Flexible Circuit Technologies (FCT) is developing a flexible liquid-leak detection assembly for AI servers and high-performance computing systems, targeting cooling components such as cold plates, manifolds, quick disconnects and coolant distribution units. The Minnesota-based manufacturer says the sensor can be customized to fit the geometry of liquid-cooled electronics, bringing leak detection closer to potential failure points.
As AI servers push data-center rack densities higher, liquid cooling is moving from a specialized technology toward a core part of high-performance computing infrastructure. That shift is creating another engineering requirement: detecting coolant leaks before they reach high-value electronics.
Flexible Circuit Technologies (FCT) is addressing that problem with a Liquid Leak Detection Flex Assembly designed for liquid-cooled AI servers, HPC systems and associated thermal-management equipment.
The Minnesota-based flexible-circuit and advanced-electronics manufacturer says its sensing assembly can be engineered around the mechanical geometry of cooling equipment while maintaining a thin, lightweight profile. Potential applications include GPU and CPU cold plates, manifolds, coolant lines, quick disconnects, coolant distribution units (CDUs) and server trays.
The approach is aimed at moving leak detection closer to the source rather than relying exclusively on facility-level water or liquid detection systems.
That distinction becomes more important as direct-to-chip cooling expands. In these systems, liquid is delivered directly to cold plates positioned against high-power processors, removing heat from components before the coolant returns through the system. JLL says rack densities can reach as high as 100 kilowatts for AI and HPC workloads, making conventional air cooling increasingly difficult for high-density deployments.
Uptime Institute’s 2026 research similarly shows how direct liquid cooling is becoming a central infrastructure consideration as AI clusters increase computing density. Its April 2026 analysis said planned AI hardware could drive rack power beyond 200 kW, increasing the need for liquid-based thermal management.
The growth of liquid cooling also changes the risk profile inside the server environment. Instead of keeping liquids primarily within facility mechanical systems, direct-to-chip architectures introduce coolant delivery hardware into or immediately around IT equipment.
FCT’s flex-based sensing concept is designed for that environment. The company says sensing patterns, circuit geometry, connector interfaces and the number of detection zones can be customized for individual cooling architectures.
Multiple sensing zones could allow equipment manufacturers to monitor different locations independently. For example, detection points could be positioned around a quick disconnect, cold plate or manifold, potentially helping operators identify the location of a problem rather than simply receiving a general liquid-presence alert.
The physical flexibility of the circuit is central to the proposition. Conventional rigid electronic assemblies can be difficult to place around irregular cooling components or inside increasingly constrained server designs. A flexible circuit can instead be shaped around the available mechanical space.
FCT’s broader business is built around this type of customized interconnect technology. The company manufactures flexible circuits, rigid-flex assemblies, flexible heaters, membrane switches and other electronics, alongside assembly and product-build services. Its published design guidance also describes manufacturing and assembly considerations specific to flexible circuits, including forming, material handling and SMT assembly.
For AI infrastructure manufacturers, the potential advantage is integration. A leak-detection assembly could be designed as part of the cooling hardware rather than added as a separate monitoring system after the server architecture has already been established.
FCT says it plans to work with server OEMs and ODMs, liquid-cooling companies, cold-plate manufacturers, thermal-management suppliers and data-center infrastructure providers. Its engineering support covers concept development, prototyping, design for manufacturing, flex-circuit fabrication, assembly and scalable production.
That supply-chain positioning matters because liquid cooling is not a single standardized architecture. Uptime Institute noted in June 2026 that direct-liquid-cooling deployments still vary in system design, coolant chemistry, resiliency approaches and maintenance procedures. The organization said operators and IT OEMs are increasingly looking at standardized approaches that can support different generations of liquid-cooled hardware.
The same fragmentation creates an opportunity for customizable sensing hardware. A leak detector designed around a particular cold plate or manifold may be more practical than a one-size-fits-all component, particularly where space, connector placement and detection coverage differ between server platforms.
The issue is also connected to maintainability. Uptime Institute has identified resiliency design and operational practices as continuing challenges for direct liquid cooling, even as AI workloads accelerate adoption. Early detection can potentially become one element of a broader maintenance strategy, alongside pressure monitoring, flow monitoring, temperature sensing and automated shutdown or isolation mechanisms.
FCT’s announcement does not disclose detection thresholds, response times, sensor chemistry, false-positive rates or independent validation results for the new assembly. Those specifications will be important for determining how the technology performs in production AI infrastructure, particularly in environments where coolant exposure could threaten expensive GPU and CPU systems.
Still, the product highlights an emerging layer of AI infrastructure engineering that is easy to overlook. As computing density increases, cooling is becoming tightly integrated with the server itself. That means monitoring the cooling loop increasingly needs to happen at the same level of granularity.
For FCT, the Liquid Leak Detection Flex Assembly extends flexible electronics into that emerging infrastructure. For data-center and server manufacturers, it points toward a model in which thermal management, sensing and electronics are engineered together rather than treated as separate systems.
Market Landscape
AI infrastructure is driving higher rack power densities and accelerating adoption of direct liquid cooling. JLL says AI and HPC systems can require rack densities of up to 100 kW, while Uptime Institute expects some future AI systems to push rack power beyond 200 kW.
That transition creates new requirements around maintenance, resiliency and monitoring. Uptime Institute’s 2026 research identifies unresolved operational issues around direct liquid cooling and notes variation in system designs, coolant chemistry and maintenance procedures.
As a result, AI Infrastructure is expanding beyond GPUs, networking and power systems to include specialized thermal-management hardware and sensors. Leak detection is one component of this broader shift toward instrumented, serviceable liquid-cooled computing environments.
Top Insights
- FCT is developing flexible leak sensing for AI servers, HPC systems, cold plates, manifolds, quick disconnects and coolant distribution units.
- The customizable flex architecture is designed to place detection zones close to potential coolant failure points.
- Rising AI rack densities are accelerating direct liquid cooling and increasing the importance of thermal-management monitoring.
- Uptime Institute identifies resiliency, maintenance and operational complexity as continuing challenges for direct liquid cooling deployments.
- FCT plans to support OEM and ODM customers from prototype development through flex fabrication, assembly and scalable production.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
