Google is reusing older DDR4 memory recovered from decommissioned servers in newer AI equipment as a worsening supply crunch hits memory chips, driven by the spread of AI data centres. Google said it remains hard to secure enough chips and called faster capacity expansion by manufacturers the most sustainable solution. Meta and others are also extending older memory use via CXL, though performance limits remain. Researchers and companies forecast further price rises and prolonged shortages.

KT said it held a special session at the Korean Institute of Electromagnetic Engineering and Science summer conference on Aug. 20 to discuss directions for developing AIDC networks and infrastructure with industry, academia and research institutions. The session in Pyeongchang included participants from ETRI and AI infrastructure companies. KT said future AI infrastructure competitiveness will depend on efficiency, and highlighted next-generation networking, CXL-based AIDC architecture and modular data centres.