciation expense of $29 million, including $10 million for accelerated depreciation on older miners, compared to the prior quarter depreciation of $57 million, of which $46 million was accelerated depreciation. In Q3, reported financial income and expense include the $6 million non-cash gain and the revaluation of the financial liability for warrants issued in earlier financings, compared to a $1 million non-cash loss on the revaluation of this financial liability in the second quarter. Under IFRS, we are required to recognize the liability for these warrants, even though they cannot be settled for cash.
Net loss for Q3 was $37 million, or a loss of $0.08 per share, and we had a net loss of $27 million, or a loss of $0.07 per share in the second quarter. Now let's turn our attention to operating performance and per Bitcoin metrics. Gross mining profit was $17 million, or 38% of mining revenue, down from $21 million, or 51% in the prior quarter. Our overall corporate electricity rate for the quarter was $0.047 per kWh an increase from $0.043 per kWh in the second quarter. The increase was driven largely by higher winter electricity rates in Argentina and a rate increase in Paraguay instituted at the beginning of the third quarter.
Importantly, with our improvement in electrical efficiency from an average of 28 watts per terahash in the second quarter to 23 watts per terahash in the third quarter, our electricity cost per terahash decreased over the quarter by 10% to $0.027 from $0.03 per terahash per day. I'll add that in September, we achieved our year-end target of 21 watts per terahash, three months ahead of schedule, which equates to about $0.024 per terahash, an additional 11% reduction in electricity cost per terahash. Our direct mining cost for Bitcoin in the second quarter was $36,000. Our total cash cost to mine Bitcoin was $52,400, and our revenue per Bitcoin was $61,000, resulting in a cash profit per Bitcoin of about $8,500. Turning now to slide 16.
For the third quarter, our adjusted EBITDA was $6 million, or 14% of revenue, compared to $12 million at 28% of revenue in the second quarter. The lower adjusted EBITDA largely reflected the impact of the halving, along with higher G&A expense associated with the expansion of our operating activities. As we've noted in previous quarterly earnings calls, our adjusted EBITDA is very straightforward, being purely a measure of the cash profitability of our mining operations and the profit contribution of our Volta subsidiary. We do not reflect mark to market adjustments of our Bitcoin holdings or any other balance sheet valuation adjustments in our adjusted EBITDA. Stated simply, our adjusted EBITDA of $6 million in the third quarter equates to cash profit per Bitcoin of $8,500 multiplied by 703 Bitcoin mined during the quarter, plus approximately $4,000 of profit for a Volta electrical subsidiary. Turning now to slide 17.
At September 30, we had total liquidity of $146 million, consisting of cash of $73 million in Bitcoin also valued at $73 million. At October 31, we held 1,188 Bitcoin, up from 1,147 Bitcoin at the end of September and 905 Bitcoin at the end of June. Our higher BTC treasury balance reflects a solid cash position and strong cash flow from operations. Further, our synthetic HODL continues to grow, increasing from 208 equivalent Bitcoin at the end of June to 602 Bitcoin at September 30 and to 802 Bitcoin as of October 31. As a reminder, our synthetic HODL strategy enables us to utilize excess Bitcoin generated each month to fund our growth at a far lower cost of capital than external funding sources while maintaining upside potential by applying a portion of the proceeds towards the purchase of long-dated Bitcoin call options.
In regard to our ATM facility, which we initiated in March and used solely to fund our growth initiatives and fleet upgrade, we raised net proceeds of $66 million in the third quarter and a subsequent $39 million since September 30. Lastly, I wish to point out that the growth and diversification initiatives that Ben has spoken about are all done with a keen eye on accretive growth, and we have and will continue to weigh financing options that provide the greatest shareholder return and minimal dilution. Before I hand the call back to the operator for Q&A, I'd like to echo Ben's enthusiasm about the strong position Bitfarms is in. We've laid a solid foundation in 2024 that sets us up nicely to execute on our growth strategy into 2025 and beyond.
Our entire team, including the recent leadership additions, are aligned and working towards these goals, and we look forward to keeping everyone updated on our progress. With that, I'll hand it over to the operator for Q&A.
