Utilizing Tech - Season 7: AI Data Infrastructure Presented by Solidigm - 08x11: Long Live Immersion Cooled Servers with Hypertec

Episode Date: June 9, 2025

More episodes and seasons of Utilizing Tech: https://utilizingtech.com/Immersion cooling requires specialized servers designed to operate submerged in a tank of coolant, but there are many benefits. In this episode of Utilizing Tech, sponsored by Solidigm, we continue our conversation on immersion cooling with Patrick Scateni of Hypertec, the leading manufacturer of immersion-cooled servers. Most of the current demand for immersion cooling is in the datacenter, but edge computing is rapidly adopting this technology thanks to the demands of AI applications. Hypertec calls their servers immersion born, since they are designed for this specific application rather than being modified for use in a coolant tank. Sustainability is growing in importance and immersion cooled solutions are much more power efficient, much closer to a PUE of 1 compared to 1.5 for a conventional air-cooled server. It also uses no water, while air-cooled data centers often use evaporative cooling equivalent to an olympic-sized pool every day or two. Immersion cooled servers can be packed closer, enabling smaller datacenters, and are cheaper, more reliable, and longer-lived than conventional equipment.Guest: Patrick Scateni, VP of Global Sales at HypertecHosts: ⁠⁠⁠Stephen Foskett⁠⁠⁠, President of the Tech Field Day Business Unit and Organizer of the ⁠⁠⁠Tech Field Day Event Series⁠⁠⁠⁠⁠⁠Jeniece Wnorowski⁠⁠⁠, Head of Influencer Marketing at ⁠⁠⁠Solidigm⁠⁠⁠ ⁠⁠⁠Scott Shadley⁠⁠⁠, Leadership Narrative Director and Evangelist at ⁠⁠⁠Solidigm⁠⁠⁠Follow Tech Field Day ⁠⁠⁠on LinkedIn⁠⁠⁠, ⁠⁠⁠on X/Twitter,⁠⁠⁠ ⁠⁠⁠on Bluesky⁠⁠⁠, and ⁠⁠⁠on Mastodon⁠⁠⁠. Visit the ⁠⁠⁠Tech Field Day website⁠⁠⁠ for more information on upcoming events. For more episodes of Utilizing Tech, head to ⁠⁠⁠the dedicated website⁠⁠⁠ and follow the show ⁠⁠⁠on X/Twitter⁠⁠⁠, ⁠⁠⁠on Bluesky⁠⁠⁠, and ⁠⁠⁠on Mastodon⁠⁠⁠.

Immersion cooling requires specialized servers designed to operate submerged in a tank of coolant, but there are many benefits to this approach. In this episode of Utilizing Tech, sponsored by Solidae, we continue our conversation from last week on immersion cooling, this time with Patrick Scantini of Hypertech. Welcome to Utilizing Tech, the podcast about emerging technology from Tech Field Day, part of the Futurum Group. This season is presented by SolidIME and focuses on AI at the Edge and related technologies. I'm your host, Stephen Foskett, organizer of the Tech Field Day event series, and joining me from SolidIME is my co-host, Janiece Naroski.

Welcome to the show, Janiece. Thank you, Stephen. It's always a pleasure to be back. Well Janiece, in the last episode you and I spoke to Doug, a company that is really pushing forward with immersion cooling and we talked about many of the benefits of immersion cooling. I guess, catch us up on those benefits, and then we'll introduce our guest. That would be great.

