Goldman Sachs also supports "storage PE valuation," raising the target prices for the three giants SK Hynix, Samsung, and Kioxia.

Goldman Sachs also supports "storage PE valuation," raising the target prices for the three giants SK Hynix, Samsung, and Kioxia.

The global storage chip industry is undergoing a historic paradigm shift: it is transforming from a traditional “boom and bust” cyclical commodity into a highly predictable strategic resource for AI infrastructure. The most crucial impact is the complete overturn of the valuation framework—an evolution from the traditional price-to-book (P/B) ratio to the price-to-earnings (P/E) ratio.

On June 1, according to the Wind Trading Desk, Goldman Sachs pointed out in its latest deep research report on the global semiconductor storage industry that the current upcycle of storage chips is unlike past ones. The sustained AI-driven demand, constrained supply growth, and structural changes in long-term supply agreements (LTAs) are pushing the storage industry from a highly cyclical commodity track to a predictable profit AI infrastructure track.

The report notes that four major disruptive changes are taking place in industry fundamentals and valuation logic:

First, the supply-demand gap is comprehensively revised upwards, with supply shortages in the DRAM, NAND, and HBM markets in 2027 expected to surpass those in 2026, and the shortages to continue into 2028;Second, the valuation framework undergoes a historic switch, with the industry benchmark formally shifting from price-to-book (P/B) to price-to-earnings (P/E), driving large upward adjustments in target prices for the "big three" (SK Hynix with an implied 53% upside, Samsung with about 60%);Third, HBM pricing logic is re-evaluated, with HBM's average price in 2027 expected to catch up and rise 44% compared to regular DRAM, and its global market size (TAM) revised up by 54% to $116 billion;Finally, long-term operating profit forecasts for the big three are fully revised upward, with high profit margins set to persist throughout the forecast period.

Coincidentally, before Goldman, Wall Street investment banks Morgan Stanley and JPMorgan pointed to the same conclusion in their latest reports: storage giants like Samsung and SK Hynix are at a historic turning point in valuation paradigms.

Morgan Stanley and JPMorgan believe that the widespread implementation of long-term supply agreements (LTAs) is likely to prompt the market to reprice these companies from "highly cyclical products" to "tech infrastructure with stable cash flows." Currently, the storage giants’ forward P/E is only about 7.3x, with a valuation discount of 50%–80% relative to TSMC.

Three Structural Anomalies: Why Will This Cycle Remain High for the Long Term?

The current cycle has completely deviated from the historic trajectory driven by cloud data centers in 2017-2018. Goldman Sachs believes the fundamental logic is being reshaped by three structural forces:

Demand Side: AI Servers Take Absolute Dominance

The cyclical drag from consumer electronics has been thoroughly marginalized. Data shows that servers will account for about 50% of DRAM and 40% of NAND demand in the industry in 2025; by 2028, these proportions will further rise to 61% and 43%. The global server memory market in 2025 (about $449 billion) is 7.4 times that of 2017.

With large language models (LLMs) evolving into enterprise AI agents (Agentic AI), by 2030, token consumption is expected to be more than 24 times current processing capacity, making memory bandwidth and capacity the core bottleneck restricting AI development.

Supply Side: Intensifying “Swallowing” Effect of HBM Capacity

Traditional storage expansion is being physically constrained. HBM production requires 3–4 times more wafer capacity than regular DRAM. As HBM iterates toward HBM4 and HBM4E, the wafer consumption ratio per unit product continues to climb.

Between 2026 and 2030, out of approximately 1.39–1.54 million monthly available cleanroom wafers from the big three, about 30% will be forcibly locked for HBM production. This will cause traditional DRAM supply growth’s annual compound rate to shrink sharply from 19% in 2017-2018 to 15%.

Business Model: LTAs Reshape Profit Predictability

Memory manufacturers and hyperscale cloud companies are systematically suppressing cycle volatility through LTAs. Clear financial evidence has emerged: Sandisk’s Q3 2026 financials reveal its new business model agreement includes $42 billion in deferred revenue obligations (RPO) and $400 million in prepayments, with breach penalties attached.

History in the silicon wafer sector shows that widespread adoption of LTAs can grant oligopolistic industries extremely high profit stability, forming a core foundation for the storage sector’s higher valuation multiples.

