Morgan Stanley Maintains Bullish SK Hynix, Samsung Outlook Despite Slowing Memory Price Gains

Woofun AI reports that Morgan Stanley released its Asian technology report titled "Memory – A Small Wrinkle" on August 6, maintaining a bullish stance on South Korea's leading memory manufacturers SK Hynix and Samsung Electronics. The investment bank left target prices unchanged, signaling confidence in the sector's fundamentals despite emerging signs of cyclical moderation. This strategic positioning underscores a belief that structural demand drivers outweigh short-term price volatility.

The valuation framework remains anchored by aggressive target prices, set at 2.6 million won for SK Hynix and 381,000 won for Samsung Electronics' ordinary shares. Based on the benchmark stock prices utilized in the analysis, these targets imply potential upside margins of approximately 74% and 65%, respectively. Such significant upside potential does not suggest that Morgan Stanley anticipates memory prices will continue to rise at their previous accelerated pace.

Instead, the report acknowledges that price increases are slowing, channel inventories are recovering, and new production capacity is gradually entering the market. Consequently, the memory industry is predicted to enter the later stage of its cycle by the fourth quarter of 2026. Despite this cautious outlook on price momentum, the firm remains optimistic, relying on three key factors: ongoing increases in AI-related capital spending, Long-Term Agreements that improve profitability visibility, and the fact that recent valuation declines in memory stocks have already reflected some concerns about a peak in the cycle.

Profit forecasts reflect this nuanced caution, with adjustments made to account for one-time events and sector-specific weaknesses. Morgan Stanley raised its forecast for SK Hynix' EPS in 2026 by 13%, but this increase was primarily driven by a one-time investment income of 63.27 trillion won recorded in the second quarter.

Meanwhile, the forecast for operating profit in 2026 was actually lowered by 7%. For Samsung Electronics, the EPS forecast for 2026 was reduced by 10%, primarily reflecting weakness in consumer businesses such as smartphones. The profit forecasts for both companies from 2027 to 2028 show smaller adjustments, indicating a stabilization of expectations. According to the revised forecasts, SK Hynix' EPS is expected to grow by about 25% year-on-year in 2027, while Samsung Electronics' is expected to grow by about 49%, roughly falling within the 25% to 50% range mentioned in the report.

The most direct driver of the previous round of gains in memory stocks was the sharp increase in product prices, fueled by AI server demand for HBM, server DRAM, and enterprise-grade SSDs. Limited capacity expansion failed to keep up with this surge, leading to significant price increases for DRAM and NAND. Citing data from TrendForce, Morgan Stanley noted that month-on-month increases in contract prices for overall DRAM and NAND reached record levels of 96% and 88% in the first quarter of 2026. Further increases of 61% and 58% were expected in the second quarter, before dropping to 16% and 13% in the third quarter, and further to 6% and 3% in the fourth quarter. This trajectory highlights a clear deceleration in price momentum as the market adjusts to new supply dynamics.

Woofun AI data shows that latest channel surveys cited in the report reveal that early transactions in DRAM contract prices in the third quarter saw a month-on-month increase of about 15%, slightly lower than the previously expected 20%. NAND contract prices rose by about 20% during the same period. PC DRAM contract prices are expected to rise by 15% to 20% in the third quarter, a significant decline from the 45% to 50% increase seen in the second quarter. Therefore, it is more accurate to say that prices are still rising, though at a slower pace, rather than that memory prices have already peaked. Morgan Stanley predicts that by the fourth quarter of 2026, the industry will gradually enter the later stage of its cycle, during which the operational leverage provided by prices will weaken, making it harder to achieve earnings that exceed expectations.

The slowdown in price increases is not necessarily entirely due to improved supply; structural shifts in demand play a critical role. The report points out that the slowdown in price increases for some consumer DRAM products is because buyers are approaching their cost tolerance limits. At the same time, demand from AI-related clients remains strong, and suppliers are shifting some of their consumer-grade capacity toward products such as enterprise-grade SSDs. This reallocation of resources underscores the growing importance of AI-driven demand in sustaining high-margin production, even as traditional consumer segments face price pressure. The interplay between these two demand streams creates a complex market environment where price dynamics vary significantly by product segment.

The first major factor supporting Morgan Stanley's bullish stance is that demand for AI data center hash rate remains unmet, as evidenced by statements from major cloud providers. Citing earnings reports from four major U.S. cloud providers, the report states that Alphabet and Microsoft's cloud computing demand still exceeds their internal available capacity. Amazon expects its capacity to remain insufficient to meet demand in 2026, with a significant portion of capacity already booked for 2027. Meta predicts that the supply of computing power in the industry will remain tight in the foreseeable future.

