Analyst Jukan relays a note from Morgan Stanley's Joseph Moore, who checked in with several data-center buyers last week. His message: the memory crunch is not letting up. Moore says prices look set to rise at least 25% on a like-for-like basis from the second to the third quarter, above both Morgan Stanley's own and outside estimates. Memory here means the DRAM and high-bandwidth chips that AI servers need in bulk, and buyers are so worried about getting enough that cloud customers are reportedly paying premiums over the expected price just to lock in supply. That is the behavior of a market where demand outruns what factories can make. Moore also keeps his longer-range warning intact: the shortage is likely to get worse again in 2027 and 2028, because there simply is not enough memory being built relative to what AI needs, and he does not see that changing. For investors the read-through is straightforward. If a top semiconductor desk is telling clients the squeeze runs for years rather than quarters, the pricing power sits with the memory makers, and the risk to the bullish case is less about demand fading than about whether these steep price forecasts prove too optimistic if AI spending cools.
Why it matters · A top chip desk telling clients the memory squeeze lasts years, not quarters, hands lasting pricing power to the memory makers.
Worth asking · A multi-year memory shortage into 2028: durable pricing power for the memory makers, or the kind of forecast that breaks the moment AI capex slows?