Memory price hike expectations have been significantly raised, but Morgan Stanley warns: short-term momentum may have peaked

TL;DR

• Morgan Stanley raised its price forecasts for multiple memory categories in 3Q26 while cautioning that short-term momentum in memory stocks may weaken.
• According to the report, PC DRAM price hike expectations have risen to 15-20%, and the breadth of DRAM earnings revisions is nearing 89%.
• Samsung and SK Hynix remain supported by AI demand, but earnings guidance, LTA terms, and capital expenditure commentary will influence the sustainability of the rally.

In a research report dated July 7, Morgan Stanley significantly raised its memory price forecasts for the third quarter, but simultaneously warned that memory stocks could face near-term correction pressure.

This is not a turn bearish on the memory cycle. The report maintains an "attractive" view on the Korean tech sector, continues to favor Samsung Electronics and SK Hynix, and projects over 35-40% earnings growth for related companies by 2027 in its models. The real caution lies in the fact that memory prices, earnings expectations, and investor positioning have all reached elevated levels, making it uncertain whether stock prices can continue rising at the same pace as in recent months.

The most direct figures are the price forecasts. The report raised its 3Q26 PC DRAM blended ASP qoq growth expectation from the previous 3-8% to 15-20%, server DRAM to 13-18%, GDDR6 and GDDR7 to 15-20%, and enterprise SSDs to 18-23%.

Public pricing agencies are also increasingly aligned in their direction. In a July 3 article, TrendForce stated that the 3Q26 DRAM market remains extremely tight, with contract prices expected to rise 13%-18% qoq, and NAND Flash contract prices expected to rise 10%-15%. However, TrendForce also noted that while server DRAM remains in short supply, long-term supply agreements could moderate the pace of increases.

Prices are still rising, but the difficulty of trading memory stocks is increasing. According to Morgan Stanley's report, the breadth of DRAM earnings revisions has recently approached 89%, nearing historical highs. The memory market rally over the past two years, driven by AI capital expenditure, HBM, and server demand, has already priced in much of the positive news.

The Most Aggressive Price Hikes Are in Q3, and So Are the Concerns

This round of upgrades is broad-based. Beyond PC DRAM, price expectations for 3Q26 have been significantly raised for server DRAM, graphics DRAM, conventional DRAM, and enterprise SSDs. Among these, the 15-20% increases for PC DRAM and graphics DRAM are the most noticeable price signals for the market. The 18-23% increase for enterprise SSDs indicates that memory price increases are not limited to AI server-related categories.

Herein lies the problem. The faster price expectations are raised, the easier it is for stocks to enter a phase of "good news being priced in."

Memory stocks have experienced corrections over the past two years. Morgan Stanley's data shows that since the start of the generative AI wave in November 2022, DRAM-related stocks have experienced three notable pullbacks, including profit-taking, specific event shocks, and multi-week corrections. Each pullback did not break the long-term trend driven by AI capital expenditure, but each served as a reminder that deep, phase-based declines can occur even within a strong cycle.

The three major pullbacks in DRAM stocks were approximately -25%, -25%, and -35%, respectively, but the overall trend still rose from the 2022 lows to 2026 highs.

Spot prices also support the bullish thesis. The price tracking in the report shows that DRAM spot prices have surged since early 2025, and NAND spot prices have recently rebounded significantly from their lows. While contract prices lag, they are trending in the same upward direction. In other words, the short-term correction warning is not due to deteriorating prices, but because prices and expectations have risen too quickly.

DDR5 16Gb spot price rose to around $47, while MLC 64Gb NAND spot price recently climbed to $31.10.

Earnings Revisions Near Peak, Memory Stocks Need to Digest Expectations

For stocks, what matters more than the price itself is how much room remains for upward earnings revisions.

The breadth of DRAM earnings revisions has recently reached approximately 89%, hovering near historical highs. This indicator essentially means a growing number of analysts are raising their earnings forecasts. When the majority of analysts have already revised earnings upward, it becomes more difficult for subsequent results to exceed expectations.

