Post

闪迪、西部数据双双杀跌：股价跌了，但存储超级周期的证据全坐实了闪迪跌7%，西部数据跌11%。很多人看见股价大跌，直接判定存储周期结束。但把两份财报拆开看，再结合马斯克、高盛、三星代工满产的线索—— 产业层面看多存储的论据，反而进一步被证实了。闪迪SNDK财报核心看多证据： Q4营收89.7亿美元，同比暴涨372%，非GAAP EPS 39.25美元，大幅超过市场预期，毛利率冲到84.6%。手握8份多年长协，保底总金额939亿美元，部分大客户签约之后还主动追加采购订单。 AI推理带来NAND需求是真实刚性，不是短期炒作。管理层明确NAND市场2026超3000亿，2027接近5000亿，数据中心占比持续抬升。推出140亿美元大额股票回购，现金流极其充沛，供给端没有无序疯狂扩产，资本开支保持克制。唯一的问题： 2027Q1营收指引区间103-108亿，上限刚好贴住市场最乐观预期，没有再超出去，直接引发卖事实抛压。西部数据WDC财报核心看多证据： Q4营收37.47亿，同比+44%，non-GAAP毛利率54.4%，EPS3.56美元，同样高于一致预期。云端业务占总收入89%，大容量24TB以上HDD出货持续走高。客户需求规划能见度已经延伸到2029-2031年，EB级冷存储需求每年增速25%以上，HAMR、ePMR大容量硬盘持续放量，云厂商冷归档的需求实实在在。 HDD属于AI冷数据底座，和闪迪NAND热存储形成完整搭配。一个管热推理。一个管海量冷数据归档。共同组成AI存储的完整链条。股价杀跌原因：指引达标，但没有给出类似希捷那种极其激进的远期订单表态，市场情绪不满意。把几条主线合并印证：马斯克在SpaceX电话会：存储供给每年+20%，AI需求每年200%以上，供需剪刀差巨大，价格理应上行。闪迪、西部数据财报，就是这句话的商业兑现。长协锁价、毛利率飙升，就是供需失衡的结果。高盛研报：市场八大担忧大多过度悲观，HBM价格2027有望大幅上行，长协、低库存、企业级需求三道供给护栏。两家财报里大量多年长协，正好印证护栏有效。三星4nm晶圆代工产能全部满载，AI芯片订单排到明年。 AI服务器持续放量，就会持续拉动HBM、NAND、HDD全链条存储消耗。现在市场的矛盾点已经看得非常清楚：产业基本面很强。但二级市场的预期已经被拉到天花板。现在不是“业绩差就跌”。而是： “仅仅达标一致预期就会跌，必须击穿最乐观幻想才会大涨。” AMD、闪迪、西部数据全部在演绎这套定价逻辑。股价大跌是情绪、预期差导致。不等于AI存储的产业逻辑证伪。长协订单、毛利率、客户追加采购、远期需求能见度，全部写在财报和电话会文字里。我坚持存储行情没有结束。这里就是低位分批买入的位置。一切情绪扰动，最终都会回归。

Goldman Sachs Breaks Down 8 Major Concerns: Storage Hasn't Peaked, It's the Start of a Misjudged Super Cycle SK Hynix dropped 35%, Samsung dropped 23%. Many people are already scared, thinking the AI storage rally is over. At this moment, Goldman Sachs directly released an in-depth research report, refuting all eight major market concerns about storage: The current valuation has fully priced in the expectation of a "cycle peak," and has even fallen much further. The core logic is broken down into three points: First, HBM prices haven't peaked; they are set to double. Goldman Sachs decisively states: By 2027, HBM prices will rise 87%-100%, 24% higher than market consensus. Don't think the current rise is enough. Demand for AI servers hasn't stopped. Each AI server requires several times the amount of HBM compared to traditional servers, and HBM capacity expansion can't keep up. Long-term contracts heavily favor suppliers. Prepayments, locked volume and prices. This is not the treatment of a cyclical stock; this is the pricing power of AI infrastructure. Second, there are three supply-side safeguards; prices simply won't fall. Long-term contract coverage is 50%-70%. Major clients prepay to lock capacity, so the "price crash" seen in the past won't happen. Inventory is only 2-4 weeks, far below the historical normal of 4-5 weeks, let alone the 10+ weeks before previous downturns. Enterprise SSD demand is expected to grow 66%, 31%, and 22% in 2026-2028; NAND demand hasn't reversed. Those saying "overcapacity expansion" haven't seen clearly— Most of the current capacity expansion is occupied by HBM. There is very little new capacity for traditional DRAM and NAND. Manufacturers are focused on producing HBM, ignoring cheaper products, causing spot prices to rise. But if by 2027, ChangXin also expands capacity, the global shortage might ease. At that time, HBM will return to a scarce state. Third, the market and fundamentals are telling two different stories. The market is trading on "cycle peak," but fundamentals are trading on "sustained AI demand." Goldman Sachs directly says: Past storage cycles were "up 1-2 years, down 2-3 years." But this time is different. AI-driven demand is structural, not short-term fluctuation. Look at Samsung and Hynix's forward guidance: both have over 5x PE, ADR premiums maintain over 30%. This is not a cyclical stock valuation. This is a growth stock valuation shift. Don't be misled by short-term stock price fluctuations. This Goldman Sachs report is not a "defense" for the market. It lays out the real supply and demand: the storage super cycle is not over. It has been over-interpreted by the market. No need to overcomplicate. Just keep an eye on the next round of HBM long-term contract quotes from Samsung, Hynix, and Micron. Price is the only judge. $SKHY $SAMSUNG

Floating PnL-68.57%·Average sub-order price150.53·Mark price145.37