Developing Data Centers from the Ground Up | Bitcoin 2026

So, welcome to our panel, everyone. Today, we're going to talk about developing data centers from the ground up. We have with us three experts in their own areas within this whole process, and we're going to walk you through how how the process works from the site selection until the the the data center is energized. So, my name is Yair Alon from E2C Partners. I will moderate this panel, and now the panelists will introduce themselves. Uh hello, my name is Adam Wiggins. I work for uh a company called Camfil, which is large uh global air filtration manufacturing company, and uh I manage our uh data center vertical for uh for Camfil. Everybody, my name is Jay Sabat. I am the co-founder and CEO of Satoki. Um we're a data center developer and operator, primarily uh developing data centers in Texas and Oklahoma. And I'm uh Glenn Miller. I am a uh partner at Katten Muchin Rosenman. I'm head of the national data center group, and uh I've been I've been doing this since the uh the early 2000s. So, it's exciting to see what's going on. So, first over to Glenn. He's a lawyer and also all about contracts and due diligence. So, what are the main risks and uh things that you look for in the due diligence process when when you select land? So, so there are some key considerations whenever you're trying to do site selection for a potential uh data center project. Uh first and foremost is is uh is is land. Uh the the permitting can be very different from jurisdiction to jurisdiction. Some jurisdictions are are uh friendly toward uh uh data centers. They appreciate the tax uh uh benefits that that can be generated by data center development. Um uh and some jurisdictions are are are not friendly at all toward data centers. Uh the second the second key consideration is is power. Uh as you know, the data centers are are are fairly heavy users of of power. Uh and depending on the location, uh uh some grids are very constrained like PJM, which is the grid uh uh consortium of multi states in the in the East Coast that is Virginia, uh Pennsylvania, etc. Very tightly constrained. Where other jurisdictions are are have more power. And uh and sometimes you have the ability to tap into natural gas pipelines and and do on-site behind the meter power, which is increasingly being encouraged by the by FERC, the Federal Energy Regulatory Commission and others. Um and uh and then the final issue that I would describe is is uh uh is the type of data center. Certain data centers like enterprise data centers, uh there's a ten- tendency for them to congregate around uh uh areas where there's large use like the government or certain agencies. Whereas AI data centers can be pretty much anywhere that power is available. There's there's no uh need for them to be close to a particular uh area because because they don't have the the latency concerns that you see with enterprise data centers. So, uh Jay, as a data center developer, what what does a good site look like for you? What are you looking for? Uh that definitely depends on uh what the business model that we're running is, right? So, um on the AI side, um for off-take you have Neo Clouds and hyperscale customers. Um on that side, you need a larger facility, and so it's usually anywhere between 50 to 500 MW. Um we also run Bitcoin mining data centers, and so those are more distributed facilities that we can um start development from anywhere between 5 to 15 MW. Um and so when when when we're really looking at uh site selection, we do a lot of work on power origination in parallel to the land uh origination. And so we have a whole team that, you know, does a great job at doing that, and we're trying to match, you know, our land option and our purchase agreement to getting um something back from the utility saying, "Hey, there is access to power here either through a point of interconnection or through distribution." Um so that that really is kind of the initiation of looking at a site, and then we can go from there depending on the, you know, the power availability um and just the the the load and the location, we can decide if it's it's applicable for HPC use. And uh Adam, uh you are expert on filtration and the systems, how they holistically work together in a data center. So, does site selection ever account for infrastructure constraints like cooling or filtration, or is that an afterthought that is kind of dealt with later? Uh well, yeah, environment plays quite a lot in in site selection and and really uh the environment that the data center is going to be around uh should be evaluated as far as um air quality, um any kind of uh potentials for corrosion breakdown, um also as well as uh you've got to look at potential uh any type of environmental concerns on your uh backup generation systems. If there's certain, let's say you're building a data center and right down the the street or across the field is a chemical plant that might off-gas a certain chemical, you know, six, seven times a week, and if the wind catches it right, it comes over and it's raining that day, and now you have acid, you know, trying to egress into your data center. So, yeah, uh environment plays a really big part in it when you're looking at, you know, the long-term total performance of your your cooling system, whether that's air, hybrid, liquid, rear door, um any type of environments, you know, concentrations that could cause breakdowns even potentially long-term should be, you know, considered at the very beginning of the site selection. So, the next step in the development process is the permitting contracts and raising the capital. So, Glenn, what does the permitting and contracts look like at that stage and where do projects usually get stuck? So, so, the permitting side on when it comes to land is it depends on the jurisdiction. Some jurisdictions are heavily regulated where you need to get you need