Critical Resources Limited (CRR) Earnings Call Transcript & Summary

Hi, everyone, and welcome. Today, we're examining one of the defining infrastructure challenges of the AI era, how data centers can support rapidly increasing compute density without placing unsustainable demand on electricity, water and existing cooling infrastructure. I'm joined by Timothy Wither, Managing Director of Critical Resources; and Eric Martinez Gurrea, an experienced mechanical and thermal engineer and CRR's technical adviser for data center cooling. Tim, how are you today?

Yes. Good morning, Wei. Thanks for having us on Sharewise.

Yes. My pleasure. Eric, you do?

Yes. Good morning, everyone. Happy to be here.

Excellent. So Tim will explain why CRR has invested in 2-phase spray cooling and how the company intends to evaluate and advance the technology, while Eric will bring the industry and engineering perspective, examining the scale of the cooling problem, comparing the available solutions, and importantly, identifying what CRR's technology must demonstrate to achieve real-world adoption. We'll begin with the industry challenge, move through CRR's approach, and finish with the pathway from laboratory evidence to independent rack-scale validation and commercial deployment.

Eric, let's start with the scale of the problem. So Eric, what has changed in AI and high-performance computing that is making cooling a strategic constraint rather than simply a facilities issue? Before you start, actually, let me just bring up the presentation just so that we can talk through some slides. All right. Hopefully, everyone can see that. So back to the question, the industry problem. What's changed in AI and high-performance computing that is making cooling a strategic constraint rather than simply a facilities issue?

Well, the biggest thing that happens in AI and HPC basically has been the evolution of the silicon. Basically, AI and HPC processors define themselves as the amount of calculations that they can do per second. The more calculations, the better. So basically, the processors have been designed to push the number of transistors to the limits. More transistors then mean more compute. And then you might be thinking why I'm talking about transistors. Basically, it's because the number of transistors implies the power that they draw, and the only way of dissipating this power is through heat -- sorry, basically, all the power drawn by the transistors is dissipated by heat. So this comes to the question, we need higher cooling technologies just to keep this compute keeps on running, the processors keep on running. So basically, just to simplify, the chain runs like this, more compute, more transistors; more transistors, more power; more power require more cooling just to keep the processors keep on running.

Excellent. And Tim, when CRR first examined this industry problem and this opportunity, what convinced you that data center cooling represented a genuine and investable technology problem?

Well, I guess I'll go -- start from the end and go back to the start. Over 2025, $1 trillion was invested into data centers. Over the next 4 years, it's projected to be $7 trillion being invested into it. And as sort of Eric has mentioned, the rapid advancement of these models have really -- and you see it in the paper every day of how much work is going into these data centers and into AI. So Critical Resources, as the name suggests, we're actually a resource company. And we've got quite a significant lithium resource in Northern Ontario in Canada. And so from that lithium resource, we were investigating upstream technologies and to differentiate us. So we've been in conversations with a university in South Dakota, and we've got a partnership there where we're investigating solid-state lithium-ion batteries. And through that advancement, real thermal management has been a challenging problem for that industry. And through our network, we had been introduced to this 2-phase cooling technology. So this is a U.S. patent and also 2-peer reviewed. So been operating -- or it has an operating test rig in Singapore, and the government had spent -- partnered with the university there and spent a considerable amount of time and money on this project. And so from that, we've been able to move from the thermal management of the lithium-ion batteries and really then identifying that it has a real unique application to high-performance computing -- compute, and that's where we've positioned this technology.

Yes. Okay. Very interesting. And Eric, looking at the market from an engineering and industry perspective, do you agree with Tim's assessment? How urgent is the need for new cooling architectures? And what evidence are you seeing from data center operators, hyperscalers and equipment manufacturers?

Well, I do agree with Tim. Basically, as I was saying before, that -- I mean, based on the urgency, the urgencies relatively, as of today, with the current technologies, we have the problem solved for today. Liquid cooling has brought a lot of headroom just to deal with the heat flux that we have as of today. But today, we have a gap between transistors about 3 nanometers. This is the level -- basically, just to put some numbers, we have now currently gaps between transistors about 3 nanometers. NVIDIA's latest chip, it's about 280 billion transistors per 50 millimeters. So this leads us to a density that even though today, we can still deal with that, air cooling is not enough for that. So we have to move from air cooling to liquid cooling. Liquid cooling are a lot of different technologies, but just as a broad term, we have to move from liquid cooling. But the road map that we have ahead is going beyond -- sorry, going below 3 nanometer of node size for the transistors. In this case, as of today, we are having prototypes at 2 nanometers. But on the road map, we can see that the OEM manufacturers are starting to try technologies to go to 1 nanometer, even down to 0.5 nanometer. This is a road map that extends 10, 15 years. So as I was saying, for today, we have the solution solved, but for the coming 10, 15 years, it's something we need to develop. And this technology, as we can see here, it allows a heat flux at a very low -- sorry, at a very high efficiency, thermal exchange. So we need to explore today this technology so we can tackle the problems of tomorrow. So yes, definitely, it's not forcing today, but we need to do it so we can be prepared for the coming years to come.

