Long-term agreements (LTA) are revolutionizing storage valuation, and JPMorgan has significantly raised the target prices for Samsung Electronics, SK Hynix, and Kioxia under its P/E framework.

```

Something unprecedented in the past thirty years is happening in the memory chip industry: buyers are actively seeking to sign long-term contracts with sellers.

Behind this reversal of roles is a structural surge in memory demand driven by the AI computing power arms race. Cloud service providers are beginning to realize that DRAM and NAND are no longer bulk commodities readily available, but are now vital strongholds of AI infrastructure.

To secure critical capacity for the coming years, they have to use prepayments and long-term commitments for guarantees—this is the Long-Term Agreement (LTA).

This shift in business model is precisely the core argument of Morgan Stanley's semiconductor research analyst Jay Kwon’s latest report: As LTAs increase their share in memory manufacturers’ shipments, the decades-old P/B (price-to-book) valuation tool is outdated, and must switch to a P/E (price-to-earnings) framework.

Based on this logic, Morgan Stanley has raised the target price for Samsung Electronics to 480,000 won (applying 8x P/E based on expected EPS for 2026-2027), SK Hynix’s target price has jumped from 1.8 million won to 3 million won, and Kioxia’s target price has doubled from 38,000 yen to 80,000 yen, marking the highest in the market.

The real underlying logic for this wave of revaluation is an extremely difficult-to-reverse supply-demand gap. Estimates show that even under aggressive expansion assumptions, between 2026 and 2030, AI storage supply will still fall short of cloud providers’ demands—the gap is equivalent to the capacity of 450,000 wafers per month. Buyers have nowhere to retreat, which finally gives sellers genuine pricing power for the first time.

From "manufacturing and waiting for buyers" to "work after getting paid", the memory industry is moving toward the foundry "make-to-order" model. The earnings visibility brought by LTA is the surgical knife striping off the "high cycle" tag. The recent EPS upgrades are just appetizers; what truly determines the height of memory giants’ stock prices is the fundamental qualitative change brought by these long-term contracts.

A gap of 450,000 wafers/month flips negotiation leverage completely

Every past round of price hikes in the memory industry has proven fleeting—expansion inevitably pushes prices back down. But this time is different, with constraints coming from both sides.

On the supply side, memory giants emerging from the last bleeding downturn have developed "expansion phobia". Without certain orders underwriting, no one wants to invest heavy capital with uncertain marginal gains.

Kioxia’s capital expenditure as a proportion of revenue this year is only 5%, whereas the five-year average exceeded 20%. This restraint is no coincidence, but an industry-wide collective choice. Samsung and SK Hynix will also keep capital expenditure percentage in the mid-single digits for the next two years.

The demand side is accelerating in reverse. The requirements for server memory throughput from generative AI are rising exponentially, while agentic AI’s spread is triggering an explosion in enterprise SSD demand. Kioxia’s ASP in a single quarter rose by more than 100% quarter-on-quarter, and DRAM prices have tripled overall in the past year.

The result of supply and demand twin squeeze is that cloud providers lose all bargaining power at the negotiating table. Micron has already disclosed progress on LTAs; one US NAND manufacturer’s five LTA contracts cover over a third of its FY2027 bit demand. Industry expects Samsung and SK Hynix to soon announce the largest LTA deals in semiconductor history.

Valuation paradigm shift: From "cyclical stock" to "foundry-like", LTA remakes pricing power

For the past decade, capital markets have firmly anchored memory chip valuation in price-to-book (P/B). The reason is simple: products are highly homogenized, weak pricing power, enormous capital expenditure, and extreme cyclical performance swings with macroeconomics.

But in the AI era, this logic is being thoroughly overturned. In its report, Morgan Stanley throws out a forward-looking view: The memory industry is undergoing a “cyclical to secular” structural shift. The core catalyst is the comprehensive adoption of LTAs.

As AI servers’ demand for HBM (High Bandwidth Memory) and enterprise SSDs explodes, CSPs, for supply chain security, are proactively signing 3- to 5-year LTAs with memory manufacturers.

According to Morgan Stanley, several LTA cases have emerged in the industry, with Samsung and SK Hynix expected to disclose the largest LTA contracts in semiconductor history. Some LTAs not only fix prices (or set price floors) but also include prepayment mechanisms.

This "make-to-order" model makes memory giants increasingly resemble foundries like TSMC. Enhanced profit visibility and certainty are prompting Wall Street to shift memory stock valuation from traditional P/B to price-to-earnings (P/E).

Morgan Stanley bluntly states that LTA is the key driver for rerating the industry, and sets valuation benchmarks based on expected 2026-2027 earnings at 8x P/E (historical downturns were typically 6x).

Giant differentiation: Hynix leads, Samsung "wealth on fire", Kioxia breaks through

Despite overall industry improvement, positions among players have quietly changed.

