Keeping Your Cool in a World of Hot Compute | Bitcoin 2026

Hello everyone and uh welcome to our panel. Uh my name is Yaran. I work for ETC E2C partners a data center consulting company and I came all the way from Cyprus to moderate this panel. So first um I'll let panelists introduce themselves. So Matthew please please start. Hi I'm Matthew Carson chief business officer of Hash House. Uh we design, engineer, manufacture immersion and hydrocooling infrastructure for mining. >> Hi everyone, very nice to meet you here and at lunch time you are here. You must hunger for the future and hunger for the knowledge. I'm very honored and I'm sin from China heat transfer technology company. >> Hello everybody, my name is Jeremy Singer with Exxon Mobile. I'm here representing our immersion fluid line. My history involves um many years in HPC and AI compute and also lowcarbon data centers. So um Matthew, you are a system integrator of data center cooling solutions. So can you give everyone an overview of what is uh data center cooling and what is it important? What are the um challenges? Yeah. So these days there's really uh three main technologies. There's air cooling which I think we're all very familiar with. Uh hydrocooling which is uh using water direct to chip with uh large central distribution units, large dry coolers uh to cool down the servers and equipment. And then there's immersion cooling which is the actual physical uh dunking of the servers in fluid uh of course for the purpose of cooling them down. Is uh cooling fully solved or is there still a lot of inefficiencies that could be solved over time in the future? I think there's still a lot of inefficiencies that are being worked on. I know uh especially in mining a huge one is uh heat reuse uh especially say in Europe with uh district heating power plants with the heating and drying of wood other uh you know vegetable products. So I think there's a lot of lowhanging fruit out there for let's say better use of the byproduct of these cooling systems which is heat. But uh in terms of the cooling themselves uh I think the systems have gotten very very advanced very very quickly depending on your needs and uses. >> Yeah. And adding on to that a little bit um you know technology matures at a very rapid pace. We're all familiar with Moore's law and how chips improve their speed, you know, on a year-over-year kind of cycle. Cooling infrastructure is is very similar. It's a physics game where efficiency matters, right? Every single dollar, every jewel of energy being used is is most advantageously applied if you are as efficient as possible. And there's a lot of technology opportunities to continue to mature there. And eventually there will be head uh a roof line where u these challenges are going to face us head-on. And so I think there's some continued work to go there in the cooling space. >> And uh another question um have miners being the one who really uh made all the innovation in the data center cooling industry over the past years or where who has been the the drivers of that innovation? Is it miners or traditional data centers? I think it's still split pretty heavily in favor of the traditional data centers. They have uh lot larger R&D budgets, a lot larger capital expenditure budgets. But um on let's say the the lower or leaner end of things is probably where uh the innovation happens in the Bitcoin mining space. You know, we're all technically competing against each other globally. So we need to get the absolute best efficiency, the absolute best uh pricing of our infrastructure down to I think a much more detailed level than at least was traditionally in the data center space. Now that uh AI and HPC has exploded, I think they're starting to see the same sort of constraints around capital efficiency that they're starting to learn some of the lessons that we've learned over the last couple of years. And uh Serin you do um uh heat exchangers so it's a very specific part of the cooling infrastructure. So can you talk about uh the different components in the cooling system and what are the most important parts that maybe data centers overlook? Uh Sh definitely I'm very uh pleased to introduce aluminum plate fin heat exchangers in the heat in the mining sector and the AI sector application. In the past the old traditional structure is stainless steel structure or copper brass structure. The heat transfer efficiency is not so high and use aluminum with a very complete size and the high heat transfer efficiency can reduce a great deal of weight and the heat transfer efficiency is very high and it is already has a very big applications in different sectors and it is matured technology in this sector And uh I think it will it is a revolutionary subject for this kind of application in the mining and in the AI sector especially for the uh GPU CPU. they have have already started to use this kind of product and for the mining as well and it most of the people no matter they choose the immersion type uh mining um method or that hazel cooling method they are start to use the alum minimum and I think this is very good to the future >> and Jarren just going back to your prior question just for a second around uh the maturity of the various cooling technologies. I I'll say some I think you can always trust a Bitcoin miner to do something a little radical in pursuit of efficiency and optimization and uh for that reason I give a lot of credit to the early Bitcoin miners trying immersion and trying things in weird places right uh typical data centers at scale are going to be in you know high uh capital intense areas with a raised floor and lots of good air conditioning and perfect environmentals and that's good for certain tech but you can trust a Bitcoin miner to innovate