Long-Term Storage Agreements Redefine Industry Pricing Power

The pricing power in the storage industry is undergoing a systemic realignment.

On August 5, SanDisk released its FY2026 annual earnings report, delivering strong results, yet its stock fell 8% after hours. The market’s disappointment was simple: guidance didn’t exceed expectations. But hidden in the earnings call was another number far more significant than the quarterly performance.

SanDisk disclosed that it has signed eight long-term supply agreements, with a minimum total revenue of $93.9 billion at floor prices. The weighted average term of these contracts is over four years, equating to approximately $20 billion annually—while SanDisk’s current annualized revenue is around $42 billion. This means that, even in the worst-case scenario, SanDisk has locked in nearly half of its revenue for more than four years at prices significantly below current market levels.

This isn't just about SanDisk.

On August 4, Goldman Sachs’ Giuni Lee team released a research report presenting a comparative table of long-term agreement (LTA) terms among four memory manufacturers: Samsung Electronics, SK Hynix, Micron, and SanDisk. The report’s key conclusion is that LTA terms are shifting in favor of suppliers across four dimensions: longer durations, broader coverage, more favorable pricing structures, and stronger enforceability. Each of these shifts points to the same trend: pricing power in the memory industry is transitioning from buyers to sellers.

(Translated from the Goldman Sachs table)

This change means that the definition of the "cycle bottom" in the storage industry has been rewritten, with the floor of this cycle raised above the previous ceiling.

In the past, the standard contract cycle in the storage industry was one year. Each year, parties would sit down to negotiate, adjusting prices and volumes based on supply and demand. Buyers held the upper hand—signing long-term contracts to lock in prices during favorable markets, and pushing for lower prices and reduced volumes during weak markets.

From Goldman Sachs’ comparative table, it can be seen that most vendors indicate contracts are primarily five years in duration, with some clients opting for three-year terms. This means that even the shortest contracts are now three times longer than previous industry standards.

Samsung clearly stated on its recent earnings call that its LTAs are based on a five-year term with an annual rolling renewal mechanism—renewing one year at a time, ensuring the contract never expires. Samsung’s co-CEO, Kim Yong-hyun, previously expressed this more directly: “Given the supply and demand uncertainties brought about by expanded AI investments, we are transitioning from traditional short-term agreements to multi-year contracts of three to five years.” This sentiment was echoed by SK Hynix’s CEO, Kwon Oh-jung, who noted that customer demand for LTAs is increasing.

Hynix has covered the top ten LTA customers and key clients, with most contracts having a five-year term; Micron has signed 16 strategic customer agreements, with most customer contracts at five years and automotive customers at three years; SanDisk’s eight LTA customers have a weighted average contract term of over four years, with a maximum of five years.

Four manufacturers are simultaneously extending contract durations, systematically reducing buyers' flexibility—once signed, procurement strategies cannot be adjusted for five years regardless of market changes.

Signing a long-term contract is one thing; covering how much capacity is another. If only 10% of the volume is locked in, even a long duration doesn’t matter much. But the numbers here are not that.

From 2023 to 2025, the industry standard for long-term contract volume ranged between 20% and 30%. By 2026, this figure saw a significant increase.

Samsung has signed agreements with the top five global data center customers and is in final negotiations with five additional major clients. Management expects that, after all contracts are finalized, multi-year order volumes will reach 60% to 70% of planned capacity. The commitment rate for advanced HBM capacity is even higher—over 90% has already been secured.

SK Hynix has completed negotiations on approximately ten long-term agreements, with long-term agreements accounting for about 50% to 60%. Micron has signed 16 strategic customer agreements, covering approximately 20% of DRAM shipments and one-third of NAND shipments, but management has clearly stated that the ultimate goal is for LTA revenue to exceed 50%.

SanDisk's data is more intuitive. Signed agreements cover over 50% of shipments for fiscal year 2027, and this figure will rise to about two-thirds for fiscal year 2028. Three months ago, the coverage for FY2028 was only one-third.

When a company has 60 to 70 percent of its capacity locked in long-term contracts, its pricing logic changes. The remaining 30 to 40 percent of capacity generates excess profits on the spot market, but its core business is anchored by contracts. Buyers lose the leverage they once had—“If you don’t lower your price, I’ll go elsewhere”—because everyone else is bound by the same contracts.

Buyers have certainly done the math. Signing a long-term contract means accepting unfavorable terms—but only if the alternative is worse: not signing a long-term contract and failing to secure enough storage chips in the AI computing arms race. When supply constraints are structural, getting overcharged is far less costly than getting no supply at all. This is the source of suppliers’ confidence in pushing their terms to the limit.

This is the most critical change among the four dimensions.

Past LTA agreements were primarily fixed-price contracts. A single price was agreed upon, covering one or two years, with both buyer and seller sharing the risk equally—profits went to the buyer if prices rose, and losses were borne by the seller if prices fell.

The structure of the new contract is completely different.

