· 个股论点
由于人工智能(AI)，内存需求是结构性的。但价格上涨可能并非如此。 当摩根士丹利(Morgan Stanley)预测： ~1329亿美元（2026年）+ ~1670亿美元（2027年）= 未来两年SK海力士(SK Hynix)的营业利润约为3000亿美元，而SK海力士的市值约为4000亿美元。 他们将坐拥大量现金。我甚至没有引用麦格理(Macquarie)预测5000亿美元营业利润的牛市情景，因为那看起来太荒谬了。 但在基准情景下，如果DRAM/NAND价格下跌，创纪录的营业利润回归到早期水平。 由于AI需求使内存需求具有结构性，它看起来仍然被低估。 AI从根本上改变了像GPU或内存这样被标记为“大宗商品”的运作方式。 这不再是2022年智能手机/消费电子需求下降导致价格崩盘，你可以简单地引用低市盈率(P/E)做空像以前那样的大宗商品周期。
Memory demand is structural due to AI. Price hikes might not be. When there's Morgan Stanley projections of: ~$132.9B (2026) + ~$167.0B (2027) = ~300B operating profit from SK Hynix in the next two years and SK Hynix's marketcap is ~$400B. They will be sitting on piles of cash. I'm not even citing bull-case projections where Macquarie projects $500B operating income because it looks absurd. But in the base case event that record operating income goes back to earlier levels if dram/nand prices decrease. It still looks undervalued as AI demand makes memory structural. AI fundamentally changes how labelled "commodities" work like GPUs or memory. This is not 2022 where smartphone/consumer electronic decreases crash prices and you can just cite low/pe = short like commodity cycles before.