The Future of DRAM Supply and Demand: Challenges Ahead

The semiconductor industry has never been at a more critical juncture. Understandably so, with total demand for DRAM that is running rampant—particularly in data centers driven by artificial intelligence (AI). In somewhat electrifying news, Samsung Electronics recently revealed plans to build a new semiconductor plant in Pyeongtaek, South Korea. They want to start production out…

Tina Reynolds Avatar

By

The Future of DRAM Supply and Demand: Challenges Ahead

The semiconductor industry has never been at a more critical juncture. Understandably so, with total demand for DRAM that is running rampant—particularly in data centers driven by artificial intelligence (AI). In somewhat electrifying news, Samsung Electronics recently revealed plans to build a new semiconductor plant in Pyeongtaek, South Korea. They want to start production out there by 2028! This announcement is the latest in a series of moves by the company to meet the growing demand for memory chips. Yet, experts warn that massive investments and groundbreaking innovations are needed to address today’s supply bottlenecks.

According to the Semiconductor Industry Association, the global capacity is only about 9,000 semiconductor fabs. They think this figure would increase by 20 percent if all facilities currently planned are constructed. Despite these promising figures, Thomas Coughlin, an industry analyst, notes that the high costs of constructing new fabrication plants often exceeding $15 billion make companies hesitant to expand capacity. The new hesitation comes amid shootthrough-the-roof prices fueled by year-over-year historic demand for all DRAM, especially that used in GPUs and AI accelerators.

Innovations in DRAM Technology

In a groundbreaking development, researchers at Samsung demonstrated their capability to produce a 16-high stack of memory dies using hybrid bonding techniques in 2024. This invention could lead to memory chips in stacked configurations that far surpass today’s standards. The industry’s current specification for High Bandwidth Memory (HBM4) allows for 16 vertically integrated DRAM dies. By contrast, today’s HBM chips typically use 12 dies at most.

Samsung’s announcements don’t stop at stacking though, with far-reaching aspirations for upcoming releases. The company implies that 20-die chips could be within reach as soon as 2024. The newly launched B300 model has eight of these HBM chips. Each chip is a complex stack of 12 dies, reflecting the manufacturers’ long-time dedication to creating smaller, faster, smarter memory.

“Relief will come from a combination of incremental capacity expansions by existing DRAM leaders, yield improvements in advanced packaging, and a broader diversification of supply chains,” – Shawn DuBravac, chief economist for the Global Electronics Association.

As businesses compete for resources in this rapidly changing environment, creative solutions will be needed to both diminish the potential for shortages. Speaking of yield, the increased adoption of advanced packaging techniques would directly improve yield rates and in turn help create a more stable supply chain.

Data Center Demand and Economic Implications

One area where demand for DRAM has skyrocketed is due to the explosive growth of data centers. In fact, there are almost 2,000 new data centers planned or under development around the world right now, based on the website Data Center Map. This expansion across the sector is indicative of explosive cloud computing needs and growing artificial intelligence applications.

NVIDIA’s data center business exemplifies this trend. That company’s revenue exploded from less than $1 billion in the fourth quarter of 2019. By the quarter ending in October 2025, it had climbed to a staggering $51 billion. This rapid expansion further emphasizes the massive demand for DRAM necessary to fill out the support for GPUs and other accelerators within AI-centric datacenter environments.

As demand intensifies, prices have skyrocketed. As advocates like analysts Mina Kim make clear, it’s important to talk about addressing supply. To accomplish this, we either need to find creative solutions or fabricate more fabrication factories.

“There are two ways to address supply issues with DRAM: with innovation or with building more fabs,” – Mina Kim, an economist with Mkecon Insights.

Still, despite these big steps, Kim warns not to expect miracles when it comes to price drops.

“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.

Future Outlook and Timeline for Relief

Still, industry leaders are averted about how long it will take for relief from existing supply bottlenecks. As Intel CEO Lip-Bu Tan recently noted, even in the best case, huge gains wouldn’t be realized until 2028 when Samsung’s new facility comes online. This aggressive timeline adds to growing worries, particularly among manufacturers, about their capacity to answer rapidly rising demand in the meantime.

When Micron Technology first announced its plan, the company had targets to achieve production milestones by certain dates. Now, they’ve lowered the bar and are set to meet those goals two full years ahead of schedule! This exciting announcement gives further evidence to the desire and desperate need in the industry to scale up production capacity quickly.

The current environment has created a challenging dynamic of increased demand coupled with decreased supply. As companies like Samsung prepare to invest in new production facilities and innovate existing technology, stakeholders across the semiconductor landscape remain hopeful for improved conditions.