Core Scientific, Inc. Q3 FY2024 earnings call

• CoreWeave exercised final two options, committing to all 500 megawatts of critical IT load. - Allocated 100 megawatts of infrastructure from Bitcoin mining to HPC, increasing HPC hosting capacity to ~570 megawatts. - Leased with option to buy an Alabama data center with 11 megawatts of critical IT load, potential for 55 more. - Completed 100-megawatt expansion at Pecos, Texas Bitcoin mining data center. - Completed miner migrations and partial demolition at two data centers for HPC hosting. - Successful $460 million convertible note offering to refinance debt, improve capital structure.

Total third quarter revenue was $95.4 million. Digital asset self-mining revenue was $68.1 million, digital asset hosted mining was $16.9 million, and HPC hosting was $10.3 million. Segment gross margin was negative 9% for digital asset self-mining, 29% for digital asset hosting, and 13% for HPC hosting. Power costs were favorable, declining to $0.038 per kilowatt hour from $0.045 in the prior year. Operating expenses totaled $40.3 million, net loss was $455.3 million, and non-GAAP adjusted EBITDA was $10.1 million.

• Expect 16 megawatts critical IT load of revenue generating HPC hosting infrastructure. - Average fleet power price for active bitcoin mining data centers of $0.042 to $0.044 per kilowatt hour. - Contracted all 500 megawatts of critical IT load to CoreWeave, projecting aggregate total potential revenue over 12-year contracts of approximately $8.7 billion.

Q: Hey guys, thanks for taking my question. Congrats on the results here. I have one, and then I'll use one of my follow-ups, if I can. First one being to the 100 megawatts that you're going to allocate away from bitcoin mining to HPC, are we already in conversation with potential customers there? And then I know you can't give specific CapEx, as you mentioned. But should we just still be thinking that $5 million to $8 million to retrofit that site? And then I have my follow-up.

A: Yeah. Thanks, John. Yeah, those conversations actually began earlier in the year, looking at the site in particular. We go back to what we mentioned on the previous earnings call, it really comes down to whether a site be bitcoin mining or HPC, it comes down to really power environmentals and latency. Based on conversations we are having with some potential clients, we went back and solved the power issue that we need to solve in order to convert that site. And so these conversations have been ongoing for quite some time, and it's looking like a competitive process for those 70 megawatts. Sorry, Joe [ph], I'll just comment quickly too on your megawatt question. Yes, we're looking at really that $5 million to $8 million per megawatt for the retrofit.

Q: Got it. Okay. And then just a follow-up. So to the HPC revenue, it looks like it came in a little bit higher than we at least expected. I think if it's just the 16 megawatts I'm seeing, that implies almost $2.5 million per megawatt annualized on a revenue basis. Not sure if we're missing something. I guess, in other words, I think we're thinking more like $5 million to $6 million, and it came in at $10 million for the quarter.

A: Yes. So thanks for the question. I think the -- as we suggested in our prepared remarks, that was really driven by a onetime adjustment associated with the fact that we had actually pulled forward the delivery by 30 days. And so你 are seeing an additional month of revenue, which is why there is a delta in Q3.

Q: Yeah, just a follow-up on the 100-megawatt transition. So in terms of the cost, are you guys -- is your plan to go ahead with construction and then try to sign a lease along the way? Or are you going to sign a lease before you start spending capital on that transition?

A: Yes. Thanks, Jon. Right now, the way we're viewing it is we're hopeful to get to a client and a final contract in short enough time period where we know exactly what they want to build. We know exactly the sizing and the requirements that they're looking for. And so from our perspective, we're not looking to spend capital prior to having that contract in hand.

Q: Got it. And then maybe I'm slightly getting ahead of myself, but I guess my follow-up is what kind of financing structures are you -- would you be looking at for future deals? Is the CoreWeave model repeatable? Or would we see more of a traditional data center financing type structure with a construction loan and you guys spending all the capital upfront?

A: Yes. The way we see it right now and where the market is moving to is clients are coming around to spending some portion of the CapEx to help buy down their rate. We are seeing higher rental rates right now. And so really, the way we're looking at it is that 20% to 30% that's generally spent by data center companies as the equity check, the rest is funded on a debt basis on a project financing level. We're looking at many of these clients looking to really cover that portion of the equity check, and we would look to project finance the rest of the build-out.

Q: Hey guys, good afternoon. Nice to see all the incremental progress here. Just any updates on -- I know our previous discussions, you were looking to expand your power capacity at some of your existing HPC sites. Any more color to provide there on progress in going above the 500 for CoreWeave, and then I have a quick follow-up.

