All right. Good morning, everyone, and welcome to TD Cowen's 53rd Annual TMT Conference. My name is Michael Elias, and I'm the Communications Infrastructure Equity Research Analyst here at TD Cowen. For this session, it's structured as a fireside chat. We have about 30 minutes. I have a bunch of questions prepared, but I'll do my best at the end to pause and take questions from the audience. For this session, we have Core Scientific, and from Core Scientific, we have their CEO, Adam Sullivan. Adam, thank you very much for being here today. Really appreciate it.
Thank you for having us. Really appreciate it as well.
All right. Let's kick things off. I think it'd be helpful if you could share just a little bit about the company's journey since inception to where the company stands today. I know there have been some ebbs and flows, so I'll turn it over to you to give us the high-level overview.
All right. We'll do the two-minute version of this. The company was founded back in 2017 by a group of folks that wanted to build Bitcoin mining facilities. They found their first CEO, which was the former Chief Operating Officer of Microsoft for 11 years, Kevin Turner, to help run the company. Kevin came in, identified a number of different sites to acquire. All of those sites are still in our portfolio today, and we built out one of the largest Bitcoin mining businesses in the world. By 2021, we were on a pathway to go public. At that time, we had 800 megawatts of infrastructure fully operational. We had power contracts for nearly 2 GW of power, and we were well on our path to maintaining our leadership position in the Bitcoin mining industry.
Now, fast forward to December of 2022, we had gone public, and we faced a number of significant headwinds between Bitcoin price falling and energy prices rising. Obviously, the war in Ukraine caused a lot of difficulties for regulated utilities at the time. We saw power prices nearly double in many of the regulated markets. We filed for Chapter 11 in December of 2022. I joined the company in April of 2023, so just about two years ago, and we had reemerged in January of 2024. We came out of the gates pretty hot. We knew we were going to transition to a data center business. That was part of the original pitch to the board when I joined. We signed our first contract with CoreWeave in February of 2024. We signed our next large contracts for 200 megawatts with 300 megawatts of option agreements in June.
Over the course of the remainder of 2024, we've now taken that contract to 590 MW with them. It is a very unique contract structure, as many people know. They're giving us just over $5 billion to build out 590 MW, and they're paying us essentially what our investment-grade lease rates are for that.
That I would argue is one of the most unique contracts I've ever seen in data center land. Kudos to you guys. With that history and context in mind, let's transition and talk about the future roadmap. Where does Core Scientific go from here? As part of that, what's the near-term and long-term strategy and vision for the business?
Yeah. I think it's easier to talk about the long-term, and then we'll weave in the short-term, how we get there. The long-term vision is we want to build out a large-scale data center platform that rivals that of Digital Realty and QTS. We believe, given the transition that this industry is currently going through between low density to high density, co-lo, we believe this is a very large opening for our business to grab a significant market share in. We have a very unique capital structure as well that's enabling that. We have the ability to leverage our CoreWeave contract up to about $4 billion. With about $800 million in cash and liquidity in our balance sheet today, plus potentially $4 billion in cash to pull out of those contracts, we have nearly $5 billion in just equity capital to apply towards new deals.
We have 700 megawatts available today in our existing portfolio to sell of critical IT load. Our goal is to get CoreWeave below 50% by 2028. We believe we'll be able to surpass that by not only doing organic deals, but also looking at inorganic opportunities right now, which are plentiful in the market. It's probably one of the more interesting parts about where we sit. There's really every process we go into, we're either the only bidder, or there's one other bidder in the process. It puts us in a really unique position to acquire assets in the United States.
Interesting. You know, I want to build on that point about building a business comparable to Digital Realty and QTS. When I think about those businesses, they've historically focused on the major markets, serving, let's say, the swath of customers, enterprises and hyperscalers. What I want to focus on first is the major markets. As I think of your footprint, Denton sits within, if I'm not mistaken, I want to say 38 mi from the Infomart Dallas, which is the key network node within that market. As you think of your footprint, how do you think of the viability of your sites for not only training, but also inference that is latency sensitive and needs to sit within the availability zone?
