Red Alert in Memory Markets: AI Demand Triggers Full-Blown Supply Crisis

A recent analysis from Morgan Stanley, a key bellwether in global finance, has sent shockwaves through the semiconductor industry. After discussions with multiple data center procurement heads, analyst Joseph Moore paints a picture of a memory market far more strained than previously understood.

Prices Skyrocket Past Forecasts: Q3 Increase Could Exceed 25%

The report's core finding is striking: there are no signs of easing in the current memory supply crunch. On a comparable basis, prices for memory chips, led by DRAM, are projected to rise by at least 25% sequentially from Q2 to Q3. This figure significantly outpaces prior market consensus and exceeds earlier estimates from both Morgan Stanley and third-party research firms.

This surge implies rapidly escalating hardware costs for downstream server manufacturers, cloud giants, and consumer electronics companies.

The Core Driver: AI Is "Devouring" Supply

The root cause of this crisis is squarely pointed at artificial intelligence.

  • Explosive Demand: Large language models and AI training/inference are creating exponential growth in demand for High Bandwidth Memory (HBM) and conventional DRAM, voraciously consuming finite global wafer capacity.
  • Capacity Crowding Out: To fulfill more lucrative AI chip orders, memory makers are shifting resources toward premium products, severely squeezing supply of standard DRAM for PCs, smartphones, and other consumer electronics.
  • A Self-Reinforcing Bottleneck: The report highlights a dual challenge: while AI development fuels soaring memory demand, the insufficiency of memory supply itself has become a critical bottleneck constraining the expansion of AI infrastructure.

The Long-Term Shadow: Shortages May Persist Until 2028

More concerning is that Moore's analysis doesn't frame this as a short-term blip. He warns that the memory shortage could intensify further in 2027 and 2028. Given that building and ramping up new chip fabrication plants often takes years, the current demand-supply imbalance is unlikely to be resolved fundamentally in the medium term.

For the tech industry, this is a clear signal: behind the heated AI race, a full-scale battle for core hardware resources—especially memory—is underway, with repercussions that will ripple through every layer from the cloud to the device.