Exploring Liquid Cooling Pipes: Extreme Production Control For Zero Leakage Performance

With AI/GPU rack heat densities reaching 50 – 150 kW or even higher (based on the NVIDIA GB300/Rubin), thermal management demands for data centers sharply increase, driving the more strict requirements for the pipelines which carrying and distributing coolants in a liquid cooling system.

So, which piping systems does this “zero leakage” requirement specifically apply to? And how do manufacturers control designing and engineering processes to ensure that every pipeline is completely leak-free? Let’s find answers from the following sections.

Which Liquid Cooling Pipelines in Data Centers Must Zero-Leak Requirements?

Check the following chart to find critical pipes included into liquid cooling system for AI data centers.

▶ Liquid Cooling Manifolds

Liquid cooling manifolds, especially for in-rack manifolds, is an area with a high risk of liquid leakage since frequent plugging and unplugging operations. They’re usually positioned inside server racks to precisely distribute coolants to every chip’s cold plate

▶ CDU Internal Pipes

CDU internal pipes refer to entire stainless steel piping integrated into coolant distribution units (CDUs), to connect every two components within CDU such as heat-exchanger, pumps, primary filters, secondary filters, valves, etc.

Dense network, compact deployment, restricted spaces, make the maintenance for CDU internal pipes more difficult. If a leak occurs and repairs are not carried out promptly, it could cause the entire circuit to shut down completely.

▶ Secondary Loop Piping

Secondary loop piping is positioned between CDUs and IT server racks, typically installed beneath the raised floor of the server room or laid horizontally along cable trays on top of the server racks. It features a dynamic adjustment capacity, optimizing heat dissipation efficiency by adjusting flow rates in real time based on server load, thereby maintaining system stability.

Once the leakage happens, not only could cause short circuits or contaminate expensive server equipment, but it could also severely impair the system’s cooling efficiency and disrupt the normal operation of the entire data center.

Why is It So Difficult to Achieve Zero Leakage in These Pipelines?

Every tiny leak in the pipes that transport and distribute coolant is magnified in AI liquid cooling systems — because they always operate in extremely demanding environments.

• High Pressure: pipelines should work at high pressures usually in 43.5 – 87 psi, with frequent start-up and stop operations and impacts.
• High Temperature: the temperature of chips can reach to 140 – 194 ˚F(60 – 90 ˚C) during operating, piping must withstand multiple thermal cycles every day.
• Hard Maintenance: multiple components should be connected with pipes in a compact space, so the pipeline deployment is often complex, so that the maintenance difficulty is unfortunately increased.

Under such harsh working conditions mentioned above, even the slightest “flaw”, i.e. a 0.1mm pitting defect, a micron-sized porosity in a weld, or a minor scratch on an O-ring — can become the source of a leak, resulting in serious consequences and losses.

How to Ensure These Pipelines Are Leak-Free? 

So, what can we do to make sure your liquid cooling pipelines far away from any potential leak risk, during our designing & manufacturing processes?

▶ Simplified Design

The best approach to achieve leak-free performance is reducing connection nodes and simplifying the construction design. Thanks to our professional and experienced technical team, we can optimize the overall piping structure of your liquid cooling system with minimum connection points, reduce the likelihood of leaks at the source.

▶Material Selection

From tube, fitting, to sealing ring, materials of all these components should be selected strictly based on their chemical compatibility, pressure resistance, temperature resistance, fatigue resistance.

The main body of liquid cooling piping must be stainless steel, while the 316L is the first choice as its Mo content ≥ 2.0% for increasing pitting resistance in pren ≥ 25. Furthermore, 100% positive material identification (PMI) test and material certification (i.e. EN 10204 3.1/3.2) are required for each batch of raw materials.

▶ Welding Control

Laser welding, TIG welding (Tungsten Inert Gas Welding), vacuum brazing are the most frequently used weld technologies to weld a varied types of pipelines for data center liquid cooling. And virtually all welding processes involve automated, unmanned operations to enable the most precise process control.

