Honeywell tests thermal module for Equinix AI cooling

Honeywell Aerospace has installed its Thermal Response Module at Equinix's Co-Innovation Facility in Virginia, where Equinix tested the system as part of its work on cooling for AI data centre infrastructure.

The module is designed to support liquid-cooled AI servers during power disturbances, sudden changes in compute demand and other cooling fluctuations. It sits between the coolant distribution unit and servers, allowing operators to add it to existing cooling loops without redesigning the wider system.

Demand for denser AI infrastructure has increased pressure on data centre cooling systems, particularly where operators use direct-to-chip liquid cooling for graphics processing units. In those settings, even short interruptions in cooling can affect performance and increase the risk of damage to expensive hardware.

The product uses phase-change materials to store and release thermal energy, keeping coolant temperatures stable while the rest of the cooling system adjusts. According to Honeywell, the module delivers 7.6 times more cooling power than chilled water, can stabilise the cooling loop within seconds and can be expanded to provide several minutes of thermal buffering.

The approach reflects a broader effort by Honeywell Aerospace to apply thermal management systems developed for aircraft and defence to adjacent markets. It says the heat loads seen in aerospace and defence can resemble those found in modern AI data centres.

Research firm Omdia estimates that some form of liquid cooling is installed in fewer than half of data centres and expects the market to grow by 20% a year over the next five years. That growth is being driven by the rising power demands of AI processors, prompting operators to move beyond conventional air cooling across more of the market.

Test results

The installation took place at Equinix's facility in Ashburn, Virginia, where the companies worked on development and testing. The module is designed to be vendor-agnostic, so it can be used with different coolant distribution unit architectures.

Equinix said the tests examined how the technology handled rapid changes in GPU cooling loads. It also compared the module's footprint with that of buffer tanks, another way of managing thermal variations in cooling systems.

"We are pleased to have collaborated with Honeywell Aerospace on the development and testing of the Thermal Response Module," said Pawel Wlodarczak, Design Innovation Director, Global Design, Equinix.

"Testing at our Ashburn Co-Innovation Facility demonstrated the technology's capability to manage rapid GPU cooling load fluctuations while requiring a smaller footprint in comparison to buffer tanks. As AI deployments continue to scale, innovations such as this will help address critical cooling and infrastructure challenges," Wlodarczak said.

Wider push

For Honeywell, the project is part of a wider attempt to expand the commercial use of systems originally built for mission-critical environments. Aerospace and defence platforms depend on tight thermal control to keep equipment operating under variable conditions, and that experience is now being applied to data centres handling heavy AI workloads.

The module includes state-of-charge sensing alongside a heat-exchanger design intended to support both rapid and sustained cooling response. Honeywell says the system can be tailored for deployments ranging from smaller edge sites to larger AI clusters.

Data centre operators have been looking for ways to reduce the risk of service interruptions as AI workloads push racks to higher densities. Systems that can absorb short-term thermal shocks may also help operators run with wider margins in cooling loop temperatures and reduce strain on chillers and upstream assets.

Rich DeGraff, President and Chief Executive Officer of Control Systems for Honeywell Aerospace, said the product was developed in response to those operating conditions.

"AI workloads are pushing data center thermal systems harder than they were designed to run," DeGraff said.

"We have taken phase-change material technology developed to protect mission-critical aerospace and defense systems and applied it to a similar challenge in AI infrastructure. Our work with Equinix shows how proven aerospace technology can help data center operators protect critical hardware, sustain performance and reduce the risk of disruption," he said.