Madrone is building hyper-efficient data center cooling that can save 30% of grid capacity — already closing first investor check

Speaker 1: How are you doing? Welcome to the show. Thank you for taking the time.

Speaker 6: What's up? Hey, guys. Huge fans, so appreciate the opportunity to talk to you guys.

Speaker 1: Fantastic. Great to have you on the show. Introduce yourself. Introduce the company. I have tons of questions, but let's start with the basics.

Speaker 6: Sounds good. So I'm I'm building my drone. We're building hyper efficient cooling for data centers. Mhmm. So specifically, we've engineered a thermodynamic process that works really well in hot and dry climates, which is where most hyperscaling is happening right now.

Speaker 9: Okay.

Speaker 6: And, yeah, it's we have our demo that's running live at demo day today, so we're actually really excited to show people in person. But the big problem really is that data centers are limited today by power and not by chips or land. So cooling and power are two sides of the same coin. If you can save power on cooling, that's more power for your flops, and that's where our real value unlock is. So people can install more GPUs given the same grid permit, which takes years to get. To give you a sense of the scale of the problem, last year in Texas, Semi Analysis said that there were a 150 gigawatts worth of load request to the Texas grid. Texas grid on a good day is 60 gigawatts and only 1.5 gigawatts worth of those requests were actually approved by the local utilities. So it's a crazy scrambling time. Like, Elon Musk was probably the first person to kinda come up with a method of putting engines on the back of trucks, rolling them onto data centers for power. You know, Boom Supersonic, whose ultimate goal is to be building jet planes, has had a slight pivot to selling supersonic jet engines, and apparently, they're crushing it. So it's it's a crazy time, and it's a it's the first time in a long time where a new company come could come into the market and and build a thermal industrial company.

Speaker 2: So what is the bottleneck for for you? Is it figuring out where your system, you know, what data centers your system can fit in? Is it actually just how you have the demand and you're trying to meet it? Like, what does your supply chain look like? Like, what's standing between you and let's say, not this round right now, but, you know, the next Yeah.

Speaker 6: It's a great question. So we actually scaled our one kilowatt prototype a 100 times during the batch to a 100 kilowatt prototype. We're gonna be deploying our one megawatt version very, very soon. Mhmm. The 100 kilowatt prototype that's actually running live at Demo Day right now is gonna be running our first pilot in two months. Mhmm. So really, it's gonna be between the 100 kilowatt and the one megawatt prototype, it's gonna be figuring out how to elongate it and make it taller. So we've done the really hard work so far, but really it's just building a physical big version of it so that we've talked to hyperscalers, their feedback because they just want a bigger version so they can because they have such a huge demand. They're not really interested in a 100 kilowatts. They're interested in a 100 meg one megawatts.

Speaker 1: Yeah. And

Speaker 2: Okay. So one megawatt will be enough to get your first real, you know, orders, revenues with hyperscaler type customers.

Speaker 1: Yeah. All the data center campers campuses would say, I'll take 10,000, please, because I'm building 10 gigawatts. I mean, basically, you chain them together. Right? But this is I mean, we saw in the video, it looks like a shipping container. And essentially, the the input output the output is just cold water. It takes in water. So you are agnostic to the design of what happens in the data center, I imagine. You're not you're not dependent on any particular chip or rack design. Yeah. But then what energy is actually going into your system?

Speaker 6: Yes. Exactly. It's a great question. So are the energy is used to be the fans and the pumps. So actually circulating the fluids through the system. If I may brag about my cofounder a bit, we don't just design the the physical box. We actually he actually designed the thermodynamics, the mechanical structure, the electronics to power it all, and then also the software to control it. We do this all in house. We're big Tesla fanboys. All our competitors are outsourcing every single level of this engineering out, and they're just integrating that and selling that as a product. So been able to get a lot of efficiency gains by just designing this all in house with two people so far. Mhmm. We're fundraising, and we're gonna hire some more people, hire our friends in the Bay Area and building the one megawatt one with our own hands.

Speaker 1: Fantastic. Amazing. And so there there there's already a lot of energy on these campuses that you can draw from. So whether it's natural gas powered, powered from the grid, powered from nuclear or solar, whatever's going on, you're drawing from that. But your goal is to just draw more efficiently and take less of the power so that can probably just go into the chips. Right?

Speaker 6: Exactly. Yeah. You get a lot more inference value out of redirecting that power to your flops.

Speaker 1: Okay. And then in terms of how you're pitching yourself at at Demo Day, are you on the path to LOIs with hyperscalers? Are you working backwards from or are you doing business with neo clouds? Like, what's the walk crawl run to actually create the economic flywheel to raise money and then go build this thing?

Speaker 6: Yeah. Great question. So we've already started conversations with two hyperscalers. We're actually talking to a third one this week after demo day, which I'm super excited about. Let's go. One of our investors actually, we closed the check today. They're actually building $20,000,000,000 worth of data centers in Texas. So that's been really nice. The really first yeah. Fantastic. The first thing for us is really gonna be focusing on that pilot, just demonstrating that we can plug into an AI data center at scale, do it reliable reliably, and at scale with our investor that we just partnered with today.

Speaker 1: Yeah. And then what type of efficiency number actually

Speaker 8: Yeah.

Speaker 1: Draws the attention of a potential buyer? Are we is 1% enough? Is it 10%? Is it 50%? Like, what what what's reasonable? Because I imagine we come we're 99% more efficient. Everyone's like, yeah, that's not real. But is is one per are we doing basis points here? Like, how Yeah. How how much of an impact do you need to have to actually move the needle? Because some of these projects are so big, their bill for electricity might be, you know, hundreds of millions of dollars.

Speaker 6: Yeah. Exactly. Like, guys, it's so crazy. My competitors are 100 year old companies that have been designing the same thing for decades.

Speaker 2: There's this

Speaker 6: thing called the mechanical chiller. Yeah. It's literally the same thing as your refrigerator at home and it costs $500,000,000 per gigawatt. So people are really desperate to get away from these chillers and look for different types of cooling Mhmm. Because that's actually eating up your grid permit. So that's a thing that actually costs 30% of your grid permit. And we can do it for cheaper and we can actually have better OpEx as well. Mhmm. So really any percentage is okay. I think that like, you you can't quote, like, 0.1% gain, but, like, you know, people are thrilled if you have a 10% efficiency gain.

Speaker 1: And then also, we're just in a supply crunch. So even if you just said, hey. We're we have the exact same thing, but it actually will ship in the next couple months, a lot of people might jump at that. So that's the benefit of being in the problem

Speaker 2: What what were you doing before this?

Speaker 6: Yeah. So I was actually scaling manufacturing at QuantumScape. I was doing computer vision, working directly under the CTO, Tim Home. I was founder CTO, took it all the way from inception to IPO. And I also did grad school at Berkeley studying material science. There's a material source of this company I we didn't get to talk about today, but we're we're doing something cool.