All right, well, welcome, everyone. Thank you for coming today, whether you're here in person or online. My name is Ben Gagnon. I'm the CEO of Bitfarms, which is one of the largest and oldest publicly traded Bitcoin mining companies out there. We've been publicly traded and mining Bitcoin since 2017. Started in Canada. Now we have an international portfolio across the U.S., Canada, Paraguay, and Argentina. Just a real quick, I'll say, safe harbor.
Starting off, I want to go from a big 20,000 ft view in the sky. Where we starting this year is with a really, really strong footing. We have a large, diversified portfolio of 324 active megawatts. We completed 10 of our site upgrades last year. We completed a whole new site construction. We completed one new acquisition. We ended up doubling our corporate hash rate last year and improving our overall energy efficiency by 40%.
In 2025, what that means is that we're really, really set up for a very exciting growth path this year. We've got a pathway between 324 MW-955 MW this year. We're going to be tripling our active energy capacity. That's split up into a couple of different sites that we'll get into over the course of this presentation. Beyond 2025, we have a pipeline, mostly in the United States, to grow that 955 all the way up to 1.6 GW over a couple of years. We've got multiple sites in development. We're on track to complete an acquisition here, the largest public M&A deal between two publicly traded Bitcoin miners, between us and Stronghold. It's anticipated to close this quarter. We're on our path to 21 exahash and 19 watts per terahash in the first half of this year.
But we're not just focused on growing the energy and growing our Bitcoin mining footprint. We're really focused on diversifying beyond just Bitcoin mining itself, not in a way that's going to distract from the Bitcoin mining, but in a way that's hopefully going to make us better Bitcoin miners. And this means that we can do things that are better returns on investment and can capture synergies across these different business lines, which means that instead of just self-mining, we can also do hosting, heat recapture, energy generation, as well as energy trading. And of course, HPC and AI, which is presenting a lot of interesting opportunities given our robust energy portfolio. We are aggressively expanding in the United States. And our focus in the United States is largely in the PJM Region, which is the largest energy trading market in the United States.
We are going from a portfolio that had only 6% of our footprint in the United States to a portfolio that has almost 50% of our portfolio in the United States this year. Dramatic rebalancing from the largely international to a still an international, but a largely North American focus. We're also well positioned now to capitalize on market headwinds, or sorry, market tailwinds and macro tailwinds, which are going to be capitalizing on with all of this growth that we have for this year. There are five areas where we're really focused on in our continued growth for this year. As I said, the first focus is on that U.S. expansion diversification beyond Bitcoin mining, with a massive strategic pipeline of almost 1 GW and beyond this year and in the years to come.
Once we have that portfolio, our focus is being good portfolio managers and figuring out how do we get the maximum value and utility out of those energy assets by leveraging our low-cost power, our operational excellence, diversified business models, with a focus on not growing hash rate, but on delivering better returns on our invested capital. We also have a very exciting success here with our BTC leverage program, which we're going to be expanding this year after having proven results throughout 2024, which we'll be getting into, and we've got really strong liquidity here to fund our growth pipeline in 2025 and beyond, so starting off with the most exciting thing is the Stronghold acquisition. As I said, this is the largest merger between two public Bitcoin mining companies to date.
And it has a really, really unique Bitcoin mining and HPC AI potential here at these two sites. And there's a lot of reasons for this. But the first is that it gives us the ability to integrate vertically by acquiring these two strategically located power facilities. And when you're acquiring a power facility in a strategic location with strategic grid connections, you have dually redundant sources of energy, which makes this not only attractive for both Bitcoin mining, but it also makes it very, very attractive for the HPC and AI side for having that level of reliability and having that massive source of potential cost savings on the build-out for an HPC AI site. As I said, this is expanding and rebalancing our energy portfolio with over 300 MW of capacity.
