Cooling Without the Pump: How Mitsubishi Electric Is Rethinking Electronic Cooling with Microbubbles
As artificial intelligence workloads grow, chips become denser, and power electronics push toward ever-higher frequencies, one constraint keeps resurfacing: heat. Electronic cooling has quietly become one of the limiting factors in the future of electronics, from AI accelerators and data centers to laser systems and high-frequency power devices.
Traditional cooling methods—fans, liquid loops, and miniature pumps—work well, but they come with trade-offs. Pumps consume energy, take up space, introduce mechanical complexity, and become increasingly inefficient at microscopic scales. In a world where performance gains are often measured in single-digit percentages, energy-efficient cooling matters more than ever.
In the first week of December in 2025, Mitsubishi Electric Corporation announced a research breakthrough that challenges one of the basic assumptions of liquid cooling: that you need a pump to move coolant. Working with Kyoto University’s Suzuki & Namura Laboratory, the company demonstrated what it describes as the world’s first technology capable of generating millimeter-scale flow inside a channel using microbubbles just 10 micrometers in diameter as the driving source.
At first glance, the announcement sounds highly academic. But beneath the physics lies a practical ambition: eliminate external pumps from microchannel cooling systems, reduce energy consumption, and enable more compact, carbon-neutral electronics.
“This technology is expected to reduce power-consuming external pumps for water cooling in electronic equipment,” a Mitsubishi Electric spokesperson told us. “Going forward, Mitsubishi Electric aims to enhance the energy saving of next-generation cooling systems by eliminating the need for external pumps, thereby contributing to carbon neutrality.”
Why Pumps Are Becoming a Bottleneck in Electronics Cooling
As electronics shrink and power densities rise, cooling systems are moving closer to the heat source. Microchannel cooling—where liquid flows through tiny channels etched into substrates—has become an increasingly attractive approach because it allows heat to be removed directly at the source.
The problem is that forcing liquid through these narrow channels requires pumps that must overcome significant friction. At small scales, fluid resistance rises sharply, meaning pumps must work harder for diminishing returns. They also add bulk, complexity, vibration, and points of failure—factors that become more problematic as systems miniaturize.
In data centers, pumps and cooling infrastructure already account for a substantial share of total energy use. At the device level, they limit how compact or integrated systems can become. Mitsubishi Electric’s research targets this bottleneck directly by asking a fundamental question: what if fluid flow could be generated inside the channel itself?
Microbubbles as Microscopic Engines for Pump-Free Cooling
The answer lies in a physical phenomenon known as the Marangoni effect. Marangoni forces arise when surface tension varies along an interface, often due to temperature differences. In everyday life, this is why a drop of detergent can cause water to move across a surface.
At the microscale, the same principle can be applied using tiny bubbles.
“In this study, we conducted a proof of concept for generating flow in a microchannel using Marangoni forces around microbubbles and the self-oscillation of a bubble,” the Mitsubishi Electric spokesperson explained.
Kyoto University developed the core technique, using localized heating to generate microbubbles and exploiting temperature differences at the vapor–liquid interface. These gradients induce fluid motion, while the bubble’s natural oscillation reinforces the flow.
Mitsubishi Electric focused on adapting this laboratory phenomenon into a controlled microchannel cooling environment relevant to electronics and semiconductor thermal management. The result was a system in which fluid moves continuously—without any external pump—driven purely by the physics around the bubble.
In the future, the company sees an even more self-sustaining approach. “Going forward, Mitsubishi Electric aims to develop technologies that generate microbubbles using waste heat from electronic devices,” the spokesperson said. In effect, the heat that needs to be removed would also help power the cooling flow.
From Physics to Measurable Flow in Microchannels
Turning elegant physics into measurable, stable flow inside a microchannel is far from trivial. In its initial demonstration, the joint research team achieved a flow speed of 100 micrometers per second in a 3 mm × 3 mm square channel with a cross section of just 100 µm × 400 µm—all without using an external pump.
While those numbers may seem modest, they are meaningful in context. “As a research-level comparison, it was shown that 100 µm/s is comparable to the flow speed achieved by the conventional EHD pump,” the Mitsubishi Electric spokesperson noted.
By optimizing temperature distribution around the microbubble and redesigning the channel geometry to suppress vortices, the researchers increased the flow speed to 440 micrometers per second—underscoring the scalability potential of pump-free microfluidic cooling.
Scaling Microbubble Cooling Toward Industrial Use
So how did they translate Kyoto University’s laboratory setup into something resembling an industrial electronic cooling system?
“The key challenges were to define the target scale required for industrial applications and to optimize the arrangement of laser spots and the channel geometry to expand the flow to the millimeter scale,” the Mitsubishi Electric spokesperson said.
Scaling remains the critical hurdle. Practical systems will require multiple microbubbles operating in concert, stable flow over longer channels, and compatibility with existing semiconductor materials and manufacturing processes.
“The research is still at the fundamental stage, and there are many challenges to be addressed before practical implementation,” Mitsubishi Electric acknowledged.
Implications for AI Cooling and High-Power Electronics
Despite growing interest in AI data center cooling, Mitsubishi Electric is careful not to overstate near-term applications.
“Currently, the research is at the fundamental stage, and the practical form has not been specifically determined,” the spokesperson said. “In the future, we aim to generate larger flows and apply the technology to cooling electronic devices such as high-frequency components and laser light source devices.”
These use cases—high-frequency electronics and photonics—often demand extremely compact, localized cooling, making them logical early candidates for pump-free thermal management.
Energy Efficiency and Carbon-Neutral Cooling
Even without precise projections, the energy logic is clear. Eliminating pumps removes a continuous source of power consumption.
“Since cooling can be achieved without a pump, this leads to energy savings and space reduction, contributing to the enhanced performance and miniaturization of electronic devices,” Mitsubishi Electric said. “Since cooling can be achieved without a pump, there is a potential for space savings.”
At scale, such improvements could contribute meaningfully to energy-efficient AI infrastructure and lower-carbon electronics.
Early Research Signals and What Comes Next
The research has been published in Applied Physics Letters. While Mitsubishi Electric says it has not yet received formal feedback on the manuscript, it notes growing academic interest. “We have given invited presentations on microbubble-related research,” the spokesperson said.
Commercial timelines remain open. “Currently, the research is at the fundamental stage, and the schedule for practical implementation has not been specifically determined,” Mitsubishi Electric added.
Rethinking Thermal Management from the Inside Out
In discussions about sustainable technology, attention often gravitates toward large systems—renewable energy, grids, and hyperscale data centers. Yet some of the most consequential innovations happen at the microscopic level.
Microbubble-driven flow challenges the assumption that active cooling requires mechanical pumping. Instead, it points toward self-driven, pump-free cooling where waste heat helps remove itself.
For Mitsubishi Electric, this work signals a longer-term strategy in advanced electronic cooling, power electronics, and energy-efficient AI infrastructure. The humble microbubble may still be a research tool—but it could become a building block of next-generation thermal management.