Jul. 4 at 4:21 PM
$DRAM the memory (DRAM/HBM) bottleneck is not going away soon. It looks structural and is expected to persist through 2027, with meaningful relief unlikely before late 2027 or 2028 (and some forecasts extending tightness into 2029–2030 in certain segments).
The Tencent deal and CXMT's rise validate China's domestic push and add a credible #4 supplier—good for diversification and China supply security. It creates manageable pricing pressure in mainstream segments over 2–3+ years, but does not dismantle the Big 3's dominance in high-end/HBM or immediately crash the supercycle. Technology + equipment gaps, plus strong AI demand, keep the threat contained for now. Monitor HBM progress and actual Shanghai ramp yields for longer-term shifts.