to have the right zoning or rezone property, you need to get site plan approval and go through certain processes to get the get the land. So, that's one element of permitting. Another key element of of of it's a type of permitting is essentially you need to be able to to line up power. And some some you'll one of the reasons that that Bitcoin miners are in a sweet spot right now is that some of the some of these processes are have already been completed. They already have sites, they already have the necessary permits, and they're already powered, right? So, that's really attractive, but if you're not in that situation and you want to go get power, you typically have to go make application to the to the public utility, you have to do an engineering study, you have to you you typically have to to in some sort of a batch queue, where you pay permits towards substations, and then you have to do substation engineering, and it's it's a it can be a certainly a multi-month and and sometimes yearly multi-year process in order to get to the point where where where you're able to provide the data center. But but in this industry and and in the current environment, timing is everything because these big hyperscalers, the the the artificial intelligence modelers, they're they're all about timing and they slot their compute within within you know, within time frames, and then they look in in the into the environment and try to find places that they can provide the necessary compute within their time frame as they which which constantly change as as these companies have grown and got and and accelerated in terms of growth. So, those are the the the key things, but it all comes down to timing. So, as you can imagine, it's a very complex process to build these data centers, and the technology is ever-changing, and a lot of these chicken and egg problems as well. So, AJ, how do you raise financing in such an environment? Like what do lenders and investors look at and how to manage it with all the permitting and all these chicken and egg problems everywhere? Yeah, it's it's a lot of parallel processing, right? So, you do encounter the chicken and egg situation. It's a lot easier to raise equity and debt financing when you have an off-take agreement already. A lot of off-takers don't want to sign an off-take agreement until they see certain level of commitment on the development side, right? And that includes, you know, having land under control, it includes having the power allocated, and having a schedule from the utility. That's rather firm, right? So, it needs to be approved, you know, like in Texas through ERCOT, right? The compliance organization. And so, if if we're raising specifically for like say um you know, a 20 MW gross, you know, facility, that could be anywhere between, you know, 50 to 100 million dollars. And you're going to need to bring in probably between 15 to 30% equity on that, right? So, you have to, you know, have very good very good networking with capital partners that are in the space that believe in in that site, right? And you typically want to do that through project level financing rather than, you know, raising funds just through your holdco. And so, for for that, it's it's become much more friendly, especially in the last like 24 months, right? There's a lot of credit funds out there that are willing to fund these projects. But the hypercritical component is the equity piece, right? You need to You need to have that sponsor equity committed so that the lenders feel comfortable utilizing the equipment as collateral and actually, you know, paying for up to 70 to 80% of the entire project and infrastructure. So, the next step is the construction and infrastructure buildout. So, Jay, another question for you. Can you walk us through the buildout? How does that work? It definitely depends on the site. You know, we we started Zidoki a little bit under 4 years ago and we were we started in the Bitcoin mining space. And so, it's kind of a cowboy-esque development process, right? You can function as your own general contractor and coordinate with with vendors on the MEP side, right? For mechanical electrical and plumbing. But now that you know, we've shifted our focus to developing HPC sites and and AI data centers, we really have spent a lot of time you know, talking to different partners that can help us on the design side, on the engineering side and get us to a level that we feel comfortable on making our own equity investments in these facilities. And so from from our standpoint, right, the development process looks like a lot what Glenn mentioned earlier is we got to get our permitting. We need to make sure our land is is under contract, the power is allocated, and then we move into the design phase, right? We get to 20% it's called like 20% of design, and then we can kind of go to market on that property and go and and, you know, sell it to off-takers as well as raise the capital associated to be able to develop the infrastructure. And Adam, how does the cooling and filtration stack actually work? What What are the components that it involves? And what are maybe the most overlooked components that can be very very important in the long term for a data center? Well, we only have 10 minutes left, and that's pretty complex in this environment right now, but I'll I'll try to cover a large majority of that. So, I think the first thing on, you know, that maybe in the transition over is really doing a an evaluation building envelope of your current infrastructure, right? And then you have to look at what are the requirements of what your client is asking you for. You know, are you needing two N on everything? Are you needing you know, how much redundancy and uptime? So, when you build your infrastructure, you have to look at it as not only you're really building it as a a protection device, which is the asset. Assets, and the assets are really