5.16.24
Tokyo Electron (TEL) is a WFE company I have not previously followed. They are a competitor of Applied and Lam Research. These results are for their full fiscal year 2024. In DRAM, sales to Chinese customers increased in the last year, characterized as “significant investment.” NAND sales fell as capex continues to be restrained, citing inventory adjustments by customers. The company forecasts CY2024 total WFE to be around $100B. They expect investment in leading-edge DRAM to recover “from H2 on higher demand for DDR5, HBM, etc.” They don’t see NAND investment to return until CY2025. Unlike their peers, the company forecasts coming quarterly revenue by market segment. TEL sees DRAM equipment sales in the first half of FY2025 (April – September 2024) to be down 17% from the prior half. For the following period (October 2024 through March 2025), the company is predicting DRAM spending back up to the level of the six months between October 2023 and February 2024. Based on their comments, and those from Applied Materials, the rebound forecasted to come near the end of calendar 2024 is the recovery of leading edge DRAM spending, coming in at a similar magnitude to the sales to indigenous China that are rolling off now. In NAND, tell is forecasting the next six months at a similar low investment level to what has been seen in the last year, then a doubling of spending on NAND equipment in the period from October 2024 through March 2025. I have confidence in forecasts out twelve months when they come from the equipment companies, because their long lead times mean they have data on equipment bookings going out several quarters. TEL does not expect new investment in NAND capacity to happen before the end of September of 2024. The company believes NAND has more inventory than DRAM currently, and that inventory adjustments for NAND will continue “throughout the year.” In DRAM, they expect “active investment” this year as DRAM is in shortage for AI applications. Some NAND production may be displaced by DRAM production. In other words, they see customers with flex fabs shifting capacity from NAND to DRAM. With my trend charts of WFE sales by quarter by segment now updated with TEL’s data, the increase in DRAM sales in the last three quarters is even sharper. Like Applied, TEL has also seen high DRAM sales to indigenous China in the last two-to-three quarters. They are also predicting this will drop off in the next few months. The company is predicting new investment in DRAM to begin in the fourth quarter of calendar 2024. None of the other WFE companies provide this level of forecast detail. In summary, NAND investment has been historically low for five quarters, a trend that is forecasted to continue until at least the fourth quarter of 2024. In DRAM, TEL believes new investment in leading edge capacity will commence in the fourth quarter of 2024, to meet the constrained supply-demand situation. From TEL’s historical sales to DRAM customers, we appear to be three quarters into the upturn, as measured by WFE investment. That means there is likely another year of investment before the market goes into oversupply. This is the hardest thing to predict in memory because every cycle is different. My working theory today is the DRAM market will peak in early to middle of calendar 2025. The NAND makers continue to invest little in WFE. This may end up being a surprising area of strength in this cycle. Based on timing so far, the current upturn may have offset timing between DRAM and NAND, with the latter seeing later investment and thus a later peak.
-S. Hughes (cyclical long MU)