Omen AI raises $31M to monitor cooling water in AI data centers

Startup Omen AI has raised $31M for a system that monitors coolant in data centers. The niche is narrow but critically pressing: as GPU racks get denser and hotter, cooling water can turn into a breeding ground for bacteria, and until now nobody has tracked this systematically enough.

Bacteria as a side effect of the GPU race

The AI boom has an unglamorous underside: some of its hardest problems are, quite literally, plumbing. While the industry debates language model parameters and inference costs, engineers on data center floors face a very different reality. As data centers pack more and more GPUs into each rack and run them at peak loads, the coolant overheats, and microorganisms begin to multiply in it.

Contaminated coolant is not a theoretical risk. Bacterial films clog the tubes and channels of heat exchangers and reduce cooling efficiency. In severe cases, this leads to emergency server shutdowns. In an environment where downtime for a single rack of H100s costs tens of thousands of dollars a day, this is a critical vulnerability, and until now it has mostly been covered by manual checks every few weeks.

The problem has intensified as the industry moves from air to liquid cooling. Racks drawing 30–100 kW have become the norm in next-generation data centers, and air systems simply cannot handle that thermal load. But liquid cooling brings new operational risks that the infrastructure market was not adequately prepared for.

What Omen AI is building

The company is developing a platform for continuous monitoring of cooling loops. Sensors track water parameters in real time (temperature, chemical composition, signs of biological activity) and feed the data into a machine-learning-based system. The goal is to replace infrequent manual checks with constant automated oversight.

Key capabilities of the platform:

• Continuous measurement of coolant quality without manual sampling
• Early detection of bacterial colonies, while they are still easy to eliminate
• Thermal analytics broken down by individual racks and data center sections
• Integration with DCIM data center infrastructure management systems
• Automatic alerts when parameters move outside acceptable thresholds

Data center operators have historically treated cooling water as an auxiliary system with no active management. Omen AI wants to change that by making coolant quality control as standard a practice as monitoring CPU temperature.

The $31M raised will go toward expanding commercial pilots and developing the product.

Why the market was waiting for this

Cloud providers and hyperscalers are investing billions in physical infrastructure, yet monitoring of the coolant itself remains a weak link. Most systems record temperature and pressure, but not water quality, and that is where the risk lies. Omen AI is betting that as rack density grows, this gap will become unacceptable.

"The hardest problems of the AI boom turned out to be plumbing

problems": there is more accuracy in the founders' phrasing than it seems.

What this means

Omen AI is an example of how the physical infrastructure layer of the AI industry creates niche but critically important markets. While attention is fixed on models and chips, the real operational competition is unfolding over the reliability of cooling, power and water supply, and over the companies that can monitor all of it before a problem turns into an outage.