When I saw the title, I thought a silicon chip in Xerox Alto has stopped functioning due to electromigration. Fortunately it was a macroscopic electromigration on its backplane and has an easy fix. I wonder if it could become a serious issue on computer history preservation in the future. Can a 7nm chip today survive an uptime of 30-50 years? Once it burns out, there would be no replacement... Or is this problem negligible compared to EEPROM losing critical firmware and data if you have proper core voltage and cooling?
Indeed, this is one the largest challenges in semiconductor fabrication today. The narrower our wires become, the more susceptible they are to tiny voids formed by electromigration.
The article seems to imply that the effect is highly dependent on the "amount" of electricity going through a connection. With that in mind, how much does this occur on the scale of a CPU where the connection is only a few tens of atoms across but the power is measured in only a couple volts?