Yes, it's good to know. I mean, you're really planning for the future here 3 to 5 years out rather than just talking about what's happening in the next 6 to 12 months. Now Eric, this term liquid cooling, it is often used as though it describes a single solution. But I think as you've alluded to, that's not really the case. So could you explain the practical differences between, say, direct-to-chip immersion cooling and 2-phase spray cooling?

Yes. I would like to mention first because sometimes we -- typically in the mainstream it's commonly known as direct-to-chip and immersion cooling. But what I want to point out is that direct-to-chip and liquid immersion cooling typically are known in the single-phase operation. Single-phase, basically, I would like to go a bit above and just explain first what is single-phase and 2-phase. This way, we can -- I can expose the technologies we have in both technologies. Basically, single-phase is leveraging the sensible heat. So it's within the same phase of the material, can be air and remains air throughout all the cooling loop or can be liquid throughout all the cooling loop. In this case, what we are changing is just the temperature of the liquid or the air. So we are just capturing this sensible heat just to upraise or down the temperature of that same fluid.

Wei, can I just ask you to move 1 slide up, and that really explains the physics about what Eric is trying to -- what Eric is explaining? This one, yes. Brilliant.

When we move to 2-phase, and this is what we can leverage a lot the physics of the materials or the fluids in this case, is because we are moving from sensible heat to latent heat. Latent heat is what is all the heat that can be absorbed by the process of exchanging this phase, in this case, passing from liquid to vapor. And just to put a few numbers here, we can go in the order of tens to hundreds of more heat capture. In this case, we can see 1,000 times, which is aligned with the fluids that we are using here. But just to put to the audience an example, and this is something that everyone can experience at home. If you put a pot of water to start boiling, during all the process before it's boiling and you pass the hand about 10 centimeters above the pot, you will see that the heat that you are feeling is not really strong. However, when you are at the point that the water is already boiling, you cannot extend the hand for a long time. So all this is basically this principle, and it's just a fun example that everyone can experience at home. In this regard, let's go back to the technologies. So now, I have explained single-phase and 2-phase. Let's go to direct-to-chip and immersion cooling. And I would like also to add one more, which is commonly used, which is the rear door heat exchanger. Rear door heat exchanger, I would like to start with that because it's like a midterm -- or like a midpoint between air cooling and liquid cooling. Basically, this is a series of coils that you put just behind the IT rack cabinets. And basically, you have coils in which you are circulating fluid at lower temperature, and through the action of fans if it's an active rear door heat exchanger or just with the fans of the IT compute trays, you are just exchanging all this thermal energy between -- before -- sorry, before the air comes down to the room. This is the first one I would like to explain. Then let's move to direct-to-chip. Direct-to-chip, basically, it's about cold plates or heat sinks, liquid-cooled heat sinks that you directly put on top of the heat-intensive components of the compute tray. Basically, direct-to-chip focuses on the heat-intensive sources, mainly only the processors, GPUs, CPUs, DPUs, only the heat-intensive. So you can extract or reject the heat very efficiently from those components, but not from the rest of the components. So you have about a split of heat, 70%, 75%, through these heat-intensive components, which is directly liquid-cooled. You are rejecting the heat through liquid cooling, but you still need a part or a portion for air cooling because all the rest of the components within the tray are -- also have the need to reject the heat. Then we move to the next step, which is immersion cooling. In immersion cooling, you go directly to liquid cooling. In this case, you change the full medium of all the IT compute tray by liquid. In this case, you cannot use -- well, you have to use dielectric fluid because the goal of that is submerging the IT compute directly into the fluid. And in this case, almost 100%. It's only a few inefficiencies about -- just to be sure -- or just to be certain, about 99% of the heat is fully transferred to a liquid-cooled loop. And the next stage is the spray, the 2-spray system or the 2-spray -- 2-phase liquid spray that we are exploring with Critical Resources. In this case, the technology is spraying micro-droplets. So all that superficial tension that we have in liquid, we are just spraying it just by going to the minimum or atomized minimum particle of liquid, and we are spraying them onto the surface of the processor. In this case, the energy just needs a bit of energy just to turn to this change of state and -- sorry, change of phase, and we can capture very efficiently because all the energy directly goes to this latent heat phenomenon that I've been explaining before. And we convert these micro-droplets into vapor, capturing very efficiently and at very high rates, about hundreds of orders against the sensible heat. So we are turning it...

