Good afternoon. Thanks, everyone, for dialing in to another session as part of our Needham's 4th annual crypto conference. I'm John Todaro, Senior Equity Analyst at Needham, covering the sector. For those who are aware and those who aren't, we just initiated coverage on Core Scientific, ticker CORZ. I believe we got that done yesterday, so that's out there. With that, Adam, I wanna thank their CEO, Adam Sullivan, for joining us. Thanks so much for being here.
Thank you for having us on, John. Appreciate it.
Of course. A lot of questions to get through, a lot of exciting ones. So would love to kind of get into it. So I have certainly a lot of HPC questions, some mining ones as well, but maybe just in your words, give us a sense of, you know, the progress Core Scientific has made over the last year, and kind of how that HPC opportunity came up, and just frame that within context of the what's. It's actually quite a large Bitcoin mining segment you have as well.
Yeah, so I think it's helpful to take it back to the company's origins. You know, being the first institutional Bitcoin miner in North America, we were trying to figure out how to solve that problem. So brought in a lot of traditional data center folks. We were on path to build almost two gigawatts of Bitcoin mining infrastructure, something that had never been done before in North America, until we went through our Chapter 11 process, beginning December of 2022, finishing up in January of 2024, so just about eight months ago. Now, the history of our company was that it was built by traditional data center folks. It was led by, many folks from the traditional data center industry.
And when we were evaluating the opportunity, looking at 2024, while we were sitting in our Chapter 11 process, we were looking back at kind of the previous iterations of Core Scientific, and one of the things that people don't realize is we are actually one of the largest hosters of GPUs in North America from 2019 to 2022, when we were running GPUs for the Ethereum proof, when it was proof of work network. So right as we emerged from our Chapter 11 process, we signed, just a few weeks after that, a contract with CoreWeave. It was an immediate need that they had for 16 MW. So we leased a facility, and right after that, we began discussions on a much broader opportunity.
We had publicly said that, you know, we have a certain number of megawatts that are convertible to HPC, and that's because our facilities were designed as powered shells, that we essentially were utilizing for mining Bitcoin until the opportunity existed for us to convert those to another form of compute, and what we're seeing today is just, you know, obviously an extraordinarily high amount of demand for, AI and really data centers that are focused on the next generation of GPU compute, and we've been able to fulfill that demand. Of the 700 MW of total contracted power that we've dedicated to HPC, of the total 1.2 gigawatts, we've signed up 382 of those.
Now, one thing, one little nuance there is, of the 700 , it's five 500 MW of IT load, and so we've signed up 382 MW of IT load, which actually totals about 550 MW of total power. So we've put ourselves in a very strong position. Since we announced that first deal on June 3rd, we've actually been able to clean up our balance sheet. We raised a convertible note that helped fully clean up our balance sheet of all of the previous Chapter 11 paper that we had. And now we're a brand-new company sitting here in September.
That's great. Thank you for that, for that overview. On the contract you mentioned with CoreWeave, where you got to the 382 and that critical net IT load-
Mm.
Is there a deadline to upsize that contract to any additional megawatts?
Yeah, so there's 118 MW of additional potential contracted power inside of our existing portfolio that's still under negotiation with CoreWeave.
Okay. And so that's kind of an ongoing process, and there's no kind of, at least publicly disclosed, deadline on that, of when that-
Yeah, it's just something that we're. You know, it's building 382 MW and ensuring that, you know, all of those timelines are nailed down, is something that's incredibly important. And so, you know, we're now refocusing back on, you know, ensuring that we can deliver on the next 118. So that's working through supply chains, working through designs, working through all the necessary components to actually get to definitive documentation.
Got it. Understood. So you guys have one thing that none of the other miners in HPC has yet, and that's a real agreement, an official agreement. So talk to us a little bit about how long did that CoreWeave agreement take? And then on the back of that, what has been kind of, you know, your... I guess, what have you learned on the back of that that's gonna help you moving forward? Is there some almost operational know-how that you're picking up from this that gives you an advantage over competitors?
Yeah, so we began really in earnest on the conversions of the bigger contracts, shortly following, actually, it was actually following the completion of our delivery timelines for the 16 MW facility. So, that facility, we fully delivered in under 30 days, and that was actually a retrofit of a previous generation data center, to fit some new gen GPUs inside. And that's when really the conversations began in earnest for that next 500 MW slug. And so, that was signed in February, announced in early March. And John, what was the second part of your question?
