Memory Chips: The Interplay Of Equipment Spend, Process Migration, Capacity And Stock Prices

by: Robert Castellano


Memory chip companies' semiconductor spend in 2017 grew 68% from 2016, which will result in capacity increases of 45% by the end of 2018.

DRAM equipment spend in 2016 dropped 34% from 2015, which has resulted in undersupply in 2017 and capacity increases of only 20% by the end of 2018.

Semiconductor equipment spend for Q1 2018 reached a historic high, which will further result in memory capacity increases and creating an oversupply situation in 2018 and 2019.

The introduction of new capacity by Chinese DRAM and NAND companies will further exasperate the oversupply of memory chips.

According to The Information Network's (Market Research for Microelectronics | The Information Network) report entitled "Global Semiconductor Equipment: Markets, Market Shares, Market Forecasts," DRAM and NAND memory suppliers' wafer front end (WFE) semiconductor equipment spend grew only 1.5% between 2015 and 2016 to $16.7 billion, as shown in Table 1, whereas WFE equipment spend for non-memory ICs grew 20.7% between 2015 and 2016. The small equipment spend increase in 2016 has resulted in undersupply of DRAM and NAND chips through 2017 and into 2018.

While the overall equipment spend increase for memory considerably lagged non-memory, equipment spend for DRAMs was also significantly less than for NAND. DRAM spend decreased 34.1% between 2015 and 2016, while NAND spend increased 44.8%.

Figure 1 illustrates WFE equipment spend segmented by NAND and DRAM for the period 2013-2017.

Chart 1

The decrease in equipment purchases for DRAMs in 2016 led to DRAM shortages in 2017. This in turn led to an increase in average selling prices (ASPs) for DRAMs, and resulted in significant stock gains for the memory companies.

I reported in a June 2, 2018 Seeking Alpha article entitled "Micron: First Price Fixing, Now Antitrust Allegations By The Chinese Government," the correlation between ASPs and stock prices for Micron Technology (MU) and I'll repeat the chart from that article below.

Chart 2

Memory companies reversed their equipment spend strategy in 2017 as memory spend increased 68.0% compared to only 13.9% for non-memory spend. Chart 3 compares Memory versus Non-Memory equipment spend.

Chart 3

Semiconductor Equipment Growth in 2017

The record growth in memory equipment spend in 2017 resulted in record breaking revenues for semiconductor equipment companies. I noted in a March 7, 2018 Seeking Alpha article entitled "Lam Research Rises In Semiconductor Etch Sector On Heels Of 3D NAND Memory," that the complexity of 3D NAND chip production was responsible for large revenue growth of processing equipment manufacturers in 2017 that will continue in to 2018.

Lam Research (LRCX) in 2017 gained market share against its main competitors Applied Materials (AMAT) and Japan's Tokyo Electron.

Chart 4 further illustrates the ramp in NAND semiconductor equipment spend in 2016, which increased from 42.5% of the global WFE equipment spend in 2015 to 60.6% in 2016 (decreasing to 56.9% in 2017). In contrast, DRAM spend decreased from 56.0% in 2015 to 36.4% in 2016 (recovering to 39.2% in 2017.

Chart 4

Effect of Semiconductor Equipment Spend on Memory Capacity

The time frame for building and equipping a new semiconductor fab or production line varies based on the size of the project. Typically, it takes about one year to construct the fab shell, another six months to install equipment, and another six months until production ramp.

But this is very important to the un-initiated. Equipment purchases do not necessarily result in increases in capacity. Yes, it's true, and here's why.

For DRAM production, the three companies have been migrating to smaller dimensions. Samsung is currently migrating to the 1ynm process, while SK Hynix and Micron Technology are switching to the 1xnm process, as shown in Chart 5.

Chart 5

These transitions are proving difficult to achieve with high yields. Also, these migrations increase the number of processing steps used to make the chip, resulting in what is termed a "natural decline" in wafer throughput. In general, movement from one node to the next results in a 5-10% decline in capacity.

To counteract this "natural decline," capacity needs to be increased, which is achieved by building new fabs and lines and purchasing equipment. Thus, the 60% increase in equipment purchases will not result in comparable increases in capacity.

Table 2 shows that CY 2018 DRAM capacity (in wafer starts) will increase between only 5.2% to 19.0% over CY 2017. In contrast, DRAM equipment spend increased 81% in 2017 from 2016.

Interestingly, MU is forecast to show the smallest growth in DRAM capacity in CY 2018. Yet I pointed out in the above SA article that its Operating Profit Margin increased 3,637% between FY 4Q 2016 and 2Q 2018 (equivalent to CY 1Q 2017 and 1Q 2018).