A: Yes, absolutely. And thanks for the question, Joe. I mean what we're looking at right now is about 300 additional -- potentially above 300 additional megawatts across our existing HPC sites. Those conversations are still ongoing, and we really view this as our fastest path to getting additional megawatts at each of those sites. So we're especially excited about that process. It could result in a very significant uplift in the number of megawatts at each of our existing sites.

Q: Yes. That would be really good progress for sure, Adam. And then, I mean, obviously, we have a Presidential Election here and Bitcoin is up. Obviously, that's good kind of for all miners. But any other further thoughts here on the results of the election and implications broadly for the Bitcoin mining industry and maybe anything more specific for Core? Thanks a lot, guys.

A: Yes, of course. I mean, I think this -- the Trump in the past has expressed strong support for both Bitcoin and energy production as well as US leadership in Artificial Intelligence. Those are strong tailwinds. And I think his three priorities there really align well with our business and our future goals.

Q: Hi. I was hoping you can give some more color into the potential deal structures of these new sites and just new customers in general. Should we expect like kind of in the similar colocation range of the CoreWeave deal? Or do you see this like changing on a site-by-site basis, like some being build-to-suit, some being colocation? Any color you could provide there would be great.

A: Yes. Really for what we're looking at for the new sites is really focused on single tenant buildings. So really, we're doing a lot of -- looking at a lot of build-to-suit here for potential clients. The size and scale of what clients are looking for, they're really looking for as many megawatts as they can possibly get at any individual site. And so as we evaluate each of our sites, these are based on deep conversations between the design engineering teams of both companies, both our company and the potential clients and really going through each of their requirements to ensure that we can meet exactly what they're looking for and the time lines that they're looking for. So just to give a little bit more color on that, too, to get to a finalized contract, it really requires us not only to have a finalized design, but also for us to evaluate where we're at in the supply chain to ensure that we can meet the time lines that both sides agree upon.

Q: Great. Thanks. And -- we've seen a number of new stories about CoreWeave recently. Hoping, maybe you could provide color on your recent conversations you've been speaking about diversifying. But ultimately, as companies continue to add megawatts and build out data centers, do you see an opportunity to return kind of to that cloud market? Thanks.

A: Yes. Yes. I mean we -- obviously, we have a lot of respect for the CoreWeave team and what they've been able to accomplish. I mean, really, what we're seeing across the news is very impressive in terms of what they're working on. We're having continued conversations with both CoreWeave as well as other cloud providers here because we view this as a big growth vector in the market. The GPU cloud development that we're seeing going on, is going to capture a significant amount of compute and the overall compute market over the next few years, and we look to continue to ride the tailwinds that, that provides.

Q: Hi, guys. Thanks for taking my question and congrats on filling out the remaining capacity with the CoreWeave. Adam, maybe just it's now been, call it, five, six months since you first announced the initial CoreWeave deal. What do you think has happened to the market in terms of rental rates or lease rates or however, you want to kind of quantify it? Have they gone up since that deal? Just how should we think about where the market is at now for maybe new capacity coming online?

A: Yes. Thanks, Brett. I mean really, the difference in the CoreWeave deal is 100% funding of the CapEx. They were able to significantly buy down their rates. And I think as we look forward, what we're seeing for 2025 is frankly rather unique. If you're able to deliver capacity in 2025 and 2026 right now -- we're definitely seeing those lease rates be much higher than we expected, especially given that many of these folks are willing to cover some portion of the CapEx of the build-out. So we're excited about where lease rates are going, and we believe we'll be able to extract a significant amount of value from the demand that we're currently seeing over the next few years.

Q: Perfect. Helpful. And then maybe just on the new site acquisitions. I guess you kind of outlined three different options. On the distressed side, what -- if it's $5 million to $8 million kind of retrofit in an existing facility, what are you seeing in the market there on a megawatt basis on the distressed data center side?

A: Yes. And Brett, I just want to make sure我 clarify the question. Are you asking about the -- to retrofit those facilities for newest generation compute?

Q: Yes, retrofit or retrofit or what you have to pay to acquire the sites and then the additional CapEx needed to get them up and running?