What have you learned from that agreement?
Oh, yeah.
Just the operational aspect as you're starting to retrofit those sites, that maybe gives you some operational know-how in the future versus competitors.
Yeah, I think the big one for us is the fact that, you know, who we're designing this data center for is just it has specifications that are needed and requirements that are needed for the newest generation of GPUs. And so we've been working very closely with the design engineers in NVIDIA, working very closely with CoreWeave, to actually figure out what are the exact product or data center specifications that are needed. And so when we're talking to many of these other companies that are looking at deploying these large GPU deployments. We understand the product specifications that are necessary from working with the folks at NVIDIA and CoreWeave, so that we can help design a facility that actually meets the specifications of many other clients.
And the big part about this here is many people no one's deployed this many GPUs before at scale, and running these newest generation GPUs that represent power densities that really haven't been seen before in the market. I mean, data centers historically were built at most at, like, 20 kW the rack. Now we're seeing 65 kW the rack. Next generation's gonna be 130. And so just as you look at this power density increase and how you design a facility for that, most folks have never designed a facility like this before. Now, for us, this is something where having a traditional data center design team that's also worked on power densities much greater than that, many multiples of that, for Bitcoin mining, gives them very unique insight into actually how to design that facility around the specifications that they're giving us.
Thank you for that. That's helpful. The CoreWeave deal is a little bit unique, where they're putting up a majority of the CapEx. Given that the debt market seemed to be getting a little bit more robust around Bitcoin miners building an HPC site, talk to us about the trade-offs between going down a route where you provide, you know, you get the financing on the back of an agreement. Maybe there's better economics to it, so maybe it's closer to $2 million per megawatt-
Mm-hmm.
But I would love to get your answer on it, versus customer providing the CapEx, and maybe that top line economic number is not as good. Walk us through a little bit, the dynamic and how you would think about it, and maybe down the road, does it still make sense to have a customer put up the CapEx, or would you look down that debt financing route?
Yeah, so if you look at a traditional data center company, it's about 70%-80% loan-to-cost, so that's the debt portion, and then 20%-30% equity check. The unique part about our deal with CoreWeave is essentially they're paying for what would've been the debt check themselves. We don't have to pay that back. And then the equity check they put up front, and then we pay it back via a credit that we credit against their bill that will take about two years, no greater than 50% of revenue, but it'll take about two, just over two years to pay that back. And so very unique structure, obviously very high multiple on invested capital, right in the high single digits over the first 10 years of the contract.
And then as we think about this going forward, there is a mix of CapEx being paid for by the client, and as a trade-off versus going down the traditional financing route. But you have to remember, that does buy down the rate. The more CapEx that's paid for by the customer does buy down the rate. So you're absolutely right on that point. And so we're evaluating all the different opportunities in front of us from a debt perspective, because over time, this will continue to move towards it being a traditional data center company.
What we're expecting to do is, you know, start actually utilizing the flywheel that other data centers companies do, which is, once these contracts get signed, projects get announced, we'll actually be able to go to the debt markets. You know, you'll be able to tap on the development side. And then also, as you start to cash flow contracts, you can actually pull some of that capital out of those contracts in order to fund the growth going forward. But those rates are much lower than the rates that you'd see on the development side. So you're looking at, like, if you're, if you have a cash flowing contract, you should be looking at, like, SOFR + 300-400, something much lower than what's normally seen in the market.
That's interesting. That's helpful color. And the CoreWeave deal, you know, before it came out, were you guys talking to major public hyperscalers that had this interest, too? And I guess just how many CoreWeaves would you say are out there that would provide the CapEx?
Yeah, so we talked about previously on our Q1 earnings call that we were in negotiations with just over a handful of clients, and at the time, we had a 100 MW minimum contract size, and the customer had to pay for the CapEx. And so, you know, that really eliminates a lot of folks, 'cause if they're going after a 100 MW deal, that means it's at least $500 million-$800 million on the build-out costs, and then they're paying another $2.5 billion for GPUs. So generally speaking, about three, at least a $3 billion investment that these companies would be making. So these were very large technology companies that were really the main ones that were interested in taking down a deal of that size.
