All right, guys, we're gonna go ahead and get started here. Next presenting company is Bit Digital, trading on Nasdaq under the ticker BTBT. Bit Digital is a sustainable platform for digital assets and AI infrastructure. Bitcoin mining operations are located in the U.S., Canada, and Iceland. Presenting today is Cam Schnier, Head of Investor Relations.
Thank you. Yeah, I'm just gonna go through our presentation, not in great detail, and there will be some terms that might be domain-specific, so feel free to interject. It's a cozy room, so, like, we don't need to wait till the end for any questions that come up, and I know we have one industry expert in Kevin, so he could certainly guide some of the conversation as need be. But I think the size of this room is kind of indicative of our valuation. We're very undercovered and misunderstood. We'll get into that. Let's pause and observe. Okay, a little bit about Bit Digital. We have two key, very distinct business lines. One is Bitcoin mining, which we've been doing since 2020. The other is newer.
It's high-performance computing, which we've got into at the end of 2023, and now this is quickly becoming our main focus as a business. Since 2020, we have mined a little under 7,000 Bitcoin. Part of our ethos in mining has been sustainability. As of a few quarters ago, we were at 99% carbon-free. We're now at 86% carbon-free as of last quarter. Sustainability used to be a much bigger focus for investors in general, but now Bitcoin mining, the economics, have become so depressed post-halving that a lot of the industry is just looking for the cheapest power and has become somewhat agnostic to, the ultimate carbon percentage of that. I think it's kind of a step change in the industry, and it just shows how bad mining economics have become.
And realistically, we are one of the few companies now that can be pretty blunt and upfront that mining is not a very good business right now. Fundamentally, it doesn't make a lot of sense to invest material sums right now. Hash price is at virtually an all-time low. Very difficult to invest in new mining equipment right now and pay off that equipment in a reasonable timeframe at current economics. Fortunately, we installed a new durable revenue stream in high-performance computing, which we'll get into as we go along. These are just a quick stats about the business. Draw your attention to the right. So our high-performance computing business, we announced that in October of 2023 .
We announced it with a firm client, which is a differentiation within our sector 'cause a lot of Bitcoin miners have begun discussing HPC, but more in an aspirational sense. More in the sense of we have access to power, and potentially we could monetize that. We went a different path, actually announcing a business with a firm customer. One of the biggest differences between mining and high-performance computing is HPC is a customer-facing business. Miners have never had to acquire a customer. It's a completely different skill. Every other core competency between the two businesses are generally aligned in terms of procuring very specialized machinery, putting it in a very specialized data center, running the machines. It's just the difference is mining, anyone in this room could effectively do in your garage. HPC requires, like, a much higher level of competence and domain expertise.
But right now, what's contracted from an ARR perspective is $92 million of revenue, $275 million over three years. $50 million of that is already in effect. $42 million represents a contract expansion from a customer. That was originally going to start in right around now, but our customer asked us to pause that so they could potentially upgrade the contract to provide for newer NVIDIA chips, the Blackwells. If you follow NVIDIA, they reported earnings last week. It now seems like this new generation, based on the manufacturing runs, probably available in February. When we have clarity on that, based on the customer, the contract's likely to be upgraded, the margins will likely be higher, and the returns profile will be a lot greater.
For example, if this $92 million contract were to go in effect based on H100s, and 2,000 H100s are running, but an additional 2,000 H100s, based on how we would finance that, would probably be around $55 million of annualized EBITDA. This is our total liquidity position as of June 30. So two months ago, it was $192 million total, $61 million of cash. As of August 31, I think it is about $215 million, with about $110 million of cash. Which is important because our total market cap right now is about, I think, $400 million when I looked. Over half of our market cap is comprised of a liquidity position.
So an adjusted EV of about $200 million, with about in excess of $50 million of contracted EBITDA and a free call option on a Bitcoin mining business. Oh, and importantly, we have absolutely zero leverage right now, zero debt. We've never taken on an ounce of debt, but it is something that we have expressed a certain openness to in the future, specifically only to finance our HPC business. GPUs, you can finance them because they have long-term contracts where you could actually model out the revenue and cash flow. And then specifically, also possibly to fund some infrastructure expansion, because there's a very vibrant debt market for that. These are a few stats highlighting our second quarter performance. Revenue increased 220%, mainly due to the. That was the first full quarter contribution from our HPC business.
So that contributed $12.5 million revenue, 63% gross margins. That contributed to the over 1,000 basis points of margin expansion on the total gross margin side. This is despite less Bitcoin mined, which generally due to the halving. The halving took place in April. So we had one month of 900 Bitcoin produced a day market, and then as of the end of April, now only 450 Bitcoin are produced every day. So that's what the market is competing for. On the EBITDA and EPS, like, on a highlight basis, the numbers didn't screen very well, but that's entirely driven by a like $12 million unrealized gain on our digital asset position.