From that 324 to the 955, this is going to represent almost half of that growth and is anticipated to close in the first quarter. We've got a massive energy trading and demand response opportunity there, which not only helps us to increase our energy portfolio, but drive down our average energy costs. Previously, what we've largely focused on is tariff-based rates, whether that be in Quebec or Paraguay or Washington State. But by getting into PJM and these energy trading opportunities, it gives us a lot more levers to pull in order to control our primary cost, which is energy. The average cost at Pennsylvania is really going to be determined by how much energy we use and how much uptime we have.
But by controlling that variable, we're able to either bring our costs down and optimize our margins and our free cash flow that we generate on a daily basis in line with current macro and economic conditions, or we can increase our uptime, increase our power, but still optimizing around that free cash flow position according to the economic conditions at the time. This transaction is going to position us to be the leader in the region. We are largely the largest player in all of the markets that we operate in. We're the largest Bitcoin miner in Canada, the largest Bitcoin miner in Argentina. We're the largest Bitcoin miner in Paraguay. And with this transaction, we'll be the largest Bitcoin miner in the largest energy trading market in the United States.
To break down Stronghold and what this power looks like over time, effectively, we've got two power plants that each have roughly a nameplate capacity of 80 MW. Now, in addition to the capacity to generate the power, they have a substation, which historically has always sold that power into the grid, but also has the ability to draw down an equivalent amount of power from the grid. So utilizing infrastructure built for 80 MW, we can actually get 160 MW worth of load consumption on the site, dramatically reducing our costs and improving a lot of our operational efficiencies. That brings us to an immediate short-term capacity of 307 MW, which matches almost entirely our 324 that's currently active today.
Beyond the 307, we've got numerous studies and applications in to expand those grid connections with various utilities through various infrastructure upgrades and new connection lines, which can increase the total amount of power to each site to almost 500 MW apiece. What this means is that these sites, which are not only very strategically located near major metropolitan areas on the east and the western coast of Pennsylvania, but they also have massive growth pipelines, which is very, very attractive for HPC and AI customers who are looking for scale, not only for scale immediately, which is primarily the most valuable thing, but scale that can continue to deliver over time. 955 MW potential capacity there by combining the current existing power and those power applications over time.
There is now over a gigawatt of potential power capacity there with our completed acquisition in Sharon, Pennsylvania, which is about 45 minutes away from the Scrubgrass facility here on the western side of Pennsylvania. A very, very big footprint here in the United States and a strategic shift here tapping into that very attractive PJM market. This is reflecting our growing and diversified energy portfolio. What you can see here is that we had an energy, or we ended this year, started this year with an energy portfolio of 324 MW, which was very distributed across Latin America and Canada, with just the United States representing a tiny sliver of about 6%.
With these transactions here at Stronghold and the expansions that we're doing, we've got this pathway here so that we're going to be almost two-thirds North American as opposed to sub 50% North American at the end of the year before the Stronghold transaction. As we continue forward executing on those growth plans, we have the pathway here to be in almost 80% North American and 20% international with a very, very attractive footprint, as I said, in Pennsylvania specifically. Not taking away from the existing footprint in Canada, Paraguay, or Argentina, but really emphasizing that the growth is happening primarily in the United States. That growth is largely spread across these three different sites here. And what I'm referring to on the figures here is not total megawatts because at Stronghold we do have the 307, but this is flexible capacity.
Now, the most valuable thing to customers is timeline to energization. And so having immediate capacity is obviously worth significantly more. Having capacity and energy assets that can service more than one market profitably is incredibly valuable and a strategic shift for us. So our assets here are capable of supporting both Bitcoin mining or HPC. The two higher value sites for HPC are going to be in North America, but all three of these locations do support both kinds of compute. The Sharon location could do either 8 EH/s or 92,000 GPUs, given the latest generations of GPUs and Bitcoin mining hardware. With the Stronghold site beyond what we already have currently planned there, there's a further 15 EH/s potential or 150,000 GPU potential at that site. Same thing with Iguazu and the 200 MW there.