Now, that has continued to evolve over time, because working with those folks made us realize that they don't all need 100 MW at every site that they go to. They also need sites that might be 25 MW or 50MW or 75 MW. And so our conversations have moved more towards that direction for certain clients on the smaller contract size, but with opportunities to expand, 'cause once you get customers in the door, it's a very sticky business. And so we view that as a really strong opportunity, where we can turn on Megawatts much more quickly than some of our peers can, and then expand that site from maybe 25 MW to maybe up to 75 MW over the course of the next 12 to 18 months.
And that's a really interesting opportunity for us and for a number of the clients who might have an immediate need for a smaller number of megawatts in a certain geographic location, but want to continue to have the opportunity to expand at that facility as well.
That's helpful. And has retrofitting started at some of the sites right now towards HPC?
Yeah. Yeah, so we've talked about in some of our previous operational updates, we talked about it a little bit more in this one that went out this morning. But we are turning off machines right now and shipping those to our dedicated Bitcoin mining facilities in order to actually allow for the selective demolition process to begin at a number of our sites. So, we've publicly said that we have five sites that are or five facilities that are currently contracted with CoreWeave, and so we've begun the process at many of those sites to actually begin the process of converting to HPC.
. Got it. Okay. And so that leaves you how many sites for mining?
Three.
Three for mining. And so should we expect kind of hash to come down, at least in the near term, as this takes place?
Yeah, but we're also developing another 100 MW at our Pecos facility. So we are trying to mitigate the amount of downtime that we have, so that's why we've staged the developments the way we have. But we, you know, we completed 72 MW at our Denton site. We're going to complete another 100 MW at our Pecos location. So we are still continuing to develop Bitcoin mining infrastructure to mitigate that downtime. So we went from over 800 MW dedicated to Bitcoin mining, of which about 600 were for self mining. And on a look-forward basis, our portfolio is gonna be about 500 self mining, and we're gonna have no Bitcoin hosting business anymore.
No hosting. And when would the hosting roll off?
2025. So it's gonna roll off, much of it's gonna roll off over the course of 2024 . There might be some small contracts that may exist into the beginning of 2025 .
Got it. Okay. Are there new machines being added to the fleet as well that would give some efficiency gains?
Yeah, so mainly the efficiency gains are gonna come through the new Block chips that we've ordered. So we've placed an order for 15 exahash of the Block chips. You know, that contract is mainly around our vision for what the future of Bitcoin mining's gonna look like. Historically, 20%-25% of all revenues have had to go into maintenance CapEx just to maintain profitability on existing machines. We think that's too high on a run rate basis, and being able to develop a facility around a chip instead of around a shoebox, we believe is gonna be the future of mining.
Because not only will we be able to save on a dollar per terahash basis for this first deployment of these chips, but on a look-forward basis, we're gonna be able to save even more dollars on a dollar per terahash basis, which we view as a long-term competitive advantage for our business. But we also believe that more companies, more large-scale mining companies, will continue to move in that direction as well.
Is that something that only Block was offering, where that final assembly you guys are doing, or could you strike that arrangement with a Bitmain, a Canaan, et cetera?
Not necessarily a Bitmain at this time, but there are many other chip design firms that are working on being able to deliver just chips. However, they will also sell full machines as well to certain clients who'd like that type of machine.
In your view, what you can say now, Jack Dorsey's spoken very highly of his chips. Are you seeing that advantage gains? First off, have those been delivered yet, or is that still coming?
Not delivered yet, no.
Okay, so I guess almost time will tell on that one, but I just broadly would be curious on so far if those chips if you thought they're quite attractive out there versus the ones that, you know, Bitmain has with the final assembly, though.
Yeah, I mean, we wouldn't have signed a contract if we didn't believe, you know, their designs and where they're at in their process and the specifications that we were looking at were not attractive compared to where we were expecting to see, you know, the new Bitmain S21 come out, and where we expect machines to be efficiency-wise in the first half of 2025 .
Got it. Understood. Talk to us a little bit about the risks you see with the CoreWeave deal. 'Cause one of the questions we get from investors is, "Look, you guys are first to it, and it's a pretty good deal, but if X, Y, Z miner comes out and has a Google, Amazon, Microsoft, do you start to get some risks that, hey, a lot of the CapEx is getting funded, and there's reliance on this one company, and they are private," even, you know, they're decent size. They're what, $19 billion valuation. But still, is there too much concentration in one company as a private versus someone else that maybe comes out with a Microsoft, Google, Amazon-type deal?