Yeah. All of these sites that CoreWeave contracted are sub 10 milliseconds. Many of them are less than 5 milliseconds. They meet the requirements for their end customers. What we are developing for CoreWeave are single-tenant, built specifically for CoreWeave's end customers. All of those sites had an eye towards low-latency facilities. Within our existing footprint, we feel very confident in that. You take a site like Grand Forks that's in our portfolio as well that we're currently in negotiations now. That connects back to the Equinix fabric up in Winnipeg. There are opportunities even outside of just being low-latency back to a major mart in a metro that we have to help reduce some of the latency constraints around more remote locations in our portfolio. Going forward, I mean, we are focusing on low-latency sites.
Now, for people to get the power that they want in low-latency sites, they can't necessarily be 1 millisecond latency back. They're going to have to be more comfortable with 4 or 5.
Interesting. You know, on the panel we were doing earlier, one of the comments that was made is that if you look at the availabilities on that 100 mi radius around the major metro, there seem to be pockets of opportunity 70 mi, 50 mi away from the core pocket. So I'd use Northern Virginia as an example. You have a dense pocket in Ashburn, but then on the periphery, there are opportunities in Leesburg, Manassas, and stuff like that. As you're thinking about the inorganic growth or even organic growth, do you see opportunities to get capacity there within the AZ?
I definitely think there's opportunity there. I mean, some of the deals that we're looking at, what we're thinking about are, is this site going to be viable for the next 20 years? Do we think that the customer is going to go through all of their customer or all of their extensions that they have in their contract? Those are all things that play a significant importance to how we evaluate some of these deals because we recognize the fact that for us to be able to increase our multiple, for us to be able to move towards identifying ways to optimize our tax structure, those are all going to be things that require blue-chip investment-grade customers. That is, investment-grade customers historically have been at sites that are very tight inside the AZs. That has definitely been a part of our inorganic growth strategy.
Perfect. All right. You know, one of the things you mentioned on the earnings call, I believe, is that the traditional data center providers are struggling to meet the density requirements that AI workloads demand. And I, on my own, have seen an evolution of rack densities in this sector, not only driven by NVIDIA, but people having to think about their legacy assets. So with that in mind, I'm curious, how do you see yourself positioned to handle those higher rack densities? And is there a standard spec that you're building to as you think of these incremental data centers you bring online?
Yeah. I mean, right now for 2025, all of our deliveries are nearly 150 kilowatts of the rack. For next year's deliveries, which are mainly being delivered for GB300s, those are going to be 200 kW of the rack plus. As we've identified sites that are going to come online in 2027, I mean, those are a lot of developments that we're looking at on the Rubin side that could be potentially significantly higher densities than the 200. We feel like we're at the forefront of power density inside of our facilities. Right now, we're developing the most capacity for GB200s over the course of 2025 that will be delivered this year. We feel like that gives us a little bit of advantage, not only developing some of the largest capacity, but also developing for multiple end clients.
Got it. Now, let's talk a little bit about the customer mix. So targeting hyperscale as well as enterprise. You know, if I think of just QTS as an example, when they were a public company, I want to say it was like 67% of their business was enterprise and 33% was hyperscale. For Digital Realty, I'd say it's more 40% hyperscale, the rest being enterprise. As you think of your long-term aspirations, how do you think about the appropriate relative mix between hyperscale and enterprise? And as part of that, how do you think about customer concentration amongst the hyperscalers? Because given their deal size, they can end up taking a large swath of capacity that moves the needle.
I know this is probably the biggest problem is that hyperscalers can be, they're going to be chunkier, but they're going to be further apart. Chunky might be a 200 MW contract or it might be a 400-megawatt contract. That mix is really hard to say. We're going to have a firm line on whether it be 50/50 or 60-40 because you don't know how much power or how much demand is going to come out of the hyperscale channel. Those deals are going to be much further apart in terms of when you compare it to enterprise because enterprise can be a much more consistent channel to tap into. You might be getting more consistent 50 MW or 75 MW contracts coming out of the enterprise channel, but you might get one or two hyperscale contracts a year that could be much larger than that.