So it's a lot of strategic flexible capacity, which is available in the near term and comes at a very, very high value to a lot of customers out there looking for growth this year. This helps us to capitalize on these macro tailwinds that are currently surging across demand for access to power and infrastructure. And when you look at compute markets, the overall bottleneck for growth in compute markets has never actually been the compute hardware itself. It's almost always the power infrastructure in order to operate that compute. And so when you look at where are the opportunities here and where is going to be the key player, the key element on anyone's growth strategy, it's not on who has a contract with NVIDIA or who has a contract with Bitmain.
It's who has the contract with the power utilities to turn that equipment on and to keep it operational. There's a huge amount of power demand for these sites and for this further growth, which has a compounding effect because you can only use 1 MW for one purpose. So either this growth is going to go to Bitcoin mining or it can go to HPC and AI. If it goes to HPC and AI, that means that there's going to be less growth in the Bitcoin mining sector, helping to keep those profits higher and those revenues higher. If that growth goes to Bitcoin mining, then that means that the HPC and AI customers are going to have that many fewer megawatts to go after and fulfill their own growth obligations.
So this gives us a very balanced and diversified approach, which helps us to tap into the surging demand for power in both sectors. The North American portfolio here is pretty well suited according to its geographies, its proximities to major metropolitan areas, its proximity to major fiber trunk lines, and the fact that we do have that secured, cost-effective, high uptime power across these sites. So what I've identified here are five different sites where we've got significant potential beyond Bitcoin mining and to tap into HPC and AI. Starting on the left with Moses Lake, Washington, which would be the smallest of these sites. We have about 20 MW that's closely located to a data center hub in central Washington. There's about 14 .5 GW of installed hydroelectric capacity within a one and a half hour drive of our site.
So this is driven by AWS, Microsoft, Cisco, Oracle. They all have data centers in this area. And just last year, Amazon bought a data center one mile down the road from our Moses Lake site. The Sharon, Pennsylvania site, just continuing kind of left to right, the Sharon, Pennsylvania site represents 120 MW of capacity that we're building out the substation for right now. And that's very closely located to major metropolitans like Cincinnati, Pittsburgh, and Cleveland.
And it's also very close to our site, Scrubgrass, which has that almost 500 MW of potential capacity. And there is a large Amazon data center about. I think it's about one and a half hours away from there. These data centers like to cluster, and they're looking for growth in campuses, campuses that are close together, campuses that could be even potentially connected with direct fiber access, data center to data center.
So having that proximity makes those sites incredibly more valuable. Just continuing further right through Pennsylvania off to Panther Creek, we have our site, which is closer located to Philadelphia and New York City. And all of these sites here in Pennsylvania are also very close to the Data Center Alley in West Virginia, which surrounds a lot of the defense infrastructure and a lot of the data center infrastructure. And so these are all very, very highly sought-after areas, not just for Bitcoin mining because of the energy trading, but because of their proximity to all the other infrastructure out there on data centers. Very attractive for HPC AI customers as well. Going a little bit further to the north, we have our Sherbrooke campus. Now, we have 170 MW in Quebec, but it's largely spread out over numerous facilities, some of them a bit smaller.
But in Sherbrooke, what we have is 96 MW in a campus within one and a half miles of each other. So it's spread out over three properties, but that power can be consolidated into one single location, which then makes it go from a Bitcoin mining site to potentially a very attractive HPC AI site that's close to Montreal. Moving beyond the success of our HPC, I want to focus a little bit here on the Bitcoin mining side of our business. What we did last year, actually started at the tail end of 2023, but we continued this program throughout 2024, was we launched what was called our Synthetic HODL program. This was a program that was born out of a couple of different objective driven objectives that we were trying to drive.
The first was, how can we be more effective with using our proceeds out of our actual operations to fund our business and to fund our growth while still maintaining upside exposure to Bitcoin that our investors want? And what we did was we went out there and strategically bought long-dated call options in the market in order to compensate for some of the Bitcoins that we sold to fund our business. So that enabled us to maintain that upside over the foreseeable nine to 11 months while still using that cash flow to fund operations, which is obviously going to be our cheapest source of capital. We ended up buying roughly $13 million worth of long-dated Bitcoin call options, which came out to generate we sold for $30.5 million, generating 135% return in U.S. dollars or normalized out for the increase in Bitcoin price.