You're right. There's a lot about immersion cooling nowadays, right? I think 2025 is a toggle between AI and just, what can I liquid cool? How can I do it? And I'm excited to have Hypertech on the show today, because we will take a deeper dive into the value pillars of liquid cooling overall. And how do all the parts and pieces come together

to make it a reality and make some real innovation happen all the way from the core to the edge? So with that, I want to turn it over to Patrick to introduce himself and give us a little more detail on your background Guys, thank you so much for having me with you today For the folks on the other side of the screen. My name is Patrick Scotini. I'm the global VP of sales at hypertech with the company for 20 plus years, so and Unfortunately, unfortunately a veteran of the IT industry with the company for 20 plus years and unfortunately, or fortunately, a veteran of the IT industry. Quick background, as you can hear,

I'm not from Texas. I'm a French guy from the south of France, Nice, and I've been in the IT industry for three plus decades, working for GE, IBM and Hypertech, like I said, for 20 years. And for the non-modern Hypertech, Hypertech is a company who just celebrates its four years. We have a multiple segment in the four years of the company, but if we focus on the last one, they said the past 10 years was to grow our business where IT manufacturer, global IT provider. And on the manufacturing side, we do our own server, storage, PC, laptop, and so on and so forth, before we go into immersion cooling.

I think that that's the reason that it's interesting to speak with you after speaking with Doug, because where we talked there about a lot of the benefits, I guess many people are unfamiliar with immersion cooling. I think that's part of what we talked about. As the manufacturer of the servers, I guess maybe first give us sort of a lay of the land because I know that you work closely with them and you also work closely with the companies that actually make like the tanks that things are. So when a customer is approaching you about implementing advanced AI or some other kind of power and heat hungry

demanding workload that needs immersion cooling, who are the players at that table? So the players are multifund because you need to look at that as an infrastructure, as a menu, right? You go from the appetizer to the main course to the cheese and I'm French, right? I need to bring cheese into it, to the cheese plate, to the dessert. And usually, you have the people who know about emergent cooling and the people who don't, right, who are interested for what they heard about the benefit to the less space, 50% less power, up to 50% or more of cost reduction on building a data center, and so on and so forth. So the tank vendor are usually the one who promote emergent cooling. They are the base of the infrastructure of emergent cooling. And is usually a tank or what they call a CDU

to cool the tank. And you don't need much more. There is no chiller except in a very specific case. And that's the step one. Step two is imagine the tank as a pot in a kitchen, right, to make a plate of pasta. You need to fill that up with water. That's the liquid vendor, right, Exxon, Shell, BP, Castrol, and so on. And after that, you need to put equipment, right? And today, as you just mentioned, Stephane, I don't need to say that as a surprise.

AI is taking over the world, right? I think you need to live under a rock to know that. And AI on the training side, on the inference side, on maybe a generative AI or to do some other workload, the power consumption of the ASIC, what we call a CPU, a GPU, the memory, and now the hardware, so we're going to come as quickly to Janice after that, right? I've achieved a level of temperature that air cool cannot do, right? So now you are left with liquid cooling. And liquid cooling usually is differentiated in two bigger tasks,

is the direct to chip. That is fairly common now. And the immersion cooling, and there you have the single phase and the dual phase, but we focus today solely on single phase for all the benefit without the problem. And that's help us cool the system. So whatever amount of machine you want to cool, right?

And I will give you some example later. In efficient, sustainable way, very important the sustainability aspect. We can maybe dive a bit more after that. And help the customer realize the benefit. I can put more machine in a particular amount of square foot. I can reduce my power consumption, like I just mentioned, so on and so forth.

And you add some switches, the cable, and suddenly you have a complete infrastructure. The only thing I didn't mention, you still which is the cable and suddenly you have a complete infrastructure. The only thing I didn't mention, you still need to have a few dry cooler on the top of the roof to evacuate some of the, not the heat, but the cooling of the cooling system. You cool the cooling system. And you have, in my opinion, the best solution the planet could sustain is not just at the equipment level, it's the best system for the planet, for the customer, for on the capex, opex model model is all better than we saw.

Yeah, no, Patrick, I think you did. Just curious, I mean, I agree with you. I think it is one of the most effective ways to cool technology for the planet. Tell us a little bit about where you see your solutions going, right? Is it just at the edge in some of these, you know, smaller containers like the last episode we did with Down Under Geocomputing? Tell us a little bit more about

where you're really deploying this hardware for liquid cooling. So I think today, most of the market for immersion cooling is at a data center level, because most of the customer demand we have require high density, large quantity, so it's a pure data center play. But since day one, I believe that Edge is the future. On top of that, there is multiple reports from the Gartner or Forrester or whatever, who show AI, the AI usage at the edge will be bigger than the AI usage in the data center by 2030.