A: Yes, absolutely. So what we're seeing today in the market, it depends on what type of load studies those facilities come along with. For instance, the site in Alabama coming alongside of a total of 100 gross megawatts available at the site. Really, what we're seeing is somewhere在 the range between $3 million to $6 million per megawatt on the existing megawatts. And really on a retrofit basis, that could range from really $1 million to $3 million depending on the type of infrastructure that's currently in place and how much infrastructure needs to be acquired to accommodate the newest generation GPUs.

Q: Thank you very much, operator. Good afternoon, everyone. Following on a similar theme to the prior question on the potential additional sites. Any desire to enter exclusivity? Or has that maybe already occurred on some of these sites? And could you speak a little bit to the competitive dynamics? How many other companies might be in the room, I would appreciate your thoughts on that. Thank you.

A: Yes. I would say where we're at right now is there is extraordinarily strong demand. There's frequent meetings on a number of clients, and I'll kind of outline that. I mean, really, we're trying to focus on about 10 clients right now where we have consistent design engineering meetings between the teams, hosting site tours for many of these clients and really trying to get to a deal here that makes sense for both parties. So there's definitely a competitive process on both the 70-megawatt site as well as the new site in Alabama. And so we have high confidence in our ability to execute on both of those sites.

Q: That's helpful. Thank you. And then, Adam, taking a step back, you have an HPC AI business that is growing very quickly where you have a very robust pipeline and then a BTC business. And I imagine some sites overlap and do both. But does it make sense? Could it make sense to maybe separate these businesses either now or sometime down the road? Thank you very much for your perspective.

A: Yes. It's really hard to speculate on the future. But really, what we see right now is significant crossover between our Bitcoin mining teams and our HPC teams and our ability to operate high-power digital infrastructure. And as we continue to evaluate each of our sites, we definitely feel strongly that we have a very strong Bitcoin mining business. And as we look towards 2025, our ability to execute on that block transaction represented a significant -- not only -- it's going to represent a significant increase, not only in our efficiency, but also given us the ability to continue to grow our hash rate. So, we have two very strong businesses. And so for us to speculate on what may happen in the future is difficult today.

Q: Great and thank you for the time, and appreciate you taking the question. Adam in terms of, I guess, the sourcing process of new customers on the HPC front, what do you think has changed now, now that you have the full CoreWeave contract? Is there anything notable on that front as you go to secure new customers now that the full 500 of capacity is contracted with CoreWeave and you kind of have that anchor customer?

A: Yes. Things are definitely different. I appreciate the question, Tyler. Things look much different. I mean we're building out one of the largest infrastructure bases for high-performance computing over the next two years. That definitely brings with it a significant amount of credibility. What we're seeing in conversations that we're having is we're the thought leaders on the design and the requirements that are necessary to operate these GPUs. That puts us in a very unique position walking into the room with hyperscalers and with Neo cloud providers. So, we feel like we have -- we're really在 the driver seat here in these conversations. We're excited about bringing new clients into the business.

Q: Okay, great. And then my follow-up here,我 know you laid out some of the power cost forecast. But more broadly speaking, as you kind of look to next year and bringing on more HPC capacity and we have some of the new sites. I guess how do you kind of think about the power strategy as you're building some of these custom-built facilities for HPC customers? And kind of what does that look like across the footprint as you kind of look out into bringing some more capacity online?

A: Yes. So, the power expectations that we laid forward, that's really just on the Bitcoin mining part of the business. But one of the unique parts about our business and our data center business is that we have a power team in-house, a power team that is a very close relationship with the utilities in which we operate and many other utilities as well. That helps us significantly in our negotiations with potential clients where we can walk them into the room with utilities who are excited about the new capacity coming online. So, as we think about power pricing, that's something that us at Core Scientific do handle on behalf of our clients. And so we're excited about the additional potential megawatts that we have at many of our sites. And I think not only Core, but other potential clients are very interested in that power as well.

Q: Thanks so much. My first question is around the block chip delivery. Given the timing and how that lines up with your progress around the HPC infrastructure rollout and your conversion of 100 megawatts from BTC to HPC. I was just wondering if those block chips, if you envision that they'll be predominantly net new capacity that you're going to find and build out or if it's going to be in part fleet upgrade. Thanks.

A: Yeah. Thanks for your question, Paul. Those new block chips are really going to be focused on a refresh of existing fleet. It's going to give us the opportunity to not only increase exahash, but also increase our efficiency and so we're very excited about once those landing, getting them up online in our facilities.