Yeah. There's two things. One, I think CoreWeave going public is gonna be a big catalyst for our company because people will recognize their creditworthiness. So I actually view that. You know, they, I believe it was Bloomberg said that they are looking at a Q1 IPO. I actually think that's a big catalyst for our company, when people actually recognize the scale and size that CoreWeave has achieved. So, you know, that's really the first one. In terms of if another company or another mining company necessarily comes out with a, you know, Google style type client, you know, I think that that's great.
I think it will further catalyze the thesis that Bitcoin mining companies have the necessary infrastructure to continue to supply data center capacity to this industry, and so I don't view it as a net negative. Also, we announced our first deal about 93 days ago.
Mm.
Client concentration, absolutely it's a concern today. We've actually identified that as a twenty twenty-four catalyst to diversify our client base and acquire additional sites. You know, give us a little time.
Yeah, no, that's great. That's helpful. As you think about, I guess I do wanna get into the pipeline, 'cause that's an interesting piece, and there's maybe not time to share. So, well, I guess, a lot of the peers are coming out, and they're announcing, you know, one gigawatt-plus pipelines. Obviously the near-term focus is on retrofitting those sites, getting them up and running. But talk to us a little bit about your pipeline, and then within that, I would wanna, like, to drill down whether you're looking at Texas or outside of Texas. We've heard good things about MISO, PJM, so just, what are the thoughts about the different power markets, especially for HPC?
Yeah, so we have 550 MW in Texas today. There are some opportunities that we're seeing in ERCOT. I would say our focus is generally in regulated utility markets, so we're mainly targeting tier two-type markets with low latency, low, you know, single-digit millisecond latency to a major metropolitan market. We're finding more power availability. Our pipeline, in terms of what we've vetted so far and been able to tier that out into, you know, most actionable, cheapest land along, you know, low single, single-digit latency. You know, we've evaluated many, many gigawatts of sites over the course of the past, you know, three to four months.
We've been working very closely with utilities across the United States, some of the ones that are in existing markets, some new utilities that we've been in communications with for many years, mainly on the Bitcoin mining side. So, we feel very strongly that the pipeline in front of us is robust and actionable. So we're not necessarily putting any numbers on it today, but, it's something that, you know, over time, the market will be able to see as we begin to sign clients and, alongside of new site developments.
Is there any risk to Texas versus some of those other markets? You know, is... So obviously, CoreWeave, I imagine, like a public hyperscaler, can't have any downtime. Does ERCOT say, "Hey, look, you're not gonna curtail, this is too many megawatts for that"? Any risk to that?
No, 'cause, first of all, all of the power contracts are passed through to the client. Second, in ERCOT, well-worn path, Dallas is a massive data center market. Fixed-price PPAs are 24/7 power, so not really a tremendous amount of risk, and you're looking somewhere, you know, $0.08-$0.09 range, which is well within the range of what data center companies are completely comfortable paying across the United States.
Oh, that's interesting. Okay, so those Dallas area sites, $0.08-$0.09 power. Okay. Okay, great. Let's turn a little bit over to the Bitcoin mining segment, and then we can do questions for the last 10 minutes or so, but we still got about 10 minutes to do some Bitcoin mining.
I guess, so you got some hash, as we talked about, rolling off. If you could just remind us again, walk us through hash rate today, where you see it ending 2025, and then if there is any capacity down the road you see for 2026.
Yeah, so we're at about 19 and a half exahash right now. Our year-end target's 21.8 exahash. Some of that has to do with miner migrations and machines that are coming offline. We haven't given forward guidance for 2025 in terms of where we expect to be, but we'll have a very good opportunity with the 15 exahash of Block chips that we've ordered to be able to increase hash rate even within a footprint that, from a Megawatts perspective on the self-mining side, is shrinking, and that's mainly due to greater joules per terahash and greater terahash per machine. It helps us fit a lot more terahash in a smaller footprint.
Understood. How do you think about the trade-off between the, the CapEx towards new machines, new chips, but also obviously lower cost, better margin, on the mining side?
Sorry, John, can you just repeat that question?
Yeah, just the trade-offs between the CapEx dollars towards new machines, new designs that are more efficient, but obviously there's an additional CapEx spend, and your Bitcoin mining costs, where your break-evens are gonna improve on the more efficient machines, but there are obviously there's just more upfront capital you have to put up.
