Memory Summary for Q1 2024

5.6.24

We have reached the point in the cycle where inventory no longer dominates the commentary from the memory companies. I barely noticed what they did say and had to go back through Micron’s FQ2-24 analyst call to see that the company did indeed mention inventory. They said server inventory will reach normal levels by the end of the first half of calendar 2024 and that all other memory types are at normal levels now. Micron sees bit demand growth in calendar 2024 to be in the mid-teens of both DRAM and NAND, an increase from their previous position. Their bit supply is going to be low teens percent in 2024. The other companies did not give similar detail on their outlook for 2024. They were universal in the view that 2024 will be a strong year for memory and that supply will be tight. In the first quarter, DRAM ASPs were up around 20% sequentially and NAND prices were up 30%, ex. WD. Western Digital’s NAND pricing was up 18% in the quarter but their gross margin in NAND is higher than the other manufacturers, so they were probably coming off a higher price. Based on Hynix, the company with most of the market share in HBM today, the premium for HBM is massive. Specifically, Hynix’s gross margin was 2500 bps higher than Micron’s, for a similar period in time. Both companies have similar cost structures, so the biggest difference in financial performance in the top section of the income statement is pricing. Besides a higher premium, the upward turn in pricing started five to six months before the rest of the DRAM industry, as measured by Hynix’s gross margins turning upward. Hynix’s gross margin turned positive in the second quarter of calendar 2023, meaning they just finished their fourth consecutive quarter of increasing gross margin. Micron, Samsung, and Nanya have all had two quarters of expanding gross margins. If HBM is like the rest of DRAM and has upturns that last about two years, the HBM upturn will peak about a year from now, in the second calendar quarter of 2025. All companies said their current capital expenditure plans are for node transitions and not higher capacity. They did add some spending to their 2024 plans and said it is for expanding HBM capacity and for more cleanroom space. Companies are clearing the decks to add wafer capacity when they believe the market justifies it. However, all the companies said that pricing currently does not justify investment in more wafer capacity. Based on WD’s results, the bottom of the NAND market was in the July-September period of 2023. This is about the same time period the DRAM market reached its nadir, excluding HBM. Thus, we are now in the third quarter of the 2024-25 memory upturn. While the memory IDMs aren’t publicly announcing expansion plans, they are certainly working feverishly behind the scenes preparing to expand capacity, when they deem ASPs justify it. The sharpness in pricing changes during cyclical transitions that are underestimated by the markets. Even now, following two quarters of DRAM prices up a total of over 40% and NAND prices up more than 60%, analysts are, I believe, under-calling how much ASPs will go up from here. I see overall memory industry supply growth in 2024 to be well under 10% while demand will be in the mid-teens, meaning further tightness is in the offing. In the upturn of 2016 to 2018, ASPs more than doubled, and that was off of a typical downcycle. I think most observers and investors are going to be shocked at how rapidly prices and profits in memory grow during 2024 and into 2025.

-S. Hughes (cyclical long MU)

3 Likes