What Are Thermal Orphans? A Data Center Cooling FAQ

A couple of months ago, we presented some work on a problem that’s hiding in plain sight inside the liquid-cooled data center: thermal orphans. These are the NICs, SSDs, and other supporting modules that used to get cooled almost for free by bulk system airflow, which is disappearing with the increased use of full liquid cooling. I thought I would take a minute here to answer some of the most common questions about what thermal orphans are, why they’re showing up, and what we think can be done about them.

Q: What is a thermal orphan?
A couple of months ago at Data Center Cooling Days 2026, we presented some work that looked at targeted cooling of what we described as “thermal orphans” within the liquid cooled data center environment. These orphans aren’t the main server components (xPUs, Memory, VR, etc.), but rather the supporting modules often present on the same server platform. Common examples include NIC modules (highspeed or optical), SSDs for local/boot storage, and a host of other specialized PCIe functions. The power in these modules is often low (<50 W) and cooling them in an air-cooled server environment was trivial—just make sure some of the system airflow finds its way across or through the module. It is the liquid-cooling revolution that is turning these once-viable submodules into thermal orphans.

Q: Why do thermal orphans occur in data centers?
In today’s emerging liquid-cooled paradigm, the power-hungry main server fans are being removed as increasingly complex liquid cooling systems are being integrated into the platform to cool all the major components directly. This means that primary system airflow we used to rely on for cooling the orphan modules no longer exists in many systems. OEMs and ODMs now face a choice of further increasing the cost, complexity, and leakage risk by bringing coolant into each of the orphan modules or equipping them with their own dedicated air moving thermal solution within each module.

Q: What components are considered thermal orphans?
As mentioned above, many components could be considered a Thermal Orphan. What matters most is not what the component is, but whether or not it receives adequate cooling within the platform environment. In truth, any system component that is not directly cooled by the liquid-cooling solution could be a thermal orphan – the forgotten child of an outdated thermal design that has always assumed there would be bulk airflow to take care of it, but that bulk airflow is being taken away.

Q: Won’t advances in liquid-cooling technology just take care of thermal orphans naturally somehow?
Probably not. In fact, as the industry moves towards denser liquid-cooled rack solutions, the problem of effectively cooling thermal orphans is likely to become more challenging. There will be fewer system fans and much denser structures within the server platform that restrict airflow movement.

Some liquid-cooling technologies like immersion have great promise for managing thermal orphans, because the coolant in that case is present throughout all parts of the system, but many immersion systems struggle at higher xPU powers and market adoption for this technology remains relatively low.

Q: Can’t a heatsink with an integrated fan solve the thermal orphan problem?
In some cases, absolutely. I know that these are the cheapest, simplest solutions that I reach for first whenever I’m doing open board development work. And in some cases, they could solve the thermal orphan problem just fine. The main challenge is not so much the cooling, but the space available for that thermal solution. Integrated fan/heatsink solutions tend to be bulky, and both blowers and the downward-firing axial fans need extra vertical space on top of the already sizeable form factor to serve as an inlet plenum to allow air to get into the cooler. This vertical space is at a severe premium in today’s dense rack configurations. There is a lot of pressure to design blades that are 2U or smaller to meet density targets, and many, if not most, of the thermal orphan modules we see are only 0.5U in height. Finding a viable traditional fan solution can be challenging in those highly constrained geometries.

Q: How does ionic cooling help solve the thermal orphan problem?
Ionic cooling air movers have two distinct geometric advantages over traditional axial fans and blowers: First, they’re not just small and thin, but they also have a high aspect ratio that allows them to fit into crevices within the server platform that might otherwise be hard to access. They’re good at getting into all those little nooks and crannies to provide cooling right where the heat tends to build up and accumulate (heat building up in these spots is something commonly seen even in some hybrid liquid-cooled servers that still have bulk air movement – it is just the build density that prevents the bulk cooling air from getting everywhere it needs to go). Secondly, ionic cooling modules have a unique side-in, side-out airflow pattern for a thin air mover. This eliminates the need for an inlet plenum above the device and provides better system integration flexibility.

Q: Is ionic cooling a replacement for fans or liquid cooling in data centers?
No. The incredibly high-power components in today’s AI data centers will continue to rely on liquid cooling, and perhaps even 2-phase liquid cooling, for the foreseeable future. The nominal flow rate for an ionic cooling device being around 1 CFM makes it a poor fit for thermal management of a 200-kW rack. But because that airflow can be targeted precisely within the server platform with minimal mechanical accommodation, we believe they are a viable option for addressing the thermal orphan challenge that designers are facing with today’s high-density systems.

If you have further questions about our thermal orphans, don’t hesitate to reach out directly. We look forward to continuing the conversation.