Google re-tested cold fusion for four years. Here's what it found

What the researchers found

In 1989, two chemists announced that palladium metal soaked in heavy water could produce fusion power on a benchtop. Other labs couldn't reproduce it, and "cold fusion" became a byword for science gone wrong. Thirty years later, Google quietly funded a group of researchers — starting in 2015 — to re-examine the claims with modern tools, on the theory that the question had been abandoned rather than answered.

The result, published openly in Nature by seven lead authors across the University of British Columbia, MIT, the University of Maryland, and Lawrence Berkeley National Laboratory: no evidence of cold fusion. In the authors' own words, their investigations "have yet to yield any evidence of such an effect."

How they found it

The teams attacked the three main versions of the claim. They tried to pack deuterium into palladium electrodes — using water-based, polymer, and ceramic chemistries — to reach the extreme loading levels where the 1989 effect supposedly lives. They built purpose-designed calorimeters to hunt for unexplained heat in metals saturated with hydrogen at high temperatures. And they ran plasma-discharge experiments on nickel and other metals to look for nuclear signatures. Across all three lines of attack, nothing nuclear showed up.

How sure can we be?

This is about as credible as a negative result gets: multiple independent university labs, modern instrumentation, and publication in a top peer-reviewed journal — by researchers who would have become famous overnight had they found the opposite. The honest limit is the one the authors state themselves: some of the extreme material conditions where believers say the effect hides proved genuinely difficult to reach and measure. A search that finds nothing can close off the territory it covered; it cannot certify that all territory has been covered.

What this doesn't say

It doesn't prove cold fusion is impossible. The authors deliberately frame the remaining extreme-loading conditions as an open, underexplored parameter space — a reason for continued careful measurement, not a verdict of impossibility.

It doesn't rehabilitate the 1989 claims either. Three decades on, no lab has reproduced the original effect with unambiguous nuclear evidence. Extraordinary claims still lack the ordinary proof: reproducibility.

And "no fusion" wasn't "no value." The program produced new measurement tools and new insight into hydrogen-saturated metals — and the discipline it demanded seeded what came next: the same UBC lab later built a benchtop reactor that measurably enhanced real, beam-driven fusion using electrochemistry. We covered that result in our brief on the Thunderbird experiment. That is how science is supposed to metabolize a wrong claim: not by forgetting it, but by testing it until something true falls out.