Nvidia's New Product Accused of Memory Capacity 'Shrinkage', Domestic and Overseas Storage Stocks Plummet in Response 英伟达新品被指内存容量“缩水”海内外存储股应声重挫

The story the capital market had woven around the storage industry ultimately unraveled before an NVIDIA presentation. Storage companies' valuations surged as the market rushed to label them "core assets of the AI era," as if cyclical volatility had instantly become outdated history. Yet, a SemiAnalysis report on memory downgrades for NVIDIA's Vera Rubin platform acted as a splash of cold water, waking everyone from this valuation frenzy. Stock prices corrected swiftly, creating an awkward moment.

The absurdity here lies in how the entire market’s nerves were manipulated by a single report—or even a misinterpretation of it. Industry insiders rushed to clarify: the downgrades pertained to system memory on the CPU side, not the GPU’s HBM-based computing core. This was merely a routine balancing act by NVIDIA regarding supply chain, power consumption, and costs. But why did the market ignore these details? Because what it craves is never technical truth, but a grand narrative that can justify—or vent anxiety about—current sky-high valuations.

Is the "core asset of the AI era" label truly sustainable? The current boom in the storage industry owes less to its own technological barriers or business model excellence and more to being strapped to NVIDIA’s AI computing chariot. NVIDIA remains the rule-setter of the game; a mere adjustment in memory prioritization sends tremors across the entire supply chain. Storage vendors may seem flooded with orders, but where is their pricing power or strategic autonomy? Once the market realizes that the supposed "infrastructure" resilience hinges on another company’s decision-making for its own commercial interests, its fragility becomes starkly evident.

Redefining storage stocks from "cyclical" to "growth" or "infrastructure" assets reflects the capital market’s eternal hunger for low-interest, high-growth narratives. However, the DNA of cyclical industries does not mutate simply by being branded with "AI." Memory chips remain commodities heavily influenced by supply and demand—now with the added variable of "AI training/inference." Once this variable fluctuates due to shifts in technology roadmaps (like NVIDIA’s recent adjustment), inventory cycles, or efficiency innovations, cyclical behavior will only become more intense and unpredictable, not disappear.

Those hitting new market capitalizations might need a reality check. Current valuations are inflated with optimistic fantasies about AI, yet lack sober reflection on their own industry positioning. In the grand game of AI computing, storage is more akin to a critical pawn than the king. Real power still lies with platform players who define architectures and set standards. Rather than rushing to sell stories to the market, it’s wiser to focus on building irreplaceable technological moats amidst giants like NVIDIA—instead of enduring sleepless revaluations with every rumor of a platform shift.

Thus, this correction is less a negative signal and more a stress test. It revealed the market’s speculative exuberance and the industry narrative’s lack of substance. For storage to truly secure its place as an "AI core asset," downstream cyclical tailwinds alone are insufficient. It requires an irreversible technological or business model revolution of its own. Until then, any surge in valuation may simply be cyclical forces donning an AI costume for another round of performance.

资本市场给存储行业编的故事，终于在英伟达的一张PPT前露了怯。存储上市公司市值一路高歌，市场急不可耐地要给它们戴上“AI时代核心资产”的桂冠，仿佛周期性波动一夜之间就成了上个世纪的旧黄历。但SemiAnalysis那份关于Vera Rubin平台内存降配的研报，像一盆冷水，浇醒了这场估值狂欢的美梦。股价应声回调，场面一度尴尬。

这事儿荒谬的地方在于，整个市场的神经被一份报告、甚至是一个报告的误读所操控。业内赶紧出来澄清：降的是CPU侧的系统内存，跟GPU的HBM算力核心没关系，这是英伟达在供应链、功耗和成本上的常规平衡操作。但市场为什么不管这些？因为它需要的从来不是技术真相，而是一个能给当前超高估值提供支撑（或发泄焦虑）的宏大叙事。

“AI核心资产”这顶帽子，戴得稳吗？存储行业此刻的繁荣，与其说是自身技术壁垒或商业模式的胜利，不如说是被英伟达的算力战车绑在了后座。英伟达才是那个定义游戏规则的驾驶员，它调整一下内存配置的优先级，整个产业链就得跟着抖三抖。存储厂商看似订单拿到手软，但议价能力和战略自主性在哪里？当市场发现，所谓“基础设施”的坚固程度，竟然系于另一家公司基于自身商业考量的某个决策时，脆弱的真相便暴露无遗。

将存储股从“周期股”重新定义为“成长股”或“基础设施股”，背后是资本市场对低利率、高增长故事的永恒饥渴。但周期性行业的基因并不会因为沾上“AI”就突变。存储芯片的本质，依然是受供需关系剧烈影响的大宗商品，只不过它的需求侧多了一个“AI训练/推理”这个变量。一旦这个变量因为技术路线变更（比如英伟达这次调整）、库存周期或能效革新而波动，其周期性只会表现得更剧烈、更不可预测，而不是消失。

那些市值新高的公司，或许该清醒一点。当前的估值里，装了太多关于AI的乐观幻想，却太少关于自身产业地位的冷静思考。在AI算力这盘大棋里，存储更多是那个重要的“兵”，而非“帅”。真正的权力，仍然掌握在定义架构、制定标准的平台方手中。与其急于给资本市场讲故事，不如多想想，如何在英伟达等巨头的夹缝中，建立自己不可替代的技术护城河，而不是在每一次平台更迭的传闻中，经历一轮提心吊胆的价值重估。

所以，这次回调与其说是利空，不如说是一次压力测试。它测出了市场情绪的浮夸，也测出了产业叙事的底气不足。存储要真正坐稳“AI核心资产”的位置，光靠下游的景气度传递是不够的，它需要一场属于自己的、不可逆的技术或商业模式革命。在此之前，任何市值的飙升，都可能只是周期律换了一身AI的马甲，在重新登场表演罢了。