Yeah, I mean, this is something that we've been evaluating, which is there is extreme disconnects between the existing or the newest-generation machine and the T-minus-one generation machine. That disconnect has grown post-halving. You know, you're seeing XPs trade in the single dollars per terahash versus S21s that are trading in the high teens. Now the opportunity there, and for a company like ours that has developed our own energy management, our own fleet management, and our own firmware, where we can much more dynamically tune machines, means that it's probably better, even in a down market, for us to be able to purchase T-minus-one generation machines.
So our view is that, yes, from a, if you're a very large Bitcoin miner that has a bunch of sites that are five- or six-cent power, yeah, you're gonna have to go after the newer generation machines. You're gonna have to pay up for those machines. The profitability of that machine may only last two to three years at a site. Our goal is to buy machines that we believe will be able to return... 'Cause this is a return on invested capital business, which people should be focused on. It's not necessarily a focus of most mining companies.
But we actually wanna be able to produce a return and produce a very good ROI on that machine, which means that oftentimes targeting some of these previous-generation units is actually a much better value opportunity for us that can still produce very high returns in an up market, and we can tune that machine accordingly in a down market, and that's something that's very unique to Core Scientific, and that's the reason why you don't see us going out and making very big, splashy S21 purchases, because from a returns perspective, it just doesn't make as much sense.
That's fair. That's helpful. Thanks for that. You know, you mentioned, or it's even kind of just within your own portfolio you can see this, but so as certain sites come online for HPC, mining gets moved to, I would imagine, more remote locations, cheaper power, but you don't need some infrastructure components, like fiber and water access.
Mm-hmm.
I believe all your sites are already outfitted with fiber. But do you see a world where Bitcoin mining almost gets moved to the fringes, eventually maybe even outside the U.S., where available capacity, if there's the fiber infrastructure, that goes towards HPC?
... Yeah, I mean, I think this is a broader term trend. I mean, think about where Bitcoin mining was seven years ago. It essentially does not exist in North America, and then you saw the migration to kind of smaller sites, five, 10 MW, 15 MW sites. Then was the hype, and then there was the mega site trend. I think the mega site trend in the United States has kind of played its, you know, run its course. I think on the Bitcoin mining side, people are gonna be evaluating much smaller sites, where you'll find economies of scale in the number of sites you're operating, not the size of a single site. And so that's gonna begin to push people towards smaller sites in North America, and then you're gonna begin to see a lot more migration internationally also on the smaller site scale.
So that's something that I think we're at the early innings of today, but something that if you look forward seven years from now, the Bitcoin mining network is gonna be even further decentralized from a site-by-site basis.
That's interesting. Do you see a world where hash rate actually starts coming down? I mean, obviously, the machines are more efficient, so that's gonna add to greater hash. But if more resources are getting allocated to HPC, does hash rate start thinking of a ceiling?
Not gonna. It's not gonna put a dent in it. Hash rate is on a rapid move higher. You know, we're at a $0.04 hash price, and we saw difficulty increase.
Yeah.
It's not gonna slow down. As the break-even hash price for the newest generation units continues to move lower, what we're looking at is just continued growth and difficulty and continued growth in network hash rate. There's not gonna be a period of time... We're never gonna see, you know, a China shutdown move any more from the perspective of, "Oh, there's too many megawatts turning on for HPC and moving away from Bitcoin mining." It's just not gonna happen. Bitcoin will always seek out low cost of abundant power, and where that exists, Bitcoin mining will find it and continue to turn on new hash rate.
In the U.S., would you say there is, though, say, it almost seems like there would be an equilibrium where those margins would even out in the U.S., where you go, "I'm gonna allocate dollars towards Bitcoin mining or HPC," since it seems somewhat available to do, right?
You're saying the margin between-
Between, yeah, your margin between Bitcoin mining business or going down the HPC route. But, so I guess, in other words, would there be... You know, my initial thinking was hash rate would start to go down because, you know, if the HPC margins are so much better than Bitcoin mining margins, everyone starts going towards HPC, that comes down, they almost even out, and you reach an equilibrium. But I guess to your point, it would just move outside the U.S., it would find the cheapest power markets out there.
Machines always find a home. Very rare, unless they're very old generation units, do they not find a new home. So that hash rate really, truly never goes away after it shuts off, let's say, at a mega site in the United States, it might be converting. It's not like that machine just goes into a pile and never gets turned on again. That machine will find a new home somewhere.