So always keep that in your mind. And the Doug, I think, interview you did, that I was part of it, but not for this one in particular, but for the system as a whole, is a very good example of what an edge system could look like, right? Is rugged, you can drop that from an helicopter if you want. Or you bring that from a truck. It's very simple to put in place. It's all in one. You have the cooling at the back.

You have the server at the front. And that's a pure edge play. And really, in edge play, you have compute, you have storage, and you have AI, right? Because the AI can pump some data until we can take it anymore, and we need to store that somewhere. And we were the only, today we are the only immersion-born. That's mean the product we do have been designed solely for immersion. That's mean you can take the server and put it in a rack.

It doesn't work. You work only in a tank with liquid. We are the only one with a storage server for immersion cooling. And we choose SolidIMA as our partner for the greatness of their product, the capacity of their E1S, E1L, or U.3, for example, in some of our product. And we make a great combo to be able to cater to all the needs of a customer. You talked there about a key phrase, and that is immersion born.

And I think this came out as well in our previous conversation. You can't just take a server and stick it in a coolant and expect it to work. There's a lot of engineering that has to be done. Now it's not impossible engineering. It's something that we've been doing as an industry and that you've been approaching for a long time. But it is not the same as just dunking a server in coolant. Talk to us a little bit about what makes a design immersion-born. as one of my favorite subjects, right? So we started immersion five years ago already.

So we are not, and we do direct to chip since 2010. So we are liquid cooling aficionados at Hypertech. The reason we went to immersion, for the reason I gave you, and when you look at the server, our first implementation was with what people call today immersion-ready. That means you take the airpool server and there is a bunch of steps to do. Remove the heat sink, change the thermal paste by an indium foil, take off the fan,

modify the bios, make sure that all the labels are covered with an acrylic or they will fall at the bottom of the tank. Use some cable, all the cable from power cable to network cable need to be immersion certified or they're going to come brittle and you push on them after three months they break. It's something crazy. And it is feasible, right? We have people today who do that.

But it's not optimal. I'm going to give you another analogy. I went to race, right, on a Sunday race, the Sunday Cup, and I take my car, right, let's say I have a Chevy or whatever, and I put bigger tire and try to crank up a bit the horsepower and make sure the car is cooling properly, and I go race. I guess what? I never win. For one good reason. There is a bunch of people on the track who have a Porsche GT3 RS, Arke for example, has been designed to do one thing, is to race.

And the analogy, forget the price side, because it's not the case, right? But the analogy is our system, when they come to the customer and when we manufacture them, they are ready to be in immersion. That means the label are the label who are made to be there from 3M by the way, who are made to be in the state of immersion. All the cables we use are certified for immersion. The power supply we use are immersion cooling certified power supply with all the changes we have applied to it. The chassis, as you can imagine, you see a server like pizza box, right?

You have motherboard, the CPU, everything. The chassis have been designed, thermally designed to be in immersion, to conduct the flow of liquid. In immersion is always from a hold at the bottom and hold at the top and you recycle, right? In a perpetual movement. The chassis have been designed to conduct the liquid

in the most and fastest efficient way and so on and so forth. We have oversized, people say, oh, immersion cooling is messy. To avoid that, we built oversized endo on the server where you grab the server. You don't put your hand in the oil to start with, right? You can still have a train, you bring everything. But we thought about ease of use, ease of installation, ease of serviceability,

as a product to have one reason to be is being immersion cooling. That's what make the big difference between the two. I can tell you that one of the main three vendors, you know I call them DHL, Dell HP Lenovo. I like the DHL acronym. One of the three, maybe the first one, do some AI server for immersion cooling. They need to take,

one other company who prepared that for them, more than 150 step to prepare the system. Think about that for a sec. To convert it, 150 step. There is no step. The product was designed to be like that, right? So that's what makes it. I hope I answered your question, Stephane. That's why we make an immersion, what we call immersion-born product. And that makes sense because I think that most IT people are familiar with the airflow

of a conventional server and the fact that there are baffles and there's front to back or back to front. Everything needs to work right in order to get that airflow working in order to cool that server. And I think if you think about it and wait, this server is going to be in a coolant, in a liquid, then all of those things are going to be different. And you need to be engineered to be able to be effective in a material that behaves differently than air.

