Omen AI raised a $31 million Series A on June 29, 2026, to solve a problem almost nobody outside data centers knows exists. The liquid coolant that keeps AI chips from melting grows bacteria. That contamination then clogs the cooling system. To run chips harder, operators add more water to the coolant mix, because water absorbs heat better. But a wetter mix invites bacterial growth that gums up the flow. The current fix is to flush the whole system. That can mean taking a rack offline for five or six hours, at a cost that runs into the millions.
Omen’s product is a small spectrometer. It reads the coolant’s chemistry in real time and flags trouble, bacteria, metal wear, fluid breakdown, before it forces a shutdown. The round was led by Nava Ventures, with CRV, Vanderbilt University, and executives from GM, Bridgestone, and Johnson Controls. The founder is the story too. Zach Laberge, 20, dropped out of high school and started his first company at 14. He pivoted Omen from monitoring tractor fluid to monitoring the most expensive computing infrastructure ever built. It’s a small device aimed at a small-sounding problem. And it sits directly in the path of AI’s entire physical buildout.
Best for anyone curious about the unglamorous reality behind the AI hype, or who likes a good underdog founder story. Not ideal for readers wanting model benchmarks, because this is about the plumbing, not the intelligence.
Everyone’s busy worrying about AI getting too smart.
The thing actually threatening the machines that run it is dumber than a goldfish. It’s bacteria.
Specifically, it’s bacteria growing in the coolant that keeps AI chips from cooking themselves, and this week a 20-year-old high school dropout raised $31 million to kill it before it kills your data center.
The Problem Nobody Talks About
Here’s something the AI hype cycle never mentions. The chips that run modern AI get blisteringly hot. They got hot enough that the old method of just blowing air across them stopped working. So the industry switched to liquid cooling. Now fluid runs directly past the hardware to carry the heat away. Water is the workhorse, because water absorbs heat far better than air.
But there’s a catch, and it’s a biological one. The coolant that keeps these chips alive is a mix of water and an additive that suppresses bacteria. To push the chips harder, operators can dial up the water content, because more water means better heat absorption. And pushing the chips harder is exactly what everyone wants to do when GPU time costs a fortune. The problem is that a wetter mix is also a friendlier home for bacteria. Warm, wet, and steady is ideal for them. And bacteria, given a warm, wet, nutrient-adjacent environment, do what bacteria always do. They multiply.
That growth turns into contamination, and the contamination clogs the flow of coolant through the system. Picture the gunk that builds up in a neglected aquarium or a pond in summer. Except it’s forming inside the circulatory system of a machine worth hundreds of millions of dollars. The most advanced computing infrastructure ever built gets throttled by pond scum. The same green slime that grows on a forgotten swimming pool.
Why Pond Scum Costs Millions
Why does this matter beyond the ick factor? Because the fix is brutally expensive.
When the contamination gets bad enough, the only solution is a flush. You drain the contaminated coolant, clean it out, and refill it. That flush can mean shutting down a rack for five or six hours. In a world where compute time is the scarcest resource in tech, taking a rack offline for most of a workday costs a fortune. The bill runs into the millions, per incident.
And here’s the part that makes it genuinely maddening for operators. They mostly can’t see it coming. Most of the industry monitors coolant the old-fashioned way. Pull a fluid sample, mail it to a lab, wait several days for results, and hope nothing breaks while you wait. For a system cooling hardware worth more than most office buildings, that’s absurd. It’s like checking your car’s oil by mailing a teaspoon to a lab once a quarter and praying the engine holds. As Omen’s founder put it, operators risk huge amounts of downtime. They have no insight into what’s going on chemically inside the loop.
So the status quo is a slow, blind, reactive process attached to the single most expensive buildout in computing. That gap is the entire business opportunity.
It’s worth pausing on the scale of the stakes. A single modern AI rack can hold hardware worth more than a luxury home, and a large data center holds thousands of them. When one rack goes down for a coolant flush, you’re not just losing the cost of the cleanup. You’re losing every hour of compute those chips would have sold, on infrastructure where demand massively outstrips supply. Multiply an unexpected six-hour outage across a fleet, a few times a year, and the math gets ugly fast. The contamination isn’t a maintenance nuisance. It’s a direct hit to the most valuable thing a data center has: uptime.
What Omen Actually Built
Omen’s answer is refreshingly physical in an industry obsessed with software. It’s a tiny spectrometer. The device reads the chemical makeup of the coolant by analyzing how it interacts with light. And it does this continuously, in real time, right there on the equipment.
Instead of mailing samples to a lab and waiting days, the device watches the fluid’s health second by second. It flags trouble before it becomes a flush. It spots bacterial growth early, when it can still be corrected cheaply, rather than after it’s already clogged the system. And bacteria aren’t the only thing it catches. Because it reads the fluid’s full chemistry, it can also detect metal wear. Copper or chromium in the coolant signals a pump wearing out. Silicon points to a failing seal. The fluid, it turns out, is a running diagnostic of the machine’s entire health. You just have to be able to read it in real time.
