Goldman Sachs: HBF Could Break AI ‘Memory Wall’ With Up to $150B Market

The interesting signal here is not the headline number but the implied shift in where value accrues inside the AI stack: if flash can absorb part of the memory burden, the bottleneck moves away from DRAM and HBM, which is precisely why Goldman's rating split falls where it does. That distinction matters most for memory makers whose roadmaps are concentrated in one camp rather than spanning both. The open question is whether GPU-level adoption of HBF actually materializes at the low end of that range or the high end, since the market sizing is entirely contingent on it.