We outperformed Bitcoin by 63%. And so this is the way that we were able to actually deliver that beta to Bitcoin that investors have been searching through the mining operations was actually less risk, higher efficiency, and a higher actual outcome at the end of the day. We are pretty excited now to be talking about how we're going to be expanding that program this year and making it more sophisticated. So we are going to be launching something called Bitcoin One. Bitcoin One is an expansion of the Synthetic HODL strategy, which systemizes a lot of what we were doing last year into codified strategies. So over the last year, we have been testing out different investment strategies. We've been doing some things ad hoc and opportunistically according to market opportunities.
But we also developed a portfolio of, I think, 12 different trading strategies that we've been testing out with the objective to deliver that outperformance to Bitcoin with less risk and higher returns, adjusting our risk-reward ratio to something that actually delivers much better expected results than getting that upside exposure via Bitcoin mining. So this is simply positive convexity. What we want to achieve is something that's going to outperform Bitcoin on the way up and something that's not going to go down as much as Bitcoin when Bitcoin goes down. Asymmetric upside where we can outperform Bitcoin while also providing some risk mitigation to downside. We do this by having a diversified portfolio that consists of direct Bitcoin as well as actively managed leverage. And by combining the two, we can manage a target leverage ratio between 1 and 3.2x to Bitcoin.
The benefits, of course, with leverage is going to be enhanced returns as well as getting actually a better risk-reward ratio through that diversification and also through the utilization of derivative and option contracts, which are a lot more efficient exposure to Bitcoin's upside because they have no exposure to actually growing network hash rate, which is going to offset a lot of that upside to Bitcoin. So I'd say with your, it's like a butterfly strategy. You're buying calls, you're buying puts after the tail. Is that the graphic? It's expensive overall, right? You're spending a lot of money on the premium. We've got a variety of different strategies. Sometimes it's going to be something like a covered call or something what you've described. Sometimes you're going naked on long-dated calls. Sometimes you're just going to be selling an out-of-the-money call.
So for instance, not a very material transaction, but a couple of weeks ago, we sold out-of-the-money calls when Bitcoin was a little over $102,000. We sold out-of-the-money calls for January. So it's a right to buy Bitcoin at $110 or $120. And Bitcoin pulls back. I mean, it was a naked exposure, but if Bitcoin goes up to $110 or $120, we're going to sell the Bitcoin to fund the business anyway. Now we've just locked in a higher price, would be the $110 plus the premium or the $120 plus the premium. If it doesn't hit, we capture the premiums. So there's a variety of different strategies that we're implementing in order to provide better returns and get greater exposure and more efficient exposure to Bitcoin's upside through utilizing that cash flow stream of Bitcoins that we have coming from our mining operations.
So in addition to that, we get a lot of capital efficiency from this because it does provide that more direct exposure. And we're able to manage our risk better through a systemized approach of diversified portfolio strategies. The financial performance from Q3, and I apologize that obviously in Q1, all these Q3 figures are usually a little bit dated. But in Q3, what we saw was we had rising revenues quarter -over -quarter and year -over -year, which reflects our increasing hash rate growth. It also reflects some increase in the Bitcoin price. And we also have pretty strong gross mining margins of 51% and 38%, which takes into consideration the improvements that we've had in operational and energy efficiency over the course of the year. This means that we have quite strong and robust liquidity to fund our growth this year.
Our 21 exahash targets for the first half of this year are fully funded. The remaining growth that we have for this year to the 955, we're focused purely on the CapEx portion of this. It's roughly $125 million, for which we have $147 million of liquidity, or we ended the year with $147 million of liquidity.
We currently have zero plan to buy additional miners, with the focus being on monetizing that increase in energy infrastructure through strategic partners and strategic relationships, whether that be hosting for Bitcoin mining or hosting for HPC and AI, which have very similar profiles, but it enables us to focus our investments this year on the hard energy assets, the assets that we think have the longest value, the best ability to accrete value to shareholders over the next, not just the next 12 months, but over the next three, five, 10 years and beyond, and to minimize our capital raising needs this year so that we're not looking at running a lot of equity issuance to fund a lot of Bitcoin miners. We've got a lot of Bitcoin miners already. We've got a lot of energy assets.
