Memory prices don’t usually double in the span of two quarters without something deeper breaking in the system, and that’s exactly why the current situation deserves the word crunch rather than the softer language of a cycle or correction. In segments like smartphones, the numbers are stark, with DRAM and NAND costs jumping sharply in a way that catches even seasoned hardware planners off guard. This isn’t driven by a single spike in consumer demand or a short-term logistics failure. It’s the result of memory quietly shifting from a background commodity into a form of critical infrastructure, something every modern device, service, and platform depends on simultaneously. Phones, cloud servers, AI accelerators, cars, industrial systems, all of them are drawing from the same underlying pool of memory production, and that pool is no longer deep enough to absorb shocks.
On the supply side, the industry is not standing still, but it is moving far more slowly than demand. Memory manufacturers are investing in new fabs and capacity expansions, yet those announcements can be misleading if you read them as near-term relief. A modern memory fab costs tens of billions of dollars and takes years to build, equip, and tune. Even after the building is finished and the tools are installed, yields don’t instantly reach usable levels. Early production is often inconsistent, with large portions unsuitable for high-end products. At the same time, manufacturers are deliberately cautious. The memory industry has been burned repeatedly by overexpansion, where too much capacity floods the market, prices collapse, and profits disappear for years. That institutional memory now shapes behavior. Companies would rather endure a tight market than risk another devastating glut, which means supply growth is real but restrained, almost conservative by design.
Demand, meanwhile, has changed in character. In the past, memory forecasts revolved around PCs and smartphones, products with relatively predictable cycles. Today, some of the biggest consumers of memory are AI training clusters, hyperscale data centers, and automotive platforms, and they don’t behave like normal buyers. They lock in capacity years in advance, sign long-term contracts, and are willing to pay premiums to guarantee supply. Once that capacity is committed, it effectively vanishes from the open market. Consumer electronics then feel the squeeze, not because total production is falling, but because the flexible portion of supply is shrinking. That’s a big reason analysts expect the shortage to persist until 2027 or even early 2028. The backlog created today doesn’t unwind quickly when so much memory is pre-allocated far into the future.
This broader crunch also explains why products like SD cards, which might seem peripheral, are not immune. SD cards mostly rely on NAND produced on mature process nodes, which gives them some insulation from the fiercest battles over cutting-edge memory. That’s why their prices haven’t exploded in the same dramatic way as mobile DRAM. Still, they draw from the same NAND ecosystem, and when supply tightens, manufacturers prioritize higher-margin products like embedded smartphone storage and enterprise SSDs. SD cards tend to sit lower on that priority list. The result isn’t an obvious shortage but an uneven market, where certain capacities or speed classes quietly disappear for a while, or where professional-grade cards creep up in price while basic models remain relatively stable.
Inside the SD card market itself, the split is becoming clearer. Entry-level cards used for casual photography, dashcams, or IoT devices are likely to stay affordable, with mild fluctuations rather than shocks. High-performance cards, especially UHS-II and newer standards aimed at video shooters and professionals, are more exposed. They require higher-quality NAND, stricter binning, and better controllers, all of which compete more directly with SSDs and premium mobile storage for the same components. Add in cautious inventory management by distributors, who know SD cards are easy to stockpile and therefore risky to overstock, and you get a market that feels unpredictable even when it isn’t truly broken.
At the core of all this is a hard truth that’s easy to underestimate: making more memory has become genuinely difficult. The physics are unforgiving. NAND relies on stacking dozens, now hundreds, of layers vertically, and each added layer complicates manufacturing and reduces margins for error. DRAM cells are already so small that leakage, heat, and reliability become constant challenges. Every incremental improvement demands more exotic materials, more precise tools, longer validation cycles, and more engineering effort than the last. You can’t just add a shift or buy a few extra machines and expect output to scale linearly. Memory manufacturing has crossed a threshold where progress continues, but slowly and expensively.
Taken together, this is why the memory crunch feels persistent rather than dramatic. Prices rise sharply in some segments, drift upward in others, and availability becomes uneven across the board. SD cards don’t explode in price, but they stop feeling boring and predictable. Smartphones absorb cost increases or adjust configurations. AI and automotive quietly lock in capacity years ahead. The industry isn’t broken, and it isn’t asleep at the wheel, but it is operating at the edge of what’s physically and economically sensible. Until demand growth slows or a real manufacturing breakthrough appears, memory will remain tight, expensive, and strategically important, less a temporary problem to be solved than a new background condition everyone has to design around.
