Samsung and SK hynix are making a major shift in how they sell DRAM, and the move is raising eyebrows across the memory market. Instead of relying on the traditional one-year supply agreements many big tech buyers have used for years, reports out of South Korea say the two memory giants are now moving almost entirely to three- to five-year long-term supply contracts.
On the surface, long-term agreements can sound like a stable, sensible strategy. Locking in multi-year deals can give suppliers clearer visibility into demand, help buyers secure supply, and make it easier for manufacturers to plan capacity expansions without overbuilding. In a market as cyclical as DRAM, that kind of predictability can be valuable.
But the timing is what’s fueling speculation. When companies aggressively push long-term DRAM contracts at today’s pricing levels, it can also suggest something else: that they may not expect much additional upside in DRAM prices from here. The logic is straightforward. If suppliers truly believed DRAM pricing was about to surge significantly again, they would have strong incentives to keep more sales flexible so they could benefit from higher spot and short-term prices later. Choosing instead to lock customers into multi-year terms could be interpreted as a signal that the market is approaching, or has already reached, a near-term peak in DRAM pricing.
That concern comes at a moment when DRAM prices have already moved sharply higher. Samsung reportedly raised DRAM prices by about 30 percent quarter-over-quarter for Q2 2026, following a dramatic year-over-year increase in Q1 2026. The next key test will be what happens in Q3. If Samsung does not implement another meaningful price hike, it will strengthen the argument that DRAM’s strongest pricing momentum may be fading and that the push for long-term contracts is a way to secure elevated pricing while it’s still available.
NAND flash, however, may be on a different trajectory. Market expectations now indicate NAND pricing could peak much later, with one major forecast pointing to the third quarter of 2027 as the high point. Two important dynamics are helping explain why NAND could have more room to run even if DRAM is cooling.
First, Google’s TurboQuant has been cited as a factor that helped moderate spot DRAM pricing recently because it compresses DRAM-based key-value cache. The important detail is that this effect primarily impacts DRAM demand, not NAND. In other words, even if certain AI and data center optimizations reduce the amount of DRAM needed for specific workloads, they don’t automatically reduce demand for NAND storage.
Second, supply conditions in NAND could tighten if Chinese manufacturers shift capacity toward DRAM. Large players such as YMTC are reportedly allocating more incremental production toward DRAM rather than NAND, which could constrain NAND supply and support higher pricing for longer.
All of this matters because the stakes are enormous for the world’s top memory makers, especially Samsung. Some analysts are projecting staggering operating profit figures for Samsung over the next two years, driven in no small part by memory pricing and shipment strength. If Samsung is indeed on track for that kind of profitability, small changes in DRAM pricing strategy, contract structure, and Q3 pricing decisions become critical signals for where the memory cycle is heading next.
For investors, industry watchers, and anyone tracking semiconductor trends, the takeaway is clear: the transition to long-term DRAM supply contracts isn’t just a contract formatting change. It may be a sign that the DRAM market is trying to lock in the best part of the cycle while it lasts. The next major clue will come from Samsung’s Q3 DRAM pricing approach and whether the industry can sustain momentum after a period of rapid increases.