Our focus this year is on building out those energy assets and working with the right strategic partners in order to monetize those more cost-effectively so that we can deliver higher returns on invested capital and minimize the capital that we need to raise this year, focusing on the shareholder value piece as opposed to just growing our hash rate to grow our hash rate. And with that, I think that's it. Yeah. Hold on one second. Can I invite Jeff Lucas, our CFO, to come up here for Q&A, and then we'll open the floor up to questions. Morning. Francois?
Financial strategy you describe, how do you track your risk? Because I was thinking as soon as you enter there are things that come to pass, it becomes, I guess, a big question mark.
Well, Jeff Lucas is overseeing that program, so I'll let Jeff Lucas.
So let me answer that in, actually, in three levels. I'm going to expand a little bit upon your question here. First of all, we actually spent about eight months making sure that we had the appropriate controls in place because when you're working with derivatives, there are so many factors beyond your control. And if you're leveraged, you do have a higher level of risk and exposure there. So we manage it very carefully. The second thing we do from a structural standpoint here is that we actually have what's called the Risk Committee . It's comprised of around five of us in various functional areas, whether it's finance, treasury, Ben's part of that group, and others who are actually in the corporate development side.
So that we really do a very thorough assessment of these opportunities and remove the subjectivity of our analyses to really focus on what the quantitative numbers are telling us. The third thing that we do, actually, when we take a position here in terms of how we protect that, is that obviously we track it religiously, needless to say. We also have various models that we have in place here to understand the various outcomes, and we attach sort of probabilities to those outcomes. So those three measures and very active diligence on our part, we're actually able to control them very effectively.
As you think about investing money on the HPC AI side, can you just think about how do you think about the demand supply? I mean, we all know AI is going crazy, but how do you think about the demand supply over the next three to five years?
So as I said earlier, the primary bottleneck to growth in any compute market is not really the compute, it's actually the energy itself. The evolution in chip design and efficiency that you see through Moore's Law is just rampant and unrelenting, which means that over time, the supply of compute just increases exponentially, which is driving down those unit economics. That creates kind of an incentive here for people to constantly invest in more growth and more growth and more growth. For companies who are trying to focus, I think, on the compute side, it becomes a really, really tricky game because you now need to focus all of your capital expenditure on constantly upgrading and buying new computers.
But even if they're getting more efficient, it still means that there's rampant demands for that power consumption growing over time. And what we see here is that it's pretty easy to forecast based on just kind of the market trends that you've seen historically and rolling them forward, how you think that network or how you think that industry is going to evolve when it comes to increasing energy consumption. And the timeline is really the key thing, right? When you're looking at how do you grow your energy over five, 10 years, that's a very different question than when you're trying to look at how you grow your energy over the next maybe 12 to 18 months. And it's more of a just forecasting question when you're looking three to 10 years out.
When you're looking 12 to 18 months, really it's driven by a market demand that a customer is saying, "I want to have this plugged in or I made this investment." And I think there's no guaranteed way to look at that. And I think a lot smarter people, a lot more smart people, people who are a lot more smarter than I am are doing that work. We have a couple of figures here that I can just scroll back to, which shows the projected figures here. This came from KKR, who I said, a lot of people a lot smarter than I am probably put together this data and this analysis. And we've got roughly 16 GW-18 GW, according to them, for U.S. power consumption.
What we have is a similar kind of amount of power consumption just for Bitcoin mining globally, but this is something that's growing every single year at a rough rate of 4x for capital expenditure. This is something, this is a massive, massive market to tap into. It's not really sure how you can tap into the precise demands for power other than just through the conversations that you have with potential customers for that power, trying to gauge their demand and their pricing sensitivity. Yeah?
