Can Data Centers Deliver on the AI Promise? | Bitcoin 2026

Hey, amazing job loose. Thank you very much, guys. Thanks for being here. And I I love hearing that there's front row energy. We we we will start with just some brief introductions. So, thank you very much for uh for joining us. I was raised by a southern mama, so I can't help it. So, I'll start with Francesca. But uh Francesca, you you uh named uh Forbes Italia 30 under 30 and recently received the Gamadana award as the top entre entrepreneurial person in in Italy. That's amazing. Would you tell us more about your background? >> Thank you. Uh yes uh I founded uh Alps blockchain um in 2018. Uh we are a mining company. Uh we are operations around the world. Our biggest one is in Oman where we have 150 megawatt and then we have um two sites in Iowa where we have 70 megawatt and then South America between Paraguay, Bolivia and Ecuador. So we are quite uh geographically diversified around the world. uh but we started from Europe so from Italy uh where is our headquarters uh with hydropower facilities and uh we are an infrastructure company we build mining farms and we build data centers. >> Oh that's amazing. Thank you very much and near I don't I don't have any other rhythms so we'll just go down the line with introductions but I uh so I've been a big fan of Mara for many years and when interesting about you I loved hearing about your background as an aerospace engineer and that also you have a paragliding as a hobby but tell the audience more about who you are and what you do. >> Sure. Yeah, thank you. Uh glad to be here. Um I joined Mara about a year ago. Um for those uh who don't know Mara um uh quite a big uh Bitcoin miner um and emerging as a digital infrastructure company um personally my background is actually in robotics, aerospace, autonomous vehicles. So kind of the intersection of hardware and software and I realized that the uh uh emerging uh need for intelligence and democratizing it is something that requires energy and that led me to Mara. >> Well, that's amazing. Uh Erin, I'll give the floor to you next. I I also enjoyed learning a lot more about Hydro Host and the impressive work that you guys do as well as a little bit about your background. Would you tell the audience more? >> Yeah, so I'm CEO co-founder of Hydro Host. So we help data centers essentially make money off of their megawws. So we are a software company that automates the facility to make it possible to access uh AI customers and resolves all the components needed at the data center level to make it much more easier for them to bring their comput capacity to market. Uh, so you think of us as like a headless Neocloud. Uh, we are the largest quoteunquote neocloud in Nvidia's orbit that actually doesn't own hardware. Uh, so the data center owns a hardware. They get the upside. We're a management co and a software company that helps accelerate the adoption of their infrastructure. >> Oh, that's fantastic. Thank you, Xander. All the way down to you. But I I just have the real privilege that you and I were able to share a meal together in Los Angeles two and a half years ago. Well, authentic Sichuan hot pot. One of the hottest meals I've ever had. I love spicy food. I was sitting there holding my breath trying to stay alive. Uh Xander, I'm a hardcore Bitcoiner and you're famous in our industry. So, it's a real privilege to be on this panel with you, but I understand not only have you been building Zillian networks, but you're also launching Tesser and I hope that you'll tell the audience more. It's really my pleasure being here meeting good old friends and also actually my company collaborated with uh hydro with Aaron's team quite a lot. Uh so I'm Xander I'm the founder and CEO of Zillian network. Uh Zillian actually is AI compute side you know company. So we aggregate and fit out the GPUs and lease that to AI customers the AI offers. Uh before that I uh built data centers the traditional tier three data centers uh but also I took the uh role in bit main as the global mining data center development uh head. Uh so we act as the designer but also the offtaker and even backstopper to some big uh projects and now we are launching uh sister company named the tessa uh which is you know to build uh uh design and build teslated AI data center to speed up the adoption of the AI. >> That's fantastic. Well, I I am going to take the first question back to Xander, but then we're not going to go down the line with the same question. I really encourage this panel to engage with each other. Obviously, you heard from their backgrounds, tremendous uh experience for this exact topic, but Xander, I I will I will give it back to you and then uh I keep the conversation going. Would you do two things? Would you start by just painting a big picture for what is going on in the AI data center buildout space? like what sort of large capacities is are are is really happening and then is it actionable like what you know what what's actually out there that can be built and be developed >> oh I just think this uh industry is amazing you know the AI demand is booming so based on McKenzie uh research uh by 2030 there will be 7 trillion US dollars invested in the global uh data center and 5 trillion of which will goes into the uh AI art the AI data center right and also 150 something gawatt demand you know aggregated and also you see from the news that 300 uh billion US dollar will be spent by hyperscalers in AI this year so I just say there's a huge demand but then the problem is you know everybody say hey we are short in shortage of chips the powers but actually in my mind you know how to turn the power into as you said actionable or you know uh the the AI capacity I think that's the issue then I think you know find that the AI data center the build out the conversion rate actually is not as big as you thought everybody says gigawatts level compass but how many megawatts actually been you know break broken ground and and why you know there is always like a delay from the market you see you know the new class the hyperscalers got project delayed you know so uh in our point of you the whether you got the project designed uh from day one to be contractable and bankable is the key. So I can explain that a little bit more at a