I don't think we have a target mix in mind today, but it's definitely something that we're keeping in mind, especially as we think about further down the road when we are going to start looking at getting rated by an agency.
Got it. That makes sense to me. You know, one of the questions I have is on the hyperscale side. We've seen a lot of activity over the last, call it, two years. I think first quarter of 2023, we saw a meaningful re-rating of data center demand. Actually, second quarter, my apologies. As you think of the near-term opportunity set with the hyperscalers, how would you characterize it? As I mentioned to you earlier, I do think you got one of the most unique deals in data center history, just to call it what it is. Do you think that deal structure is replicable with other hyperscalers?
I would say yes, it was probably the most unique deal ever done in history. We had achieved 3,700% IRRs on a data center deal, which I don't think has ever been done in the history of the data center industry. You would probably know because you're practically a historian on the industry. Right after we signed that deal, we were having conversations with hyperscalers, and some of them said, "Yeah, we're willing to pay for infrastructure, but here's the lease rate that we're willing to pay." We're like, "You know, they're essentially, they're paying for infrastructure, and so you're essentially just getting a PowerShell deal at that point." The rates were not as attractive, obviously. I think from what we're seeing today, it's very straight down the fairway in terms of what lease rates are generally for hyperscalers.
Their demand has gotten much more specific to the first part of your question. Their demand last year, last summer, was, "We need 200 megawatts, and we need it by X date." Now it has shifted to, "We need 40 MW outside of Atlanta, within 50 mi of our existing cloud facility." Those have become much more specific on the hyperscale side, where 12 months ago, they were just trying to put megawatts in the ground.
It kind of goes back. This is something that I've, in my conversations in the last, call it, three weeks on the road, it's like location, location, location. That's what I'm getting out of the hyperscalers. They have very specific geographical requirements, and either you can solve them or you can't. It's almost like the days of the anywhere deal are gone.
I agree with that. Also, the time of any utility is kind of gone as well because they'll say, "We want to be within 50 mi of our existing data center, but we won't be with that utility." You say, "You look at the map, you're like, 'There's only one basket of land that you could possibly fit into.'" Then you're like, "We need to go see if we can buy land right there." They've gotten a lot smarter. They have much more specific requests. I think in terms of what we're going to see over the course of the next two years for development on their side, it's going to be smaller facilities, not necessarily some of the larger facilities like we're seeing out of Google today.
Yeah. That to me is indicative of a shift away from training more into inference, right? Because the way I've thought about it is that the latency aperture around training, call it 8-12 milliseconds or something like that, the inference needs to sit one millisecond. If you have location-specific deployments, it implicitly infers, "Okay, this is the workload that they're looking at." It makes complete sense to me. I have a question for you along these lines. When I think of companies like Equinix, they're very clear about what their North Star is, right? For them, it's AFFO per share growth. As part of that, you know the returns that they generate on their investments, which is why you've seen them do hyperscale off-balance sheet. I'm curious for you, what is the North Star as you think about growing the business?
Is this oriented around gaining market share? Is it oriented around maximizing returns? What's the North Star as you build this business?
Yeah, I mean, right now, it is growing market share with return profiles that we find attractive. I think one of the biggest issues that we face right now is we trade at, let's call it, seven to eight times forward, which is significantly lower than our peers in the colo industry. The way we're thinking about the business right now is there's significant opportunities for us to create value for our shareholders by increasing our contracts, diversifying our customer base, increasing our footprint. We recognize that right now, we have a very specific use of capital, which is to continue to grow and diversify this business. Over time, yeah, absolutely. We're going to have a much more straightforward business that we're going to be focused on some of the metrics that you mentioned in terms of AFFO, et cetera.
Right now, we have a mandate to continue to grow our multiple and put ourselves in position to be a competitor with Digital Realty and Equinix in the public markets.
Makes sense to me. Now, as part of that, one of the things you mentioned, your multiple, one of the things that I think about is that given the customer concentration of the platform, the way I've thought about it historically is that your equity is essentially a proxy for the unsecured credit of your largest customer, right? I think that's kind of reflected in the multiple where you guys trade. You talked about diversification, which is clearly a focus for the business. As you think about going about that diversification, is it just largely adding incremental customers to then bring that percentage down?
