New Atlas ran the University of Illinois Urbana-Champaign copper cold plate work again on May 20, this time with the projected PUE attached to the headline. The piece puts the demonstrated reduction at 32 percent better cooling performance and 68 percent less pumping energy versus conventional cold plates, and projects a facility PUE of roughly 1.011 against the theoretical floor of 1.0. The framing data point is that 2025 global data center electricity consumption sat near 485 TWh, with cooling accounting for around 30 percent of it, which works out to more annual energy than Sweden draws.
Same paper. Same researchers, Behnood Bazmi and Nenad Miljkovic, published in Cell Reports Physical Science. Same topology-optimized channels at 30 to 50 micrometer resolution made with electrochemical additive manufacturing. We covered the result when the paper landed in early May and again when OilPrice headlined the 97 percent figure.
A lab result that hits mainstream press three times in five weeks, with a science publication, an energy-markets publication, and a general-tech publication each picking it up independently, is a result the procurement audience is going to ask their cooling vendors about. The vendor answer of choice this morning is some version of polite caution, because additive manufacturing throughput in pure copper at the required resolution is not yet a commodity supply chain. That answer is fine for now. It will not be fine for long.
The reason this result keeps getting republished is the size of the energy slice it claims to address. Cooling at 30 percent of 485 TWh is roughly 145 TWh, which is on the order of a mid-sized European country's total annual consumption. If a production cold plate can credibly compress even a fraction of that, the savings are a budget item large enough to fund a multi-year procurement program, not a research line. That scales the direct-to-chip cooling market projections in a direction the current forecasts do not capture.
We said it in May and we are still waiting. The single most informative event in the next twelve months for this story is a major cold-plate manufacturer announcing a license, joint development, or acquisition of the UIUC ECAM process. That move converts a research narrative into a procurement reality. Until it lands, the cycle of mainstream coverage will keep producing the same headline with a different percentage in front of it.