TetraMem memristor survives 700 degrees Celsius as startup moves AI inference chips to production wafers

"Every probe humanity has sent to Venus has died. The Soviet Venera landers survived between 23 minutes and two hours on a surface where the temperature exceeds 460 degrees Celsius. Their electronics, designed to endure heat that would melt lead, still failed. The longest-lived mission in the history of Venus exploration lasted 127 minutes. Then the chips stopped working and the data stopped flowing."

"A team at the University of Southern California has built a memory chip that operates reliably at 700 degrees Celsius, hotter than molten lava, and more than 200 degrees beyond anything Venus could throw at it. The device, published in Science on 26 March 2026, held data for more than 50 hours at that temperature without refresh, survived more than one billion switching cycles, and ran on 1.5 volts with a switching speed measured in tens of nanoseconds. Seven hundred degrees was not the device's limit. It was the limit of the testing equipment."

"The device is a memristor, and the company commercialising it, TetraMem, has already built working AI chips that perform machine learning inference at speeds and efficiencies that conventional hardware cannot match. The extreme-temperature version was an accident. Its implications are not."

"A memristor is a nanoscale component that stores information and performs computation simultaneously. The device that Joshua Yang's team built at USC consists of three layers: tungsten on top, hafnium oxide ceramic in the middle, and a single-atom-thick sheet of graphene on the bottom. Tungsten has the highest melting point of any metal. Hafnium oxide is a standard insulator in semiconductor fabrication. Graphen"

Venus lander missions have repeatedly failed because electronics cannot survive extreme surface temperatures above 460°C. A USC team created a memristor that operates reliably at 700°C, far beyond expected Venus conditions. The device stored data for more than 50 hours at that temperature without refresh, survived over one billion switching cycles, and operated on 1.5 volts with switching speeds in the tens of nanoseconds. The memristor uses tungsten, hafnium oxide, and a single-atom-thick graphene layer. A company commercialising the technology, TetraMem, has already moved room-temperature AI inference chips to 300mm production wafers with SK hynix and support from the CHIPS Act.

Read at TNW | Artificial-Intelligence