later time but I just say you know this is a big picture and also the problem uh is there and how to you know turn the AI project you know into um you know design that into contractable and a financable things that's the that's first the topic in my mind. >> Can you expand more about the financability and contractability? Oh yeah. Uh so there are uh three folds right. So first if you do the AI data center build out you need to exactly know who your customers is right and what kind of AI workload are they work running what kind of GPU clusters you know the models they are running. So you know today we are doing you know the Nvidia like the black well GPUs you know the racks density you know is now already 130 kW today but but it will soon ramp up to 230 KW you know in half year or year's time and it could be even higher density. So if you cannot design your AI data center you know to fit that needs right that will be a disaster and also if you have the super lonely time items if you you know have like thousand people working outside but you have a shortage of the uh the the labor forces you know when the your data center is ready then the new generation GPUs will come. So I think first you need to exactly know your c who your top customer is and you need to know the timing and the second thing is you need to understand you know who your lender is what they need right the lenders definitely they need high SLA the the service level agreement assurance but also they want to seek you know repeatable and credible design that they can easily to do the due diligence they can easily see hey the track record from others hey whether this design has been built and and and and met the timeline before and the third thing I think is the sequence. You need to first make the your the correct design and then make your uh project contractable and then the lender will you know see the offtaker and start to finance your project. If you do this you know sequence wrong this project will be a non-starter. >> Hope that explains. >> Thank you. I'll take that. >> Erin, I I might uh I heard something interesting there that I think that you could add some context to. You know, Xander led with understanding the the type of machines that the customer is running as well as their power draw and I learned a lot about hydrohost. You guys have actually solved for that so that the data center operator doesn't have to understand those details. Would you add some context? >> So on the the question that we're supposed to be here for is like can data centers deliver? The answer is like there's no other option like AI has to AI is basically pressing forward and it presupposes the need to convert power into tokens and and I I I think that yes there's like lots of questions around how to actually move by you know from like will this land work or this power draw work or this etc. But the overall kind of arc to give like people confidence of what's actually happening is it it's why Bitcoin is such like a great metaphor for AI is like essentially a democratized access to money and as you said it's like this democratized access to intelligence. It's like it's not like oh I'm moving my web app to GPUs. This is like a completely different way of thinking how energy can be translated into a good and that good is inelastic. Like how do you tell someone if they're fully intelligent or not fully intelligent? Like if you have kids, do you tell them all right you're a six sixth grade now enough intelligence for you? Right? It's an it's an inelastic good. There is no real price on it. We create all these like abstractions. College. So Harvard says $80,000 a year is smart. Or another school says 60 or 30 or 40. We say a degree, we say, you know, being a chess champion. Like we create all these abstractions because the human the human itself is actually a really horrendous test subject. We have really bad benchmarking because we're all so uniquely different. We're all skilled at different things. And so this is why I'm so bullish on the space is because it's not like anyone said no to intelligence. Just like everyone, if you look at your phone, go actually, you can look this up. Go actually look at the number of gigabytes on your phone of data you actually have deleted. It's usually like less than 5%. total data stream. You actually something you deleted like intentionally. So, so it these are all great things because it it allows compute to move down stack to make it where previously just like money that Bitcoin moved money accessibility away from the Federal Reserve to everybody. So now we have the ability with technology to move intelligence away from let's say universities, elites or kind of however whatever benchmark you want to create to literally a machine and that raises the profile of so many other countries or let's say specifically this data centers that were previously unaccessible that were not interesting to a hyperscaler or to a tech company. Now they're salient and as well as the the the profile that miners bring on the data center side or the exact type of cost profile that will win which is low cost, no I'm the Costco of tokens. That's that's literally what every single AI company wants. They don't want hyperscaler like tier 4 magical thing of look what this box can do. They're like but does that cost like 40% that Mara does? cuz I'm like I don't know if that's worth it, man. Like cuz you just give me Jensen and Mara's going to give me Jensen. So what's the difference of your Jensen? Oh nothing, right? And and that's just really hard for hyperscalers to compete with. But miners are already programmed that way to say the only way I win is better costs. And that that ultimately is the end entropy of this entire space is like if you can convert power to tokens at the most efficient rate you will always win and everything else is just noise. And I'll add to that just a second. Like I love what you said about the kind of the energy. If you look at the horizon, the cost of the cost of intelligence converges to the cost of electricity, right? And so similar to how I think um the uh telecom companies have enabled us all to reliably consume so much media and content online on our phones. The same way you're now going to be expecting your providers to deliver the same