The memory market is experiencing one of those moments when the industry shifts from treating chips as interchangeable screws to recognizing them as a strategic resource. Amid soaring prices, the three major manufacturers — Samsung Electronics, SK Hynix, and Micron — are altering how they sign their DRAM and NAND supply contracts, with a change that until recently seemed unthinkable: shorter agreements, more frequent renegotiations, and a new “post-settlement” pricing model that adjusts the final cost based on market conditions.
This transformation is significant because it affects the core of the business. Traditionally, these contracts operated under relatively stable logic: a reference price was agreed upon at signing, and although the market moved, subsequent revisions were limited within narrow negotiation margins. It was a kind of commercial “gentlemen’s pact” that provided visibility for both sides: the buyer secured volume, and the supplier guaranteed continuity.
But that stability is cracking. According to information published by ETNews and industry sources, contracts with a “post-settlement” clause have begun to emerge—a mechanism where the final price isn’t completely fixed at the outset but is recalculated to reflect the market price even after the product has been delivered.
What does “post-settlement” mean: paying more… after the chip has been received
To understand in plain language: in the classic model, if a company signed an annual agreement to receive DRAM at a set price, that price remained relatively stable with minor adjustments. An example often cited in the industry goes like this: a contract at 100 won per unit could be renegotiated quarterly, typically within a ±10% range, meaning 110 won or 90 won for the next quarter. The goal was to absorb some volatility without breaking the commercial relationship.
The new approach breaks that ceiling. With post-settlement, if during the contract period the market jumps sharply, the buyer might pay an extra to “catch up” with the actual market price. The example described by ETNews states a contract signed at 100 won could end up costing an additional payment if the market price doubled: delivering as agreed, but settling the difference afterward.
For suppliers, the incentive is clear: in a rising cycle, sticking to prices fixed months earlier could mean leaving money on the table just when capacity is tight and demand is high. For buyers, meanwhile, accepting this arrangement is a lesser evil if the primary goal isn’t the price but securing supply.
Why now: AI has turned memory into a bottleneck
The context is the rapid expansion of Artificial Intelligence infrastructure. Data centers are growing at such a pace that they stress the supply chain and push manufacturers to prioritize higher-margin, higher-demand products. In this environment, memory stops being a “controlled” cost line and becomes a factor capable of blocking full deployments.
In fact, analysts already foresee a rally with alarming figures: market projections have indicated steep price increases into the first half of 2026, with upward revisions in DRAM and NAND as competition for inventory intensifies. When the market moves this way, the traditional annual contract becomes a straitjacket for manufacturers… and a risky gamble for buyers.
The result is a shift in bargaining power. Available information suggests that the main recipients of these new contracts are large North American tech firms, precisely those expanding their computing capacity fastest and prioritizing delivery predictability above all else. Industry sources quote that for these clients, “securing memory” now matters more than the contractual format itself, even if it entails higher costs later.
From long-term agreements to shorter ones — from a year to a quarter… and even to a month
The second major change is equally significant: contract durations are shortening.
On paper, customers would prefer the opposite. In an AI expansion environment, it makes sense to lock in 2-year or longer contracts to ensure continuity, plan investments, and avoid surprises. However, suppliers are tightening their stance: accepting long contracts during shortages poses two clear risks.
- Opportunity risk: committing to a client at a fixed price may prevent capturing better conditions from other buyers if the market continues rising.
- Volatility risk: in a market where demand shifts quarterly, committing capacity for years limits the ability to react.
According to industry reports, this tension is already manifesting in concrete cases: data center operators requesting 2-year agreements being rejected, ultimately securing shorter supply commitments from other providers, often with post-settlement clauses. The message is twofold: shorter terms and less “locked-in” prices.
What does this mean for the market: more uncertainty… and more discipline
This shift brings immediate consequences. For buyers, especially those building AI infrastructure, memory costs become harder to forecast. If prices can be settled afterward, procurement teams no longer negotiate just “price and volume,” but also how the reference market is measured, how often adjustments are made, and what happens if the cycle turns downward.
For manufacturers, the move also carries risks. Post-settlement acts as a safety net during price rises but could become problematic if prices fall: the same mechanism that today protects margins might strain relationships if buyers feel they are paying “too much” for an unfavorable adjustment. Still, industry consensus among analysts suggests that the likelihood of sharp declines is lower than the current upward pressure.
Lastly, for the entire tech ecosystem, these contracts are a sign of the times: when infrastructure leads, commercial rules are rewritten. And memory — DRAM, NAND, and everything fueling AI — is making it clear that, by 2026, buying it will no longer be the same as before.
Frequently Asked Questions
What is “post-settlement pricing” in DRAM and NAND contracts?
It’s a contract model where the final price can be adjusted after delivery to reflect the market rate, instead of being completely fixed upfront.
Why do major clients accept paying a post-hoc adjustment?
Because in a shortage scenario, securing supply takes priority. For AI infrastructure, running out of memory can delay deployments and be more costly than a price adjustment.
Do shorter contracts mean memory will be more expensive?
Not necessarily always, but they tend to favor suppliers in up-market conditions: renegotiating more often allows manufacturers to capture market increases sooner and avoid being “locked in” at outdated prices.
When might this situation normalize?
Industry sources suggest this favorable trend for suppliers could last until at least the second half of the year, as price upward momentum is expected to moderate.
via: Jukan