Is there an element of, "Hey, I can perhaps crystallize some of the value that I've created with these data center builds and then in turn recycle that into new projects, which then allows me to not only get a new customer that brings down my percentage, but also I'm reducing my exposure to that large customer"? How do you think about that?
Yeah, I mean, the one thing I would note, though, we are fully first lien against all the infrastructure that Core pays for. And at the end of the contract, we receive it. If someone told me that Core was trading at $120 and we were still sitting at $11, I would be scratching my head in terms of why they've gone up nearly 300% since their IPO, and we've sat essentially flat. I guess in terms of how we're thinking about the future growth of this business and how we really build value here, I mean, I mentioned some of those numbers earlier. We can finance against the Core contract up to about $4 billion.
If you have $5 billion in equity capital to deploy on a fully financed basis, that means if we're thinking about, let's call it 75% loan to cost, we should be able to do deals upwards of $12 billion at the low end. We think we can build out, at least if we use round numbers here, like $10 million a megawatt, we should be able to build out at least a gig organically over the course of the next three to four years. That will bring our customer concentration below 50% for CoreWeave, which we think is critical in order for us to have a much more stable story to the market because CoreWeave's story is going to take a few years to play out anyway.
I think there's a lot of question marks about their ability to continue to execute and sign new contracts. I think that's going to take three years to play out before people really start to recognize the value that Core Scientific brings to the market. I think three years from now, people are going to view Core Scientific much differently than they do today.
Fair. I want to just stick with that point for a second and maybe say it a different way. It's that after this large facility delivers or after this campus delivers, you're effectively going to be underlevered against a large NOI base, which then allows you to add additional leverage and then applying that incremental, applying that debt financing and then taking the equity that you have, that's what gets you to the gigawatt. There is runway for you to grow out the platform incrementally. Now, what I'm curious about is how are you thinking about the relative split as you're deploying that capital between inorganic and organic, right? There are data center platforms out there that have capacity in the major markets that you could potentially acquire. How do you think of the relative balance between the two? From there, I have a follow-up.
Yeah, I mean, I would say, again, we do not have a specific mix that we are necessarily targeting because the deals are so different. We are looking at a number of different multiple middle market platforms that essentially are currently capital constrained to be able to continue to grow their platform, looking at some one-off sites that have a customer, an anchor tenant, but have additional capacity to grow into, which we think are great tuck-ins for our portfolio. We think some of the organic side over time will actually come through the inorganic channel of buying some sites that are potentially some stabilized, some under development, and some greenfield. Being able to continue to grow our portfolio into new geographies that way as well. We recognize the need that we need to put new pins in the map.
A lot of that, I think, will come through inorganic in terms of getting a larger spread of sites and touching more metros. We think that's probably going to come through inorganic. We'll probably do larger, chunkier sites on a single site for a lot of the existing organic stuff that we have in our portfolio today.
Got it. Yeah. How do you think about when you're negotiating deals, how do you think about the appropriate contractual rent escalator for these? Because as you think about long-term AFFO, that obviously provides a degree of organic AFFO per share growth, right? As you think of it on a cash basis, how do you think of the appropriate escalator?
I mean, I think the industry is pretty standard on rent escalators at this point.
I feel like there's been some dispersion recently.
It's like 2%-3.5%, I would say, is the general range.
Yeah.
Obviously, we would love to target the higher end of that, just.
You know, the CPI component too.
Yeah, it's all part of the negotiation of MRC, NRC, and then rent escalators because all three of those factor into what the return profile of a deal looks like. Everything comes down to how much pricing power do you have on essentially the initial year? If you know there's a trade-off, would you prefer to have higher initial revenue in the front six years of a contract or the back six years of a contract? Some of that comes down to how we feel about the site, how we feel about the customer, how we feel about their longevity, et cetera.