Thank you. If you'd like to ask a question, please press star one one. If your question has been answered and you'd like to remove yourself from the queue, please press star one one again. Our first question comes from Mike Colonnese with H.C. Wainwright & Co. Your line is open.
Hi, good morning, Ben, Jeff, and team, and thank you for taking my questions today. First one for me, if you could provide more detail on the discussions you're having with prospective customers for the HPC AI vertical. I know you'd announced your intentions of entering the space last quarter, so it'd be great to get more color there.
What could investors really expect as it relates to the potential economics from an HPC AI hosting deal?
Hey, Mike, thanks for the question. We're in active conversations here with multiple different parties. Really, when you look at what it takes to deploy an HPC and AI site, there's a few different things that we need to bring into our wheelhouse from other parties. And really, it's the specialized expertise on what is effectively very, very different infrastructure and very, very different equipment. We have a number of different parties calling us. Our phone is very busy right now because we have a huge pipeline of megawatts available for us in the near term.
And really, when you look at the revenues that you can get on an HPC and AI deal relative to Bitcoin mining, the effective value I think that most of these deals are looking at, we're going to be in the $100 a MWh plus kind of revenue before you start adding on these other services and fees. So it is more revenue right now than Bitcoin mining. But one of the things that we've been talking about for several months now, Mike, is that we believe, and we still believe, that the next maybe 15 to 18 months, we're going to have better economics and a better return on our invested capital with our Bitcoin mining efforts than we will on the HPC and AI.
And it's really towards the tail end of 2025 or towards the tail end or the beginning of 2026 that we want to begin integrating HPC and AI into our portfolio in a meaningful way that's contributing to revenues. When looking at the 1 MW HPC and AI site, that's really for us not to be a meaningful contributor of revenues, but for us to properly develop this technology and bring these expertise into our company and into our portfolio.
That's great color, Ben. Appreciate that. And just one more for me. How should we think about the cadence of hash rate expansion from where we are today at 11.9 exahash and as we progress through the first half of 2025 to get to the 21?
Yeah, we're continuing to plug in miners. Regularly, we've received and deployed the first miners as the hosting agreements that we've announced with Stronghold just this week. So we're receiving the first miners and plugging in. We should see continued progress on the hash rate. But it's really that towards the end of the new year and with the holidays and everything else, as well as us just trying to be a lot more conservative in our guidance strategy, what I think you should just expect is kind of regular linear growth that you can track with our monthly production reports. And the progress is going to be, or the, sorry, the S21 hash target is going to be hit in that first half of 2025.
Great. Thanks, Ben.
Thanks, Mike.
Thank you. Our next question comes from Mike Grondahl with Northland. Your line is open.
Hey, thanks, guys. And congrats on the HPC progress. In terms of the two pilots, the 1-2 MW , are there megawatts behind those so those sites could be used for larger projects once the pilots have proceeded and you've learned what you needed to learn? Any number you can give us at each site, like how many other megawatts are available, if any?
Hey, Mike. Good question. So we've got the two different sites. The goal would be to finalize on one of the two as opposed to building out two similar pilots. Both sites have expansion capacity in different quantums. One of the sites has an expansion capacity of about 20-ish MW, and the other one is in the triple figures.
Got it. Okay. So one's small and one's large. Okay.
Correct.
When do you think you'll decide on one site or the second site? And when you do that, will you communicate if it's the 20-ish or the triple-digit site?
That's going to be dependent, I think, largely on the strategic partner that we end up working with. The different sites have different advantages for the different strategic partners based on their proximity to different metropolitan centers. So hard to determine which site would be the final choice because it really is dependent on which strategic partners we're going with.
Fair. And then it sounds like you've made some progress with strategic partner or customer. Have you been able to explore the financing availability kind of terms, and do you feel like you understand that well?
Yeah. I mean, I'll add a quick comment here and then maybe pass it off to Jeff Lucas. But we've had numerous conversations on the finance side. I think there's no shortage of capital that's available at the moment, especially for infrastructure projects like this, and especially with Bitcoin pumping right now. The opportunity here for both Bitcoin mining and HPC AI is very, very hot, and I think capital is very much available right now. Maybe I can pass it off to Jeff Lucas to provide a little more color.
