Kuo: AI Demand Will Widen the DRAM Supply Gap Through 2027
TL;DR
- Kuo forecasts 15 to 20 percent of consumer electronics DRAM capacity will shift to AI data centers in 2027, widening the supply gap.
- Apple's A20 chip procurement in H2 2026 through Q1 2027 may fall 10 to 20 percent below original targets due to tight LPDDR supply.
- Apple's lobbying for CXMT access is driven by supply security, not cost savings; even success would not materially fill the global DRAM supply gap.
The pressure on Apple has shifted, according to Ming-Chi Kuo's latest industry note, from soaring memory costs to a widening supply gap. That shift matters because a pricing problem is something purchasing teams can budget around, but a supply availability problem cannot be solved with a bigger check.
Kuo estimates that 15 to 20 percent of memory capacity currently allocated to consumer electronics in 2026 could be redirected to AI data centers in 2027, and that the supply-demand gap will keep widening through that year. The downstream consequence for Apple is specific: actual A20 chip procurement in the second half of 2026 through the first quarter of 2027 could fall 10 to 20 percent below original targets because of tight LPDDR supply. A shortfall of that magnitude, during a product launch window, shows up in shipment forecasts and market share numbers, not just gross margin.
That is also the context for Apple reportedly lobbying the White House to keep CXMT, the Chinese memory maker, off the Commerce Department's Entity List. As Kuo noted on X, even a successful lobbying effort would not materially lower costs or fill the supply gap because CXMT's capacity is far below domestic demand. The strategic logic is simply adding an additional source at a moment when the global memory imbalance is worsening. Kuo draws an explicit contrast with Apple's 2022 YMTC evaluation, noting that YMTC was mainly about lowering NAND costs, while CXMT is about managing DRAM supply risk.
The honest caveat is that this is one analyst's forecast, and the piece does not specify how much of Apple's LPDDR demand is already covered by existing suppliers, or at what pace memory makers could bring new LPDDR capacity online. If AI infrastructure spending moderates or new capacity arrives sooner than expected, the gap could narrow before 2027.
For OEMs without Apple's lobbying leverage or procurement scale, a structural DRAM shortage is a harder problem than a price spike. Companies that secured long-term LPDDR supply agreements before the gap widened are in a meaningfully different position than those competing on spot markets through 2027.
Originally reported by medium.com
Read the original article →Original headline: Ming-Chi Kuo: 15–20% of Consumer Electronics DRAM Capacity Will Shift to Data Centers in 2027, Widening Supply Gap Beyond Price — Apple A20 Production May Fall 10–20% Short