You know, I wanted to ask you this is, when I think of the hyperscalers, there are a lot of people who are chasing that opportunity set. I have a friend in industry who comments on every single day there's like, "Oh, we're building a gigawatt here. We're building a gigawatt there." You add all of it up, and it just doesn't quite make sense. One of the questions that I have for all the new platforms that I engage with is like, have you gone through the vendor pre-qualification process with the hyperscalers? I come to find sometimes that they're like, "Oh, we didn't know there was one." From that perspective, have you gone through that?
If so, I won't ask you to name specific hyperscalers, but just like how you feel in terms of your positioning to be able to win deals and be in their system.
Yeah. No, I mean, we're engaged with multiple hyperscalers today, have gone through onboarding process, which is a time-consuming process for a number of different teams inside the company. I think that's probably the best question I like to ask people as well when people are saying, "Oh, yeah, we have a gigawatt site. We're currently in talks with multiple hyperscalers." I'm like, "How was your onboarding process? How did that go?" They're like, "What's the onboarding process?" You're like, "Okay, so you're not actually talking to the hyperscalers." I think you mentioned people announcing every day that they have a new gigawatt site. I mean, as we've gone through our process of evaluating different deals in the market, we talked to one company that said they had two gigawatts worth of sites.
We're like, "Where are your power agreements?" They're like, "We have load studies in." We're like, "Okay, you're probably 24 months away from hearing back. You have no idea how much power you're actually going to get, but you're telling the market and marketing to the public that you have 2 GW of sites." I think there's a big dislocation between what the utility is actually providing people and what people have announced publicly.
I would agree with that. I was having a conversation with one large data center operator, and I asked them, and I'm going to ask you the same question, so be ready. I asked them, "What keeps them up at night?" They said, "Finding powered land." I said, "How is that the case?" Every single person I talk to, it seems like there's a gigawatt here, a gigawatt there. It's like, what happens is that when you dig in with them, you find out that they have a load request in or some load study in process. For the ones who've made it far enough when they have to put up the hard dollars to get the power, they don't have it because they ultimately need someone else to come in behind them and put that money up. I completely agree with you there.
To the question I was asking, for you, what keeps you up at night? What is the major bottleneck in your mind to delivering incremental data center capacity, particularly on an organic basis? Let's put the inorganic aside.
It's utilities changing their mind. Utilities have the ability to essentially, even after you've signed a contract, they can still change power agreements. It's something that we've seen historically, even on the mining side. We were one of the largest power consumers in the United States. You see it happen at the utility level. It's hard because you can't go up against the utility. Good luck finding a law firm in the jurisdiction to go up against the utility. You can't hire anyone. No one will ever go against the utility. You're kind of a taker in a lot of markets, which makes it a big challenge because the most sought-after markets, the utilities know they hold the cards.
They are charging two things upfront, one of two things, either a CIAC, which is a cost-native construction, which could, if you're building 200 MW, they might just ask for $50 million. Even though it might cost them $10 million to change up the transmission lines, they want $50 million. They might ask for capacity payments if they're looking to secure new generation for your load. They might say, "We want five years of capacity payments," or they might say seven. They are just trying to price people out of the market and find essentially their efficient frontier of how much pain they can inflict on somebody before someone walks away from a power contract.
You know, not to rough on this philosophically, but what I find interesting about that, I would agree with you. I call it demand destruction of the queue. And I think what I've seen is that when you look at some of the requests for data centers in the queues of utilities, I look at Dallas as an example. It's like it pretty much implies that every single GPU for the next five years will be shipped into Dallas. That's what that queue implies, right? And none go anywhere else. To that point, I do think that they're kind of trying to call the demand. As part of that, you mentioned that they could change their mind. It's like there is no recourse to the utility at the end of the day.
You have no recourse as a data center operator, which then brings me to the next question, which is that creates risk for a data center operator, which is that you have a contract that says you need to be ready for service by this date. If you miss that, depending on how you structure the contract, that opens up the potential for the customer to walk, right? Not always, but quite a bit of the time. Against that dynamic, how do you think about protecting yourself so that you're not in a position where the utility pushes power out, says, "Hey, sorry, plans changed," and now you're holding the bag for a contract and you're now at risk of stranding the capital that you'd put into the data center?