While this does not guarantee that all AI investments will yield expected returns, it at least indicates that large cloud providers have not significantly scaled back their infrastructure construction. As a result, Morgan Stanley's tracking model for cloud-related capital spending raised its forecast for year-on-year growth in 2027 from 14% a month earlier to 29%. This is crucial for memory manufacturers, as AI servers require not only GPUs but also HBM, server DRAM, and enterprise-grade SSDs. As long as cloud providers continue to expand their data centers, memory demand will not be solely determined by the cycles of traditional consumer electronics such as PCs and smartphones.

However, "structural AI demand" and "cyclical price adjustments" can coexist, and Morgan Stanley's core judgment is that AI demand may extend the earnings cycle for memory companies, but it will not completely eliminate memory prices and inventory cycles. The second supporting factor comes from Long-Term Agreements (LTA), which provide greater visibility into future demand.

Compared to traditional quarterly purchases, multi-year contracts allow suppliers to confirm part of their demand in advance and plan accordingly their production capacity and capital expenditures. SK Hynix' official announcement of its second-quarter results showed that the company has completed LTA negotiations with about 10 key customers and is continuing discussions with others.

Morgan Stanley further reported from the company's earnings call that these agreements typically last around five years, but the duration and pricing mechanisms vary depending on the customer and product; some agreements include deposits, and prices are adjusted according to market fluctuations. SK Hynix did not disclose the specific proportion of capacity or revenue covered by LTA, only stating that it will keep such agreements at an 'appropriate level' to provide downside protection while reserving capacity for new demand.

Therefore, it should not be assumed that these long-term agreements lock in most of future revenue.

Samsung Electronics' specific terms are based on Morgan Stanley's summary of its second-quarter earnings call, revealing a more aggressive approach to long-term contracting. Samsung plans to include 60% to 70% of its capacity under long-term agreements, having already reached agreements with five large global data center customers, with another five in the final negotiation stage. These agreements follow a rolling five-year structure, requiring customers to pay upfront payments and setting price floors for some mainstream products.

Potential customers listed in the report's table, such as AWS, Microsoft, Google, Meta, and Oracle, are clearly marked as media reports and should not be considered as officially confirmed trading partners by Samsung. The main purpose of LTA is to improve profitability visibility, support capacity investment, and reduce some price fluctuations, rather than completely locking in future profits. Different contracts vary in terms of the capacity they cover, duration, price adjustment methods, and breach protection mechanisms; if AI development slows down, product specifications change, or market prices fluctuate significantly, the protection provided by these agreements still has limits.

The market has already started pricing in a slowdown in memory company earnings growth, with valuations adjusting to reflect this new reality. Morgan Stanley notes that the P/E ratio of DRAM-related stocks over the next 12 months usually leads the EPS forecast by about two months, and recent valuation declines reflect investors' pricing in slower earnings growth. This also explains why memory companies may have strong current earnings, but their stock prices may not continue to rise in tandem.

Investors' focus is shifting from current profits in 2026 to sustainability from 2027 to 2028: whether AI-related capital spending can continue to grow, whether LTA can withstand downturns, and whether prices can be maintained after new capacity comes online. High profits themselves can also attract additional supply, posing a long-term risk to margins. Morgan Stanley points out that the current gross margin for DRAM is close to 90%, which is exceptionally high historically.

If high returns encourage leading manufacturers to accelerate capacity expansion or attract new suppliers, existing profit margins may revert to their long-term average. It is important to distinguish between these two figures: the near 90% refers to the DRAM gross margin discussed in the report, while SK Hynix' 76% in the second quarter represents the company's overall operating profit margin, and the two cannot be directly compared. SK Hynix officially reported that its revenue in the second quarter of 2026 was 79.

3187 trillion won, with operating profit of 60.5426 trillion won, resulting in an operating profit margin of 76%. Capacity expansion by Chinese manufacturers is also listed as a long-term risk. The report states that the new supply from CXMT and YMTC may reduce the scarcity of some products; CXMT plans to start supplying HBM in the Chinese market as early as 2027. This is Morgan Stanley's description of the companies' roadmaps and does not mean that Chinese HBM has already achieved mass production or can immediately replace high-end Korean products.

Additionally, the report predicts that major manufacturers will continue to bring new capacity online from 2027 to 2028. If AI demand continues to grow rapidly, this new supply may not cause shortages in the short term; but once cloud providers slow down their capital spending, increased supply could accelerate price declines. Therefore, the key assumption behind the target price of 2.6 million won is not that memory prices will keep rising indefinitely, but that AI demand can extend the earnings cycle and LTA can reduce volatility, with recent valuation declines already accounting for a significant amount of cyclical risk. What truly needs to be verified going forward is whether these long-term agreements can withstand a price downturn and whether memory companies can find new drivers for EPS growth after 2028.