reliability on intelligence, it's like your own chief of staff, your own partner, your own like you know um thinking uh apparatus in a way. So that's something that >> yeah like like you're exactly right that like it's deflationary. That's why it compresses down to the marginal cost of producing energy is that just as telome telos uh any other like major infrastructure provider it's all deflationary and hyperscalers are not deflationary hyperscalers are extremely expensive and the rents that we've been paying to three providers that monopolize compute for two decades Jensen just drove a bus through and you know the marketing is AI AI right that's the marketing pitch but but literally Jensen drove a bus through public cloud and now data centers that previously were just relevant for you enterprise colo mining. Now they have a completely new market they can go into that is a very very lucrative market and you can think Jensen Jensen completely changed this the infrastructure um topology to where you can be just a power provider and you can just provide compute directly and you can just walk away. It's really it's really amazing. >> Thank you uh Francesca. I want to uh um pull you into this conversation when when people say that we need more data centers for AI capacity. What are they missing? >> So uh for sure one thing to consider is that AI runs on a stack not only on power that means that um it's not so easy to convert uh power into AI data centers. That's what we are saying from the beginning. Uh because you need reliable power and this is a really important key. Um that's completely different from like the mining activity. Um we as miners are used to build data centers uh from scratch in bad conditions. We uh for example with Alps we have data centers in the Amazon forest um that are not like they are off-grid and they don't have a um a reliant uptime as it should be for an AI data center. So the first thing to consider is for sure is the source of energy that uh is going to be used the right one in order to run an AI data center because we are going on to develop um renewable energies but renewable are not stable. That's the problem. And so uh all the things that um we as miners are working in a grid in order to be a kind of uh stabilizer of the grid that doesn't work with AI. So if we convert our mining farms into AI data centers, there will be a problem in the grid to be handled. So power is the first thing that is really important. Second is connectivity. Um a mining data centers can work with Starlink. We have some examples um can work with uh satellite connection uh AI data center need the right fiber connection. This means that not all the source of power um and this the available site can be switched again from mining to AI. So uh I think that um it can be a synergy um that can work. Uh for example, if you have a big facility that is purposed for mining, um you can repurpose a part of that um into AI in order to have um only a part of the load that is 100% uptime with the right connection with like the right keys because you need redundancy for example that for Bitcoin mining is for sure less. Um yes there are sites that can be 100% converted but not all while the demand is contin uh continue growing the infrastructure is ready for that we don't know uh because um if we read the data um there are plenty of GPUs available to be installed but both by the hyperscalers but the bottleneck is you need the infrastructure you need the data center working and building that is not easy. We as miners have a lot of knowow on how to convert power into computing power but uh the knowow for the for the AI part is different. We also need to learn that uh and it's not easy to switch and not all sides can be switched uh easily. Xander, one of the things that I heard uh Francesca from Franchesca there was about reliable power and cooling and a few other variables, but as we as we look at can data centers deliver on this AI promise. Is is there a variable that we're missing about delivery timeline like that these hyperscalers have an immediate need or a future need but can we transition? What can you talk to us about delivery timelines? uh so uh so I just want to rephrase that you know the we check the box of the technical requirements and but then we need to also understand you know what their immediate demand for now what's the demand for the midterm that long-term right so I think so actually uh you know what we see from the market is that uh people will just be no-brainer you know take whatever the the capacity available online by this year by you know by first half of this year but when we ask hey if there's any uh you know data center uh capacity online by 2028 20 2029 what's your what's your idea so actually I got different answers but most of the people just say hey give whatever you have today and we do have no time to think about you know for two years three years later so actually that's also my question ask you know franchise and Aaron you know based on your experience you know is the contract terms really matters I think the answer is yes but how much that matter to your you know decision to turn this you know facility into a data center and to to Aaron uh you got a lot of hosting deals right and did you really see seven years eight years long-term you know GPU releasing or just the three years to five years I think that's also my question yeah so I'll just answer uh quickly on the uh hosting kind of contract terms like a aka like offtaker demand I so like I I think that the way you got to think about making money or like returns is that anything long-term means you push risk to to the end customer. So everyone here flew right here logically uh let's say with within you know unless you're from Vegas how many of you bought your tickets like last minute versus bought farther out and I think if you reframe the problem statement away from risk management which is a bank question not an in customer question just like none of you cared what the financial situation of Southwest or United American Airlines is. you paid based upon the convenience of when you wanted to transact and if you bought nearer term you paid more versus you bought longer term you paid less. If the entire market of the airline industry was driven by you paying 