Microsoft unveils microfluidics cooling to cut AI data center power use amid AI’s soaring energy demands

AI’s hunger for power has turned data centers into some of the world’s most energy-intensive facilities. Cooling alone can account for nearly 40% of electricity use. Microsoft thinks it may have found a way to keep chips from burning up without draining the grid: a cooling system that pushes liquid directly through channels etched inside the silicon itself. On Monday, Microsoft announced “a microfluidics breakthrough goes straight to the silicon to cool up to three times better.”

The approach, called microfluidics, is still in prototype stages, but early results suggest it could reshape how servers run AI workloads. By tackling the heat problem at its source, Microsoft hopes to make data centers more efficient, cheaper to operate, and less harmful to the environment.

“The chips that datacenters use to run the latest AI breakthroughs generate much more heat than previous generations of silicon. Anybody whose phone or laptop has overheated knows that electronics don’t like to get hot. In the face of rising demand for AI and newer chip designs, the current cooling technology will put a ceiling on progress in just a few years. To help address this problem, Microsoft has successfully tested a new cooling system that removed heat up to three times better than cold plates, an advanced cooling technology commonly used today,” Microsoft said in a blog post.

How Microsoft’s Microfluidics Cooling Breakthrough Could Save Power and Boost Performance

Microfluidics relies on hair-width channels carved directly into a chip. Liquid flows through these veins and pulls heat away where it’s generated, instead of relying on external methods like air fans or cold plates. This allows chips to operate at higher temperatures—up to 70°C (158°F)—without throttling performance.

Microsoft’s engineers say the inspiration came from biology. Veins in leaves and butterfly wings distribute nutrients and fluids with remarkable precision, and the company used AI to mimic those patterns for cooling. Husam Alissa, who leads Microsoft’s systems technology group, explained that some chip hot spots can reach heat fluxes comparable to the surface of the sun on a micro scale.

To build the prototypes, Microsoft teamed up with Swiss startup Corintis. Together, they developed a design that prevents clogs, maintains flow, and keeps the silicon intact. Tests have shown the method removes heat up to three times more effectively than current cold-plate technologies, cutting maximum temperature spikes by as much as 65%.

“One of the major reasons why AI data centers are sucking up so much power is the need to cool processors that run very hot. But Microsoft Corp. is trying out a possible solution: sending fluid directly through tiny channels etched into the chips. The technology is called microfluidics, and it’s being used in prototype systems in test conditions at the company,” Bloomberg reported, citing Husam Alissa, who oversees Microsoft’s systems technology.

Why Microfluidics Matters

AI’s explosive growth has driven a surge in energy consumption. Companies are scrambling to meet demand, and cooling remains one of the biggest bottlenecks. By embedding liquid cooling directly into chips, Microsoft could pack more computing power into smaller spaces. That means denser server racks, fewer buildings, and lower operating costs.

Another advantage is overclocking—running chips hotter and faster when workloads spike. Jim Kleewein, a Microsoft technical fellow overseeing Office 365 services, described how Teams traffic surges at the start of meetings. Instead of spinning up more servers, microfluidics could let Microsoft push existing chips harder for short bursts, cutting down on hardware needs.

The potential environmental impact is just as significant. Microsoft estimates the approach could improve power usage effectiveness, reduce greenhouse gas emissions by 15%, and lower water consumption by up to half compared to traditional cooling.

Beyond Prototypes

The system has already been tested on Intel Xeon chips and GPUs handling AI workloads, with prototypes supporting Office cloud applications. The company plans to extend microfluidics to its own custom Cobalt CPU and Maia AI accelerator. Engineers are even looking at 3D-stacked chip designs where coolant flows between layers, potentially delivering huge leaps in performance.

Getting here wasn’t simple. Engineers iterated through multiple designs, experimenting with etching methods, coolant formulas, and packaging to prevent leaks. Judy Priest, Microsoft’s corporate vice president and CTO of Cloud Operations and Innovation, said reliability was a top concern: “But we needed to prove the technology and the design worked, and then the very next thing I wanted to do was test reliability.”

Microsoft is also working with Corning Inc. and Heraeus Covantics to scale production of hollow-core fiber for faster networks, and with Stegra on low-carbon steel for data center construction. These partnerships point to a broader effort: rethinking the entire system, from chips to buildings, to meet AI’s demands.

The Bigger Picture

Satya Nadella praised the breakthrough on X, calling it “an important breakthrough from our teams: a new approach to liquid cooling that uses microfluidics, opening the door to more efficient, sustainable, and power-dense datacenters than conventional methods.”

The stakes are high. AI’s energy needs are projected to surpass Bitcoin mining by late 2025. Without innovations like microfluidics, the power required to train and run large models could overwhelm grids and inflate costs. Ricardo Bianchini, a Microsoft technical fellow, put it bluntly: “If microfluidic cooling can use less power to cool the datacenters, that will put less stress on energy grids to nearby communities.”

With Microsoft pouring billions into AI infrastructure, this step is part of a larger bet: that cooling technology will be just as important to the future of AI as the chips themselves.