On the flip side of that, is there a demand to getting HPC sites outside of the U.S.? So with AI, there's a fair degree of regulation where EU-domiciled companies maybe aren't able to train their LLMs at data centers in the U.S. Is there internal discussion from you guys that maybe it makes sense to spin up an HPC center in the EU, U.K., anything like that, down the road?
Yeah, absolutely. I mean, we're looking aggressively at a number of opportunities internationally right now. We're in discussions around, you know, certain sites that we think are very interesting from a growth perspective. Even from some of our potential clients that we're in discussions with right now, they're looking at the U.S., but they also have needs abroad as well. And so we're looking at a number of different countries. You know, EU is definitely at the forefront of that. Obviously, on the much smaller scale, the EU isn't a place that's known for a high amount of power availability, but it also represents an opportunity to go build out, you know, maybe a 15 MW-20 MW facility, that could service clients in a market that's very hard to find power.
We're looking at some other countries as well, that we've been approached by, either companies that are there already looking for data center partners, or from governments. Governments have also been very aggressive at trying to figure out ways to attract data center companies to their country, and we think there are a number of opportunities that exist today that could be very actionable for us to capture, a new set of clients that aren't U.S.-based.
That's great. That's, that's super helpful. We have about 10 minutes here. So let's turn it over and take a couple questions from the audience. Seen a few come in here. First one asks, "Core, does CoreWeave have the capital set aside today? What gives you confidence that they can deliver on the CapEx needed to retrofit those sites?
Yeah, the capital is set aside, and then part of the capital is already in, the $1.5 million per megawatt's already in an escrow account as well. So, from a capital perspective, we know the capital is there and ready to, execute on the contracts for the CapEx in front of us today.
Up to the 382 MW?
Yeah.
There we go, answered that question. The next one is, "Which one of your peers do you see as either the biggest competitor in HPC or one that you think is, you know, best suited to go down that route?
Of the Bitcoin miners, is that the question?
Or broader. It could be the Equinix's, Digital Realty's, if you think they're going down that route. Applied, if you know, much about them.
Yeah, I mean, we haven't seen in any of the processes that we've gone through, we know who our competitors are. Our competitors are Switch, Digital Realty, QTS, you know, really some of the other larger data center companies. None of the other Bitcoin miners have showed up in any of the processes that we're currently involved in.
That's really interesting. I heard Switch didn't have a ton of capacity. Are those guys still going out and finding sites and able to build that, or now is it where the Bitcoin miners step up 'cause they have the power capacity?
Yeah, I mean, Switch has been very good at capturing a pretty significant amount of the initial tranche of the GPUs that were landing in the United States. You know, they'll continue to build. They have their own growth pipeline that they're executing on. But, you know, Bitcoin miners do represent an opportunity. I think the biggest opportunity or the biggest point for many of the Bitcoin miners is, you know, being able to find the right contractors, 'cause they don't have the necessary teams in-house. We were very lucky to have a traditional data center team already running our operations. We've been able to add a lot of talent, too, since even June 3rd to our team.
Because a lot of folks at the more traditional data center companies feel like their equity story's a little bit played out, meaning they're not gonna see a two X or a three X from current valuations. Whereas they look at Core Scientific, they recognize the value of just our existing CoreWeave contracts and the potential valuation just from this existing contract, let alone them bringing their large-scale data center expertise to our company and having the potential to, you know, get a significant multiple on any equity that they receive.
Interesting. Okay, thank you for that. Next question asks: "Is there any additional operating costs or G&A associated with HPC?
No, we laid it all out on page 12 of our Q2 investor deck, or page 23 of our June 12th investor presentation. It lays out, you know, what our expected margins are on a go-forward basis, for the existing contract with CoreWeave. We also outline all of the associated costs that we'll be in charge of. So some personnel, maintenance, CapEx, security, and some ancillary costs, like certain insurance that is necessary to pay for. On a run-rate basis, after the CapEx credit's fully paid for, we're expecting between 75% and 80% gross margin. That's on a cash basis, by the way. So that excludes G&A and stock-based comp.
Yeah. Yeah, notes. Okay, next question, this is back to the Bitcoin side. Bitcoin's been range-bound. Is there any catalyst you see on the horizon to get it going from here?