I'd love to get some more color on closing of the Stronghold. I think they announced maybe last year or two, some kind of quarterly regulatory milestones that both.
Yeah, sure. The expectation is that's going to close in this first quarter. Right now, everything is going along on track. There are two key milestones that come into play. First of all is our registration statement, what's called the Form F-4, that allows us to issue shares for the acquisition. We're buying with stock here. You get approved by the SEC. That is very far along in the common process here. As a matter of fact, we're really down to just two questions that we received two days ago that are very straightforward. No accounting issues.
Those are all behind us here. So that's moving along. And then once we have that, actually, we're allowed to go effective, as they say. The next step here is just to issue the proxy and hold the vote among the Stronghold shareholders. Now, a lot of the management team, the ownership, or the active ownership, I should say, are shareholders of this. So we have every expectation that it's going to go very smoothly and don't envision any problems there.
So again, just to repeat, we expect this to happen in the first quarter.
Wonderful. Just a follow-up on that. Given that maybe to explain why Stronghold's so important. Riot announced that it's going to be an acquisition offer for Riot. So how does shareholders feel about the fact that things turned out and the company is being acquired by Riot at a much higher price for Stronghold?
I think what you have to look at is there's a lot of different forces that are driving shareholder price and not just one transaction. Obviously, the Stronghold transaction is a very meaningful part of what we're doing. We think it's going to be generating a ton of value over the next not only 12 months, but over the next three, five, 10 years plus because these are assets that not only have a lot of opportunity right now, but have a tremendous growth pipeline for the company that is largely secured.
The forces that I think have been driving our share price, candidly, I think are less to do with the Stronghold execution and more just about having to push back the compute targets of 21 exahash into the first half of this year. That has largely to do with a couple of things on construction, which puts us back, changing our allocation of miners across our portfolio in order to optimize for a longer-term view as opposed to getting the hash rate on immediately.
But how do you get more utilization out of the energy assets and the compute assets themselves over the next 12, 24 months, as well as a lot of minor RMA warranty servicing that we've had to do that has been beyond what we've forecasted? So we announced that we swapped out 3,000 miners in July in kind of an expedited fashion. Those things are still happening now, and that's been pushing us back. All of those things, though, have been navigated through, and now it's really just about continuing to execute on the delivery here. And so I think as we show those milestones and as we show this progress over the next couple of months, I think that underperformance is going to turn into overperformance in the relatively short period.
As a potential investor or as a current investor, obviously, the historical is important to understanding kind of what happened and where you are today. If you're looking at what are the opportunities today, the opportunity is not in the companies that have already done 40%-50%. It's in the companies that didn't do that 40%-50% and have that ramp ahead of them over the next couple of months. I think that's the key thing to focus on is what is the opportunity that comes out of that relative underperformance. As we execute over the next six months, I think we're going to do incredibly well. I think we're going to prove why this was a better path for all of our shareholders.
On that note, we hadn't talked to Riot about putting out, but what can you talk to about the 90%? Can you expect that?
Well, I can't speak for Riot's management as to their position or their strategies or their intents. What I will say is that they have taken us as a Canadian issuer. Even selling one share requires a filing. So there's been no filings at all since they accumulated that position. They're largely staying stagnant in this position. And I think that they have a lot of confidence that our underperformance is also going to be addressed. But right now, they're a long-term investor in us. And beyond that, I can't say. They're not active in trying to, for the past. I think we're both activists as well. But no. We came to that settlement agreement several months ago, and both companies were very aligned during that settlement agreement to just forge on as separate independent companies.
By the way, that settlement agreement extends into May of 2026, just so you're aware. Yes.
To wrap up and oversimplify, Ben, I guess where my brain goes is to deduction. You came in six months ago. You've observed kind of with your view and strategy along the board. What I think I'm hearing you say today is, "I'm not going to invest in Bitcoin mining. I'm going to invest in HPC and really try and unlock shareholder value to get some contracts that are perhaps 10 years that draw up a lot of cash flow because that's a market growing at 12%-15% on KKR." Is that fair? And then my next question is, the $750 million market cap you are now, what kind of cash flow stream?
