Several prominent AI companies are propelling the mobile phone industry toward creating devices that can support large language models boasting 100 billion parameters, with the expectation of achieving this feat by 2028. Yet, this ambitious vision is not without its hurdles: firstly, operating such colossal models demands a substantial amount of memory. Secondly, AI data centers are vying for a share of the global memory production capacity, resulting in a constrained supply and escalating prices for consumer-grade memory products.