Table 3 shows that CY 2018 NAND capacity will increase only between 5.2% and 23.2% over CY 2017. NAND equipment spend in 2017 increased 58% over 2017.

NAND companies are moving from 2D NAND to 3D NAND, requiring increased number of processing steps like with DRAMs. Chart 6 shows that all NAND companies will move to 96 layers within the next 12 months.

Chart 6

Bit Growth

Capacity increases in wafer starts will be minimal in 2018 despite the huge increase in equipment spend in 2017, but because of the time lag between equipment purchases and production ramps, sending in 2017 will continue to result in capacity increases into 2019.

Bit growth in Gigabytes - Gb is different from capacity increases in wafers per month. The migration to the 1ynm process for Samsung (OTC:SSNLF) and 1xnm process for SK Hynix (OTC:HXSCL) and Micron, is resulting in increases in bit growth as dimensions get smaller and chips get more powerful.

Table 4 shows that DRAM capacity increase in millions of 1Gb equivalents will grow 19.6% between 2017 and 2018. These values are about 50% greater than the 12.2% change in wafer starts per month capacity increases (Table 2).

Table 5 shows that NAND capacity increase in millions of 8Gb equivalents will grow 44.6% between 2017 and 2018. These values are significantly greater than the 9.82% change in wafer starts per minute capacity increases (Table 3).

Clearly the increase in NAND equipment spend in 2016 benefited bit growth between 2017 and 2018, whereas the decrease in DRAM equipment spend in 2016 is resulting in delays in increasing bit growth.

Investor Takeaways

Companies are moving to tech migrations in DRAMs and NAND because of cost savings, as shown in Chart 7. According to Micron, DRAM cost improvements per Gb is greater than 25% for 25nm to 20nm migration and greater than 20% for 20nm to 1Xnm migration.

Chart 7

Clearly the increase in NAND equipment spend in 2016 benefited bit growth between 2017 and 2018, whereas the decrease in DRAM equipment spend in 2016 is resulting in delays in increasing bit growth.

DRAM shortages in 2017 resulted in exceptionally large increases in ASPs, which, in turn, led to large increase in stock prices and operating profit margins.

This begs the question: Are consumers rewarding DRAM companies for not spending money to increase capacity, which has led to led to enormous profits for Samsung, SK Hynix, and Micron?

The reduction in DRAM equipment spend in 2016 also resulted in a class action lawsuit in the U.S. against the three companies for price fixing, and an investigation by the Chinese government. These punitive measures led to short-term stock drops for the companies, but pricing resumed its upward path.

What has impacted stock valuations for Micron in particular, has been downgrades of the company by UBS and Morgan Stanley in the past month or so on expectation of supply increases in NAND and DRAM, which will drive down ASPs.

The record-breaking spending on equipment for NAND and DRAMs in 2017 appears unabated in 2018, which will further add capacity in 2019. SEMI, the semiconductor equipment industry consortium, issued a press release on June 4 that equipment billings reached a historic quarterly high of US$17.0 billion for the first quarter of 2018, surging 59 percent in March to end the quarter with an all-time monthly high of $7.8 billion.

Table 6 shows the quarterly billings data by region in billions of U.S. dollars, quarter-over-quarter growth and year-over-year rates by region. Korea, home to Samsung and SK Hynix, grew 78% YoY in 1Q2018. Much of that equipment spend will go into NAND and DRAM, because with the exception of Samsung's foundry, there is not much of a semiconductor industry in Korea outside memory.

In addition to capacity builds, in the next six months we will witness the introduction of DRAMs by China's JHICC, ramping up output as high as 20,000 wafers per month, and Innotron. China's XMC, a subsidiary of YMTC of Tsinghua Ungroup, is currently preparing to manufacture NAND flash memory chips, further exasperating NAND price decreases that started in Q1 2018 and initiating DRAM price decreases.

Increased capacity in memory chips will result in lower prices. Since they are directly tied to stock prices (Chart 2), lower stock prices should result. NAND prices have already started dropping in Q1 2018. DRAM price drops are estimated to begin in Q3 for Samsung and SK Hynix and Q4 for Micron.

MU, for example, is trading around $60 per share during a period of undersupply of DRAMs. Yet in mid-2016, during a period of oversupply of DRAMs, the stock was trading at under $10 per share.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Editor's Note: This article discusses one or more securities that do not trade on a major U.S. exchange. Please be aware of the risks associated with these stocks.