Published in Joule on June 13, the research introduces a specialized fiber membrane that leverages capillary action and evaporation to manage heat fluxes exceeding 800 W/cm²—one of the highest on record for passive cooling systems. The membrane is composed of commercially available fibers originally designed for filtration, repurposed through mechanical reinforcement and pore-size tuning to deliver reliable, stable heat dissipation over extended operation.
“This passive system provides a compelling alternative to traditional cooling methods that rely on fans, liquid pumps, or phase-change materials,” said Dr. Renkun Chen, professor of mechanical and aerospace engineering at UC San Diego. “The porous structure enables continuous evaporation while minimizing the energy and water footprint, making it highly suitable for dense computing environments.”
Traditional air cooling is already reaching its thermal limits, especially with power-dense processors like GPUs and AI accelerators. Liquid cooling and immersion methods, while more effective, come with higher infrastructure costs, greater energy use, and environmental trade-offs. The fiber membrane, in contrast, can be integrated directly onto cold plates and microchannel heat sinks, offering drop-in compatibility with existing electronics architecture.
Importantly, the technology supports the industry’s dual mandate: enabling high-performance computing while aligning with corporate sustainability goals. Data centers already consume up to 40% of their energy on cooling alone, and with the sector projected to double its energy use by 2030, passive solutions like this are essential to mitigate emissions and operational costs.
The team is currently working to refine the membrane’s efficiency and begin prototype integration into cold plate systems for CPUs and GPUs. Commercialization is underway through a newly launched startup.
If successful at scale, this passive membrane technology could mark a significant shift in how the industry approaches thermal constraints—turning a filtration material into the next frontier in data infrastructure efficiency.
