Shutting down a GPU rack for five to six hours just to flush bacteria-clogged coolant costs millions of dollars in lost compute. Omen AI has a $100 part — a tiny spectrometer — that makes that scenario obsolete.
The $1M+ Flush Problem
Data centers running liquid-cooled chips mix water with a biocide to prevent bacterial growth. Push the system hotter (as everyone wants) by increasing the water ratio, and contamination blooms into a sticky biofilm that clogs fluid pathways. The fix: shut down the rack, flush the loop, lose half a shift of revenue. Omen CEO Zach Laberge calls it “flying blind” on fluid chemistry.
A Spectrometer That Sees the Slime Coming
Omen’s device mounts inline on the coolant loop, using recent improvements in optical components and signal processing software. It identifies bacterial colonies before they clog, and also spots copper or chromium particles from worn-out pumps, or silicon from degrading seals. Real-time alerts replace the manual “pull a sample, mail it to a lab, wait days” workflow that has been standard.
$31M to Scale from Construction to Compute
Laberge, who dropped out of high school after raising $3M for a construction-equipment sensor startup at age 14, founded Omen in 2024. The same fluid-monitoring tech that caught on with Caterpillar dealerships for heavy machinery is now being deployed in data centers. Caterpillar, a major supplier of gas-turbine generators for on-prem power, helped bridge the gap. Omen today announced a $31M Series A led by Nava Ventures, with CRV, Vanderbilt University, Mann+Hummel, and individual investors from Bridgestone, GM, Johnson Controls, and TensorWave.
A Dozen Data Centers Already Piloting
Omen is working with a dozen data center customers including TensorWave, the AMD-based AI compute cloud. “The fluid is a critical variable that most of the industry is flying blind on,” said TensorWave president Piotr Tomasik. Competitor Pyxis rolled out a coolant monitoring product this month, but Omen’s head start and Caterpillar relationship give it a direct channel into backup-power and cooling systems.
Real-time fluid analytics will become table stakes for liquid-cooled infrastructure as rack densities climb. Omen is betting that the cheapest part of the loop — monitoring — saves the most expensive thing: uptime.
Source: Omen AI's plan to optimize data centers is all wet
Domain: techcrunch.com
Comments load interactively on the live page.