Memory has quietly become one of the most important cost pressures in the tech supply chain — and this week, that pressure acquired a longer time horizon.
US chipmaker Micron Technology Inc (NASDAQ:MU) warned that tight global supply of dynamic random access memory (DRAM), the core RAM used in PCs, smartphones and servers, is unlikely to ease in any meaningful way before 2028. The company said that even with new factories under construction, supply is unlikely to catch up with demand any time soon.
For device makers and buyers already grappling with higher prices, it helps explain why memory costs continue to show up in places that once felt insulated.
A supply problem years in the making
Supply chain pressure began building in late 2024, when DRAM prices started rising as inventories tightened and manufacturers pulled back on lower-margin production.
By last year, those early signals had hardened into a clear supply squeeze. Wafer capacity was increasingly locked into long-term contracts, inventories fell, and spot prices moved sharply higher across several DRAM categories. By the second half of the year, the industry was no longer talking about a short-term imbalance, but about a market that had fundamentally tightened.
Micron’s latest comments extend that timeline, suggesting the forces that created the shortage are still very much in place.
Why new factories won’t fix this quickly
At face value, the solution sounds simple: build more fabs. In reality, expanding memory supply takes years.
A modern memory fab can take several years to move from construction to high-volume, fully qualified output. Tool installation, yield optimisation and customer certification all stretch timelines well beyond the initial build. Micron’s view is that while new capacity will arrive, it won’t do so fast enough — or at sufficient scale — to meaningfully relieve pressure this decade.
More importantly, not all new capacity is destined for consumer markets.
AI has changed the memory equation
The biggest shift sits on the demand side. AI infrastructure has become one of the most memory-intensive workloads the industry has ever seen.
Data centres and accelerators require large volumes of advanced DRAM, including DDR5 and high-bandwidth memory. These products consume more silicon, rely on more advanced manufacturing processes and generate significantly higher margins than traditional commodity modules.
That has reshaped production priorities. Alongside Micron, memory suppliers such as Samsung Electronics and SK Hynix have tilted output toward AI-linked products, redirecting wafer capacity that once fed mainstream PCs and laptops.
The result is a market where total memory production can rise, yet availability for consumer devices remains constrained.
How the squeeze shows up in tech prices
RAM may be invisible once a device is assembled, but its pricing has a direct impact on hardware costs.
For PC and laptop manufacturers, higher DRAM prices feed straight into bill-of-materials calculations. With margins already thin, companies have responded by lifting prices, scaling back default memory configurations or charging more for upgrades. The crunch is also visible in fewer entry-level deals and a shift towards lower base specs.
Smartphones face the same dynamic. While handsets use less RAM than PCs, each new generation demands more memory to support heavier software and on-device AI features, keeping pressure on component costs.
Why Micron’s 2028 outlook matters
In past memory cycles, shortages were often followed by oversupply as new capacity came online and demand cooled. This time looks different.
AI-driven demand is longer-dated and less sensitive to short-term economic swings. Once data centres commit to memory-heavy architectures, consumption tends to persist. That makes the risk of a sharp price correction lower, and the prospect of sustained firmness higher.
With its latest outlook, Micron joins a growing chorus telling the market that elevated memory pricing may be a feature of the next few years, not a passing phase.