In particular, DRAM (Dynamic Random Access Memory) demand is exploding worldwide. This spike is particularly pronounced in artificial intelligence (AI) data centers. This unexpected rush of demand has pulled supply away from other applications. Consequently, prices have soared and the memory chip market is under even more extreme pressure. Industry experts in a recent roundtable warned that real relief is likely years away — as late as 2028. Second, it takes more than 18 months to build and bring new fabrication plants—or fabs—into production.
The AI technology boom has catalyzed recent historic demand for semiconductor production capacity. This increase has placed significant strain on DRAM production, particularly for DRAM used in graphics processing units (GPUs) and other accelerators. Firms are scrambling to put up new data centers to host AI applications. The sudden spike in demand triggered a new memory chip shortage that’s been much more impactful. Then the COVID-19 pandemic threw everything into complete disarray. It set off a chip supply crisis, with manufacturers unable to keep pace with skyrocketing demand.
The Impact of Rising Demand
Right now, the demand for DRAM—especially in AI data centers—is outpacing demand in other areas. NVIDIA and other feeding-frenzy beneficiaries have seen data center businesses explode. Their combined revenues exploded from less than $1 billion in late 2019 to a projected $51 billion by the end of October 2025. This growth trajectory is indicative of the deepening adoption of AI technologies and the resulting demand for powerful memory solutions.
Currently, according to Data Center Map, almost 2,000 new data centers are on the drawing board or already being built. This breakneck growth fuels a tsunami of demand for DRAM. With every new facility accumulating incredible memory loads to inform its day-to-day operations, their importance is only growing. Data center construction isn’t just moving fast – it’s accelerating at an incredible pace. This urgency forces DRAM suppliers to be nimble and agile with respect to swift market changes.
“There are two ways to address supply issues with DRAM: with innovation or with building more fabs,” – Mina Kim, an economist with Mkecon Insights.
This proclamation highlights the urgency for a change in direction and deeper collaboration across the sector. The absence of short-term fixes raises fears over a prolonged jump in prices as producers try and catch up with exploding demand.
Challenges in Production and Supply Chain
The semiconductor industry has a storied past when it comes to boom and bust cycles. Unfortunately, it’s up against a powerful challenge—namely, these cycles are coinciding with a firehose, accelerated buildup of AI hardware infrastructure. The expense of constructing new fabs has ballooned to more than $15 billion. As a result, for good reason, firms are fearful to invest during these perilous economic times.
After a fab is built, it’s a long time before any production actually starts—up to 18 months or more. This delay can be especially egregious, though, since it makes it very difficult for manufacturers to respond quickly to changing market demands. To respond to the growing demand for these chips, Samsung said it would start producing advanced memory chips at a $37 billion plant in Pyeongtaek, South Korea, by 2028.
The company has made great progress in technological innovations, demonstrating its ability to manufacturer a 16-high stack with hybrid bonding technology. On top of that, Samsung has indicated that raising the stack to 20 dies might be possible within the same year. Even with these innovations, the new high bandwidth memory (HBM4) standard doesn’t cut it. It still only supports 16 stacked DRAM dies, though this is up from the 12 dies currently in use.
“In general, economists find that prices come down much more slowly and reluctantly than they go up. DRAM today is unlikely to be an exception to this general observation, especially given the insatiable demand for compute,” – Mina Kim.
This somewhat half-hearted line is representative of the harsh and unrelenting realities of pricing in DRAM. Given demand greatly outstripping available supply, any price cuts are likely to happen only after supply has rebounded well above more normal levels.
Future Prospects and Market Conditions
Firms are already deeply engaged in working through these challenges. According to industry experts, relief from the current shortage will arrive from three main developments: incremental expansions of capacity from current DRAM leaders, improved yields through next gen packaging processes, and greater diversification of supply chains.
Shawn DuBravac, chief economist for the Global Electronics Association, pointed to this multilateral approach as key to addressing supply constraints. Given the intense, mounting production and scalability pressures manufacturers will face, incremental improvements will still be key.
Even though there is promising outlooks for increased production capacity, industry executives warn that meaningful relief may be years away. Intel CEO Lip-Bu Tan emphasized this point clearly:
“There’s no relief until 2028.”
Such statements indicate a consensus among industry insiders that while advancements in technology and production methods are underway, they may not be sufficient to address immediate supply challenges.