Samsung: After catch-up rally, special dividends are the final prerequisite for completing valuation

Samsung’s stock price has soared 143% this year, but technological lag in HBM is a clear weakness. Its improvement in HBM validation is more about riding the sector tailwind, not really recapturing the technical moat.

The financial logic is more straightforward. Mixed ASP for DRAM and NAND will see YoY increases of 293% and 234% respectively by end of 2026, operating profit forecasts up 2% to 11%, but labor costs from strikes and tax pressures mean final EPS forecast is only slightly revised up 1% to 5%.

Bigger catalyst lies in cash return. From 2024 to 2026, Samsung’s funds pool for shareholder return approaches 160 trillion won, over 115 trillion won post-regular dividends available for special returns. Market inclines toward direct payout over buybacks, and the special dividend expectation for Q3 2026 is fully priced in. If management doesn’t deliver on schedule, the valuation rebuild will lack its last piece.

SK Hynix: Up over 200% this year, but 31% EPS CAGR shows it’s not at the top yet

SK Hynix is the fiercest player in this rally, target price jumping from 1.8 million to 3 million won. The key support is sharply improved earnings quality—EPS CAGR for 2026-2028 expected at 31%.

ASP negotiations in HBM are in fast lane, LTA proportions not only stabilize baseline prices but give Hynix absolute advantage in pricing against cloud providers. Even with new capacity plan rollout, overall DRAM bit supply growth in 2027-2028 remains capped under 20%, sustaining the high-margin shortage regime.

Cash return path is also clear: From 2025 to 2027, free cash flow up to 240 trillion won will all be distributed to investors (14 trillion in 2025, 66 trillion in 2026, 160 trillion in 2027), and shareholder total return yields in 2026 and 2027 will reach 5% and 12% respectively.

Kioxia: After 326% surge, a clear shareholder return roadmap decides if valuation ceiling breaks

Kioxia’s upside logic is most direct: Q4 ASP soared over 100% QoQ, operating profit quadrupled, capital expenditure ratio fell to 5%. High-profit SSD and memory products’ share in total sales rose from roughly 50% to 60%, and marginal contribution rate will stay at 85%–90% through FY2027.

LTA is in place—Kioxia sealed mid-term contracts covering 2027-2028 in Q4. With extremely restrained expansion, 17 trillion yen in free cash flow will be intensively released over the next three years.

Currently, a 7x P/E calculation yields a target price of 80,000 yen, still a 10% discount against the 15-year industry average (8x).

Only one trigger removes this discount: A clear, powerful shareholder return plan at investor day. Bain Capital-led consortium’s suspended stake reduction expectation is the biggest external factor suppressing the premium. Making money is just the first step; how profits are allocated to shareholders determines whether the 326% rally miracle continues.

Effective boundaries of the new model: Can the "cycle" be really left behind?

This deduction logic is not without failure boundaries. The biggest risk is embedded at the start of the chain: If LTA negotiations become opaque, or cloud providers’ AI capital expenditure declines significantly, the P/E valuation framework built on "earnings certainty" will instantly lose its support. Specifically:

First is the ultimate question of AI monetization.

Morgan Stanley points out what investors still worry about: "There may be disconnect between the AI capex race and the semiconductor ecosystem’s profitability." If cloud providers’ AI models fail to deliver matching commercial returns, or if breakthroughs occur that reduce memory demand for AI servers, the currently inflated demand expectations could shrink.

Second, LTA's spirit of contract is still to be tested.

History has precedents; during DRAM shortages in 2017, firms signed forward purchase agreements, but as demand slowed and inventories rose, these deals were easily overturned in following cycles.

This time, whether memory firms can enhance contract enforceability by introducing third-party financial derivatives etc. determines if "escape from cyclicality" logic can stand up.

Finally, structural changes in global capacity map can't be ignored.

Apart from traditional giants’ restrained expansion at new BiCS 8/10 nodes, global diversification of supply chains is accelerating. As emerging markets (like China's domestic NAND capacity) gradually rise and release, the structure of global memory supply is becoming richer.

In the foreseeable future, this regional capacity boost will challenge the existing global supply-demand balance system, and add more variables to industry competition.

Morgan Stanley believes, AI indeed delivers the memory industry an entirely new “armor”, and LTA adoption makes the moat unprecedentedly wide. But although cycles may be lengthened or smoothed, it's hard for them to be completely eliminated.

In the fervent capital feast, whoever leads technologically, and maintains restraint in expansion, will truly enjoy the long-term compounding leap from “cyclical stock” to “growth stock”.

Risk Warning and DisclaimerThe market has risks; investment should be cautious. This article does not constitute individual investment advice, nor does it consider the specific investment targets, financial status, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article suit their particular situation. Investing accordingly is at your own risk. ```