A prolonged DRAM shortage may be on the horizon for everyday PC buyers, and it could last longer than many expected. An internal market outlook attributed to SK Hynix, one of the world’s largest memory manufacturers, suggests that tight supply conditions for mainstream “commodity” DRAM could persist through 2028. If that projection holds, the PC market may continue dealing with higher memory prices and limited availability for years, making affordable system builds and upgrades harder to come by.
The assessment indicates that most of the industry’s production momentum is shifting away from conventional PC-focused DRAM and toward specialized memory used in data centers. In particular, high-bandwidth memory (HBM) and SOCAMM modules are expected to see stronger growth, while commodity DRAM supply expansion remains constrained. The reason is straightforward: major memory makers have been prioritizing the booming AI server market, where demand is surging and profit margins tend to be better than in the consumer space.
Another factor adding pressure is inventory. Supplier stock levels are described as having fallen to historically low ranges, which can amplify allocation challenges across the supply chain. When inventories are thin, even a modest demand increase can trigger outsized pricing swings, and it becomes harder for PC manufacturers and component buyers to secure steady volumes at predictable costs.
Rather than rapidly ramping new capacity, manufacturers are reportedly taking a cautious approach to expansion—aiming to protect profitability instead of oversupplying the market. At the same time, server DRAM demand continues to accelerate and is expected to grow even more sharply next year. Longer term, the share of DRAM going to servers is projected to climb dramatically, rising from an estimated 38% in 2025 to around 53% by 2030. That kind of shift would naturally leave less commodity DRAM output available for traditional consumer devices unless total capacity rises significantly.
The AI-driven data center buildout is a big part of what’s powering this trend. As cloud providers expand AI training infrastructure, demand for server memory can scale quickly—enough to fuel what some are calling a potential DRAM super-cycle. There are also indications that key production capacity has already been heavily booked out well in advance, with some reports suggesting crucial DRAM production slots for 2026 are effectively sold through. If true, that would reduce flexibility for redirecting supply back toward the mainstream PC market in the near term.
Meanwhile, the “AI PC” wave could add another layer of demand on the consumer side. AI-capable PCs are expected to take a much larger slice of the overall PC market, with projections pointing to roughly 55% share by 2026. That’s notable even if overall PC shipments remain relatively flat, because AI-oriented systems often ship with higher memory configurations—raising average DRAM demand per PC.
The same internal outlook also points to potential spillover effects in NAND flash. NAND supply growth may trail demand in the consumer market as well, largely because server-side demand is growing and tends to be more lucrative for suppliers. If both DRAM and NAND remain tight, consumers could face elevated costs not just for RAM upgrades, but also for SSDs and storage-heavy devices.
Taken together, the analysis paints an uncomfortable picture for value-focused PC buyers: the memory market imbalance that many hoped would ease by 2027 may not meaningfully unwind until 2028 or later. If suppliers continue prioritizing AI and server customers—and keep capacity growth conservative—PC memory pricing and availability could remain a persistent challenge across multiple upgrade cycles.