The key technical unlock, according to the company, is timing. Optical hardware, the sensors that make this kind of spectrometer work, has gotten cheap enough to deploy at scale. And the signal-processing software has gotten good enough to make sense of the readings. Neither was true a few years ago. The device sits right in the path of the most expensive build-out in computing. It works by watching the one variable, as one customer put it, that most of the industry is still flying blind on.
The 20-Year-Old Behind It
The founder story is almost too good, and it’s worth telling because it explains why this scrappy hardware company exists at all.
Zach Laberge is 20 years old. He founded his first company in 2020 when he was 14, raising $3 million to put sensors on construction equipment, and he dropped out of high school to do it. His parents backed the plan. That’s notable, given that his mother was a former Minister of Education for Ontario. There’s a particular irony in the former education minister’s kid leaving school to go build sensors. By all accounts the family was fine with it.
That first company eventually shut down, but it taught Laberge something that became the seed of Omen. The insight was simple. Fluid systems, oil, coolant, water, are a window into a machine’s health, if you can read them fast enough. He started Omen in 2024 with one idea. Replace the slow, lab-based process of fluid analysis with real-time, on-site awareness. The first market was heavy machinery, tractors and excavators, with Caterpillar dealerships as early customers.
Then the AI compute boom happened. Laberge saw the exact same problem inside AI data centers, just at vastly higher stakes. Fluid you can’t see inside of, attached to expensive hardware you can’t afford to have fail. The pivot from construction equipment to liquid-cooled GPUs is the whole company in one move. Same tech, bigger target. It’s the same core insight, applied to infrastructure an order of magnitude more valuable. Omen has now raised around $41.5 million total since its founding, with this $31 million Series A as the big one.
Why Investors Piled In
The investor list tells you this isn’t a novelty. The round was led by Nava Ventures, with participation from CRV, Vanderbilt University, Mann+Hummel, Starhill Holdings, and Hard Launch Capital. More telling are the personal investments from executives at Bridgestone, GM, Johnson Controls, and Tensorwave. These are people who run physical, industrial businesses, and they understand exactly what unplanned downtime costs.
Omen is already working with about a dozen data center customers, including Tensorwave, a company building an AI compute cloud on AMD chips. Tensorwave’s president framed the appeal bluntly. The fluid running through these massive systems is a critical variable that most of the industry is flying blind on. Better monitoring is how you support compute customers without nasty surprises. When the people building the data centers are also writing personal checks into the company that monitors their coolant, that’s a strong signal the problem is real and painful.
It’s also a fast-crowding space, which is its own kind of validation. Omen isn’t alone. An established water-monitoring firm called Pyxis rolled out its own data center coolant product earlier the same month. When two companies independently race to solve the same niche problem in the same month, it usually means the problem just became urgent. The liquid-cooling transition is happening now, at scale, and the contamination issue is arriving right alongside it.
The Bigger Picture
Strip away the specifics and Omen is a window into a truth the AI conversation keeps skipping: artificial intelligence is a physical thing, and physical things break in mundane, stubborn ways.
We spend enormous energy debating whether models are conscious, whether they’ll take our jobs, whether they’ll outsmart us. Meanwhile the actual machines running those models are vulnerable to heat, to corrosion, to wear, and yes, to bacteria growing in their plumbing. We’ve written before about how the AI buildout is colliding with physical limits like power and water. The coolant problem is the same story at the molecular level. AI itself is abstract. The infrastructure it runs on is wet, hot, and corrodible.
There’s a whole economy forming in this unglamorous layer. The same week Omen raised its round, a networking startup raised $190 million to get GPUs from different manufacturers to talk to each other. An observability company raised $35 million to monitor AI infrastructure. The flashy money goes to the model labs. But a quieter, arguably more durable business is being built by the companies keeping the physical machines alive. Picks and shovels, except the gold rush is compute and the shovels are spectrometers.
The lesson Omen offers is almost philosophical, dressed up as plumbing. The most advanced technology on earth still depends on keeping water clean enough to flow through a pipe. Bacteria don’t care that they’re clogging the coolant loop of a frontier AI cluster. They just grow, the way they’ve grown for three and a half billion years, indifferent to the value of the hardware they’re slowly choking.
So while everyone else stares at benchmarks and argues about the singularity, a 20-year-old who never finished high school is quietly solving a different problem. One that sounds like it belongs in a pool-maintenance manual. He’s probably going to make a fortune doing it. The unglamorous problems are the ones that actually have to get solved before any of the glamorous stuff can run. The future of AI doesn’t just depend on smarter models. It depends on someone, somewhere, making sure the water stays clean.
