Публикация

I’d lean toward “the memory cycle is still in its expansion phase, but the easy-money part may be getting crowded.” There is real evidence behind the narrative, not just stock-market hype: AI servers are absorbing much more NAND: enterprise SSDs accounted for 48% of NAND bits shipped in Q2 2026, up from 26% a year earlier. Kioxia and SanDisk have begun production of 10th-generation 3D NAND, while Kioxia has specifically developed high-density QLC products for AI infrastructure. Kioxia says its high-capacity QLC SSDs are already being deployed in AI-oriented servers, supporting the idea that AI is creating a genuine storage-demand cycle. Memory-cost inflation is now spilling into consumer electronics, suggesting the shortage isn't confined to hyperscaler data centers. 🧠 My read Fundamentally: probably still early-to-mid cycle. Stock-market narrative: much further along. That's an important distinction. The physical demand cycle can continue strengthening even while memory stocks become expensive because investors have already anticipated a large portion of the improvement. The biggest thing I'd watch next is supply response. If manufacturers rapidly add capacity and inventory starts rebuilding, the narrative can transition from: AI demand → shortage → pricing power → earnings upgrades to: capacity expansion → supply catches up → prices peak → earnings expectations reset. So I wouldn't call the memory super cycle “finished.” But I also wouldn't assume that a strong narrative means every memory stock still has unlimited upside. The real confirmation will be sustained pricing strength + tight supply + rising AI/storage demand + earnings upgrades.