TIG welding is used for regular welding process, high-pure argon(≥ 99.999%) shielding is maintained throughout the process to ensure that the back of the weld is free of oxidation and that the inner surface exhibits a silvery-white metallic luster.

While for precision components or joints welding, such as in-rack manifold joints, cold plates, laser welding and vacuum brazing will be the top options for the sake of minimized heat-affect zones and deformations.

Each liquid cooling pipe whether large or small, whether straight or bend, must experience pickling and passivating to remove the HAZ sensitized layer to restore the anti-corrosion capability.

▶ Smooth Surface & Clean Assembly

Keep inner surface of piping smooth and clean is one of the most critical action to achieve the leak-free operating.

Polishing is the first step to make it reach the target roughness of Ra ≤ 0.6 μm or even Ra ≤ 0.2 μm for certain high-precision requested pipes.

Pure water flushing to remove impurities, then high-pure nitrogen gas blowing to clean and dry these pipes. After manufacturing, they’ll be pre-assembled and packaged in a class 100,000 or 1,000,000 cleanroom, eliminate the possibility of contamination at the final stage before shipment.

▶ Redundancy Design

Redundancy designs are necessarily required when connecting pipes or pipelines in the liquid cooling system for various high-density computing and hyperscale computing centers.

• Double Sealing Design

Primary seal + backup seal, double sealing insurance adopted for every key interfaces involved thread or flange connection.

• Redundant Isolation Design

N+1 or 2N redundant design is always a reliable “Plan B” to minimize leakage risk and resulting losses. Install redundant valves in parallel at critical points (such as main inlets/outlets and major branches), or use a double-isolation design, to ensure that when one valve is being maintained, the other securely contains the fluid.

▶  Testing & Validation

Zero leakage is not only a matter of the welder’s skill or welding technology; it more depends on the entire verification system.

• Weld seam inspection

Borescope and X-ray inspection, covering all of inner surfaces and elbows.

• Leak test

100% helium mass spectrometry detection, ensuring pipe and welds with high integrity.

• Fatigue test

500,000 times of pulse fatigue test, far exceeding the requirements of standard industrial piping.

• Dimension validation

CMM (coordinate measuring machine) inspection or blue-ray 3D reverse engineering, controlling tolerances within a minimized range with 100% installation compatibility.

• Corrosion test

Strict salt spray test and crevice pitting test, ensuring integrity of pipes and welds under long term operation in highly aggressive environments.

Conclusion

In conclusion, leakage prevention for the various piping systems in data center liquid cooling is a major challenge that tests the full chain from design, manufacturing, to validation. If you have any questions on the liquid cooling pipes for data centers, hyperscale computing centers, high-density computing centers… feel free to contact Filson via sales@filsonfilters.com or directly call at +(86) 157-3695-8886.

FAQ

▶ What happens if liquid cooling leaks?

Even a single drop of leakage could cause CPU/GPU clusters to crash, result in data loss, or even trigger a fire — leading to economic losses in the millions or tens of millions. Therefore, it’s crucial to eliminate the risk of leaks, whether in the manifolds that are in direct contact with the servers or in the secondary loop piping installed over the ceiling or under the floor.

▶ Why is orbital welding often preferred for zero-leak liquid cooling pipes?

Orbital TIG welding is widely preferred because it offers dependable welding quality with low defect rates and minimal oxidation range. This allows it to achieve better consistency required for leak-critical applications, like for pipes integrated within CDUs or rack manifolds.

▶ How can buyers evaluate whether a liquid cooling pipe supplier can really deliver zero-leak assemblies?

Ask beyond “Can you make leak-free pipes?”, you could request evidences that can identify their real capabilities, such as: certificates for welding (about welding equipment, process control/record, welder qualification), leak test criteria, NDT test report, raw material traceability certification and PMI report for finished pipes, if any traceable number for each pipe for process validation, if any a cleanroom for particle-free assembly and packing.