So for air sense safety, the people, I don't know the regular folks, right, my wife, my kids, whatever, I don't think they've ever been in a data center. When you go in a data center and you go between the hot air in a back, you can survive because it's too hot, right? The noise is at a level that everybody were the noise canceling apparatus on their head. When you have them cooling,

not a lot of people talk about that. It's silent. Think about that for a sec. So there's no heat coming out of the tank. Nine, zero. Even if you open the tank, there is still no heat. Because the power of heat absorption from a liquid, it's not water, it's a liquid, is a hundred times higher,

I think is even thousands of times higher than air. So you don't have all the problem that people in the data center had. 20 years ago, data center, you have a couple of machine in a rack, right? It was not like today where you have football field of racks and equipment who become unsustainable in terms of environment. I appreciate the background. I always often forget about the sound myself after being

in so many data centers, and it is true. I don't think I've ever been in a data center with all things liquid cults, but I probably would have come to mind. But you talk about sustainability, right, Patrick? So that's obviously top of mind for everyone in the industries trying to figure out how do we save on energy and power. Can you comment on just how much a particular customer, maybe use an example of, I don't

know, can probably talk about down under Geo or even like someone like Submer, how much more efficient are these solutions for those, you know, deploying these systems? Like how much energy are they saving? Well, it's very simple. For the folks on the other side of the screen, like I said, we don't know. The efficiency of a data center is measured by what they call the PUE, is power usage efficiency.

That's mean, say a kilowatt, for example, of IT load. That's mean I have a server. It consumes 1,000 watt. In regular data center today, still the average, as I speak right now, is 1.5, 1.6. That's mean if I put a thousand watt server, I need 500 additional watt to pull the server. Is that the infrastructure and everything?

That's mean the power consumption I'm gonna pay, you know, my electricity bill is on 1500 watt, right? Now multiply that by thousands and you can have the good idea. In EmergenCooling and what we did with Doug, the Doug Nomad where you did the previous episode, Doug down in the NGO have also a data center. We made an announcement.

We did an installation last year for one of the largest HPC deployment in immersion cooling ever done, right, thousands of nodes. And they are very vocal that their PUE That means my electricity bill would be on 1,003 watts or 1,005 watts. So I just saved 500 watts, right? That when we talk about sustainability, the first one is, I'm going to reduce my electricity bill tremendously. I think sustainability doesn't stop right there. and then they reduce my electricity bill tremendously. I think sustainability doesn't stop right there. Sustainability is well above and beyond

the power consumption is what some people don't know that. Yim Hap-Nguyen have been clean, despite his name, doesn't use any water, right? So now there is another lingo, the WUE, the water usage efficiency, where you are in clean skin, with no one in sight to be able to compete with that, right? Because it's zero water, that's it, nothing else.

And how much water does air-cooled servers use and where do they use this water? That's very interesting, they take the water out of the water supply, right? And the people in California, they are solely desperate with no water at all. But mid-size data center will use in water that completely evaporates of an Olympic swimming pool size every two days. To give you an idea, it's thousands of homes, whether you read it.

You know, that's what you see sustainability come to play. Right. And the server, when you go a bit further, is all about style reduction. Right, I told you is you can reduce the size. We have the average of data center, not the most advanced, whatever. Is a ratio of 10x.

