For the full year 2021, end use of lithography systems was 30% memory and 70% logic. In the previous year memory was 28% of total sales, so the sector has grown at about the same rate as ASML’s overall sales. But 2021 was a huge growth year for ASML, so total memory revenue grew more than 40%, from €2.89B to €4.09B. This level of photo equipment sales is still below the last peak of €4.55B in 2018. Capital intensity increases every node, as does the overall size of the memory market, so the peak of this cycle’s memory equipment sales will be larger than the last one. ASML’s total sales of equipment into the memory sector should easily top €5B in 2022. In 2022, about 20% of their total EUV sales will be to memory customers. DRAM utilization is “very high” right now, so they expect extra lithography tool shipments are needed to support the high-teens DRAM bit growth they forecast in 2022. In the analyst call management said overall memory revenue growth for ASML in 2022 will be 25%. That translates to €5.1B in memory sales in 2022. I think it will be the peak year, with memory prices reaching their cyclical highs in the second half of 2022 or early 2023. When pricing turns down, photo tool sales into memory will drop rapidly. But new memory bookings rose sharply last quarter. In Q3-21 they were €989M. For the next quarter, Q4-21, total memory bookings were €1.62B. That rate of bookings annualizes to more than €6B, which I am confident would oversupply the memory market. But it is just one quarter and is swollen by expensive EUV systems as well as by extended lead times. The tools ordered for memory in this high bookings quarter will be delivered in late 2022 at the earlies.
2021 was a company record for DRAM sales. In 2022, Applied believes overall sales of WFE will increase around 15%, led by foundry-logic, so memory will undergrow the market. Management believes that memory sales in 2021 were moderate, despite what the DRAM record might indicate. They cited that 2007, 2017, and 2018 were three years in the last 15-20 years when sales were more than 50% memory. Memory was 40% in 2021 and is expected to be down in 2022. Later in the call they said overall memory is expected to be flattish in 2022. Between the two comments, I conclude Applied sees total equipment sales into the memory segment in 2022 will be slightly down.
Within memory, KLA management expects NAND to grow faster than DRAM in 2022. This is the reverse of 2021 when DRAM led NAND. The company sees foundry/logic leading the growth in WFE sales in 2022, so they must see memory growing below the overall rate. In the fourth calendar quarter of 2021, they saw memory sales of process control equipment at 29% of total. In the March quarter they are modeling this same market to be 41% memory. With the March forecasted to be at least similar in sales as the December quarter, this means memory sales will be up by about a third. For 2022 overall, they expect another year of double-digit growth supported by broad-based demand across multiple customers. KLA sees memory spending will be led by NAND in 2022 but followed closely by DRAM, which will pick up strength in the year.
In the December quarter, LAM’s revenue from systems sales was 35% NAND and 23% DRAM, with the balance into foundry/logic and other. This is a decrease from the previous quarter for memory compared to foundry/logic. Within memory, the mix shifted to DRAM from NAND. Little was asked and said about memory during the analyst call, though one of the last answer given revealed their view of the memory space is that spending is currently “balanced.”
The WFE companies all believe 2022 will be a growth year over 2021, and they all believe this growth will be led by foundry/logic. Commentary on the memory sector continues to be that investment is not overheated, though it is good to keep in mind the companies that sell equipment have an incentive to tell investors that the good times will continue to roll. The numbers provided by ASML indicate to me that 2022 will likely be the peak year for memory investment. But as I’ve said before, my confidence to predict this cycle is greatly reduced because of how the COVID pandemic has disrupted demand and supply. I can see a world where supply chain constraints easing throughout the year lead to growing undersupply in the DRAM market as 2022 progresses and gross margins for Micron increasing into the low 60% range by the first half of 2023. I’m not predicting this to be the most likely scenario, but I do believe that the DRAM market will have its best year this cycle in 2022, and the commentary from WFE companies is consistent with this view. I continue to be wary of NAND because of statements from Samsung and Hynix that they will invest heavily to grow their share. But DRAM is more important to Micron, and both Hynix and Samsung have said their top priority in this space is to protect profitability. This means they will be cautious about adding capacity.
-S. Hughes (long MU)