10.23.24
The nuclear winter in NAND capital investment continues. This quarter was the lowest ever for NAND tools from Lam; $263M. The previous low was $308M, at the nadir of the downturn in 2023. NAND manufacturers were not convinced to even accelerate tech migrations meaningfully in response to the rising ASPs of the last nine months. Now that the non-AI memory markets have weakened, the brakes are back on. This bodes well for the NAND market to not swing into severe oversupply soon, but it still says that current NAND prices, which have more than doubled off their lows, are not high enough to make the business healthy. DRAM is a better story (how could it not be?) Equipment spending on DRAM tools at Lam increased 15% sequentially in the September quarter, to $521M. This is 2x to 3x higher than what was seen at the bottom of the trough in 2023. My conclusion is this stronger DRAM spending is a combination of faster node migration from the Big Three and inefficient capital investment from indigenous Chinese entrants (mostly CXMT.) I think the bits being added by the Chinese upstarts will drag out the downturn in the legacy end of the market but will have no direct effect on the leading edge, which is where most of the action from AI investment is (DDR5 and HBM.) I don’t think the $509M per quarter average of DRAM spending seen so far in calendar 2024 at Lam is high enough to overshoot the current level of demand for at least the next two quarters. The company stayed with their prediction of CY 2024 WFE to be in the mid-$90B range, the same as last quarter. Recall two quarters ago they forecasted the current calendar year to be low-to-mid $90B range in total revenue. From the analyst call, they are in an “industry environment where NAND spending has yet to recover.” Domestic China WE will be down in the second half of calendar 2024 compared to the first half. They see “growth” in overall 2025 WFE compared to 2024 and were hesitant to give much information beyond that. The spending recovery in NAND will be driven by technology upgrades. DRAM investment is growing in advanced packaging (HBM). The NAND industry continues to be in “a prolonged cyclical downturn.” DRAM spending in the quarter was focused on technology upgrades to 1-alpha, 1-beta, and some initial 1-gamma transitions. The drop in NAND spending from last quarter to this quarter was because of timing of a specialty product that is categorized as nonvolatile but is not traditional NAND. This means last quarter’s true NAND equipment revenue was lower than the reported number ($369M) indicated. Within 2025 WFE spending, they believe China will be lower. DRAM spending at the leading edge continues to be strong, focused on node migrations. Older nodes still have weak demand and are working down inventory. They are leaning into the fact that a lot of NAND capacity is on older nodes and that it needs to be upgraded. Also, NAND fab utilization continues to be strong for capacity that is in place. The subtext here is NAND makers aren’t adding new capacity. They are just migrating the nodes they already have forward. In summary, the NAND market shows no signs of increased capacity investment. NAND makers have increased fab utilization and are meeting the need for bit growth purely through node migration. DRAM investment is also in node migration, plus advanced packaging to support the fast growth in HBM demand. All this shows little sign of pending overinvestment in either DRAM or NAND capacity.
– S. Hughes (cyclical long MU)