We get 10x less space. So space today, God doesn't give more land, right? He created the land He could create, right? And you don't use space, you don't use water, you consume less power. The server we manufacture are what we call vanity free, there is the least amount of component as possible. Even when you do a retrofit to be emergent ready, you remove the fan and everything. As it was designed, there is no fan,

the chassis is barely a skeleton of a chassis. So that's less metal. Less metal meaning less stuff we bring out of overseas. So that means less traffic on the boat or by airplane, and you can peel the sodium to the end, right? And that's why it's the most sustainable IT possible. I just am fascinated by it. That's why it's the most sustainable IT possible. I just am fascinated by it.

I think, you know, in my head, I can't stop thinking about the 150 steps, and you guys have just built like just a seamless, beautiful server. I can't believe I actually look at your servers and think, wow, they're genuinely beautiful, but they work amazing. Right. And you really have put so much thought into all the parts and the components to make it that Ferrari that's going to win the race. Right. And by the way, is a Ferrari at the price of a Chevy? I just want to say that because people have a misconception. The cost of a liquid cool server is a few percent less than a cool server.

So on top of that, you even save on the price, sorry, on the capex. Well, I would assume too, we didn't talk about this, but because you're dropping everything in the cooling solution, do the parts and things run a lot longer? Do they stay, you know, is it easier to, you know, I guess, fix parts, if you will, things don't break down as easily maybe if they're in the cooling solution? So that's a very interesting point, right? First, when we develop the server, we start to choose vendor also.

And the same way, we try to choose vendor who are sustainable at the core. So we're having one of them, right? But talking about the lifespan of the system, because there is no moving part, because there is no dust, and so on and so forth, the system can run much longer than regular system. Liquid immersion cooling, there is one customer in particular in the oil and gas. So, Doug, for example, is 10 years. They do Implosion Cooling. And if you go to their main data center in Texas, in Houston, they have systems who are 10 years old. And they are running time, except his old

generation, right, his own technology. They still have a few of that. There is the other company in the oil and gas who run in-emergent cooling for 10 plus years. And I was just talking to them months each ago. And we were talking about failure rate. And they say their failure rate is close to zero. Think about that for a sec.

Close to zero, right? Because the components are not stressed out. One of the main reasons of component dying is the heat. People don't mention that as much, but it's always a heat component in, except the manufacturing department or whatever. But this is the heat. And in emergent cooling, they are secure.

They are shielded for that, right? That means they're only in a very stable environment. The temperature is always the same. And that's why we have this long lifespan span on the hardware. Don't think that everybody's gonna run 10 years, everybody want to change, right? Every three to five years because technology advance and is more efficient and everything.

But I do feel it also because it's very, very, very low. It's interesting for the customer,'s less work at the data center. When people say, oh yeah, it's a pain, you may have been cooling, I'm gonna put liquid everywhere. It's never the case. You're not gonna grow in a tank as much as you grow in your regular rack.

To start with. Thank you. I think, as Janisse said, I think that's. Just definitely isn't right. Thank you. I think as Janice said, I think that's something that people don't think about, but heat is the enemy of reliability. And, you know, systems, the thing that kills them is, you know, when they experience thermal shocks,

or when the, you know, the dust gets in there and coats on the components and the components get too hot. It just doesn't happen. And I think that that is really an overlooked aspect of this. So, you know, in summary, I guess, if we want to kind of wrap it up, people should be looking at this. It's not more expensive. It's not more exotic. It's just less familiar. And you said that it takes up less space, it makes the servers last longer.

Why is it that more people aren't using immersion cooling or is it that it is getting used and we just aren't aware of it? Why people didn't buy the first Tesla when they came out with the Lotus, that was fitted, right? Because he was new, because he was disruptive, because people say, I don't know about that, right? So I'm used to have my regular car and previous to that is,

hey dude, I'm riding my horse extremely well. I'm afraid I'm going to go in this crazy machine here, right? And it's an evolution. And immersion cooling is extremely disruptive for that because it's an entire parameter change. I think that's one of the reason the people are not doing it. One of the other reason, I would say, is to be successful,

you need to have a very strong ecosystem of people who do. I told you, the tank the the server the liquid the switches and everything and there is a you know you never get fired to buy a Dell and HP or Lenovo right and they are not the most even if they do so right as a retrofit on per-profit basis need to sign a waiver that they do so, as a retrofit on a perfect basis, need to sign a waiver that they don't give you a warranty or so on and so forth. People will say, hey, okay, I need to buy from who do I buy from?