Q: Thanks Adam. And then a quick follow-up around power, with this conversion and maybe other conversations you're having to expand site capacity from a power perspective, how are your conversations going with utilities that you may have demand response enrollment with and not sure if you are un-enrolling or if -- as you approach these conversations, it's non-interruptible capacity. How those -- the tone of those conversations is evolving and any impact that might be having on spot rate that you're talking through with these utilities? Thank you.

A: Yeah. And that's really a bifurcated question between regulated and non-regulated markets. In regulated markets, those were voluntary programs that we are participating in to reduce our power costs. Those are some power programs that we will be exiting as these sites begin to convert to high-performance computing. And then在 the deregulated markets, those are new PPAs that we'll be looking to sign that will be on a different rate structure than the rate structure we are on for Bitcoin mining. So that's a well-trodden process. And we feel very, very strongly that we have execution capability here with our utilities.

Q: Thanks. Hi Adam. Thanks for having me on. You mentioned three components that you see locked in supply issue. I was wondering if you could just add a little more color to that. You said that you thought you could find workarounds. I don't see that happening with gensets. And I was wondering how you could size that up against meeting your, I think, end of first half target for the CoreWeave deal, which appears to be static, so congrats on that. Just maybe sort of relieve a little of the unknowns regarding the supply chain.

A: Yeah. Thanks, Kevin. I think as we look at 2025, we have high confidence in the supply chain locked up for those deliveries. For parts of 2026, those are things that we're still working on. But even on gensets, for example, that was used, this is -- these are things that -- the traditional gensets that are used in data center development. There are different sized gensets that we're evaluating to help alleviate some of those traditional design choices that have been utilized over the past 10 to 20 years. So we are also coming up with methods and looking at potential rental of certain equipment to help really bridge the gap in areas where we might need to bridge some of those gaps. So we have high confidence in our ability to execute on the supply chain even with the complexities that are presented due to the high demand for this equipment.

Q: Okay. And with regard to the block machines, do you imagine those are water cooled or immersion? And would you think that you'd have to retrofit existing Bitcoin mining facilities to manage those?

A: Yeah. I mean if you look across our infrastructure base today, we've become -- we are the leading experts in really air-cooled facilities. And so for this deal with the new block chips, we're designing a new form factor that's really focused on air cooled design that we can maximize and we've mentioned Pecos in particular, that can maximize the production of these -- the production coming out of these chips even at high temperature ranges. So no significant refresh in terms of the infrastructure inside those Bitcoin mining facilities. So we believe it's going to be a very strong process or easier process for us to install all of these machines.

Q: Yes. Good afternoon, everyone. Thank you for taking the call. The 100 megawatts that you've identified to convert from Bitcoin now to HPC that you hadn't identified prior to this. What was it about this that changed? Or did your view change about what you could do? How did that -- what was the thought process there? And is there the possibility of that same thought process being applied to other megawatts that you have currently employed in Bitcoin?

A: Yes. Thanks, Rosemarie. It really calls back to the three points that我 laid out. So power, environmentals and latency. This site, in particular, was something that we are very focused on the power aspect, ensuring that we could create a contract with our utility that would be amenable to both sides. We've got into that point today. So that's what really led to the conversion of thinking this site was going to be specifically utilized for Bitcoin mining and undergoing a process to convert that to HPC. As we evaluate the rest of our sites, there are definitely points that we would need to continue to evaluate on both the power environmentals and latency side. And it's hard for us to speculate now about where we'll end up in each of those processes. But right now, we feel we have high confidence in our ability to continue to acquire sites and bring more critical IT load megawatts to market.

Q: Okay. Thank you. And then the new site in Alabama that's currently powered at 11 megawatts, but can go to 66. What has to happen for that to actually get to 66 and what would the time frame be?

A: Yes. This is one of the sites that really -- it goes back to the distressed data centers that come with load studies and additional approved power. So on Alabama, in particular, that's what really excited us about that site, 11 megawatts in traditional data center terms would be a significant amount of megawatts. But given that ability to expand while also bringing on capacity in 2025 is extremely attractive to potential clients, which is why it's one of our target areas for finding new capacity.

Q: Thanks for taking my question. Just two, if I may. First, on the process for securing additional power with the 300 megawatts at the existing site you guys have in the deck. Is that kind of like -- just talk about the process? Is that weeks, months, years and just how concentrated is that by site? And then when it comes to additional kind of expansion on HPC, like how do you balance customer concentration versus economics? And what I mean by that is the 100 or so megawatts you have available today, if CoreWeave was wanting to strike a similar deal, would你 take those economics that want to diversify it to an additional customer? Thanks.

