Stop blindly buying storage? Morgan Stanley: Original manufacturers > module manufacturers, DRAM > NAND

AI is still extending the storage upcycle, but the industry's trading logic is shifting from "price hikes across the board" to "structural differentiation." Morgan Stanley's latest research report's core judgment is that storage is still in a favorable cycle, but it prefers original manufacturers over module manufacturers, and DRAM over NAND.

According to the Zhuifeng trading desk, Morgan Stanley updated its NAND supply-demand model in a global technology research report released on July 2. Their calculations show that AI-related NAND demand will grow 60% year-over-year in 2027, driving a global NAND market supply-demand gap of approximately 9% in 2027. Shortages persist, but the market is no longer uniformly tight.

More critical changes are occurring on both ends of demand. Server and AI-related demand remains strong, with long-term supply agreements (LTAs) providing downside price protection; however, the consumer side is already showing signs of price ceilings. Module manufacturers and distributors' inventories are rising, smartphone and PC customers are under pressure between sales volume and profit margins, and actual order cuts have begun after price hikes in 2Q26.

This means investors need to reassess the relative positions within the storage chain. DRAM is tactically superior to NAND due to better LTA terms, higher demand visibility, supply discipline constrained by EUV, and potential capacity squeeze from HBM4E; within NAND, original manufacturers are stronger than module manufacturers due to profit resilience and supply control.

AI demand remains the main theme, with NAND shortages extending into 2027

AI is becoming the core source of incremental NAND demand. AI NAND demand is expected to rise from 205 EB in 2025 to 400 EB in 2026, and further to 609 EB in 2027. AI's share of total NAND demand will also increase from 18% in 2025 to 32% in 2026, and then to 41% in 2027.

At the aggregate level, global NAND demand is expected to rise from 1111 EB in 2025 to 1250 EB in 2026, reaching 1484 EB in 2027. Supply during the same period is expected to be 1128 EB, 1058 EB, and 1347 EB respectively. This corresponds to supply-demand sufficiency rates of 2% in 2025, negative 15% in 2026, and negative 9% in 2027.

This data indicates that the industry shortage will not end quickly. Even if supply recovers to 27% year-over-year growth in 2027, AI servers, enterprise SSDs, QLC storage, and CSP inventory buffers will still be sufficient to absorb a large amount of new supply.

But this does not mean NAND pricing can rise indefinitely. AI demand and consumer demand have clearly diverged. The shortage is concentrated mainly in server and AI-related products, while consumer-grade products' capacity to absorb price increases is declining.

Price signal divergence: Strong servers, consumer peak

Channel surveys show that in 3Q26, NAND prices for TLC enterprise SSDs rose about 30% quarter-over-quarter, but consumer-grade NAND products only saw slight increases. On the DRAM side, server-grade product prices in 3Q26 rose about 20% quarter-over-quarter, while traditional DDR3 and DDR4 saw increases of 30% to 40% due to tighter supply and increased AI-related demand.

LTAs are changing the way prices fluctuate. Memory suppliers and major customers are still negotiating long-term agreements, which typically include price caps and floors. Floors help protect original manufacturers' profitability and valuations, while caps limit the potential for significant further price increases.

Customer attitudes are also diverging. Customers are more willing to pay higher prices for DRAM to secure supply, while NAND price hikes have encountered some resistance. This is related to rising profit margin pressure on consumer electronics customers.

Inventory levels also signal warnings. Supplier inventories remain at historically low levels, but module manufacturer inventories have significantly increased, and distributor inventories of consumer-grade memory are also high. Distributors believe demand has not fundamentally deteriorated, but aggressive price hikes over the past three quarters have raised costs, suppressed small and medium-sized buyer purchases, and reduced transaction volumes, increasing the pressure of holding inventory.

Why DRAM is preferred over NAND

The overall view of the storage cycle remains constructive, but the tactical allocation favors DRAM.

There are four reasons. First, DRAM's LTA terms are more favorable, with customers more willing to pay to ensure supply. Second, demand visibility is higher, with AI computing and related server demand still being the main support. Third, supply discipline is clearer, with constraints from technologies like EUV and capacity limitations restricting rapid expansion. Fourth, potential capacity squeeze from HBM4E may further tighten DRAM supply-demand.

However, storage stocks are still affected by the "rate of change." If year-over-year price increases plateau around 4Q26 and the supply-demand situation in 2028 remains unclear, short-term cyclical catalysts may weaken. But the earnings visibility brought by LTAs may still support valuation revaluation.

Why original manufacturers are preferred over module manufacturers

Within NAND, original manufacturers are preferred over module manufacturers, with the core reason being profit resilience and supply control.

In traditional cycles, module manufacturers typically hoard low-cost inventory at cycle bottoms and release inventory during upcycles, achieving higher profit elasticity. But this model also brings significant cyclicality; once low-cost inventory is depleted, profit margins peak, and stock prices often come under pressure.

This cycle has differences. If AI-driven shortages persist for three to five years through LTAs, module manufacturer profit margins may be more stable than in the past. Module manufacturers still face three constraints: low-cost inventory will be gradually depleted within the year, consumer-grade price increases will narrow in 2H26, and original manufacturers are allocating more supply to CSP customers, limiting module manufacturers' shipment growth in 2026 and 2027.

In addition, hyperscale customers tend to source directly from NAND original manufacturers or sign long-term supply agreements, which may limit module manufacturers' long-term addressable market in enterprise SSDs and AI storage. Most Asian SSD module manufacturers' enterprise SSD revenue contribution is still only about 10% to 20%, making it difficult to fully offset the weakening consumer SSD market.

Investors should focus on three types of risks

The first is a slowdown in AI capital expenditure. This is the biggest macro risk for storage stocks. As long as AI capital expenditure does not peak in the near term, storage profitability is still expected to continue beyond 2027.

The second is pressure from the consumer side. Smartphone and PC customers' capacity to absorb further price increases is declining, order cuts have already appeared, and inventories at module manufacturers and distributors are rising. If consumer-side volume continues to be weak, the logic of broad NAND price increases will be undermined.

The third is the supply reversal in 2028. If new wafer capacity is accelerated and supply discipline loosens, NAND may face oversupply risk. Conversely, if new AI inference SSD products enter mass production, they may consume about three times the capacity of ordinary SSDs, further tightening industry supply.

The conclusion is that the storage cycle is not over, but the "blind buy" phase is passing. AI demand continues to support high-end storage and server products, and LTAs improve earnings visibility; however, consumer-side prices are peaking, module manufacturer inventories are rising, and supply variables in 2028 are increasing. For investors, the key in the next phase is not to judge whether storage is still on an upcycle, but to distinguish who has stronger pricing power, more stable supply access, and clearer demand visibility.

Risk Disclosure and Disclaimer

        Market risk exists, investment requires caution. This article does not constitute personal investment advice, nor does it consider the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at your own risk.