The unglamorous truth about the AI boom is that some of its hardest problems are plumbing. As data centres pack more GPUs into every rack and run them hotter, the fluid that keeps the chips from cooking has started, occasionally, to grow bacteria.
That is the problem Omen AI has built a company around, and on 29 June it said it had raised a $31m Series A to chase it. The round was led by Nava Ventures, with CRV, Vanderbilt University, Mann+Hummel, Starhill Holdings, and Hard Launch Capital taking part, alongside personal cheques from executives at Bridgestone, GM, Johnson Controls, and Tensorwave.
The mechanics are oddly specific. Liquid-cooled chips run on a mix of water and an additive that suppresses bacteria. To push the chips harder, operators can dial up the water, which absorbs heat better, but a wetter mix invites contamination that clogs the flow.
The fix, once it goes wrong, is to flush the system, which can mean taking a rack offline for five or six hours at a cost that runs into the millions. Omen’s answer is a small spectrometer that reads the fluid’s health continuously and flags trouble before it becomes a flush.
“You’re not risking huge amounts of downtime because you have no insight into what’s going on chemically,” chief executive and founder Zach Laberge told TechCrunch.
Laberge is an unusual founder for an infrastructure company. He started his first business in 2020 at 14, raised $3m to put sensors on construction equipment, and dropped out of high school to run it, with the backing of his parents, one of whom was a former Ontario education minister.
After that company shut down, he started Omen in 2024, originally aimed at fluid systems in heavy machinery: the same idea of replacing lab samples with real-time readings, applied to engines rather than servers.
The pivot to data centres came through the back door. Caterpillar dealerships were early customers for the machinery business, and Caterpillar also supplies the turbines and generators that power data centres on site.
About six months ago, Laberge said, those dealers started asking whether Omen could monitor the buildings too. The buildings, it turned out, were full of fluid, from HVAC systems to chip cooling, and a fast-growing customer base came with them.
Omen is now working with about a dozen data-centre customers, including Tensorwave, which is building an AI compute cloud on AMD chips. The bet riding underneath the round is the same one driving every cooling startup: that liquid cooling is no longer optional.
Rack densities have climbed past what air can handle, the same threshold that drew $26m into the liquid-cooling firm Iceotope as operators scramble to retrofit.
Omen is not alone in trying to bring fluid analysis out of the lab. Pyxis, an established water-monitoring company, rolled out its own data-centre coolant product earlier in the month, and the broader category is crowding fast as the environmental cost of water-hungry data centres draws regulatory attention.
What Laberge argues sets Omen apart is timing: optical hardware has become cheap enough to deploy at scale, and signal processing good enough to make sense of the readings.
It is a small device addressing a small-sounding problem, monitoring the chemistry of coolant, that happens to sit directly in the path of the most expensive build-out in computing. The fluid running through these systems, as one customer put it, is a critical variable most of the industry is still flying blind on.
