Why Micron’s CEO Says the Global Memory Chip Shortage May Last Until 2028

In many cases, the constraint is no longer GPUs alone but the memory needed to feed them. As hyperscalers build massive AI clusters, they are purchasing millions of accelerators that each require large amounts of DRAM and specialized memory to operate efficiently.

Micron executives say the company can currently supply only about 50% to two‑thirds of the demand from some key customers, highlighting the scale of the gap between supply and demand.

High‑Bandwidth Memory Is Tightening the Entire Market

One of the biggest drivers of the shortage is high‑bandwidth memory (HBM)—the advanced memory stacked with GPUs used for AI training and inference.

HBM chips are significantly more complex to manufacture than conventional DRAM. They require:

• larger die sizes
• advanced packaging
• additional wafer capacity

Because of these requirements, producing HBM consumes more manufacturing resources than standard memory chips. As suppliers prioritize these higher‑value AI products, less capacity remains for conventional DRAM used in PCs, smartphones, and other devices.

The result is a tightening of the entire memory ecosystem. Some reports indicate that future HBM output is already largely pre‑committed to major AI customers years in advance.

Why New Supply Won’t Arrive Quickly

Even though chipmakers are investing billions to expand production, semiconductor manufacturing timelines are extremely long.

Building a new leading‑edge memory fab involves several stages:

1. Construction of the fabrication facility
2. Installation of specialized manufacturing equipment
3. Process calibration and yield tuning
4. Customer qualification and product certification

Because of this process, new plants that begin construction today typically take several years before they deliver “meaningful output” that affects global supply. Industry reporting linked to Micron indicates this impact is unlikely before about 2027–2028.

Micron’s $200 Billion U.S. Expansion

To address the long‑term demand for memory, Micron has announced a massive expansion plan in the United States.

The company plans to invest roughly $200 billion in semiconductor manufacturing and research, including about $150 billion for domestic memory production and $50 billion for R&D.

The expansion includes:

• Idaho: new leading‑edge memory fabs in Boise
• New York: a massive semiconductor complex with up to four fabs
• Virginia: modernization and expansion of the existing Manassas facility

The New York project alone could become the largest semiconductor manufacturing complex in the United States.

However, even these projects will take time to influence global supply. For example, Micron’s first new Idaho fab is expected to begin operations around 2027, with broader capacity additions arriving later.

Other Chipmakers See a Multi‑Year Shortage

Micron’s outlook is broadly echoed across the memory industry. Samsung and SK Hynix have both warned that demand for AI‑related memory is rising faster than supply, with shortages likely to continue through at least 2027.

Some industry leaders say the imbalance could persist even longer as AI data‑center construction accelerates worldwide and manufacturers struggle to add enough wafer capacity quickly.

Impact on Consumer Electronics

The shift toward AI‑focused memory production is already affecting other sectors of the electronics market.

Because manufacturers are prioritizing high‑margin AI memory, companies producing PCs, smartphones, and other consumer devices are facing tighter supply and higher component costs.

Rising memory prices and limited availability could pressure margins for electronics makers and contribute to higher prices for end‑user devices in the coming years.

A Structural Shift in the Memory Industry

The key message from Micron’s leadership is that the current shortage is not simply another cyclical downturn in semiconductor supply.

Instead, the rapid expansion of AI computing has fundamentally changed the demand profile for memory. With new fabs taking years to complete and AI workloads requiring far more DRAM and HBM than previous computing models, the industry may remain supply‑constrained until late in the decade.

For now, the global memory market is entering what many analysts describe as a multi‑year AI‑driven supercycle—one where demand growth may outpace supply well into the late 2020s.