Penetrating Supply-Demand Gap Data: Facing More Severe Shortages in 2027

Data indicates that not only has the supply-demand gap for the three major product categories in 2027 not narrowed, it has worsened compared to 2026:

DRAM: The supply-demand gaps in 2026/2027/2028 plunge to -5.0%, -5.9%, and -3.9%, respectively. Driven by strong server DRAM demand, traditional DRAM’s average price in 2026 is projected to surge by 326% year-on-year, with operating profit margins stabilizing at historic extremes of about 80%.NAND: The gaps are -4.4%, -4.6%, and -3.0% in succession. Enterprise SSD (eSSD) demand will spike by 66% and 31% in 2026 and 2027, driving NAND operating profit margins to remain at high levels in the mid-60% range.HBM: The shortage is most critical, with gaps at -5.4%, -6.0%, and -4.3%. Demand from ASICs surges (with 172% growth in 2026), and Goldman predicts HBM market sizes of $56 billion, $116 billion, and $168 billion in 2026, 2027, and 2028, respectively.

Abandoning PB, Embracing PE: Leapfrog Upward Revisions for "Big Three" Target Prices

With a qualitative leap in profit visibility, Goldman Sachs has officially anchored storage stock pricing on P/E multiples (using 9x as a baseline):

SK Hynix (Buy): Target price jumps to 3.3–3.5 million KRW. Stress tests show that even in extremely adverse scenarios with prices falling 30% for two consecutive years, profit margins can stay healthy at 40%, definitively disproving the old “cycle top equals loss” logic.

Samsung Electronics (Buy): Target price revised up to 480,000 KRW. Operating profit in 2026 is projected to grow over eightfold year-on-year, with ROE hitting a historic high of 52%. Its HBM revenue in 2027 will soar to about $44 billion.

However, it’s noteworthy that rising storage prices are biting downstream, with Samsung’s smartphone unit’s operating profit margin expected to collapse from 11% to a historic low of 2%.

Kioxia (Upgraded to Buy): With expectations for “longer highs” in NAND, using 7.8x P/E for FY3/28E profits, the 12-month target price is set at 93,000 JPY.

Wall Street Consensus: How Morgan Stanley and JPMorgan View the Valuation Framework Shift

This paradigm shift from P/B to P/E is not an isolated view—Morgan Stanley and JPMorgan’s recent analyses strongly resonate.

According to a previous article from Wallstreetcn, Morgan Stanley clearly states that memory has become the absolute bottleneck for AI infrastructure. LTAs are turning traditional cyclical businesses into long-tailed cash flows with rigid guarantees and high profits.

If the market continues to price these companies as ordinary cyclical goods, there will be severe valuation dislocations. Quantitative modeling shows that under neutral scenarios (HBM 100% LTA coverage, regular memory 70% coverage with 10x P/E), Samsung and SK Hynix’s implied overall P/E should reach 8.5–8.6x; if LTA coverage rises to 80%, implied P/E will break past 10.5x.

JPMorgan’s logic goes straight to the essence of commercial bargaining: buyers’ fear of supply disruption and sellers’ fear of demand default together give rise to legally binding long-term agreements.

This institution likewise issued aggressive bullish calls: Samsung’s target price revised up to 480,000 KRW (corresponding to 8x P/E), SK Hynix up to 3 million KRW, and Kioxia’s target price doubled to 80,000 JPY.

It’s worth noting that all three top Wall Street institutions coincidentally benchmarked TSMC: After binding Apple to long-term agreements in 2014, TSMC successfully switched its valuation framework to P/E and has maintained a 10–30x range long-term.

Currently, the storage giants’ forward P/E of just about 7.3x, and the 50%-80% valuation discount relative to TSMC, stand before a historic narrowing opportunity.

However, Wall Street also retains a final harsh bottom line: the literal meaning of contracts is not enough to fully immunize against cycles. At the end of the cycle in 2017, forward agreements became worthless within months after demand collapsed.

This time, only one iron evidence can truly support the new valuation framework: the balance sheet must have genuine cash prepayments and legally locked deferred revenue obligations. Without real cash inflows as a shield, any grand narrative about transcending cycles will remain a mirage.

 

~~~~~~~~~~~~~~~~~~~~~~~~

The brilliant content above is from Wind Trading Desk.

For more detailed interpretation, including real-time analysis and frontline research, please join the [Wind Trading Desk · Annual Membership]

Risk Warning and DisclaimerThe market is risky – invest cautiously. This article does not constitute personal investment advice and does not take into account individual users’ specific investment objectives, financial conditions, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular situation. You bear your own liability if you invest accordingly.