Sorry, there's a question which has come through, which relates to this. So I might just ask it while you're on the topic. So does the type of liquid being sprayed on the chip change the cooling efficiency, which is gained?

Can you repeat, please?

The type of liquid, which is being sprayed on this chip, if you're using a different type of liquid for the spray, does that change the cooling efficiency or how well the chip is being cooled?

In this case, we need to use dielectric -- sorry, the refrigerant fluids because we need to keep them in contact with the electronics. In this case, we are just spraying liquid onto electronics. So we need to do it with liquids that are electrically insulating. So we cannot foster any electrical conductivity there. So you're right, different fluids have different performance. What we are balancing here is we are selecting the fluids that have boiling temperature about 55 to 60 degrees. We don't want higher boiling temperatures because we don't want the chip to go beyond that. We want the chips to be working and operating at maximum 60, 65 degrees. That is just to enable the processors to operate without throttling. So we are selecting fluids in this range of operation. You can select different chips, of course, but we need to adapt the fluid to the chip and to the environment that we want to operate, which is boiling temperature about 57 -- just to put a range, about 55 to 60 degrees and condensation temperature about 35, 40 degrees. This is the boundary conditions that we select for the fluids.

Yes. Excellent. Yes. And maybe I'll just switch over to Tim quickly. So with several competing liquid cooling approaches already in development or commercial use, why did CRR select 2-phase spray cooling? Which characteristics of the technology were the most compelling from your perspective? And a follow-up question, which has come through online, which is, are there any other competing technologies being explored for next-gen cooling by CRR?

Yes. Can I just ask you to go up 1 slide again?

Yes, of course. Sorry about that.

Really -- no, that's good, and we can talk about that slide. I mean, the physics really shows where the industry needs to go. So we're hitting -- and as Eric was saying, we're hitting really nearly the limitations of what direct-to-chip can do. And that's the current phase of NVIDIA and all other OEM chip manufacturers. I mean, really, these AI data centers are just a massive amount of computers, laptops sitting in massive warehouses. And today, 80% of these are just very large air-conditioned rooms, warehouses. And 30% of that power goes into cooling the silicon. So now back to why we have to cool it. So every kilowatt that goes into these data centers comes out as heat. There's no other energy that comes out other than the light that goes down the fiberoptic cables to your phone and what's happening with this video right now. So heat, the cost of cooling the silicon, and these are very expensive silicon. So if you can think of 1 server rack now, it's around $4 -- USD 3 million to USD 4 million. So if you've got a rack, you can be talking a significant amount of capital. So you don't want these chips -- well, you don't want these chips to fail, so thermally regulating them, too. And what Eric was talking about is throttling. So if the chip goes a little bit higher in temperature, what will happen is they'll automatically reduce the power. So if you've got a $3 million chip, you want that to be running as most efficiently as possible. Liquid immersion requires you to remove certain components from that. So it's sort of -- I use the analogy, you don't remove your radiator from your Ferrari to make it run better so you can stick it into liquid immersion. So this way, the direct 2-phase, as Eric was saying, it's the phase between the liquid into vapor. It removes the heat more effectively. So I guess, are we looking at other technologies? No, because this is where the industry needs to go.

Excellent. And Eric, so you do have this experience across immersion, direct-to-chip and other cooling architectures. Now that you've examined CRR's 2-phase spray cooling technology, how do you assess this underlying engineering proposition?