A: Thanks, Darren. So for the additional power at existing sites, that process is different for each utility. Some of them require additional load studies. Some of them are in negotiations about where and when we will have our ramp-up in terms of megawatts. I think our ability to capture additional megawatts at existing sites really comes down to our proof of execution. It comes down to what we show them on the Bitcoin mining side and now in the movement to HPC, these utilities are excited about working and growing their relationship with us. So it can, in some instances, be a lengthy process. But as it comes down to our negotiations with potential clients related to some of these additional megawatts, they understand time is of the essence, and they're helping and working alongside of us to get to a final answer, which is why we said there's potential for some of these megawatts to be approved inside of 2024. Now looking at -- now talking about your second question related to customer concentration and how we're thinking about that related to economics. Obviously, that CoreWeave deal is extraordinarily attractive. It has the potential to be the best data center deal ever signed in the history of data centers. So obviously, we would like to continue to expand our relationship with CoreWeave. But we do think for the long term, it's necessary for us to diversify our client base, bring other large tech companies into the fold and really create a business that we believe is extraordinarily stable over the next 15 to 25 years. And that's really what gets us excited. And we think we have tremendous growth opportunities with new clients and also with CoreWeave.

Q: Thank you so much for taking my follow up question. I just wanted to go back to the Alabama opportunity. Adam, could you speak a little bit about the time frame from when you first recognized that side to today? How long does that take? And any indication on consideration for this asset? Is it cash? How much an earnout, profit share? And anything你 could share would be helpful. Thank you so much.

A: Yeah. Thanks, Lucas. And I would qualify this site as it takes probably low to mid-single-digit months in order to really get to a finalized deal here because we have a significant amount of time now to really perform all of our due diligence. And a lot of that due diligence is not only on the infrastructure side, but it's also spending a lot of time with those utilities. We want to ensure that we're going to have a strong working relationship and also find a pathway to growth. As we evaluated that site, the 100 megawatts is fantastic. And we also want to figure out ways to continue to grow with the utilities. You're seeing it today across a number of our sites where we have the potential to potentially expand the number of megawatts. And that's very important to us, strong working relationships with the utilities. Now it relates to the consideration, this is a lease site that we have the option to buy. So we're -- we want to put ourselves in a position where we don't necessarily have to buy the site immediately. But having the option to buy provides us a lot of optionality. And it's at a fixed price. So regardless of the upgrade that we do and perform and any additional power that accrues to the site, those accrue to our value.

Q: And just to add, Lucas, this is Steve. There is a bit additional description of that deal in our 10-Q that should be filed later today right tomorrow.

Q: Really appreciated it. Super helpful Adam to you and the team continue best of luck.

A: Thanks, Lucas.

Q: Hey. Adam, you clearly delivered on adding new sites, and you did highlight three avenues to continue to expand your portfolio. But maybe you could give us a little bit more insight on that pipeline specifically. I know you're speaking to some pretty large numbers. I'm just wondering how你 see that market and Core's ability to compete given huge demand for power and access.

A: Yes, absolutely. Thanks, Kevin. There are a lot of sites outstanding right now. There's a lot of power available in the United States, which is our focus area right now. And we're finding the most attractive sites to be ones that aren't being run through broker processes. And so that means we have many boots on the ground, speaking with utilities, speaking with landowners and finding sites that are much more esoteric than the traditional ones being run through broker processes. So we're finding ourselves with a strong ability to compete. We've evaluated over 15-gigawatts worth of sites over the course of the past, I'll call it, six to eight months, putting us in a strong position to find unique deals in the market. And we're excited about our ability to continue to execute on deals coming forward. So, Alabama is really the first one here that we've executed on, but we look forward to 2025 and 2026 to continue to grow our critical IT load megawatts.

Q: Thanks, Adam. While you're still there, maybe you could speak a little bit to the block chip deployment time line. I know你 mentioned second half next year. I'm kind of wondering whether or not that is what你 initially expected? Or was there some wiggle room in there? And do you suspect you'll be able to find all the components you need to make the full unit?

A: Yes. Thanks, Kevin. This is something that our research and development team has been working closely with the Block team on for a long period of time. This isn't something where we had only been working with them for months. This is something that we've been working on with their team for a much longer time period. So, we have all of the manufacturing set up. We have our entire game plan in terms of our rollout, in terms of our execution already put into our schedule. So, this time line is exactly aligned with what we expected, and we're excited about being able to get these up and running and continue to reduce, I would say, increase our efficiency and also increase our exahash exposure.