Yeah. Actually, I think Ben covered it pretty well here. Just a couple of comments. We are seeing a large number of opportunities here to take advantage of what we're doing in the HPC AI here for in terms of our own financing needs.
But I think it's also very important to understand here that as we go forward in establishing our partnerships here, we recognize that a lot of these parties here bring a lot more to play in terms of capital raising than we can do as efficiently and as effectively. So in summary here, a lot of opportunities that we're looking at right now and several hundred million dollars, should we pursue that path? But secondly, we'd be looking to work with the partners that then play a pretty instrumental role in whatever capital raising to take advantage of their more efficient cost of capital.
Got it. Okay. Hey, thanks, guys.
Thank you. Our next question comes from Joe Flynn with Compass Point Research & Trading. Your line is open.
Hi. I was hoping to get some more color on ultimately the decision to go forward with hosting the SDIG sites rather than I think originally was supposed to be the addition of megawatts in Paraguay. Yeah, is that more related to the machine delays or maybe unexpected problems in Paraguay? Thanks.
Thanks for the question. The decision to move the inventory from Paraguay off to the Stronghold sites was due to a number of different factors. At the end of the day, we are portfolio managers, and we are constantly looking at our portfolio of assets and our growth pipeline and working to manage our positions in order to drive an optimum outcome. When we were looking at the decision to deploy in Paraguay versus the United States, effectively by deploying in the United States, we were able to do a couple of things really quickly.
One, we were able to focus on deployments as early as this month, which started, as I mentioned a few minutes ago. We had our first few miners plugged in this week. But the second thing is really it's more optimal deployment around the T21 specifically. So we're able to save on logistics, duties. We end up getting a lower all-in power cost in the Stronghold sites. And one of the advantages with the T21, which is not prevalent in the other miner models, is the ability for them to be overclocked. And so the PSU on those miners are capable of doing well over five kilowatts. And at the Stronghold sites, we have the ability to do a lot of energy demand response programs and energy trading.
Having that greater flexibility with the T21s means that we can get a lot more utility out of those assets than we would be able to deploy them in Paraguay where we've got a static fixed cost of electricity. When we look at Paraguay and Yguazu specifically, how do we get the best value for that site? That site, we could either have 100 MW accounted for and 100 MW that we would need to fill out or allocate to some strategy, or we could have a full site of 200 MW. To have the full site clean, ready, and strategically flexible provides a lot more value to the Yguazu site when we're looking at what are the strategic opportunities to either deploy more miners or host or do something else with that site.
Thanks, Ben. Appreciate the color. And then looking forward into 2025, I think you mentioned you're still deciding ultimately how to. But the best deployment of the megawatts. But I guess related to the integration of Stronghold site, I mean, do you have all necessary approvals to ultimately generate power through the plants or take power from the grid? And just how do you guys think about that mix between importing and your own generation?
Yeah. So those sites are active. They've been active power plants for, at this point, over three decades. They also have the ability to actively draw down power from the grid. So those sites have been in operation doing both those things, generating power and drawing down power from the grid for several years.
When we're looking at how do we get the best utility and value out of those sites, we were doing things like refining the equipment deployment schedules so that we can get more utility out of those miners and out of those sites. But there's a lot of strategic flexibility there for the 200 MW which remains. When you're looking at a Bitcoin bull market and HPC and AI, really the answer is there's no shortage of opportunities to monetize those 200 MW right now. Just like the Yguazu site, to have 200 MW of near-term capacity in 2025 cost-effectively, that is a unique asset going into a Bitcoin bull market or when you're in discussions with HPC and AI hyperscalers.
Great. Thanks.
Thank you. Our next question comes from Brian Dobson with Clear Street. Your line is open.
Hi. Good morning. Thanks so much for taking my question. So I guess if we can maybe take it up to 30,000 feet and take a view of the broader industry for a second, how do you think an administration change in Washington could potentially impact regulations in the United States? And I guess how are you viewing that, and have you gotten any additional color on what regulations might look like? I know it's early stages, but if you could perhaps just contribute a little bit to the conversations going on in the industry right now, that would be appreciated.
Thanks, Brian. Yeah, it's a good question. When you look at what happened there with the election and kind of where things fell out, I think one of the clear things that happened is Pennsylvania, which is a big swing state, voted very clearly in favor of Trump and I think energy policies.