So now, under FASB accounting standards, we mark-to-market our digital asset position as of the last day of the quarter, and if Bitcoin and Ether are down, it results in a large unrealized loss. So it's a bit of an accounting nuance to be aware of. This represents the strategic priorities we laid out at the beginning of the year. The first one was that we planned on expanding our mining fleet from 3 exahash to 6 exahash. We recently walked back this target and said we're unlikely to reach that. And that was universally lauded by institutional investors. I think it's important to put that up there 'cause across our sector, I think there is a pretty good history of overpromising and under delivering, but in this particular instance, not investing in mining will, at current economics, yield better returns.
So for us right now, all growth CapEx near term is likely to be devoted to HPC. Returns profile is vastly superior. Really, like, from the mining perspective, we're comfortable high-grading the fleet, reinvesting just to make the fleet more efficient, lower our production cost, but we really see long-term value creation investing in this HPC business, and we'll get into that further. We had a target to reach a $100 million run rate for HPC by year-end, which, if you think about it from, like, a startup perspective, if you go from zero ARR to a $100 million in a year, at high profit margins, like, that business gets a very high level of VC funding. Like, that becomes a unicorn overnight. And importantly, we're at $50 million.
We had 42 contracted that might slip into 2025, but we said, even if it does slip into 2025, our pipeline is so robust with new customers that we're confident that we'll still hit that by year-end. So it's a pretty bold proclamation. We want to keep a clean balance sheet. We have the caveat is we might take debt at attractive financing levels eventually to fund GPU growth or infrastructure growth. We'll never take it for mining growth. There's zero market for ASIC-backed loans or loans backed by Bitcoin mining equipment anymore. Like, if you were to get debt for that, it would probably be 35% interest rate. Companies that did that in the past all went bankrupt.
One differentiator of our company is that we are one of the lone Bitcoin miners that converts a substantial portion of our mined Bitcoin into ETH, and we then stake that ETH. That generates about a low single-digit return. For us, we're bullish on ETH, we're bullish on Bitcoin, but we think a passive income stream that we could then reinvest into a capital-intensive mining business makes more sense than just holding Bitcoin and hoping for price appreciation and, you know, diluting on the back end to fund that sort of long-term crypto pipe dream. This just shows where our mining facilities and HPC facilities are located. We've always strongly believed in geographic diversification. Our mining facilities are in Canada, Texas, New York, Kentucky, and Iceland, and our HPC data center that we're occupied is in Iceland.
We're running about, right now, only 2 MW for HPC, which contributed $12.5 million of revenue in the second quarter. Mining is about 80 MW , and that contributed $16 million. So that just gives you an idea of the energy intensity and revenue per megawatt of HPC versus mining.
Say that one again. Repeat that one again.
For HPC, we are running 2 MW and change of electricity. Mining is north of 80 MW. For that 80 MW, that generated $16 million of mining revenue. For the 2 MW on HPC, it was $12.5 million in revenue.
Big difference.
Yes, it's substantial.
Plus margin.
Yes, plus extremely better margins. Our HPC margins were 63%, the mining margins were 37%. So from a gross profit, EBITDA perspective, it's night and day, and it paints a very. You start to understand that why we would publicly say we're gonna invest in HPC over mining, and it just makes a lot more sense. This is just as I discussed earlier, we call this the Bit Digital flywheel. We think converting a portion of our Bitcoin into ETH, staking that ETH, and then generating returns from the staking. The staking rewards provides a passive income stream that we can then reinvest into mining hardware, general corporate purposes, even fund an HPC business.
It just becomes a passive, continuous, self-fulfilling revenue stream, as opposed to if you hold Bitcoin and, you know, if it goes up or down, you eventually just have to sell that. You don't have to sell the ETH, you just generate. It's like owning a treasury versus cash. So this is how we originally branded our HPC business, Bit Digital AI. That's probably gonna go away, the AI portion, 'cause we're now exposed to non-AI end markets, but it's kind of a nifty branding for now.
What do you mean by both the non-AI and HPC?
Specifically, we started the business, and our first customer, we're training their large language model, where we're providing the computational power. Our second customer, Boosteroid, which we'll get into, is a cloud gaming company, so not related to AI at all. This is just a brief overview of our HPC business. Really, what it is right now is GPU as a service. We procure high-end GPUs, NVIDIA chips, basically design the network requirements for every customer, and it's a very highly customized offering. It's far from cookie-cutter, so, like, there is a lot of iteration, understanding a customer's needs, designing the exact specifications, the hardware, so it maximizes their ROI and our ROI. Because of that, the, like, life cycle of meeting a customer to inking an MSA, signing a purchase agreement, it can take a few months.
