The global memory-chip shortage is no longer a temporary supply-chain disruption. It is becoming a structural problem driven by the rapid construction of artificial-intelligence data centers, limited manufacturing capacity and growing geopolitical friction. The result is a costly squeeze on the DRAM and NAND flash used in smartphones, computers, gaming consoles, televisions and countless other electronic products—and consumers are increasingly being asked to pay the difference.
Artificial intelligence is the largest force behind the imbalance. Modern AI servers require enormous quantities of high-bandwidth memory, conventional server DRAM and high-capacity storage. Technology companies are expected to spend at least $630 billion on AI infrastructure during 2026, while major memory manufacturers have acknowledged that they are struggling to keep pace with demand. Because high-bandwidth memory is more profitable, Samsung, SK Hynix and Micron have directed more production capacity toward AI customers, reducing the supply available for traditional computers and consumer devices.
The expansion of data centers also affects NAND flash, the memory used for long-term storage in solid-state drives and smartphones. Sandisk recently reported strong AI data-center demand and has signed multiyear supply agreements worth nearly $94 billion, illustrating how much future production is already being reserved by large corporate customers. Meanwhile, Samsung says inventories throughout the DRAM and NAND supply chains remain below historical levels, even as manufacturers introduce denser and more efficient memory technologies.
Device makers have few attractive options. They can absorb higher component costs and accept smaller profit margins, raise retail prices, reduce the amount of memory included in entry-level products or concentrate on expensive premium models where the additional cost is easier to hide. Gartner has estimated that rising DRAM and solid-state-drive costs could increase average PC prices by approximately 17% and smartphone prices by 13% during 2026. Manufacturers including HP, Dell, Lenovo and Raspberry Pi have already raised prices or warned customers about higher costs.
Geopolitics is making the situation even less predictable. U.S. export restrictions affect which equipment can be installed at Samsung and SK Hynix factories in China, while new semiconductor tariffs and the possibility of broader trade measures add another layer of expense and uncertainty. China is rapidly expanding its own memory industry, including plans by DRAM producer CXMT for an additional factory, but political concerns could limit how freely those chips enter American and allied supply chains.
There is still some hope for consumers, although meaningful relief is unlikely to arrive immediately. Micron, Samsung, SK Hynix and emerging Chinese manufacturers are investing heavily in new factories, packaging capacity and higher-density memory. Micron alone has increased its planned U.S. investments to more than $250 billion through 2035. Analysts also expect that the current boom will eventually encourage enough production to weaken memory prices, with some projections pointing toward a significant decline by 2028.
Memory has historically been a cyclical industry, and today’s unusually high profits will eventually attract additional supply. The problem is that semiconductor factories take years to build and qualify, while AI demand continues to grow now. Consumers may therefore see occasional discounts, especially on older models and previous-generation memory, without experiencing a broad return to the inexpensive electronics of the past. Prices should eventually stabilize, but until new capacity catches up with the AI boom, memory shortages are likely to remain an invisible surcharge attached to nearly every electronic device we buy.