I'm not looking for guidance, but what type of numbers are we talking about to understand potential valuation and how it may or may not be over the next few years?
Yeah. So there's maybe two parts to that question to address kind of in sequence. The first part is we've never been a company that said, "We're just going to grow our hash rate just to grow our hash rate." We've always been focused on how do we strategically buy miners according to opportunistic conditions. And if anybody is going to go out there and tell you that you should be buying miners every month, then they just haven't run the math because if you do run the math, it'll show you that there's only a few times where it makes a lot of sense to invest in Bitcoin miners. And it's not linearly over time.
It's very strategically during the troughs when everybody is panicking and those prices are the lowest that you can capture those rising economics. Outside of those short windows, which we believe we captured at the end of 2023 with our investments in all the Bitcoin miners and the first part of 2024 with the continued purchases and the exercises of the options, we don't see that market continuing through 2025. And we think we've captured that market opportunity. We're really happy with that position. And we are looking at this from a diversified portfolio approach. And I don't want to continue to raise $500 million to buy miners that I don't think are a good investment. I'd much rather focus on raising less money or using less money to focus on the harder, more tangible assets, which is the energy.
Now, what you do with that energy, I think, has very different implications when it comes to expected revenues, profitability, cash flows, and you can kind of look at it as a spectrum of options ranging from lower CapEx, lower returns, lower risk to higher CapEx, higher returns, higher risk, and you can try and find a middle ground somewhere in between or an optimal position that you want to position yourself on that line. We think that probably a combination between using some of our assets that are already actively mining Bitcoin or are better suited for mining Bitcoin with hosting agreements with strategic partners is a really good utilization of those assets.
Because when you look at the kind of market for Bitcoin mining hosting agreements, and let me just kind of put some high-level numbers out there for you, the profit split arrangements that generally are being closed in the markets are anywhere between maybe 25% and a 40% profit split, depending on the efficiency of the machine. Now, when you look at what it costs to build out infrastructure, usually the miners are somewhere around 80%-90% of the cost to build out that compute infrastructure. So if I can cut my CapEx by, let's say, split the difference, I can cut it to 85%. And now my costs are 15% of the original, but I'm still getting 25%-40% of the profits. My ROIC now is dramatically better. My risk is much lower. I don't have those assets depreciating on my balance sheet.
They're giving me upside exposure and good positive free cash flow. Now, it's going to be less than if I invested all the money and I took all the risk on the Bitcoin miners. But my returns on invested capital are greater regardless of any underlying macro conditions. They're focused on how do you deliver that value for shareholders. That's how we're looking at it from that piece. That same logic applies to the HPC and AI side. We have no interest in going out there and saying, "I'm going to compete with Amazon on a NVIDIA contract for GPUs." I mean, we've got no reason to believe that we would be able to compete with them, have a lower cost of capital than them, have a greater chance of success.
But what we do have is what everyone wants, which is a ton of megawatts available now and in the short term. And so how do you get the best utilization of those? It's going to be focusing on opportunistic deal-making around, do you just sell them the power? Do you want to invest a little bit more money to build them a powered shell and get a higher return? But if you're going to require that capital, do you want to take it even further and build them out a fully built-to-suit site? There's going to be a different level of revenue and profitability and expected returns and risk associated with all of those.
I think what you can see is that when you look at the HPC and AI side, focusing on maybe the powered shell opportunity, you can build out that powered shell for maybe $2.5 million-$4 million a megawatt. Then the hyperscaler or the end client can come in there and they can put in their further investment to bring that closer to $8 million-$10 million and then all the compute associated with that. That can give you maybe a yield of high single digits, so maybe 7%-9% or so, depending on the deal opportunities. Those are pretty attractive returns. We look at the portfolio that we have there where we've got several hundred megawatts and, okay, it's going to cost $2.5 million-$4 million, and you can get maybe 7%-9%.
That kind of gives you an indication of ranges for what that would look like.
So I'm afraid we're out of time.