But we are solving for that, which I'll get to later. This highlights a few of our key technology partners. It's important to note that NVIDIA just doesn't sell GPUs to anyone. Like, they have a very extensive KYC process. We have a long-standing relationship with NVIDIA 'cause we used to be an ETH miner. About five years ago, NVIDIA, about, I don't know, 15% of their total revenue was comprised from selling GPUs to ETH miners. That's 0% now 'cause you don't mine ETH anymore. ETH is proof of stake, but that used to be a major revenue stream, so, and we used to mine ETH, so NVIDIA is well aware of us. We had long-standing relationships.
That sort of allowed us to, I don't know, not cut the line necessarily, but bypass a lot of the standard KYC requirements to get NVIDIA GPUs at a time where they were very tight. So our initial deployment in Iceland, like from announcing an MSA to installing a 2,000 GPU network in a super remote data center with miles and miles of highly advanced cabling, that was an eight-week process. We won this contract from a hyperscaler, a Google or a Microsoft, and that company told this customer, like, "We can't do that in that timeline." We went in, we did it in that timeline, and it was our first time ever doing this business, and since then, our customer said that we have beat the hyperscaler on every efficiency metric.
And, like, in this industry, a lot of companies are posturing, in mining, posturing that they can go out and start a business like this, mainly from the basis of we have access to power, we can retrofit this and do that. We deserve a multiple. It's a much easier said than done business. And I think if you look at the amount of words said across the industry versus the amount of revenue they've generated, like on that graph, we would beat everyone. I mean, CoreWeave maybe now, or sorry, Core Scientific.
No revenue until next year, maybe.
Well.
Be a function of the Blackwell.
At least they have a firm contract. Most don't. And I think it's just an important distinction to make. So right now, our GPU-as-a-service business, we can work with customers that have existing data centers or data center relationships, or we can go out and contract data centers on their behalf. It's a flexible model. And what this slide does is, the main retort right now against our business is that we don't own a Tier 3 data center. We don't own a specialized HPC data center, and we're keenly aware that that is an argument. And in the history of our company, we've always been we've publicly disavowed buying a mining site 'cause we didn't think there was a very good ROI on that.
But we were very interested in owning a Tier 3 data center and acquiring a team that can help us develop new megawatts for HPC. By owning our own Tier 3 data center, it would allow us to expand our offerings. It would allow us to unveil a colocation model similar to what Core Scientific does for CoreWeave, but it would also allow us to offer on-demand computing. So right now, we have a three-year contract for H100s. At the end of that, you're unlikely to sign those for a multi-year contract, but they still will have two years of operational life left.
If we owned our own data center, that would be an opportunity to put them there, offer on top, on-demand computing, and gain, like, a significant amount of residual value, as opposed to trying to sell those on the secondary market, where there still would be residual value, but it would just make us more flexible. It is something we're keenly aware of, and I would look at that as a catalyst when we eventually solve that.
Where is it? Can you talk about where it is on your priority level?
It's exceptionally high. So I think if you look at our most recent quarter, you can dig into our filings, and we didn't spend virtually any CapEx, but we raised some money, and it's hard for us to spell out why we raised the money, but there's a reason we're raising money. We have $200 million of liquidity. There's no reason, and, you know, our share price is low, but we want. There's something that we're going to spend this on, and then GPUs is one thing, but we are very keenly interested in solving this problem immediately, but an M&A process could take weeks, months.
It's hard to provide any tangible timeline on that, but it is a top, like, one of the number one priorities for us 'cause we think that would really solve a lot of our, I don't know, why our valuation pales in comparison to certain mining peers that don't have an HPC business, but they have certain megawatts that could theoretically be transformed into an HPC-capable data center. I mean, I've hit on a lot of this. What's worth noting is our first customer, at their behest, we weren't allowed to name them, which, you know, is unfortunate from a marketing perspective, but definitely helps us from a competitive perspective.
Like, certain of our mining peers have announced our customers, and I can tell you that, like, when we find out that customer, like, we have people contacting that customer and trying to undercut them or win that business. So by having that in stealth mode, we don't have to face that competition, which is pretty nice. Another nuance of this is, so our initial 2,000 GPU installation for this customer, it's $50 million of revenue, 63% gross margins, but the margins would be substantially higher if we didn't execute a sale-leaseback on 750 of the GPUs.
We did that to sort of defray the CapEx, but also not to invest too much money in a single customer, especially 'cause we knew that they had a lot of demand on the back end, which ultimately came to fruition with them doubling the size of that to 2,000 GPUs. So right now, I think a rule of thumb is like a thousand GPUs, a thousand H100 might cost $30 million plus $5 million for InfiniBand, the networking equipment. So it's pretty capital intensive. If we didn't execute any sale-lease backs, that $92 million of revenue could be $85 million of EBITDA, $80 million of EBITDA. There's virtually no overhead in this business yet. It's all gross margin, and energy is high single digits of revenue. So this is a new deal, the second deal we announced about two weeks ago.