I think it's gonna be range-bound until we have better clarity on where the election goes. I think the political risk and where monetary policy is gonna go in whatever new regime we would receive is gonna really dictate where Bitcoin price goes over the kind of the next 12 months. So it's hard to say where it'll go prior to the election. It feels like we're gonna be a bit range-bound, but it feels like we're also gonna potentially experience some higher volatility as we inch closer to the November election. So I think range-bound for now, higher volatility expected just prior to the election, and then we're gonna go one way or the other.
I'm not sure which way it's gonna go, but after the election's over, you know, we'll definitely be going one way pretty strong.
We, we've heard from other folks that they didn't wanna name names, but it seems one side very much would be good, the other side very much would be bad. So you also would hold the sentiment that if it goes a certain way, it would be very bad, or would Bitcoin find a way? Or maybe you could make the case that there might actually become more of an inflationary environment if it goes the other way, so, so maybe Bitcoin, you know, responds to that.
Yeah, I think over the long term, Bitcoin will respond to that. But I think in the short term, people are very concerned about regulators' policy that could infringe upon Bitcoin's freedom in America. So I think that's really the short-term concern over the long-term inflationary effect of very loose monetary policy.
Got it. Next question asks: "Does the 70%-80% cash margin you mentioned for HPC, does that factor in tax?
No, it's on a contribution margin basis. So that number is a contribution margin, which is more like a cash growth margin.
Okay, and then last one, is about, obviously, Bitcoin is, has a fair degree of regulation around it, as we talked about the election. Is there anything that's standing out on AI that could be kind of a black swan that negatively impacts AI and HPC?
I think some of it is, you know, there's limited to no policy around artificial intelligence at the federal level. You know, there's the potential for some sort of regulatory policy that can negatively impact artificial intelligence and training and the type of models that can be developed. I think that's kind of a black swan event that, you know, hasn't been talked about, but it's been floated amongst many of the publications, but we haven't necessarily seen anything come out from a government perspective. Politicians have kind of touched on it a little bit, but at some point, this could come into a lot, a lot more stringent focus here, and that could, you know, it could potentially be negative for certain companies, which could potentially hinder the growth of artificial intelligence.
I think the big part is, though, do you believe the growth in the GPU cloud? If you believe in the growth of the GPU cloud outside of just artificial intelligence, which is kind of in the bucket for what GPU cloud represents, then GPUs will continue to get turned online. There will continue to be use cases for the GPU cloud, which has the potential to grow even larger than the existing cloud infrastructure that's been developed.
Yep. I just got one more. We got three minutes, but.
Yeah.
You know, that reminded me of something. You know, one of the pushbacks you're starting to see now is, it almost reminds me a little bit of blockchain and crypto, where there's all these investment dollars going towards it, but are people really using the product? Do you see parallels to crypto blockchain, where you do see a lot of investment dollars, certainly a lot of excitement, but quite candidly, at this point, you know, they're not applications or uses that you're doing every day. I mean, even with Bitcoin, I guess you could say, well, maybe people are storing it for digital gold purposes. I'm not really paying with stable coins necessarily that much on a regular basis.
Mm.
I do interact with them some. So, you know, I guess, what are your thoughts on AI, kind of what the actual end use case is? Do you see some parallels with crypto blockchain?
You know, we're such early innings of how people are gonna utilize the GPU cloud. Obviously, you know, looking historically at how GPUs were utilized, you know, for rendering, or some other use cases, and you look at it, how it's gonna be utilized going forward for everything from running software, managing protocols, managing, you know, databases, and being able to gather information or extract information from databases. You know, all of these things are... we're at such early innings, where a lot of these applications that are going to be successful outside of just artificial intelligence, these applications are gonna require, you know, low latency inference, compute at a localized level, and that's gonna require a significant amount of GPU build-outs across the United States and across almost every single major metropolitan area.
So I think the applications that are being developed are still in the early stages, but there are large companies that are developing applications that are going to require GPU compute in order to utilize them, and that's something even outside of artificial intelligence that, you know, haven't necessarily been fully factored into what the growth of this industry is gonna look like.
That's great. With that, we are at time. Adam, I know there's a few conferences going on, so I appreciate you being here with us and sharing some thoughts on the story and taking our questions.
I appreciate the time, John. Thank you so much.
Of course, and thanks, everyone, for dialing in.