Omen AI raises $31M Series A to monitor liquid cooling fluid in AI data centers

the commerce platform that grows with your business and lets you sell in seconds online, in store, on mobile, on social, on marketplaces, and now with AI agents. And our next guest is in the waiting room. We have Zack Leersge from Omen AI as the founder and CEO. Zack, how you doing?

What's going on?

How are you guys doing?

Did I get your last name even remotely correct?

No, that was perfect.

Perfect.

Wow.

Wow. Let's go. Yes. Thank you. Anyway, please introduce yourself.

Yeah, I'm Zach, founder of Omen AI. We do blood monitoring for mission critical machines.

Okay. Mission critical machines. I'm thinking data centers. What do you think?

Yes, sir.

You'd be right. Um, we got our start in big mining machines. You know, like the big machines you see on site. That's where spent a long time working.

Where does the name come from?

It's a it's a forecast of something to happen whether good or bad. You know, we obviously don't want our customers machines to break and so we like to know good omens.

So, the lifeblood of the modern data center. There's water, there's liquid that's cooling the data center. uh sometimes in a closed loop flowing around getting cooled outside the data center being brought in take removing the heat from the chips. Uh what's actually in that liquid? Is it just water? Is it water in something else? And then what happens like how does it get dirty if it's in a closed loop? Why doesn't it just stay clean and fresh forever? Why do you have a job? Why do you have a business?

Yeah, it's a great question. Um so it, you know, it depends the type of fluid. Normally, it's a water-based fluid with some sort of mixture of either ethylene glycol or propyline glycol. Um, you know, you're seeing a big shift recently towards more like, you know, uh, safer fluid. So, you know, propyline glycol is more preferred since it's less, you know, uh, toxic. Um, but

can you drink it? Could you have a

You can you can definitely vape it. Both of those are heavily used in the cigarette industry. Yeah. No, no joke. The whole VGPG ratio is a big thing in uh in the ecigarette world. But good that we found a good use for

you're basically smoking data center when you hit

that's one way to put it. The other way is that uh this is the solution to the vape crisis is that drive up the price of propylene glycol so that the everyone's like ah I got to quit vaping. I I got to save it for the data centers.

I'm going to use that now every every time I see someone hit a ecigarette. It's like, oh, really?

You're smoking a data center.

Yeah. Anyway, so what happens? How how does a a liquid cooling system actually degrade over time? How long does it take? And what's actually going on?

Yeah. So, a bunch of different things can happen in the system. Um, you know, you see a lot of, you know, basic like infrastructure wear. So, you know, when your pumps wear, your seals wear, all of that stuff starts to contaminate the liquid itself. Okay. And so whether it's like copper or chromium from the pipes or we're starting to see get in. Um what also happens is like the more heat that hap you know and the more wear that happens to your fluid the more it starts to break down. So you go from this very nice like high concentration of you know glycol to water to then that starts to degrade and then you start to get bio growth and other crap that gets in there and starts to eat your chips and uh you know sit on your in your pipes.

Has anyone tried using beef tallow to cool a data center? I think that might be the solution. That's a real meet in the middle compromise policy position maybe. Now, uh what is the go to market like?

Most niche niche joke of all time.

What what is the go to market like? $31 million raised. There's some hyperscaler mega data centers. Do they pick up the phone for you? Can you can you sell into them or are you working your way up with smaller data center providers? Sort of a walk approach like what's working? Yeah, I mean we've just we've been working with almost every you know kind of neo cloud that's out there and some more pilot deployment. It's been pretty crazy. I mean you think about like 9 months ago folks were just starting to get these you know systems online and these clusters online for liquid cooling and now you started to see folks who have running dozens maybe hundred hundreds of megawatts worth of you know liquid cooled compute and they're starting to see failures and so we kind of went from hey we're selling and folks don't really know what the problems are to now everybody's like [ __ ] we've had failures or we're starting to see failures our competitors are seeing failures um you know we need to adopt something. Yeah, and it's probably pretty easy to underwrite because every hour that the data center is not functioning is leaving hundreds of thousands, maybe millions of dollars on the table. Uh what is the actual installation process? Can you just like cut the pipe and then insert your device into the flow? What how long does it take to install? How big is is your solution?

Yes, our solution just fits across the rack manifold. So, we're kind of plugging into those OCP quick disconnects, mounting if you were a GPU. Um, so it only takes a couple minutes to install. What's great about like, you know, a cabinet is that we get power, we get, you know, networking and, you know, the fluids right there. Again, coming from industrials, that was not fun where you have every machine that has completely different integration. So, you know, we get to, you know, eat a lot less pain over on this side of things.

Sure. Sure. Uh, who did the round? Uh, what are you actually going to be spending it on? Is it just staff up? like what is the cash conversion cycle like for the business such that uh because I imagine there's a lot of cash flowing around and they might be able to just pay you up front.

How much did you raise first? Tell us. We want to hit the gong. Jord's warmed up.

Raised 31 million series A.

We got a bigger mallet. Really

got a bigger mallet.

Okay.

One of the first one of the first rounds with this mallet.

$31 million. Is that is that hiring scientists to to iterate on the product? Is it just uh scale up manufacturing? What are what are you actually putting the money to work for?

Yeah, I mean it's really just growing out the team and growing up manufacturing. Uh you know, when we raised our seed like eight months ago, it was basically just me, you know, working on this problem, you know, putting sensors on on excavators. And so we've kind of gone through this interesting process going from one to 20 and now, you know, we're going to go from 20 to to 40 folks by the end of the year. Mh.

Um, and the round was mostly like just really smart data center experts. And so it was led by a fund called Nova Ventures. Um, really a lot of strategic value from the data center space, a big LP in their fund to Cheryl Samberg who is also now an investor in us. And so we're really happy to have her on board. Um, and then we also have like CRV and uh a few other uh awesome like kind of seuite folks from Core Weeb, Spencer Wave and some other smart place

is Oh, sorry Jordy, you have something? I got another question. Uh so detection is only one piece of the puzzle is if you detect something earlier, you detect microbial buildup or copper leaking into the into the liquid cooling cycle. Um what does that actually what happens downstream of that? Is it that uh you swap all the water out or you go fix the root issue? Like what what is the technician side doing with the information that you give them? Yeah, I mean that's kind of the problem right now is like you have these monster labs that like they pull a sample from every major data center you can think of. They ship it off to their lab and now it's like let me wait a couple weeks to get a result back. And normally it's pretty binary. It's like we either flush it which you know is millions of dollars in downtime. I mean you think about a loop that's like call it a thousand GPUs. It takes four to five hours to flush a rack. And so now that's 5,000 GPU hours that just went down just to one rack. And so by monitoring it every day, you can say, "Okay, what are smaller chemistry changes we can make?" Whether it's bioshocking, whether it's adjusting the propyline glycol to water ratio, those are all things that the CDU can do. We just can't do it right now because we have these silly lab tests.

That's very cool. Very cool. Um, is the business 100% US right now? What is the state of the data center buildout internationally?

We work with only US like Neoclouds to date. Um, you know, some of these folks have projects all around the world. So, we're looking at some projects in Europe that are run by, um, a few of our partners, but it's been primarily US-based. We're based in San Francisco and we have an office in Orange County right now.

Amazing. Thank you. Great to meet you for coming on the show. Great to meet you. Congratulations and we'll talk to