It's a company called Boosteroid. They're a cloud game, a cloud gaming company, the third largest in the world behind NVIDIA and Microsoft. The total term sheet provides for 50,000 GPUs. These are specialized gaming GPUs. They're not H100s. They're not used to training LLMs, so, like, the cost is much lower on a per unit basis than a NVIDIA H100 or B200. But if the entire deployment of this contract is executed, and it won't be immediately, there'll be a growth ramp. It represents a $700 million revenue opportunity over five years. Very high margins. So that's about a $140 million a year at full deployment. We've said the starting quantity this year will be an initial $3 million, $2.6 million revenue contribution or annualized.
Part of that is that they are also waiting for a newer generation AMD chip. That'll be out in the first quarter of 2025, and then we expect a larger purchase order. But based on what we've discussed with them, we think probably 30% of that deployment takes place over the course of 2025, and then scales probably linearly through 2026 and 2027. But this is a company that, they already have 5.7 million users, growing. Each user pays $7 a month. They just struck a deal with Mercedes, so now every new Mercedes car will have the Boosteroid app installed in it. I don't think you'll be able to drive and play games on the app, but, like, for kids in the backseat, it's a home run.
But ultimately, every new Mercedes car will have technology that's powered by Bit Digital GPUs, which is a fun fact for us. And so these are just a few highlights of this business in general. I mean, ultimately, we are downstream of NVIDIA. Like, we are the base layer, we are the end user, or we are the end reseller. So NVIDIA sells to OEMs, OEMs sell to us, and then we sell to the end customer. That's where we fit in the value chain. So, I mean, it's an absolutely, you know, mission-critical component of that value chain. And over, I don't know, in the second half of 2023, NVIDIA obviously went parabolic. Superm icro followed, like everything in the value chain. CoreWeave, the largest private, went from a $7 billion valuation in September to $19 billion in February.
We're one of the companies that hasn't got that bid. Many reasons. Part of it is just having a mining business. Part of it's being undercovered and underfollowed. But it's something that I think a lot of what we're gonna do in the back half of this year will solve for. It's like right now, we're trading 3x-4x contracted EBITDA on the HPC side with a free call option on mining. We're not gonna invest heavily into mining unless the economics make sense, and they're more attractive than HPC, and I don't see that happening near term. So we're a company that's getting off the capital-intensive Bitcoin mining flywheel. Like, and we will be candid that mining economics aren't great, but now we have a business where the returns profile is amongst the best I've ever seen.
The CAGR of GPU as a service is probably gonna grow at 30% over the next 10 years, and then the infrastructure layer, the software layer on top of that, is an even greater opportunity. And so we've been doing this business with just our internal staff. We just made our first key hire. We're about to make a lot of key technical hires. I don't think there's a tremendous moat beyond execution right now, but I think there will be when we add some technical upstream software layers, and that's where we're going. We're gonna own our own data center. We're gonna have a diversity of customers. I think we're gonna solve pretty much every nit on the stock over the coming months.
This has been great. We are at time, but is there a best way to get in contact with you and your team?
Yeah. cam@bit-digital.com is probably the best way.
Got it. And then I do want to take maybe one question just before we wrap, if anyone has anything. Otherwise, we will wrap. Kevin?
You, sir.
Talk about your business development side of HPC. How did you land your anchor customer? How did you land Boostronic?
Boosteroid, mind you.
Boosteroid, right. And how did you or how do you feel so confident about your, your pipeline?
Well, both of.
Especially since you don't really have a sales team.
Both of those deals were struck with our own team, the team we had used for mining. Like relationships, introductions, friend of a friend, like getting in front of someone, being able to convince them that us, as an unproven company, can handle this contract. We also look in sort of unconventional places. We're not calling up Anthropic. We're not doing the same thing that every other North America-based HPC company is doing. We have offices in Singapore, Hong Kong, Iceland. So, like Boosteroid, they're a European company. They're a Ukrainian company. Now they're headquartered in Austin, but that was just a relationship that someone was able to strike up organically. But now we've actually started hiring real, dedicated staff to run that process.
People that have come over from some of the largest private companies that already have a book of business, already have a history of building a business from zero to 1 billion of ARR, and that takes our pipeline from, you know, five leads that we've scraped together as an existing team to multitudes of that.
Is the plan to build out a sales team eventually?
Yeah, no. So we made our first hire, last month. Second dedicated sales guy we just made. We're gonna hire a CTO. Yeah, we're gonna invest pretty heavily into making this a very legitimate business.
Got it. Okay, that's all we have time for.
Thank you.
Thank you, Cam.