Well, you're right. I'm coming from direct-to-chip immersion cooling environment. I've been 7 years working on that, probably in the baby steps of the industry when we were developing liquid cooling technologies. When Tim brought me this opportunity and I started analyzing the papers, the patent, for me, I knew about 2-phase. Typically, 2-phase for me was a bit of complex. I was always seeing it like 10 years, 15 years, but it has its own limitations. But when -- at least it was, I believe, when I was starting to read the patent and I spent literally 4 hours reading every line of the patent, just pressure testing, writing a lot of questions that I had for the academic team behind the patent. I was just exploring, exploring, researching, and I completely reached -- achieved -- reached a conclusion that everything was sound, everything was grounded. So it was not much to pressure. It's just moving it to the next phase, just to testing on a physical prototype. Actually, this is one of the goals that we have now. Reality is that the system operates not only theoretically, it operates well practically. The academic team behind the patent have developed, if I'm not mistaken, 4 different prototypes. In the 4 of them, they have been testing the technology one step above, one step above, one step above. So for this project, we are mainly planning -- or at least we are developing now 2 development paths. The first one, I'd like to call it MVP1; the second one, MVP2. MVP is because of minimum viable product. And with this, we are going to pressure test how we can pass from this academic environment testing to an industrial ecosystem. We are going to move from all this theory to an industry-ready technology. And on the second one, the MVP2, we are planning on developing a real-scale rack, vertical rack with all the full capability. And we are going to -- and we are focusing on the same footprint, same scale as any other equipment in data halls, and we are going to test it in real operation within our data hall. So with this...

Right. Sorry, Eric, can I ask you move down a couple of slides?

Yes.

You can -- and then the audience can see what's actually built now. That one, move. That's the team. We got that one there. Brilliant. Thank you. Sorry, Eric.

Yes. I was already done. But basically, here, we can see the 4 prototypes I was referring to. So with these 4 prototypes, this is the academic environment. We are just moving all these oversized equipment to an industry-optimized envelope. So we can just move to a more product optimization, industrial-ready envelope. Clear?

And just coming back to that technology, I mean, cooling is one thing, but it's not really the only thing at this point in time, right? How important are -- is energy consumption, water availability and grid capacity when operators -- when an operator selects cooling architecture?

I might answer that one, give Eric's voice a bit of a break because he's just coming off a flu.

Yes.

Just 1 slide up. Please wait. One more. Yes. Brilliant. So as Eric mentioned, this is a U.S. patent and has got 2 peer-reviewed papers done on that. So typically independently verified. And so from that lab-scale unit, they've been able to show nearly 26% reduction in energy. And that is basic basis the cost savings for that is actually removing the chiller. So the technology we've been talking about, liquid immersion, direct-to-chip, the rear rack fans, they all require refrigerated fluids. So they use a liquid glycol mixture. That liquid has to be reduced down to 21 degrees, give or take. To do that, you've got to have a chiller unit, refrigeration unit. And as a 2-phase system because the boiling temperature of the fluid is sitting around 50 degrees, 50 to 60 degrees depending on the fluid, and good question before, that then removes the major expense on that unit. The other benefit, and you see it in the paper all the time, and it's because not having the chiller, we can go to a dry cooling tower. So it actually doesn't require water usage. And because both of these systems are closed loop, it dramatically reduces the amount of water that can be done. So it is chiller-less-free -- chiller-free and water-free. So both of these units are closed loop. So we keep the unit. There's minor water loss through these things. But because we can add in -- because of the power, and we're removing the power to nearly 20x more than direct-to-chip, that means we can actually put in more compute into the same area. So like I said, with these data centers, they're just massive halls. So the ceilings are nearly 5 meters high, but the racks are only taking up just over 2 meters. So there's a massive amount of white space that could be potentially used. If not, we're bringing down the size and condensing the compute into a smaller footprint so that there's massive warehouses through there. So this is where that technology can really move it. So you're saving power. So one of these server racks -- I did this in the presentation in Sydney. One of these racks is nearly 1 megawatt. That's 1,500 Australian homes. So if you're saving nearly 30% or 26% of the energy bill, and you sent to those 1,500 houses, I'll save you 26% of your monthly bill, but you give me 5%. We would have people lining up. And that's a significant cost savings. If you think about when you're talking megawatts, so some of the ones, particularly audience in Sydney, I think NEXTDC's 200 megawatt. If you can save 30% of 200 megawatts of the energy cost, these are just astronomical savings you can be making, particularly when our energy grids are already under strain. The other point with the chiller less and the dry tower is you can actually -- I mean, this was designed in Singapore, some of the most challenging places to cool. So if you think where the major populations are around the equator, and this is where it's optimized for, there is markets where this technology can move into.

Yes, makes sense. I mean, if you can chill in Singapore, you can -- well, keep it cool in Singapore, you can pretty much keep it cool anywhere in the world.