3 years for all of your travel with a deposit of 20%. The airline industry would be very small. Think about homes right home when you buy a home they you generally do that on average 10 to every 10 to every 15 years. So like the market scale is directly connected to the size of the transaction and how often that transaction occurs. So if a data center is going to make money, they have to realize that if you're just pushing risk to an end customer, you're not going to make very much money. You're going to have to lower your return profile because you're moving risk to him. If you take on more risk, you could make more money. It's exactly how you look at bookings of this hotel or bookings of an airline or really any other utilization of any heavy capex sort of product. So, so data centers have to accept that the way that which they're going to drive returns is revenue mix. If your entire profile of your book of business is three-year contracts, you're really not going to make very much money. But if your entire business is completely on demand, you may make some money, you may not, right? That's the bet you make. It's the it's the dice you roll. So our answer is it's a mix just like this hotel that the Venetian that we're all staying at has a mix has corporate contracts has long-term arrangements has on demand has flex capacity right and and the data center will evolved into that type of strategy uh over the course of it would just be forced to iterate because right now if you had an H100 cluster that you bought in the even the height of the craziness of 2023 you are making so much freaking money right now it's like unbelievable. But the only way you make that is you have to be liquid in that capacity cuz if you lock it up for a 5year long deal, you're not making any of that upside right now. In fact, you're paying a$120 dollar10 or something like that when if it's liquid on the market, you'd be making $225 right now. >> And if I can add something, um the obsolescence of the uh GPUs is growing a lot during time. So it's difficult to find out the customers that science for having like computing power for long term because the technology will change a lot during this term and uh these companies are really focused on okay I have to deploy I need GPUs for inference I need GPUs for training and that's what I need for a short time because they are fast growing but they don't have idea of their needs during time and that's like the the issue that uh the infrastructure companies are having because you are investing for example 10 times more for the infrastructure. If you are building an AI data center then a mining data center but then like the the question is okay uh how long can I have a contract contractual um computing power so how can I run my business plan and my model I don't have any certainty and that's the big issue right now >> wonderful near I I we're we're in our closing minutes and I want to uh ask a real direct question we heard um we heard Franchesca talk about you know AI spending more but If if enterprise clients are ultimately going to want to control their own capacity, could you talk to this audience who might be more familiar with Bitcoin mining about what what what's the gap between the capacity that's being built today versus what the hyperscalers or enterprise commercial clients actually need? Yeah, I think going back to what Aaron was saying, uh, a lot of these data center are actually thought of as they're infrastructure projects and you could think of them as essentially like a hotel. You are building the infrastructure, the capability and then you're going to be ideally maybe having commercial contracts for long term on some of the rooms and then some of them otherwise. And then came about other models, right? Call it Airbnb style, right? where you might want to have a temporary allocation uh with a specific unique vibe in a specific area that you so want and I think you having like and there's also another model there which is uh like ref like uh you know furnished apartment all right that might be long-term or a suite in a hotel that long-term stay so there's everybody right now is are building hotels but I think the emerging need to have your own control of your data whether it's sovereignty or sovereign grade, whether it's the modularity of it to be able to handle the planned obsolescence, then there's going to be um an emergence and you can see it here obviously in around the modularity of the build um ones that are compatible with the crazy cycle of hardware, the changes, um but also enables you to be closer to customers and what they need and how they want it, but also putting it uh where they want. At the end of the day, humans are physical entities. Even though all of us are using, you know, the hyperscalers through our applications, uh, enterprises, uh, have a sense of, uh, a need for control. And if you, I just heard an anecdote of a Spanish startup, uh, that was building their entire, um, capability, entire stack, entire product off of Claude. Um, and the harness was essentially detected at Claude, uh, saying that uh, they violated terms. They shut it off. So an entire company shut down just because they've essentially outsourced their intelligence to a third party and they had no control of the model, no control of the inference engine. Um and that's a challenge. So making sure that enterprises have that control um means that they might need to have their own data center or mini data center or a super modular one because essentially every enterprise needs their own digital manifestation and that's their access to intelligence. That's fantastic. You know, I I I said in an audience uh Pomp's event in New York uh a year and a half ago and made a note to myself that these Bitcoin mining pubos are actually pivoting to AI data center uh infrastructure and this is just proof of it. I I do I do think we had a good conversation. We're up into our closing seconds. If you guys would please thank the uh panelists and appreciate you being here. Every year this community comes together to celebrate, to debate, to build what comes next. And every year the stage gets bigger. Sound money center stage. So, where do you go to celebrate the next chapter in Bitcoin history? You come home. Nashville, July 2027.