Sunon expects cooling demand surge

Cooling used to be the supporting actor in data centers. Now it is drifting toward the center of the budget. That is the real news behind Sunon’s latest readout: the Taiwanese fan and thermal supplier says AI infrastructure spending is pushing up demand for industrial cooling gear, and it expects electronically commutated, or EC, fans to become a major growth driver in 2027. The timing matters because the rest of the AI stack is getting denser, hotter, and harder to build around. ### What exactly did Sunon say? Sunon’s message was pretty simple — global AI infrastructure investment is no longer just helping premium server parts, it is pulling through demand for the boring but necessary hardware that keeps those systems alive. The company specifically flagged industrial cooling products, and singled out EC fans as a major growth engine into 2027. That fits with how Sunon has been positioning itself in recent months, with a louder push into AI thermal management and higher-efficiency motor systems. (digitimes.com) ### Why do EC fans matter? Because they are one of the cleaner ways to squeeze more cooling efficiency out of a power-hungry facility. EC fans combine electronic controls with brushless motors, which lets operators adjust airflow more precisely and waste less electricity than older fixed-speed setups. In AI data centers, that matters twice — once in the direct fan power bill, and again because better airflow can widen the safe operating envelope for racks that are already pushing brutal thermal loads. (digitimes.com) Sunon has been marketing exactly that angle. ### But isn’t AI moving to liquid cooling? Yes — but that does not kill the fan story. Turns out liquid cooling usually adds layers instead of replacing everything. Cold plates, coolant loops, and sidecar systems still need air movement around power supplies, networking gear, storage, and the room itself. Sunon has been explicit that it wants to sell both sides of that stack, from AI fan modules to direct-to-chip and air-liquid systems. So the company is not really betting against liquid cooling. (sunon.com) It is betting that denser AI clusters need more total thermal engineering, not less. ### Why does Nvidia’s IREN deal matter here? Because it shows the scale that cooling suppliers are trying to map against. Nvidia and IREN said on May 7 that they plan to accelerate deployment of up to 5 gigawatts of AI infrastructure globally using Nvidia’s DSX AI factory architecture. Nvidia also received a five-year right that could let it invest up to $2.1 billion in IREN. That is not just a chip supply story. Five gigawatts is a giant physical buildout story — land, substations, switchgear, chillers, pumps, fans, and all the plumbing around the GPUs. (sunon.com) ### Why is cooling becoming a bigger bottleneck? Because compute density has outrun the old assumptions. When clusters scale fast, operators hit power and heat constraints before they hit theoretical chip demand. A useful way to think about it is this: GPUs are the engine, but thermal and electrical systems are the roads. If the roads are too narrow, you do not get full speed no matter how good the engine is. That is why suppliers like Sunon can talk about demand surging even if they are far less glamorous than the chipmakers. (nvidianews.nvidia.com) ### Does this change who captures AI spending? At the margin, yes. The AI trade started as a chips-first story. It is becoming an infrastructure story. Companies that can tie silicon choices to airflow, liquid loops, redundancy, and facility efficiency have more room to matter than they did a year ago. Digitimes’ framing around Sunon basically lands there: once customers commit to large AI factories, thermal hardware stops being optional optimization and becomes part of the deployment critical path. (digitimes.com) ### So what should readers watch next? Watch whether Sunon turns this demand language into visible revenue mix changes in 2026 and 2027 — especially around EC fans and AI-oriented cooling systems. Also watch whether more Nvidia-style infrastructure partnerships show up, because those deals make the hidden constraint obvious. The more AI gets sold as a factory build rather than a box of chips, the better the setup looks for cooling vendors that already have products spanning both air and liquid systems. (digitimes.com) ### Bottom line? Sunon’s update matters because it captures a shift people keep underestimating: AI demand is no longer measured only in GPUs shipped. It is measured in heat that has to be removed, power that has to be delivered, and facilities that have to run without breaking. In that world, cooling stops being background hardware and starts looking like core infrastructure. (digitimes.com)