what's inside that that white space or inside the data center. And by doing that, you know, taking that approach, you have to evaluate how much, you know, investments you're going to have to make to meet those demands. And that will vary across multiple, you know, data center sites or mining sites on how big of a leap what that looks like. But the mindset really needs to shift from particularly in the mining community of you know, moving from as much air flow as you can through a system if you're using air to more of a protection mindset of you're really now, you know, signing agreements and you have to hit those commitments of those agreements. So can your infrastructure support that? And in doing that, you know, a lot of folks like Jay mentioned you know, he was his own GC. Well, that's that's really that that isn't the way it works in the data center space when you're when you're transitioning over. You're going to work with, you know, a general contractor. You're going to work with a an MEP design firm. Your end user client is actually probably going to do a majority of the designing with you know, some partnerships with MEP firms as well as the provider working with their current infrastructure. From that step on now you're bringing in different contractors, sub contractors, drive mechanical contractors that are doing your piping and they're doing the they're doing all the installation side. So it is a multi-step process and what you have to look at is really from a holistic approach. You know, so by by putting the all of those components together, that's really on the infrastructure side causes you to be way more profitable from the beginning instead of going back and and fixing your problems later or potentially having penalties where you can't meet, you know, your your your uptime agreements and because there's something in your infrastructure that you know, your it didn't allow you to hit that hit that goal. Well, that's potentially you know, an end game mistake. It could be potentially. So, you know, I think some of those infrastructure on the front end is very very important. You know, and you mentioned that that the afterthought typically is most of the time uh you know, as we transition into HPC, you know, with liquid and and rear door, um a lot of the afterthought is the air side of that. Everybody thinks there's no air in those systems and uh the air is protecting the other the building envelope outside and all the components that are actually supporting those liquid modules, the the electronic inside, um all of the different components. So, um it is a a really a holistic approach. So, I would say that, you know, from be sure that when you are signing these agreements, when you are signing these SLAs, that you've checked, double-checked, and ran, you know, a lot of simulation modeling that you can meet the demand of what your client's asking you for. Welcome to Predict. The [music] world is a market. Everything is a market. Every headline moves the line. Every moment is your market. [music] Call the moves. Bet on your instinct. Your prediction, your edge. Dual bets. Predict, where everything is a market. And um so, I Glenn um I don't just start on the penalty aspects and the risks regarding the contracts with the with the customers. So, how do you manage all of that risk? Yeah, this is this is uh if if if anybody in the audience has uh existing sites and has uh has has has been running um you know, Bitcoin uh data centers, the opportunity to transition uh these sites and your power to be able to to service artificial intelligence AI type data centers is is amazing. It's It's It's potentially very lucrative. But it is fraught with with with peril because it's very different. The the type of of data center that is that you typically see with with with Bitcoin mining is a tier three or below data center, whereas the sort of data center that uh the AI guys of the world, the the Anthropic's, the Open AI's of the world are looking for is something called a 4/9 data center, which is that has to be up 99.99% of the time. So, it means you you can't you can't be down more than uh 52 minutes a year, essentially. Uh and and the contracts that you see with these deals, and they're often something called an MSA or master services agreement or for a colocation, is that they have a series of provisions that that require you to have interruption, what happens if the data center goes down, um and they and they and they provide uh service mechanism for you to respond promptly to to various service issues. Um and the way these these MSAs typically work is they provide that if you don't meet certain availability thresholds or if you don't meet or if you have service interruptions, uh and then then the uh then then your customer can hold back your monthly recurring payments uh in in what what's called a credit mechanism. It's uh and and what happens is they'll just say, "You were down for more than X number of minutes, and I'm able to penalize you 10% or if it's longer, 20% or 50% or 100%." Uh Uh, and and you can very quickly find yourself not getting any revenue for the month. And yet yet you still have power commitments, you have uh, debt service, etc. So, the key here is you've got three three basic ways for you to for for you to mitigate uh, that issue if you're a if you're a a a Bitcoin uh, data center uh, uh, seeking to to go into the AI space. Um, one way is is uh, through uh, through your design. You know, how do you how do you design your data center? Do you do do you have sufficient redundancy? Do you have proper systems? Uh, do you have uh, uh, uh, do you have the necessary uh, mechanisms in place to make sure that your the data center you're providing is has has sufficient redundancies that if something breaks, you can you you know, you can uh, you can fix it. Um, but uh, the the the other approach is to try to shift risk uh, to to your contractor, to