AI processors have a heat problem, making liquid cooling essential infrastructure.US firm Aviemore Motion Technologies has launched a new line of seal-less magnetic coupling pumps at the SC26 supercomputing conference in Chicago. It is designed specifically for liquid-cooled AI and high-performance computing (HPC) data centers.This patent-pending unit combines a pump, slotless brushless DC motor, and native 800 VDC drive under a single part number. The product platform delivers 0.10 to 5.0 kW of output across a 12 to 800 VDC range, offering specialized solutions for in-rack cooling units, row skids, and custom OEM architectures.“Liquid cooling is now critical infrastructure and the pump is the one moving part every rack depends on,” said Moe Barani, CEO of Aviemore. “Most pumps on the market were designed for another era and adapted for the data center. We designed ours for where AI infrastructure is going. There is no seal to leak or wear out, the drive runs natively on 800 VDC, and our team can turn a custom change in weeks instead of quarters.”Seal-less magnetic pumpLiquid cooling removes the heat generated by AI processors using closed-loop systems that recirculate coolant to save water and energy. Aviemore pumps could protect these closed loops by cutting the need for vulnerable shaft seals.Data center pumps usually use mechanical shaft seals, bolt-on motors, and third-party drives originally designed for other industries. Those shaft seals are the leading cause of leaks and premature hardware failure. Aviemore abandoned these outdated components to design a completely new solution.The firm designed a seal that transmits torque across a non-contact barrier, keeping the fluid path completely sealed. So nothing penetrates the fluid path. The pumps can even run completely dry during system commissioning or sudden loss-of-prime events without frying the hardware — a common nightmare for data center operators.“The traditional pump industry is full of dinosaurs,” said Matt Carroll, VP of Engineering at Aviemore. “They are still selling the same seal, the same separate motor and the same slow change process. Data center operators are left to design around it. We started from a clean sheet. The seal is gone and the motor and drive are built in. When a customer needs a change, our engineering team turns it in weeks.”Built for next-gen AI powerData center power architectures have been evolving faster alongside cooling demands. Most of the current AI server racks pull roughly 140 kW, but 240 kW configurations are already looming, and megawatt-class racks are deep in the planning pipeline. Rack power is simultaneously shifting from 54 VDC to 800 VDC, with initial commercial systems slated to ship in Q2 2027. To meet that horizon, Aviemore’s platform spans an output range of 0.10 to 5.0 kW across a 12 to 800 VDC bus range. It comes with dedicated product lines for everything from in-rack cooling distribution units (CDUs) and row-level coolant skids to fully custom OEM architectures.Vertical integration powers that speed. Aviemore handles winding, machining, assembly, and testing under one roof at its 156,000-square-foot ISO 9001-certified facility in Radford, Virginia. This in-house model lets the company turn around custom engineering modifications in weeks rather than quarters.