We are not, maybe with this podcast, we will. We are not as well known as our famous three colleagues here. But we start, right? I think we are in a very good place. When I see all the Tag Mender, right? Summer, GRC, NIDAS, there is a bunch. There is more and more coming. I think Vertiv is coming with one. We're going to solve this portion of the equation. You have plenty of ten vendors, you have plenty of all vendors to choose from.

And now is to have, is not to our advantage, but it will be to have more and more people as a server and all doing immersion server. Because suddenly, people will have the choice, a large choice to choose from. I can tell you that in pure bake-off, supplier-aids versus us, we always win. When people take the time to test, supplier-aids versus us, we always win. When people take the time to test, right,

to see how this works and what we do for them, if you think about people like Doug who deploy at larger and plenty of other, we have larger the project we cannot tame them, we develop, you know, we talk about sustainability, we develop eco-packaging to ship in bulk. What's it have to do? First, when we ship on one single speed, we have a large amount, a much larger amount of machine than any other one. The packaging we use The packaging we use is made out of 100% recycled cardboard, but more importantly, we designed the packaging to be reusable. Think about that, right? And the way we designed the packaging, because of its identity, we made it where it goes in phase.

When you install the system, if it's compute, you install what we call the tray after the node. You can see practically, and we cut the amount of time it take to install thousands of node by days because of that. And it's all sustainable, right? Because it's more density, you don't need to have, we just for a lot of, we went from a six truck to three. So now calculate the carbon footprint of 18 wheelers, 53 feet long truck,

who don't need to be on the road. That's sustainability. Well, excellent. And I really appreciate you sharing your passion for this technology. And helping, you know, I do hope that our listeners will consider immersion cooling, if only for the environmental sustainability benefits, the power utilization benefits, even setting aside some of the other aspects. And of course, the longevity as well. I mean, you know, every server, every year that you can use a server a little bit longer is another one you don't have to buy. And I think that that's another huge benefit here. So it's sustainable. You don't need to take this server and

God knows what happened to it. Well, thank you so much for this conversation. Before we go, where can people continue the conversation and where will they be able to see hypertext servers themselves? So the next few events will be the defined one. We will be at Supercomputing 2025, I think is in St. Louis. I will have love Las Vegas, but now we got St. Louis.

I'm joking. And we will have as a show that the best is visit our website www.hypertech.com and hypertech is H-Y-P-E-r-t-e-c dot perm, no h. And you will see all the information and you will see a solid arm as a vendor or partner vendor, for example, because there is a reason we choose them for our emergent product storage and for other regular product. And I think the best way to contact us

or you just give me a phone call and... Well, thank you so much. And yes, we did speak on the last episode a little bit about the ways that SolidIME is making their storage compatible with immersion cooling. Janice, I know that sustainability, longevity, these are hallmarks of solid I'm as well. And I really appreciate having you on the podcast this season.

It's been a pleasure. Yep. All right. So thank you all for listening. If you enjoyed this conversation, please do subscribe. You'll find this podcast in your favorite podcast application. You'll also find us on YouTube and on the new TechStrong TV app on your TV or in your portable device. If you enjoyed this discussion, please do give us a rating, a review. We'd love to hear from you as well. This podcast was brought to you by Solidim,

this whole season was, as well as Tech Field Day, my own company, part of the Futurum group. For show notes and more episodes, head over to our dedicated website, which is utilizing tech.com. We'll include links to the companies that we've mentioned here as well. Or you can find us on ex Twitter, Blue and mastodon as utilizing tech. Thanks for listening and we will see you next week.