The ability for us to tap into the energy assets that we have there at the Stronghold sites as well as the Sharon, Pennsylvania sites, I think are a lot more favorable and a lot more valuable now that the election is over and Trump has been victorious in his campaign here. That bodes really, really well for energy assets, energy infrastructure, regulatory environment, regulatory red tape, and moving forward there with our expansion plan. So we're very, very bullish on our U.S. energy assets as part of the reason why we've done such a strategic refocus on the U.S. and our North American portfolio. But surely the future for energy in the United States, especially in Pennsylvania, and really specifically at our Stronghold assets, is looking brighter than ever now under a Trump administration.
Let me add one comment here, if I can. Also, in terms of the capital raising side, stating the obvious, that's expected to be a different playing field here with the SEC. We actually enjoy a very constructive relationship and a dialogue with the SEC, particularly as we're working through our filings in connection with the Stronghold acquisition. Naturally, we only expect that to get a little easier, if anything, going forward.
Yeah. Very good. Thanks. I'm curious also to add a little bit of color on how you're thinking about potential returns on, call it, growing out or building out your mining facilities versus HPC, given the recent run-up in Bitcoin. Has that changed the math for you as far as where you'd like to take the company?
It's a great question. I think one very, very relevant today with Bitcoin almost at 90,000. The math really has not changed at all because these were the exact economic situations that we've been talking about for, at this point, six, nine, maybe even 12 months now. We've been talking about how hash price equates out to an equivalent revenue on a per megawatt-hour basis. What do we expect for hash price to do post-halving?
And what we've outlined is historically, every single halving cycle that's happened other than the first one has been that we've seen a 300%-400% increase in hash price from the first month exiting the halving to the peak of that Bitcoin bull market. And what I outlined here was if historical cycles repeat, that we should see anywhere from, on the very, very conservative end, a $0.12 hash price, on the moderate end, a $0.15, and on the high end, even as high as $0.20.
Now, that's assuming that history would repeat itself. There's no guarantee that history would repeat itself. But those are the situations that we, as a Bitcoin miner, are working towards. And when you look at the value that that hash price would equate to on a megawatt-hour and you convert it over into the T21 or an S21 Pro or what have you, you'd see that once hash price starts doing what it's doing, it's up about 10% this week with the Bitcoin price. We'll see that the hash price is actually going to drive a much better return on invested capital in 2025 than an investment in Bitcoin mining or, sorry, an investment in HPC and AI would today. But that doesn't mean that the HPC and AI is not a great technology and a great business line to add in.
It just means that in the near immediate future, the next 12, 15 months, we still believe Bitcoin mining is going to be the best use for capital. And we think that there's going to be maybe a phase shift that happens kind of at that 12, 15 mark, right, where then HPC and AI is really the more attractive place to be getting better returns on your invested capital. And so that's what we've outlined for quite a few months now. And that's still consistent today with Bitcoin price rallying.
Glad to hear. Thanks so much.
Thank you.
Thank you. Our next question comes from Lucas Pipes with B. Riley. Your line is open.
Thank you very much, Ben. And good morning, everyone. This is Fedor Shabalin asking questions on behalf of Lucas Pipes. And my first one is on HPC side, technically about slide 12. You pointed on accelerated deployment for HPC and AI side. And what exactly is this? And what is the timeline and associated CapEx for this pilot project? Thank you.
Thanks. It's a good question. What we've previously outlined with regards to HPC and AI is that the earliest that we could have something online is at the end of 2025 or at the beginning of 2026. That was a timeline that worked out well both from an infrastructure and construction timeframe, but also from most of the HPC and AI kind of client timeframes for when they would expect to get their GPUs deployed and online.
What we're looking at is, can we make that timeline even faster than the Q4 or the Q1 with the pilot project so we can get that understanding and we can get that skills and the expertise in hand before we start allocating more capital onto a more meaningful resize deployment? And so that's really what that accelerated deployment timeframe is from. When it comes to the CapEx requirements, I think I'll pass it off to Jeff Lucas here in a second. But just to reiterate something that we said before, when it comes to HPC and AI, we're really much more focused on the infrastructure side as opposed to the compute side. And so by focusing on the infrastructure as opposed to the compute, you cut out up to 75% of the CapEx associated with an HPC and AI deployment if you're only doing the infrastructure.
