AI Systems Need Far More Memory Than Expected, Creating Shortage Problems

AI systems are reshaping how much computer memory we need. Running large language models (like ChatGPT) requires 80GB to 1TB of memory per system—this is 50 to 100 times more than typical cloud applications needed before. Big tech companies are competing fiercely for limited supplies of this high-capacity memory, creating delays that affect both smaller companies trying to deploy AI and device makers. Unlike older cloud workloads that spread across many regular computer types, AI needs concentrate on specific specialized hardware, forcing semiconductor makers to rethink how they supply the market.