Samsung’s terms are the most aggressive. A Bank of America Merrill Lynch research report dated August 1 revealed that Samsung’s contract terms establish clear asymmetric pricing—quarterly price declines are capped at 5%, while price increases have no upper limit and can reach 10% to 20% or higher. Jaejune Kim, head of Samsung’s memory business, stated on the earnings call that the company employs different pricing models based on customer segments and product categories, and has set floor prices for commodity products.

During market downturns, Samsung's decline is capped at 5% per quarter. During market upturns, gains are uncapped. The buyer bears nearly all the downside risk, while the seller retains nearly all the upside potential.

Micron’s contract structure differs, but the direction is consistent. Its largest contract sets price floors and ceilings based on market prices in the second quarter of 2026. The key is the floor—management repeatedly emphasized on the earnings call that even at the floor price, gross margins remain significantly higher than the peak of any previous cycle in history. Micron’s historical peak gross margins in prior cycles were just over 60%.

SK Hynix has taken a more aggressive approach. According to TrendForce, citing Korean media reports, Hynix has eliminated the industry-standard price cap in its latest contracts. Even for long-term agreements, if market spot prices rise due to supply shortages, contract prices will fully adjust in line with actual market conditions. Goldman Sachs noted in its report that, compared to peers that have locked in price caps, Hynix has greater exposure to the price elasticity of standard DRAM—meaning its upside potential is significantly greater if prices exceed expectations.

SanDisk's pricing mechanism falls between fixed and floating pricing, employing a customized hybrid model. CFO Luis Visoso revealed on the earnings call that the agreement combines "fixed and floating pricing mechanisms," tailored to each customer's actual requirements.

Four players, four structures, but pointing to the same thing: price protection is one-sided and favors sellers.

Past LTA agreements had limited enforceability. Even after signing, buyers could reduce volumes, delay deliveries, or even default without facing significant consequences.

The biggest difference in this cycle is the advance payment mechanism.

Micron is expected to receive approximately $22 billion in cash deposits and related financial commitments. SanDisk has disclosed financial guarantees exceeding $11 billion and customer default protections of $16.5 billion. Samsung stated that the contract includes substantial advance payments, of which approximately one-quarter has already been received.

This is not a deposit. It is the cost for the buyer to secure inventory in advance. For a cloud provider to secure sufficient HBM and DRAM by 2028, it must first place billions of dollars on the manufacturer’s books. This money not only locks in supply but also binds the buyer—the cost of default is prohibitively high.

Goldman Sachs has referred to the advance payment mechanism as "the most notable feature distinguishing this LTA cycle from previous ones."

When viewed together, the changes across these four dimensions lead to one conclusion: the storage industry is using contracts to redefine the "cycle bottom."

Micron's management revealed that even if prices drop to the contract floor, the gross margin would still be significantly higher than the peak of any previous cycle. Micron's historical gross margin peaks were just over 60%.

SanDisk's $93.9 billion floor price tells the same story—although the profit margin corresponding to the floor price has not been explicitly disclosed, annualized minimum revenue of approximately $20 billion has already been locked in by contracts, while SanDisk's current gross margin is at its highest historical level. When viewed together, Micron and SanDisk suggest a directional conclusion: the floor of this cycle may well have been raised above the previous ceiling.

The costs vary. Micron’s contract sets clear price floors and ceilings—the floor is protected, and the ceiling is locked, capping upside potential. Samsung’s terms are asymmetric: downside is limited, but upside is unlimited. SK Hynitz took a different path: it eliminated the price ceiling, preserving the greatest upside elasticity for standard DRAM—Goldman Sachs specifically noted in its report that, compared to peers with partial caps, Hynix has more significant upside potential. The cost is a long-term contract coverage rate of only 50% to 60%, about 10 percentage points lower in revenue certainty than Samsung’s.

The inventory data in the Goldman Sachs report provides another perspective. As of the end of Q2 2026, Samsung and SK Hynix’s DRAM and NAND inventories stood at 2 to 4 weeks, below the normal range of approximately 4 to 5 weeks and far below the over 10 weeks seen prior to previous downturns. With no inventory buildup, manufacturers have no incentive to lower prices. Contract terms favor suppliers, reflecting genuine supply-demand tightness.

However, the contract locks in price, not capacity. Current total NAND capacity is approximately 2.01 million wafers per month, and with existing facilities, equipment, and technical upgrades, it is expected to reach around 2.15 million wafers by the end of this year. True capacity increases will depend on new fabs and cleanrooms, which have a construction timeline of 18 to 24 months. Industry estimates indicate that this additional capacity will be released mid-next year or later, accounting for approximately 17% to 19% of total capacity.

The time when production capacity is unleashed is when LTA terms are truly put to the test. When supply catches up, will buyers who signed five-year contracts and prepaid billions of dollars find themselves locked into outdated prices? The answer depends on the fine print of the contracts—and the fine print is entirely controlled by the suppliers.