Thank you for this. And second one, technical. So three or four quarters ago, you were probably the first one who described the performance of Bitmain T21 miner and mentioned it could work better than stated characteristics. And can you tell maybe the same about S21 Pro miner if you had a chance to test these machines, how efficient they are?
The S21 Pro is even more efficient and powerful. It's about 20% more efficient and about 20% more powerful. So this is a very, very capable machine. The main difference, I think, between a T21 and an S21, besides just the stated hash rate and the efficiency, which is higher in both categories, is the fact that the T21 has a lot more flexibility in terms of its power consumption. But the S21 Pro is a fantastic unit.
Everybody who's been deploying them on market has had nothing but positive things to say. And we look forward to receiving our first S21s here, S21 Pros here, next month.
Got you. And specifically, have you seen their performance kind of beyond 234 terahash per second, or they didn't cross the line?
I'm sorry. Are you asking about overclocking on the S21 Pro?
Yeah, overclocking, like you talked previously about T21.
So on the T21, the difference here is that because the T21 was specifically designed around that energy flexibility. So it had, by the manufacturer's default, a normal energy mode and a high energy mode. So on the T21, it wasn't actually overclocking. We were actually using the miner according to its specifications from the manufacturer. When it comes to the S21 Pro, we haven't done this in testing, and we don't actively do this in operation.
But yeah, most of these miners can do better performance in overclocking. And I think most of the industry numbers that I've seen on the S21 Pro, they can do well over 260 terahash quite reliably.
That's very helpful. Thank you. And if you allow me to squeeze the last one, a short one. Sorry if I missed this. What was the main reason behind increase in SG&A expenses quarter- over- quarter?
So the largest driver of the increase were actually non-recurring expenses, many associated with some of the one-time events, including the Riot requisition here. We had a settlement with the former CEO and others. So we actually incurred roughly around $11 million of expenses that are non-recurring during the quarter here, $11 million-$12 million. I don't anticipate recurring. To a small degree, there were also some of the expenses associated with the acquisition of Stronghold.
We're going to incur M&A costs as we're pursuing new opportunities going forward here, but nothing of the magnitude that we incurred in the third quarter.
Thank you, Jeff. I turn it over for now and continue best of luck for you. Thank you.
Thank you. Our next question comes from Martin Toner with ATB Capital Markets. Your line is open.
Good morning. Thanks for taking my question. I know you talked about it a bit already, but can you reiterate the source of caution around Paraguay and the build-out and exahash there and just give us a sense for where that could fit into your plans going forward into next year?
Sure. First, I'd like to rephrase. It's not that we are exercising caution, I would say, towards Paraguay. It's really more about managing our portfolio to try and create the most optimal scenarios with our miner deployment schedules and our capital allocation. The original plan with Paraguay was originally devised before the Stronghold transaction had materialized. As the Stronghold transaction moved forward and was announced, the ability for us to drive an optimal result out of our portfolio meant that you shouldn't stay static in your positions. We needed to move miners around. We needed to adjust the schedules. We needed to do things like upgrading some of the miners in order to drive that optimal performance.
I'd really strongly state it wasn't a caution around Paraguay. It's really more about managing the portfolio to try and create those optimal scenarios for our miner deployment schedules and our investors. We still remain very long on Paraguay and LATAM.
We have 280 MW about to be ready there in Paraguay, 340 MW across the entire LATAM region. And we still remain very, very bullish on that area because the power is renewable, cost-effective, and high uptime. But how do we optimize the existing miner deployments given the characteristics of the miners and various other factors? That was really what was coming into play here.
That's great. Thank you very much. That's all for me.
Thank you. There are no further questions at this time. I'd like to turn the call back over to Ben Gagnon for closing remarks.
Thank you very much. Thank you for everyone for your time today. Again, I'd like to restate. I've never been more excited about our future.
I really look forward to updating you all on our progress as we continue to execute our transformation and growth plan to capture additional market share within both Bitcoin mining and expanding into synergistic value-added business lines. Thank you very much.
Thank you all.
Thank you for your participation. This does conclude the program, and you may now disconnect. Everyone, have a great day.