As global demand for data centers surges, a growing number of startups are betting that the biggest opportunities aren’t only in servers and AI training—but in the behind-the-scenes infrastructure that keeps these facilities running reliably. One of the most expensive, disruptive risks inside modern data centers is also one of the most mundane: water.
MayimFlow, a startup that won the Built World stage at this year’s TechCrunch Disrupt, is built around a single, high-stakes mission: preventing damaging water leaks before they happen. The company says it can help operators move from reacting after a leak to predicting issues early enough to fix them—without taking systems offline unexpectedly.
Why water leaks are a big deal in data centers
Even small leaks can create outsized consequences in a data center environment. These facilities rely on complex water systems—often tied to cooling infrastructure—and the presence of water near sensitive equipment can trigger safety concerns, equipment damage, or operational shutdowns. In a business where uptime is paramount, an incident that forces systems to be powered down can quickly become extremely costly.
John Khazraee, MayimFlow’s founder, told TechCrunch that many data centers largely depend on reactive approaches to leak detection. In practice, that means teams may only discover problems once water is already where it shouldn’t be. The resulting response can involve emergency remediation, service disruption, and downtime that may cost millions of dollars.
MayimFlow’s pitch is straightforward: rather than waiting for alarms that indicate water is already leaking, predict the warning signs that suggest a leak is imminent and intervene earlier.
MayimFlow’s approach: IoT sensors plus edge-deployed machine learning
Khazraee’s background is rooted in large-scale infrastructure. He spent more than 15 years building infrastructure for IBM, Oracle, and Microsoft, experience that shaped his view of how preventable failures can cascade into major outages. Based on those observations, he built MayimFlow around a combination of IoT sensors and machine learning models that run at the edge—meaning close to where data is collected, rather than relying exclusively on centralized cloud processing.
According to Khazraee, the goal is not just to detect active leaks, but to identify subtle patterns in system behavior that often precede them. By deploying models at the edge, MayimFlow can potentially analyze sensor signals and produce alerts with lower latency while continuing to function even if connectivity to a central system is limited.
In an interview, Khazraee described the typical reality he saw in data centers: when the leak finally becomes visible, operators are forced into an expensive and disruptive scramble. Servers may need to be shut down, remediation crews brought in, and normal operations interrupted. That experience motivated him to pursue a more proactive solution.
What “edge AI” means in this context
Edge-deployed machine learning is often discussed as a way to make industrial systems smarter and faster. In a facility setting, it can enable on-site analysis of sensor data—pressure changes, flow anomalies, or other operational signals—without waiting for data to travel elsewhere for processing. While MayimFlow has not publicly detailed every input it monitors, its core claim is that the models can detect early indicators of developing issues and warn customers before damage occurs.
A team built around data centers, water management, and IoT
MayimFlow’s founding team reflects the cross-disciplinary nature of the problem. The startup brought together experience in data centers, water systems, and connected device infrastructure:
- John Khazraee, founder, previously spent more than 15 years building infrastructure for IBM, Oracle, and Microsoft.
- Jim Wong, chief strategy officer, has spent decades working with data centers.
- Ray Lok, chief technology officer, has built a career focused on water management and IoT infrastructure.
The combination matters because leak prevention in data centers is not purely a software challenge. It sits at the intersection of physical systems (pipes, pumps, valves, cooling equipment) and digital operations (monitoring, alerting, incident response), all within environments where downtime is costly and reliability standards are high.
From reactive detection to 24–48 hours of warning
MayimFlow’s promise is early notice—specifically, Khazraee believes the company can provide data center operators with 24 to 48 hours of advance warning that repairs will be needed. That window could be significant in a facility operations context, where planned maintenance is dramatically less painful than emergency response.
With an extra day or two, operators may be able to:
- Schedule maintenance during lower-usage periods
- Move workloads to redundant capacity
- Stage parts and personnel ahead of time
- Avoid shutting down systems unexpectedly
Even when redundancy exists, emergency incidents can disrupt workflows and increase operational risk. MayimFlow is essentially arguing that predictive warning turns a crisis scenario into a manageable work order.
How MayimFlow gathers and uses data
Predictive monitoring depends heavily on the breadth and quality of historical signals. Khazraee said MayimFlow has collected a glut of sample data from various industrial water systems, and that this dataset enables the company to make predictions about leak risk and maintenance needs.
On deployment, the startup can work in more than one way. It can provide sensors to monitor a water system, or it can integrate its machine learning models into existing sensor setups when a company already has relevant hardware installed. That flexibility could matter for data centers and other industrial facilities, where operators may be reluctant to rip and replace equipment that’s already part of a broader building-management system.
Why “picks and shovels” startups are attracting attention
In booming markets, many companies focus on being the “picks and shovels” suppliers—selling the essential tools that everyone else needs, rather than competing head-to-head to deliver the primary product. In the data center world, that can mean making money by improving resilience, monitoring, cooling, power management, or maintenance—areas that grow alongside capacity expansions.
MayimFlow fits that category by focusing on a specific but costly operational risk. While it isn’t selling compute, racks, or AI models, it aims to reduce the probability of outages and expensive remediation work—problems that can scale with the size and density of facilities.
A founder’s frugality and a focus on efficiency
Khazraee also framed his interest in water efficiency as personal. He told TechCrunch that he didn’t grow up in the most well-off family, and that his father frequently pushed him to conserve water—jokingly asking whether he was “singing” when he stayed in the shower too long. That upbringing, he said, shaped his mindset toward efficiency as he pursued engineering.
In college, he worked at a facility that collected frying oil from restaurants and converted it to biodiesel. He described the work as messy, but satisfying because of the outcome. Today, he says he’s combining that inclination toward efficiency with his team’s industry experience to address what he views as a growing challenge.
Beyond data centers: commercial buildings, hospitals, manufacturing, and utilities
While MayimFlow’s initial focus is data centers, Khazraee said he wants to expand the company’s technology into other environments where water issues create financial and operational risk. He pointed to commercial buildings, hospitals, manufacturing facilities, and possibly utilities. In his view, any organization that wants earlier leak detection or better water optimization could be a customer.
This broader ambition reflects a common pattern in industrial tech: start with a high-urgency, high-value niche (like data centers), then apply the same monitoring and predictive approach to adjacent sectors that share similar infrastructure challenges.
Turning down Big Tech roles to build MayimFlow
Khazraee told TechCrunch he has turned down roles at multiple Big Tech companies while building MayimFlow over the last two years. He said he’s committed to the company’s vision and believes in the impact of tackling water-related risks—especially as water becomes a larger global concern.
In data centers, water is intertwined with cooling strategies and operational continuity. From MayimFlow’s perspective, better detection and optimization aren’t just cost savings—they are resilience measures.
Conclusion
MayimFlow is positioning itself as a practical safeguard for one of the most disruptive problems data centers face: unexpected water leaks. By combining IoT sensors with edge-deployed machine learning and aiming to provide 24 to 48 hours of warning before repairs are needed, the startup wants to help operators reduce downtime, avoid costly remediation, and improve efficiency—then bring those same capabilities to other water-dependent facilities.
<<>>
Related Articles
- Why the Electrical Grid Needs More Software as AI and Data Centers Drive Demand
- Climate Tech in 2026: Investors See Data Centers, Power Demand, and Grid Resilience Driving the Next Wave
- WeTransfer Co-Founder Nalden Launches Boomerang, a Simpler File Transfer Service Without Logins
Based on reporting originally published by TechCrunch. See the sources section below.
