Morgan Stanley: The AI-driven storage "Supercycle" has arrived.

Author: Jukan

Compiled by: Deep Tide TechFlow

Morgan Stanley pointed out that a new round of memory “super cycle” driven by artificial intelligence (AI) has arrived, and its strength and logic are completely different from any previous cycle.

Morgan Stanley stated that unlike in the past, this cycle is led by AI data centers and cloud service providers, whose customers are less sensitive to price, while inference workloads have become the main driver of demand for general-purpose memory. At the same time, the latest channel checks show that contract prices for server DRAM in the fourth quarter have surged nearly 70%, with NAND contract prices rising by 20% to 30%, giving suppliers unprecedented pricing power.

The agency maintains an “Overweight” rating on SK Hynix and Samsung Electronics, expecting that the rise in memory prices will drive stock prices to new highs, and that the profits of memory manufacturers will significantly exceed expectations.

Morgan Stanley pointed out that the core driving factors of this cycle have fundamentally changed in quality. Buyers are no longer traditional customers sensitive to price, but rather AI data centers and cloud service giants participating in the arms race to build computing infrastructure.

For these companies, ensuring the supply of memory is a strategic “necessity,” with price sensitivity dropping to historic lows. At the same time, the production of HBM (High Bandwidth Memory) is structurally eroding the capacity of traditional DRAM. Morgan Stanley emphasized in the report:

“The uniqueness lies in the fact that the current demand for memory has evolved into a competition dominated by AI data centers (compute-intensive platforms) and cloud service providers, who are not as price-sensitive as traditional customers… The exponential growth in inference demand has laid a solid foundation for this, which is also the reason that makes this cycle fundamentally different from any previous cycles.”

According to Morgan Stanley's latest channel checks, the price outlook for DRAM has shifted dramatically bullish in just two weeks. The contract prices for server RDIMMs in Q4 surged by about 70%, far exceeding the previous expectation of 30%. The spot price for DDR5 (16Gb) skyrocketed from $7.50 in September to $20.90 currently, an increase of 336%. The quoting level for DDR4 also rose by 50%. Most contract deals are expected to be finalized by the end of this month, but customer acceptance seems inevitable—due to concerns over further price increases and supply shortages.

NAND has become a key component of AI computing infrastructure and video storage. In response to capacity constraints, the price of 3D NAND wafers (TLC and QLC) is expected to increase by 65% to 70% month-on-month. Nearline storage specifications are transitioning from 128TB to 256TB QLC SSDs. According to TrendForce's forecast, by 2026, demand for bit-based enterprise SSDs in servers is expected to grow by approximately 50% year-on-year. Samsung limited bit production in the first half of 2025 due to the transition from V6 176-layer to V8 321-layer NAND, and capacity will only gradually increase in the second half of the year, resulting in a restricted growth rate of bit shipments this year to 10%.

The market is often influenced by “peak fear,” believing that once stock prices hit a new high, a reversal will soon follow. However, Morgan Stanley emphasizes that in this AI-driven market, the ultimate deciding factor is earnings growth, not historical valuations.

Currently, the price of server DRAM is $1/Gb, while it peaked at $1.25/Gb during the cloud supercycle in the first quarter of 2018. Considering the scale of AI infrastructure investment and the dynamic demand from hyperscale customers, the price peak in this cycle is very likely to exceed historical highs. Memory cycles typically last 4 to 6 quarters, and profits are gradually being realized. However, the key lies in the comparison with market expectations—there is significantly higher enthusiasm in the market for general memory prices. Valuation cannot predict future returns; it reflects the supply-demand relationship rather than historical precedents.

Strong long-term drivers based on AI have pushed memory prices into the “uncharted territory,” with profit prospects far exceeding general market expectations. This means that stock prices still have significant upside potential.

“As capital expenditures related to AI continue to expand, the proportion of memory in total expenditures may continue to rise—this will drive the price-to-book ratio (P/B) far beyond historical peaks. We believe this is a story of valuation expansion layered on top of a cyclical profit recovery…”

We believe that analysts' data revisions are always lagging behind - for SK Hynix and Samsung, our earnings forecasts for 2026 and 2027 are respectively 31%-48% and 38%-51% higher than the market consensus.

Overall, the driving factors of this round of memory “super cycle” are more enduring, with price increases exceeding historical records and a more optimistic profit outlook. Coupled with strong cyclical performance, this creates a rare investment opportunity for memory manufacturers with pricing power.