Jevon's Paradox explained: when something gets more efficient to use, we often use way more of it, not less. Coal, then oil, now compute and memory chips.
WHO: every AI lab and hyperscaler chasing cheaper, faster compute. WHY it matters: efficient AI models don't shrink memory demand, they unleash it. HBM alone is now a $54.6B market, up 58% year over year. Cheaper compute means more compute, and every bit of it needs memory.
https://en.wikipedia.org/wiki/Jevons_paradox