and iterate to find new technologies as well. So um you know I think immersion is a is a good example of um miners brought that really at scale kind of earlier and it's starting to take off a little bit in the data center world here and there depending on the technology. >> So when we talk about the different cooling methodologies is air immersion and hydro. I will just throw this question out to all of you and whoever you want can answer. What are the main differences and uh what are the tradeoffs and advantages of the different and in what applications can they work better than others? >> Uh for the >> Yeah. Go ahead. Go ahead. Go ahead. >> Thank you. for the air code mining sector or for the AS sector if you use the air code and the PUE value is 1.5 to 1.8 eight. That means for one mining factories if you choose hydro cooling systems you can see save about 30 million US dollar and uh today when I in my booth and some people is checking with me about the electricity price because it is the main cost of this of the mining operation. But if you choose the hydro cooling systems and this will help you reduce have a such great deal of saving. So I think uh if you try to reduce your POE value it is maybe more important than you choose uh compare the electricity price. just use advant advanced technology to saving your cost to increase your profit profit. >> Okay. >> So some of the the other differences between um hydro air and immersion. Um air in general as as kind of Sulin was mentioning is far inferior to hydro and immersion in terms of power utilization efficiency. So, in general, hydro and immersion are going to be 30 to 45% more efficient than air. And then when it comes to hydro versus immersion, um, and we were, you know, talking about this a little bit, um, backstage and and elsewhere, it's it's really about the the customer. And I know Matt has some opinions on this as well, but, um, both have their pros and cons. Similar efficiency, similar capital requirements, but there are some operational differences from a hydro versus immersion. and it can be highly sight and customer dependent. Maybe you can throw a little bit more detail on that. >> Yeah, I would say you know in the eternal debate between hydro and immersion uh especially these days as the technologies in the mining space have become fairly mature uh really comes down to what the customer is looking for. Uh, for example, in a very cold climate, uh, I might recommend immersion as the fluid generally has a lower freezing temperature, less issues with expansion and contraction in colder temperatures, um, and less chance of, you know, a small leak on the floor of the container causing ice causing someone to slip and fall. Whereas, you know, if uh we were looking at a hot desert environment, I might recommend hydro as most immersion designs are what we call open tank designs and the chance of foreign debris entering the system like sand and dust is quite high. In those circumstances, I might recommend hydro. So, uh it's overall a tradeoff and it depends on which environment you want to deploy the infrastructure in. Um and I can I also want to ask um what are the capex differences between these three cooling systems? Matthew um from our experience uh generally of course air is the cheapest option has the least number of uh subco component sub assemblies. Um but for us for mining the price difference between immersion and hydro is actually very very close. Um, for most customers, the system quoted price ends up being pretty much within a rounding error. So, I'm not sure what uh experience is on the HPC side of things, but I would imagine it's quite a bit different. Yeah, depending on the scale of an HBC system, um they can get, you know, and how much infrastructure and what your floor white space looks like and the kit buildout costs that can that can impact whether um you go with hydro or immersion. I'll also say that your operational support staff can be uh impacted by this, right? So, you know, with with hydro, you're talking about closed loop systems. You're talking about cooling delivery units, CDUs, um, and a simple um fluid that is that circulates kind of like a circulatory system. Whereas with immersion, there are uh more variety of tank types and oil choices and um maybe some more operational support considerations in terms of training staff and um places to maybe drain for support purposes to drain your gear um to keep um when you're doing maintenance. So there are some different operational um uh considerations. I'll also say that um you know immersion, you know, speaking for immersion, right? So it is it is a little simpler, right? You are you literally have a tank full of a dialectric oil that does not conduct electricity. You put the machines in that oil. If the tank is designed appropriately, which originally they weren't as it was kind of a newer uh style, but they have matured dramatically. You put them in a tank and essentially the systems stay in a in a very temp comfortable temperature range and the reliability of the systems are increased and and you don't have to touch them as much. Whereas there could be scenarios in the hydro world where um there are a lot of interconnection points, a lot of opportunities to drip. Um hydro fluid in this case is uh does conduct electricity and it can be corrosive depending on the hydro fluid you're using. So these are some of the considerations um from an operational perspective to think about when deploying at scale. So um hydro and immersion is about the same cost in the terms of capex and you mentioned a lot of um challenges and nuances of running hydro and immersion operations. It depends on the on the specific uh geography, the climate etc. the amount of dust, but if you can give a