The breadth of DRAM earnings revisions rose to approximately 89% after 2025, approaching historical peak levels.

This is one reason why short-term momentum may weaken. It's not that memory stocks lack fundamental support; rather, rising prices, upward earnings revisions, crowded positioning, and AI chain sentiment have all converged. Once earnings guidance is not strong enough, capital expenditure commentary is not aggressive enough, or large cloud companies' stock prices come under pressure, the memory sector is more prone to amplified volatility.

Morgan Stanley still prefers DRAM and conventional memory, where actual capital flows are clearer and supply bottlenecks more pronounced, deeming them more attractive than NAND, and is relatively least bullish on memory module manufacturers. This ranking suggests the market is not simply betting that "all storage prices will rise" but is instead assessing whether price increases can truly translate into profits.

3Q26 PC DRAM price hike expectations raised from 3-8% to 15-20%, enterprise SSD to 18-23%, with simultaneous upward revisions for various DRAM categories.

Samsung Has Provided Guidance, SK Hynix Awaits End of Month

The earnings reports from South Korea's two memory giants will serve as a key test for the bullish price thesis.

Samsung Electronics released its 2Q26 earnings guidance on July 7. The company expects sales of approximately KRW 171 trillion and operating profit of approximately KRW 89.4 trillion, with an operating profit range of KRW 89.3 trillion to KRW 89.5 trillion. For Samsung, the market will focus not just on quarterly profit, but also on the recovery momentum of its memory business, progress in AI-related products, and whether the ongoing rise in traditional memory prices can continue to improve profitability.

SK Hynix's earnings date is still pending. According to the public market calendar, the company is expected to release its next earnings report on July 29. Given SK Hynix's more prominent position in HBM and AI server memory, the market will be more sensitive to its Q3 commodity memory prices, long-term supply agreement (LTA) terms, and capital expenditure commentary.

If management confirms that Q3 commodity memory remains strong, LTA commitments increase, and capital expenditure is only moderately raised, any short-term correction might appear more like a healthy adjustment. However, if guidance is not robust enough, or if capital expenditure is interpreted as signaling overly rapid supply increases, the market may reassess how long the memory price rally can last.

LTAs themselves are not risk-free signals. Historically, long-term agreements have not necessarily led to stock price increases; some agreements have been renegotiated or become constraints forcing customers to take delivery when demand changes. The market will not only look at "how much was signed" but also consider pricing, duration, customer quality, and execution flexibility.

AI Demand Persists, Market Begins Questioning Overcapacity

The long-term bullish thesis still stems from AI, particularly AI agents capable of executing tasks, using tools, and engaging in continuous interaction. In its report, Morgan Stanley projects that related companies' earnings could still grow by over 35-40% by 2027, which is why it does not equate the short-term correction risk with the cycle's end.

However, the debate around AI demand is evolving. Previously, the market predominantly bet on the continuous expansion of model training and inference scale, driving demand for computing power and memory. Now, some investors are beginning to worry that after Q3, cloud vendors might place greater emphasis on token savings, inference efficiency, open-source low-cost models, and the pressure of chip inflation on margins.

Another, more sensitive issue is whether the largest AI spenders genuinely have sellable computing power, potentially hinting that earlier buildouts might be phase-based overcapacity. This notion has not formed a definitive conclusion, but it is enough to make the market more cautious ahead of large cloud companies' earnings reports.

The divergence in this memory cycle is not about whether AI demand will disappear immediately, but about whether price increases, upward earnings revisions, and customer capital expenditure can continue to reinforce each other in the same direction. If Q2 AI supply chain earnings remain favorable but Q3 guidance begins to weaken, memory stocks could first undergo a round of valuation and positioning adjustments.

The memory cycle remains within the context of AI capital expenditure, but short-term trading focus has shifted from "how much prices have risen" to "how long the market will continue to believe in these increases." The upcoming earnings calls from Samsung and SK Hynix, cloud vendors' capital expenditure commentary, and the execution of LTAs will determine whether this correction is merely a breather in a bull market or the beginning of a real slowdown in the rally's pace.