The memory market is entering a new phase as AI moves from basic automation into advanced reasoning with large-scale deployment. Hyperscalers and data centre operators are pushing to lock down as much memory inventory as production capacity will allow. This is ultimately shifting supplier attention toward AI workloads and away from more traditional end markets including smartphones, PCs, vehicle, and industrial systems manufacturers.
Reports suggest memory availability is reshaping competitive dynamics, despite a recent slowdown in DRAM price growth. Pricing adjustments or declines don’t necessarily indicate falling demand or that the broader memory cycle has suddenly turned. Underlying demand remains high, and things like spot pricing can reflect short-term market dynamics rather than the long-term reality of contracted supply. Even if spot pricing softens in the near term, the overall market can still be constrained. In this case, it is because the amount of memory the market wants is outpacing what fabs can reliably produce and deliver.
A structural shift at play
In a February 2026 survey from the Global Electronics Association, 62% of manufacturers reported constrained memory availability or extended lead times. Likewise, 82% reported rising prices. Only a small 14% of surveyed manufacturers expected conditions to improve within six months. As hyperscalers race to obtain high bandwidth memory, capacity is being pulled away from conventional DRAM and NAND products, with HBM reigning supreme. This reprioritisation is tightening availability for everyone else.
Experts consider this dynamic to illustrate a structural shift, rather than a temporary market fluctuation. Previous years reflected the ups and downs associated with memory’s cyclical market. This is different, as suppliers are devoting their production capacity to highly specific products for AI. Building and qualifying new capacity is not immediate either, as the manufacturing ecosystem, including tools, materials, and labour must scale in parallel. Meeting the current level of demand would take years to bring additional facilities online, even if companies made aggressive investment decisions today. Some experts are warning of long-term contract commitments extending out as far as 2030.
Long-term supply contracts, and pricing impact
In this market, spot prices can move, but long-term agreements increasingly define who has access to supply when things are tight. When experts talk about contracts stretching into 2030, they are pointing to a world where OEMs and hyperscalers want certainty and where suppliers can demand stronger terms.
These memory contracts are non-cancellable and non-returnable, meaning everyone who is booked out through 2028 is booked out. If AI disappeared tomorrow, all those data centers and hyperscalers would still get those orders. The allocation is already spoken for, and the physical supply chain is already committed.
As an important nuance on why pricing power may stay elevated, the reallocation toward HBM is not easily reversible on a short timeline. HBM production requires dedicated wafer capacity, specialised packaging, and yield profiles that cannot easily flex back to commodity DRAM. At the same time, buyers in the AI supply chain are willing to pay premium prices for high power memory, which reduces the incentive for manufacturers to prioritise lower margin chips.
A second pressure point that can keep supply tight and pricing sticky is rising material and geopolitical risks, including concerns tied to energy and the flow of critical raw materials through the Strait of Hormuz. Specialty chemicals remain a potential continuity risk for fabs, with helium and bromine as inputs with no viable short-term substitutes. Similarly, labor unrest may prove to be another defining risk point. Worries around a prolonged union strike at Samsung, if an agreement is not reached before 21st May, could cause significant delays and, ultimately, raise prices once again.
Assessing OEM exposure
For OEMs, exposure starts with a straightforward reality that not all buyers enter this market with the same leverage. For example, compared to its competitors, Apple would fare slightly better because of strong contracts and strategic inventory reserves. Apple has scale, established supplier relationships, and the ability to commit early, and that tends to translate into priority access when supply is constrained.
Competitors have spent the last several months struggling to secure inventory while juggling with price adjustments. Many of their cost-effective price points, when compared to Apple, have effectively disappeared. Apple would also win in securing supplies faster than a smaller PC maker. Most smaller players do not have the capital leverage to lock in the same terms, and even when they do, they may face longer lead times or reduced flexibility. In tight markets, being late to procurement can mean paying more, waiting longer, or both.
AI-driven memory shortages are affecting smartphones, laptops, vehicles, and industrial systems. The OEMs most exposed are the ones that sit in the middle of the pack. They do not have Apple-level contracting power, meaning they cannot reliably command premium pricing that makes higher component costs easier to pass through. As tight supply persists and priority capacity continues to flow toward advanced memory modules, small-to-mid-sized OEMs are left competing for whatever supply remains.
About the author:
Kathryn Ackerman, Market Analyst, Sourceability