rough number on the OPEX difference, um maybe one of them requires less um less employees, for example. Are there some differences in in OPEX from all of this? Yeah, I would say for mining at scale generally um a well done immersion system whether it be containerized or building skidmounted generally should require less people to keep running um and is generally a little bit more I'd say uh operationally forgiving um with hydro especially the larger systems with the high pressure high flow rates that the water has to move at in order to remove and move that much heat. Um the the whole system tolerance is much much higher than immersion which has a much higher specific heat capacity. So is much more efficient at removing the heat and without each separate in this case minor needing a connection to a cooling manifold. You know you can it's much easier to just leave a couple of miners out if they're being repaired or need to be fixed. Whereas in hydro you have to be very careful about the overall system flow and pressure. So maybe if you are removing too many miners in one specific rack, uh you may actually have to power down more of the entire system in order to make sure your cooling loop is staying uh within parameters. So um you would say immersion is a bit more modular and easier to operate. Yeah, overall yes. and uh swirl in the heat exchangers. Are there any differences between the the immersion and hydro when it comes to heat exchangers? >> Uh in fact they can uh aluminum heat exchangers can be used both in the immersion type or in the hole type and the both both advantages they are very easy for maintenance. You don't need to spend a lot of time to checking about uh uh the block and the leakage because uh you know at the moment every uh units before send out they are modular typed they are already leakage tested and uh very easy connection and the fast uh uh can put in operation very past um I I think both immersion type or hazle type they are all very friend to aluminum and uh they are operation in a closed circuit don't have the evaporation of the liquid so uh maybe at the very beginning the invested cost is higher than the air cooled type but from the long run and the hazel and immersion type will be much more uh effective and the easy maintenance and have a long operation life. So um the liquid is extremely important in immersion cooling. So Jeremy, can you tell about that liquid and uh what does Exon Mobile do with that liquid? >> Yeah, absolutely. So, as I kind of mentioned at a high level before, what we're really talking about here is a oilike substance that is does not conduct electricity that you are putting the mining computer inside literally just putting the whole system inside of a tank. The tank is flowing the fluid through a heat exchanger um and is removing that heat externally. And the cooling fluids we're talking about here, so they don't conduct electricity. they are um maybe synthetic in nature. So created um with um you know chemicals, petrochemicals or other chemicals that are designed to be um to remove the heat and to be very and to not to not go on fire and to uh not conduct electricity as we mentioned and you know across the the landscape of fluids um all of them have kind of similar proc um similar properties. Some may have different viscosities, so they will flow easier or harder depending on temperature. Some of them may have different flash points. All of them should have lower flash points um than you need um in terms of compute, meaning the compute will turn off way before the fluid will go on fire. So very safe to operate. Um they should be uh non-toxic. So um and and Exommobile's fluid um hits all of these attributes, right? So very high quality fluid that will last a long time, removes heat without degrading and uh flows very well from a viscosity perspective. >> So um let's move over to Bitcoin mining versus AI compute and other forms of data centers. So um question for anyone of you, what are the differences in cooling of a Bitcoin mining data center, AI data center and a traditional data center? >> Oh yeah. So I'll start and then you guys can add some of the details here. So having um only mined personally and and not at scale at the levels that I' I'd really uh aspire to. But I have built and designed and run and supported multiple top 20 supercomputers in the world. I'll tell you that supercomputers are very finicky beasts. They need to stay cool. The workloads that we practice at XM mobile are very uh highly intensive all to all communications that require all processors and all network devices to talk to each other with full fidelity so that you can take a massive math problem split it into millions of parts. Execute those parts at in parallel sync at the appropriate times and then create a vi a work product. In this case, we're talking about subsurface imaging or or making pictures underneath the earth of what um oil reservoirs may look like so that you can optimize those. And to do that effectively, a supercomput needs to be fully online and and fully um communicative of itself. So, it needs to talk to each other with high integrity. Controversially, um Bitcoin mining is not like that. Each minor operates kind of independently by itself. Yes, you're mining together as a pool, but really it's a function of how much power are you consuming and how much power how much hash are you creating and whether a certain minor is finicky or not communicating. This doesn't matter as much as it would in the supercomputer or AI training world where a node that goes out during a training run can cause disruption across uh the entire run and and actually invalidate the run depending. So, I would actually um treat Bitcoin mining more like an inference workload. Uh more of like