Yes.

Yes. Yes. And -- I mean, to your point, like just on this water side, Albo has been talking about it quite a bit, too, right, in terms of the whole AI, not just being on the energy consumption, but the water consumption. So this 2-phase, it sounds like it actually is quite conservative in terms of being able to effectively not really consume water. Tim, I might just stick with you at this point in time and give Eric a bit more of a break. But just in terms of how you've talked about this as a capital-light and evidence-led milestone-gated approach that CRR is taking. What does that mean in practice for the cooling program?

Yes. We just completed a raise this morning. We just come out of a raise, raising $1.6 million from existing shareholders. Majority of that funding will go into our exploration. We've got some amazing exploration projects and our Mavis Lake, flagship Mavis Lake Lithium Project in Ontario. When we say capital-light, really capital-efficient, we spend quite minimal amount of money onto this. And it's -- and the reason why is we're in this development phase, and the research phase is moving through to it. So I guess the evidence base is we have a U.S. patent. So we've got a 10-year global license over this technology. We do have a period of time to move this into commercial. We don't -- we can't sit on this forever. So we do have a small moat over 2-phase spray technology. And our goal is to move this into commercialization. The other part, the evidence basis, this is not theory. This is actually practical. We have a unit -- test units up in Singapore. And the idea is we'll build our prototype here in Australia. And we've got a team operating in Sydney together with our NTU researchers in Singapore. And so the idea is to get to that prototype base. So we're moving from a pod into a shelf and then from a shelf into a full-size rack. So we'll end up -- it's really modularizing this whole thing up and into -- ultimately, we would love to be able to deployment of chips, and that's where you can really see valuation. You see Firmus' valuation in the market again this morning. That's where you get real valuation for the company. But some of these other cooling technology companies through the U.S. have been merging acquisitions, and we're talking significant money for it. So the technology, including our solid-state batteries, this is just giving optionality and potential licensing and partnership opportunities for our shareholders and for our projects.

Yes. No, that makes a lot of sense. Eric, there's a question which is probably just back to you, more on Singapore, it's very humid. How does this affect cooling efficiency in a 2-phase system?

In Singapore -- I lost there.

Well, it's very humid -- hot and humid in Singapore. So does the ambient temperatures in a region impact on the efficiency of a 2-phase system?

Yes, absolutely. In the end, you always have to think that you have the processor and that you -- what you are trying to do is just reject the heat from the processor. You always need to have an end heat sink. The ultimate heat sink in every cooling loop is always the atmosphere or at least you can have atmosphere. Typically, the conventional is atmosphere. You can have also alternative like the seawater, river water, lake water, geothermal. So you can have other alternatives, but the convention is typically the atmosphere. So basically, to perform this heat transfer with the ambient, with the atmosphere, you need to leverage directly the conditions that you have at that time that you are exchanging. More humidity typically creates a limitation. It's better if you have dry ambience rather than humid. It's always better to have lower temperatures rather than higher. So you always need to try to find that good ecosystem just to extract. In this case, Singapore is a very humid and very stable, around 30 degrees all year around temperature. So if you want to use traditional cooling technologies, as Tim was mentioning before, you need to use chiller, chiller technologies. You need to use compressor-based technologies. So you can create these refrigerant loops, try to pull down the temperature and then have a good heat medium just to transfer all that heat out from the processors. So in this case, when you are using high thermal efficient technologies, as this one, what you can do is just go closer to the temperature of operation of the processors. In this case, processors, I mentioned before, need to operate around 60 degrees, maximum temperature. When you can reach the temperature of supply of the cooling loop closer to that temperature, let's say, 40 degrees, 45 degrees, which is the trend that NVIDIA is pushing on the road map. They are trying to push supply cooling loops at 45 degrees. What this means? This means that for a location as Singapore, which has a stable temperature of 30 degrees, the ambient is always below the temperature of supply. So you don't need extra chillers, extra compressor-based technologies just to cool down the cooling loops because the ambient is already a good heat sink just to reject the heat of the system. So you can only transfer heat from 1 body that has more heat to 1 body that's less heat. In this case, all the ambient will be at less temperature than the one we need to supply to the system. And this can only -- just I want to highlight, this can only happen when you have very efficient technologies. When you have very efficient technologies in the thermal heat thermal process, you can raise your temperatures closer to the temperature of operation of the chips. So basically -- I'm not sure if I answered the question. I hope I did.