U.S. memory stocks have recently faced profit-taking pressure, with market concerns over slowing growth, rising capital expenditures, and customers reducing procurement specifications potentially impacting the sector's outlook. However, Morgan Stanley (MS-US) stated in its latest report that these concerns are not new and had already surfaced a month before the stock pullback. The firm emphasized that there has been no fundamental deterioration in the industry, and this downturn instead presents a rare opportunity to add positions.
Morgan Stanley highlighted that the core driver of this cycle is that memory is increasingly becoming a primary bottleneck in AI infrastructure development, and this structural constraint is expected to persist for several years.
Against this backdrop, the firm believes the risk-reward profile of memory stocks is rapidly catching up to previously favored names like NVIDIA (NVDA-US) and Broadcom (AVGO-US), and the current sell-off has created a strong entry point.
Joseph Moore’s analyst team at Morgan Stanley conducted intensive visits to multiple data center procurement clients last week, reporting that there are no signs of easing in memory supply tightness.
The team further estimates that data center memory prices in Q3 will rise by at least 25% compared to Q2—exceeding both prior market expectations and the firm’s own earlier forecasts.
Regarding external concerns about 'slowing price momentum,' Morgan Stanley acknowledged this is true but stressed it is an inevitable outcome of the industry’s expanding scale.
According to the Semiconductor Industry Association (SIA), DRAM prices surged approximately 70% in Q1 and rose over 40% in Q2. With the industry’s quarterly revenue exploding from around $46 billion a year ago to over $200 billion, maintaining such high growth rates is no longer realistic. Continued explosive price increases could actually backfire on end-demand.
Nevertheless, Morgan Stanley argues that such warnings were already well understood by the market before the stock peak and do not constitute unexpected negative news.
On the much-discussed long-term agreement (LTA) issue, Morgan Stanley interprets it as further confirmation of supply-demand tightness, not evidence that prices are being locked in.
The report notes that Micron Technology (MU-US) suggested during its earnings call that 'Q2 pricing may be the ceiling for new contracts,' but this is more likely a conservative statement.
Industry sources indicate that these contracts are mostly pre-agreed deals delayed only by legal processes. For genuinely new contracts still under negotiation, there will be no higher price ceiling.
AI Is the True Driver of This Cycle
Morgan Stanley emphasizes that this memory cycle is fundamentally different from past cycles. The demand engine has almost entirely shifted to data centers, and noise from consumer electronics, PCs, and smartphones are merely 'false signals' that should not be misinterpreted as signs of an economic turning point.
The report points out that cloud providers are even willing to pay premiums above Q2 expected prices for spot goods with six-week lead times, just to accelerate procurement. Morgan Stanley believes these customers are not stockpiling but are instead rushing to secure supply, further confirming that supply tightness stems from real capacity bottlenecks, not inventory-driven dynamics.
Data shows AI compute spending growth has already exceeded 50%, far outpacing the 3% to 5% annual growth in PC and smartphone markets. As AI’s share of total demand continues to expand, this gap will only widen.
Looking ahead to next year, the launch of the Rubin Ultra platform will double HBM memory usage, and the increasing complexity of HBM4 manufacturing will consume significant production capacity.
Additionally, demand for low-power DDR5 at the rack level and enterprise storage remains strong. In contrast, Morgan Stanley observes that NAND players are still maintaining abnormally conservative capital expenditures. Even with a slight rebound next year, meaningful supply expansion is unlikely.
Less About How High, More About How Long
Morgan Stanley argues that the market’s focus should shift from 'how high this cycle’s peak profitability can go' to 'how long high profitability can last.' The latter, it says, is actually more critical for supporting valuations.
The report notes that long-term agreements and customer-side engineering optimizations have flattened the cycle’s volatility but simultaneously extended the duration of the upcycle. If the industry can maintain a gradually rising profit trajectory over the next few years, it could provide more sustained and powerful valuation support than a single super-boom.
The report also mentions that NVIDIA has indeed significantly reduced LPDDR5 usage within racks and is restructuring compute, working memory, and storage architectures to alleviate memory bottleneck pressures.
However, Morgan Stanley believes the logic behind these 'down-specing' moves precisely reflects the industry’s widespread expectation that 'memory shortages will last for years'—a sign of confidence, not a negative signal. As supply gradually loosens, memory usage is bound to rebound accordingly.
FACT BOX
- Source: PR Times
- Category: News
- Products / services: DRAM / LPDDR5