Taming Heat at the Nanoscale to Satiate Energy-Hungry Artificial Intelligence
semiconductor-digest.com Sep 21, 2026

Taming Heat at the Nanoscale to Satiate Energy-Hungry Artificial Intelligence

AI-summarised brief · reviewed before publication

Korean University researchers have engineered a “double‑confined” phase‑change memory cell that layers antimony telluride with nickel ditelluride and molybdenum ditelluride. The design confines heat to specific regions, allowing two distinct data states to be written sequentially without disturbing the underlying material. This architecture reduces reset energy by up to 76 % to 115 pJ, speeds switching to 50 ns, and extends endurance beyond one million cycles. The thin layers can be deposited with existing sputtering equipment, making the technology scalable for data‑center chips.

💡 Why It Matters

  • · By slashing power and heat in memory cells, the new design directly tackles the energy bottleneck that limits AI workloads.
  • · It offers a manufacturable path to denser, cooler, and more reliable memory for tomorrow’s high‑performance computing.