2.12.25
I’ll start with commentary on bit growth and pricing for memory. Hynix has seen seven consecutive quarters of rising DRAM prices. They have the most because they were first to the HBM market. Samsung’s DRAM prices have risen for six quarters, Micron’s for five, and Nanya’s for four. Nanya is getting squeezed between China on the low end and their lack of advanced design, process, and packaging technology at the high end. I think this DRAM downturn has been muted and shortened by the emergence of CXMT into the low end of the market. Micron cited an independent researcher who estimated CXMT’s DRAM market share was mid-single digits percent in 2024. That is enough to explain why the market hasn’t been better following an historic downturn. Samsung’s DRAM pricing in the most recent quarter – up twenty percent – is a data point I don’t understand. Micron and Hynix saw DRAM ASPs rise half that much. HBM is too low a fraction of their total output to have this large an effect. Maybe they had favorable mix for one quarter? Either way, they forecasted a slight decline in DRAM ASPs in Q1.
The NAND upturn was even shorter. Samsung saw rising prices for five quarters in NAND and all the other players had their ASPs increase for four quarters. The same independent analysis that Micron cited for DRAM estimated YMTC’s NAND market share to have been high single digits percent in 2024. Just like with DRAM, that is plenty of bits to disrupt the undersupply situation and shorten the downturn. The emergence of YMTC in China is going to make the NAND market unhealthy for at least a half a decade. Even before this new player was added, NAND had one too many companies. Profitability has not been high enough for at least a decade to justify technology investment, but all the participants are pot committed. I just don’t know what happens to that market. Micron took the bold step of underloading their NAND fabs by 15% and will start taking charges from that in the spring. They are doing this in hopes that the rest of the market will follow.
All the memory companies saw weaker than expected demand in the fourth quarters of 2024. Samsung was the furthest off in their prior belief that the market would improve in Q4. They are all now saying that the DRAM market will start to improve in the second quarter of 2025 and NAND, hopefully, in the second half of the year. PC and mobile unit growth has not recovered as had been hoped for. I think the predictions that AI-enabled PCs and phones will spur demand is fanciful. Alongside their positive comments about 2025, Micron still only predicted supply will be in line with demand for the year. Last quarter the big three DRAM companies all said they would hold back inventory to support pricing. It hasn’t mattered.
What will happen now? It is over for NAND. That is clear to me. YMTC is approaching 10% market share in a business that already had one too many players. WD acknowledged that the business is in a new regime, one of lower demand growth and reduced capital intensity. DRAM is bifurcated between legacy and leading-edge/AI. CXMT is disrupting the legacy market. Nanya is at existential risk from this. The Big Three are migrating their fabs to advanced nodes faster. This is their best defense against China. The story for 2025 will be the race between AI demand holding up and how fast Micron and Samsung can bring on more bit supply to feed it. If their HBM and DDR5 supply overtakes the AI demand, then the whole DRAM market will go into free fall.
– S. Hughes (no MU position)