whoever's you know, building the data center uh, in terms of delivery uh, etc. because if you're late in delivering a data center, you can have you can have penalties of 80,000, 100,000 day a day or an escalating. But, the main way that you you you manage this um, is is in your contract, in your negotiation of your contract with your counterparty. Because these contracts um, uh, you you'll you'll be dealing with somebody who's got a big law firm in New York City who'll say, "This is the way, sign here." And sign your life away. And you have to be able to know what's market, what isn't market, and you have to just you know, you have to to to try to do your best to be able to to shift risk and control so that something goes wrong that you're able to uh, uh, to to manage it. And and the final way is just expertise. I mean, having good people who who can who can maintain and operate the system. It's not just design, it's not just contract, it's also how you execute and be able to to to run the data center. And And that's frankly the the the an area where where Bitcoin miners have a lot of great expertise and a lot of great people that are used to these these situations and you can harness that and use it to to to to to move up to AI data centers. So, now we have about 3 minutes left. So, we will go through each panelist will have 1 minute to talk about one thing the industry needs to get right in the next 12 months. So, let's start with with Adam. Uh yeah, I mean, we we touched on a lot of a lot of the topics, but I think if you're looking to make the shift to AI and HPC, um you know, along with all of the the financial obligations, the infrastructure parts meet those demands is key. And you know, if you're searching for, you know, the lowest power rates out there, um environmental does play into it. And as far as looking at it from a very controlled environment for your infrastructure, where Bitcoin mining is a not a very controlled environment, where inside that building envelope of a data center is a very precisely controlled environment. Um and so in saying that, uh you know, with the environment that it goes in, a really cheap megawatt and a dirty environment can become a very expensive megawatt um later down the road. So, uh you know, I I think that is a major part that needs your infrastructure has to be solid before you start signing agreements. And as well, model and simulate that beforehand to make sure that you can meet that demand and the and the requirements, uptime, and the redundancy. Yeah, that's great. Um on on my end, I'd say one thing that's not really top-of-mind when we're when you're developing a data center, but I think will really be top-of-mind in the next 12 months is um the community engagement aspect of it, right? Um we're seeing a lot of communities um really get more involved um with their city planning departments, and you know, voicing their concerns and opinions associated to uh data center developers coming into their communities. I think it's really on us as developers to help educate, you know, the community and help them view data centers as the necessary infrastructure that they are, right? Um just like how, you know, we need, you know, trains, railroads, we need uh cell phone towers to be able to communicate with each other, we need data centers to be able to communicate as well, right? And and fuel our infrastructure for the future. And so, that's that's something I'm going to be spending a lot more of my time on, and our team has been spending a lot more time on. Um it's something where, you know, we need to bring in some experts to to assist us with that, but I think a lot of good firms is have really gone ahead of that as well and start start already engaging with the communities way before, you know, they break ground, way before they even get their permits approved, right? I think that's hypercritical going forward. Um and then one last component, aside from the, you know, community really development side, and something we didn't really touch on too much, is um you know, time to build and like modular infrastructure, right? Um it kind of plays into the community aspect as well. Some communities don't really care. Some communities don't want to see a bunch of, you know, modular buildings. Um they're they're much used to seeing the more aesthetic, you know, commercial style um power shell build-out. So, I I I'm really bullish on, you know, let's call it Bitcoin mining infrastructure providers that are moving into the HPC space and are going to be able to figure out um how to get the modular setups um deployed effectively, right? And cut down our time to build, Um that that's pretty critical as well. And so I I'll we're over time here, but I will say power, power efficiency, behind the meter power. I think on the community engagement, I think one of the reasons that that there's so many people are up with the pitchforks, environmentalists, community activists, etc., is the the fear that their their power bills are going up. And and frankly, one of the reasons that power bills are actually going up is because of other forms of power plants being shut down, but I think it's important for the data center industry to embrace some of the guidelines that have been set forth by FERC, the Federal Energy Regulatory Commission, for data center developers and data center users to bring their own power behind the meter, build their own power plants, and and not tap into the grid, and and strain already overtaxed grids. Thank you very much to all the panelists, and thank you all for coming and listening to this. >> [music] >> Every year this community comes together to celebrate, to debate, to build what comes next. >> [music] >> And every year the stage gets bigger. >> [music] >> Sound money center stage. So, where do you go to celebrate the next chapter in Bitcoin history? You come home. >> [music] >> Nashville, July 2027.