Key Highlights

  • Wolfe Research has modeled a memory shortage extending through 2028 rather than easing in 2026 or 2027.
  • SK Hynix has separately warned that shortages in some memory segments could persist even longer.
  • Micron's 2026 high bandwidth memory capacity is already reported sold out.
  • Extended shortage forecasts support continued pricing power for the largest memory manufacturers.

Why Forecasts Keep Extending

Early expectations that the AI-driven memory shortage would ease as manufacturers added capacity have repeatedly been pushed back. Wolfe Research now models a shortage lasting through 2028, and SK Hynix has offered commentary suggesting some memory segments could remain tight even longer, as new fabrication capacity takes years to build and AI accelerator Demand keeps growing faster than supply.

The Supply Side Constraint

Building new memory fabrication capacity requires enormous Capital Investment and multi-year lead times, which is why manufacturers cannot simply flip a switch to meet demand spikes. Micron's disclosed roughly USD 200 billion capacity investment reflects the scale of spending underway, but even that level of investment takes years to translate into meaningful new supply hitting the market.

What This Means for Buyers and Investors

For AI companies and Data Center operators, an extended shortage means continued elevated costs for memory components central to Training and running large models. For investors in memory manufacturers, it supports the case for sustained pricing power and Margin expansion, which has been the primary driver behind sharp stock gains at companies like Micron over the past year.

Where Things Stand

Multiple credible forecasts now point toward a shortage lasting through 2028 rather than easing sooner, driven by the multi-year lead times required to build new fabrication capacity against AI demand that keeps outpacing supply.