a question and answer workload where if I'm asking, you know, where is the bathroom? It's okay if it takes an an extra second to answer that for me. Um or if it's wrong and it says where is the the ball room, right? These are these are little problems that I'm less concerned about versus when I'm I want it to be absolutely right in a training load. So, cooling matters here. Cooling resiliency, cooling quality. If your cooling resiliency is low and your nodes are falling out very erratically, you're not going to get very much job, very much work done in the AI world. So, um, the integrity required in AI from a cooling perspective is much higher, but the Bitcoin world eats a lot of power. In fact, the more power you eat, the more money you make. And so, you want to stay cool so you can have long-term resilience. Long answer hopefully that you guys can add on there. Yeah, I would say building off of that, uh, it really comes down to a sort of, uh, resiliency and its corresponding redundancy requirements are the primary differences. You know, uh, from a cooling system or cooling fluid standpoint, your fluid can contain, you know, x number of jewels of energy of heat that can be rejected. That's a physics. That's a known that's a known factor. We know your dry dry coolers will have a certain amount of heat rejection capability based on certain ambient temperatures, certain fans that are installed. So those those factors don't really change. What changes is, you know, like my hydrocooled Bitcoin miner might be able to accept anywhere from -20C to 80C fluid intake temperatures, where I'm sure that most uh HPC AI servers would be absolutely dying if you tried to feed them 50C fluid intake, regardless of what fluid you choose is. So, not only do you have the cooling redundancy and the cooling deltat T that can be suffered by HPCAI versus Bitcoin, but then, you know, we talk about redundant networking, we talk about redundant security systems, redundant power systems, all things that are much more important when you need to keep your whole load live and happy versus Bitcoin mining. you know, if I have a dry cooler fail on one of my containers, that doesn't take down my whole, you know, sightbased load. I just lose the proportion of my income that that particular container represents as my, you know, overall project. >> Welcome to predict. The world is a market. Everything is a market. Every headline moves the line. Every moment is your market. Call the moves. Bet on your instinct, your prediction, your edge. Dual bits predict where everything is a market. >> So um we already touched on the nuances in cooling um dependent on the geog geography, the climate etc. Um so I have a question uh for u for you guys. How how does the different geographies impact the power usage effectiveness? Usually in the cold area for example in Canada and in um the area like Canada North Pole or South Pole the temperature is very friend and uh they usually have a very lower consumption for the dry coolers and for the hot areas and especially for the places like short of the waters and like a middle east maybe the immersion type is more suitable for them. And uh uh from the corrosion uh perspective if the products is very close to the sea and they need the anti-corrosion treatment. Usually when we design the heat exchangers we will make the design based on the installation city. We will first ask them the question where will your manner farms will be located? For example, if it is uh at the shore area and at the plate and at the mountain area, we will choose different design for that situation and we will make sure the heat exchanger or the cooler will per performance perfectly and uh we will meet that very demanding situation. Yeah, just building on what Sulin said, uh the location of the client site is actually probably the most important piece of information they can give us. You know, the more detailed weather information, altitude information all goes into what dry cooler we select, what fluid we select, uh what combination of systems we select. You know, if we're talking about building a mining farm, you know, in the UAE in the middle of the desert where we have 50C ambient temperatures, that's going to be besides the fluid itself, the s whole system is going to be designed completely different than one that I might install in say Alaska or Alberta. It's they are similar only in uh function, not much in form. So, um what about heat reuse? Let's do that very quickly. Uh do you think heat reuse will be a big thing for data centers in the future all over the world or is it more like a niche opportunity like right now? I think in terms of you know absolute scale it's probably a niche opportunity just in the total amount of power consumption for compute versus you know how reasonable it is to replace some of these uh hot water systems. But you know in terms of energy requirements uh heating water is one of the largest uses of energy in across the entire world and both Bitcoin mining and HPCAI when paired with proper systems basically become complete step-in replacements for these water heating systems. So I see a very large future in uh heat reuse and heat recapture systems in our industry. >> Yeah, I would agree with that. I think um heat reuse is largely a capex and energy optimization problem. Energy is uh from my perspective it is a um it's a resource that is very precious and we should seek to optimize it wherever possible. It is kind of a niche right now but I think you know um with capex and with some uh creative um solutions there are there's future potential and I'm looking very forward to getting all the energy out of every molecule possible. >> Thank you for coming everyone. 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.