I mean, I think there's two parts to that too, right? You have the one side, you have the customer that has either the potential to walk away. For certain hyperscalers, that's a requirement. The other side is the utility. You could have minimum bill payments starting on a specific date.
Yeah, that's the take-or-pay point piece.
Yeah, you're either on a construction schedule or you're on a utility schedule, and you're trying to manage both of those. I think that is where we've seen a lot of hiccups even start to show in the market today, which is we've gotten calls from smaller data center operators that say, "We have minimum bill payments coming up. We haven't finished construction. Would you be willing to do some type of joint venture on the site to help us take power more quickly?" Most of the time you're like, "No, we're not catching that falling knife in that market." That's probably the biggest—I will change my answer about what keeps me up at night. It's probably that dance. Because in order to get a hyperscale agreement, you have to have the power contract, which means you've taken on the power risk at that point.
You've taken on the minimum bill payment requirements that start maybe 12 months or 24 months out. On the other side, you're up against the time clock of trying to sign a hyperscaler, which they dictate their own timeline.
Yeah, that's right.
They could say, "We want to move really quickly," and then drag it out for 12 months.
I've seen that so many times. To that point, I'd be remiss if I didn't ask you about this. As you think of the near-term opportunity set with the hyperscalers, I mean, is there an expectation that you can sign, or do you believe that you can sign in the near term another large deal to bring down your exposure to that one single customer? I know these things are tough to handicap, but I'm sure you have a sense of your pipeline and kind of what's in process.
Yeah, I mean, I would say from our perspective, our next deal, we want to add at least probably at least $50 million in EBITDA from a single customer per year. Thinking about somewhere between like a 75-megawatt contract, somewhere in that range. That's probably on like just over 100 megawatts of gross. Megawatts. That feels like it's kind of the right fit right now for a number of different hyperscalers in terms of markets they're looking at. I'd say that's kind of what our—that's not our optimal, but I would say that's most likely in the target zone for what the next contract looks like.
If I take that and I pair that with your commentary earlier, that would suggest to me, given the location-specific dynamic, we're talking about not one of these gigawatt-scale deals. We're talking about what you could argue is an inference-type workload sitting within a major market. Now, the derivative question then becomes, do you feel you're positioned with that capacity to sign a lease of that size?
Yeah, I think so. We have 300 megawatts at our existing sites of critical IT load. Those are mainly sites that are currently being contracted with CoreWeave, where we have additional power and additional land to build another campus next door. We think a number of those sites are very interesting, especially because some of those customers were potentially looking at utilizing CoreWeave for some of that capacity and now say, "If you guys have additional power, we would love to have our own facility there." I think that is a more direct opportunity, as well as our site like Alabama. Georgia is essentially completely power constrained at this point.
Our site in Auburn, Alabama, represents a very interesting opportunity for a number of different hyperscalers to be able to access the Atlanta market without being in Southern Company, which is frankly something that a lot of them are trying to avoid.
One other thing I wanted to ask you is that as I think of Phoenix as a market, APS gave, I believe, 500 megawatts of incremental power allocation. As part of that, what they did is they said, "Okay, we'll give it to you, but you have to curtail off-grid." How do you think about the—if you get power in some of these new markets, how do you think about handling the need for curtailments, or do you think that's something that's tied to specific markets and that's it?
It's definitely tied to specific markets, but it's all coming down to when we present a power contract to a potential client, you have to present also the design engineering to solve for it. The curtailments look different in every market. Some of them might be you have 130 megawatt curtailments over the course of a year. Others say that we need to be able to shut down 10 times for four hours. Those are much different design engineering requirements. When we're having these discussions and negotiations with the utilities, it all comes down to, is this easily adjustable on the design engineering side to design around this potential curtailment issue?
With that, we're just about out of time. I went completely off-piste with this, and you rolled with the punches, so I super appreciate it. Adam, thank you so much for being here. Really appreciate it.
Thanks, Mike. Appreciate it.