Finding The Bottom Of The Memory Trough

In announcing its Q2 fiscal 2019 results, Micron Technology, Inc. provided lower-than-expected revenue guidance of between $46 billion and $50 billion for the current quarter. However, what was particularly noteworthy was the company’s announcement to cut output by 5 percent due to weaker-than-expected market demand and its prediction that its customers’ inventory correction will last until midyear. And in an unexpected turn, Samsung Electronics disclosed ahead of its official financial announcement that the company’s Q1 results will miss consensus estimates due to a weakening display and memory business environment. Both announcements suggest that the current memory inventory cycle will bottom out in Q2 or Q3.

Additionally, forecast revisions for memory growth rates and total semiconductor sales have trended downward, reflecting deteriorating market conditions over Q1 2019. The question is: When will the memory market likely show signs of bottoming out? It’s an important query because the trough in the cycle could impact both the 300mm fab utilization and silicon wafer shipments. Following is an analysis of the current environment using the Memory Inventory Cycle Index introduced in this SEMI two-part article series in October 2018.

Figure 1, a look at historical memory revenue trends, shows that memory sales have fallen sharply since December 2018. February 2019 sales are creeping toward levels last seen in the first half of 2017, when this memory cycle started. Moreover, year-over-year (YoY) memory sales growth in December 2018 registered -3.1 percent, the first negative growth rate of this cycle, while February 2019 YoY sales growth approached the bottom rates of the previous two cycles. The trends have led to mixed expectations: Will the slowdown continue or is a rebound on the near horizon?

Figure 1

Remarks

1. Source: WSTS Bluebook
2. Sales revenue calculated based on 3-month moving average.

The Memory Inventory Cycle Index – based on financial data reported by Samsung, SK Hynix and Micron – is the difference between the YOY growth rates of sales (or shipments) and inventory. The index explains business cycle fluctuations such as expansions and contractions, trending up in expansions and declining in contractions. Figure 2 shows both historical Micron sales (blue dotted line) and the quarterly Memory Inventory Cycle Index (black solid line). To minimize seasonal fluctuations, both were calculated based on a four-quarter moving average of sales and inventory.

Figure 2

Remarks

1. All sales and inventory data calculated based on a 4-quarter moving average.
2. Memory Inventory Cycle Index = YoY growth rate of memory sales revenues – YoY growth rate of memory total inventories value
3. Calculated memory sales and inventories based on Samsung, SK Hynix, and Micron public announcements.
4. South Korea Won converted to US$ based on the quarterly average value released by FRED.
5. Memory Sales and Memory Inventory Cycle Index in Q1 2019 were estimated.

Figure 2 shows that the index has already dipped below the 2016 trough, suggesting the index will very likely continue to level off, at least in Q1 2019. In light of the two major memory makers’ downbeat comments, the completion of the current inventory correction is not expected until the first half of 2019. Also, after the previous two reversals of downward trends in the index – in 2011 and 2016 – it took a few quarters for inventories to adjust and the index to rally again.

A significant additional drop in this memory inventory cycle index is highly unlikely, unless the memory market maintains high utilization by increasing shipments in spite of the higher inventory – a risky game of chicken. The estimated index value in Q1 2019 is expected to near or drop slightly lower than the two previous troughs. Until the memory inventory cycle index rebounds in Q2 or Q3 2019, memory sales are expected to stagnate in 2019 and recover in 2020.

Figure 3 shows YoY growth rates for the proportion of both finished-goods and WIP inventory in the total inventory. The trendlines show how much growth rates could slow before we reach the bottom of this memory cycle.

Figure 3

WIP inventory growth rate will slow if the finished-goods inventory growth rate exceeds a certain level. In other words, manufacturers will likely reduce utilization until WIP inventory is cut to the appropriate level. The idea is to lower the finished-goods inventory growth rate to correspond to sluggish demand. On the flip side, the WIP inventory growth rate, which dropped to counterbalance stagnant demand in a contraction, will reaccelerate to increase finished-goods inventory once demand is expected to continue strong growth.

The red boxes in Figure 3 show the periods when demand is on the cusp of strong growth and the YoY growth rates of the proportion of WIP inventory in the total inventory turn positive. By contrast, the YoY growth rate of the proportion of finished-goods inventory in the total inventory simultaneously drops. In fact, it is evident that the WIP and finished-goods inventory situation in Q4 2018 is not like the previous two red-box periods. That means the inventory correction should last at least a few quarters unless unusual market conditions dictate otherwise.

WIP inventory adjustments, as Micron said, inevitably will drive down 300mm memory fab utilization. What’s more, utilization will likely continue to trend lower until the inventory cycle index rebounds. Compounding matters, this underutilization will erode silicon wafer usage. As a result, total silicon wafer unit shipments – which started to decline in Q4 2018 as shown in Figure 4 – will likely continue to drop until Q2 2019.

Figure 4

Remarks

1. Source: SEMI MMDS Silicon shipments data
2. Estimated 300mm utilization is total 300mm fab utilization including 300mm foundry fabs, not just 300mm memory fabs.

Figure 5 shows historical trends of the Memory Inventory Cycle Index and YoY memory fab equipment investments. The Memory Inventory Cycle Index recovered faster than fab equipment investments in the past two cycles. In the most recent memory cycle, both indexes peaked in the second half of 2017 and ultimately fell towards negative growth in Q4 2018. Consequently, the significant growth of memory sector investments, which the industry saw in 2017, is expected to resume if the inventory correction is all but complete in 2019.

Figure 5

Remarks

1. Source: SEMI WWSEMS
2. Equipment spending in Q1 2019 estimated by SEMI Fab Database and calculated based on a 4-quarter moving average.

In the two memory inventory cycles since 2008, once the cycle index hit bottom, it rebounded within a few quarters and memory sales revenue rose again after three or four quarters. In Q4 2018, the memory cycle index has already passed the trough of the 2016 cycle. It is very likely that this index will continue to level off in the first quarter of this year since this inventory correction will not likely be completed until the middle of the year – a projection supported by the YoY growth rates of the proportion of WIP and finished-goods inventory in the total inventory.

Presumably, the current index is very near the bottom of this cycle. Due to the lull, major memory manufacturers, including Micron, are controlling WIP and finished-goods inventory. As a result, significant additional downside of current memory inventory cycle index is not expected. In summary, memory sales are expected to stagnate in 2019, triggering a decline in 300mm fab utilization and total unit shipments of silicon wafers but will set the stage for a recovery in 2020.