NAND And DRAM: An Upturn Is Closer Than We Think

by: Robert Castellano

Projections of short-term oversupply of DRAM and NAND memory chips are impacting memory companies like Samsung Electronics, Micron Technology and SK Hynix.

The oversupply of memory chips is resulting in decreased average selling prices, but price erosion will improve in Q4 2018 for NAND and Q1 2019 for DRAM.

Although chip prices are dropping, bit growth of the memory chips is resulting in high revenues and margins for manufacturers.

Although demand for smartphones is flat, and capex spend for cloud servers is slowing, increased memory content per device is growing strongly.

Semiconductor equipment companies Applied Materials and Lam Research have seen stock prices plummet because of push-outs of purchases by memory companies, which will continue in 2019.

I reported in several Seeking Alpha articles that a drop in ASPs was coming for NAND and DRAM memory chips. For example, I reported In a December 13, 2017, Seeking Alpha article entitled “The Memory Sector Is Distorting Overall Semiconductor Industry Growth In 2017” that both NAND and DRAM ASP growth began dropping in Q2 2017.

I also noted in the article that these chips:

“...will drop into negative growth through 2018 for both DRAM and NAND.”

I projected in the charts that ASPs would turn to negative growth in Q1 2018 for both companies for both NAND and DRAM. In reality, NAND growth turned negative in Q4 2017 for Samsung Electronics (OTC:SSNLF) and Q1 2018 for SK Hynix (OTC:HXSCL). I’ve tabulated inflection points in Table 1.

While we won’t know Q4 2018 DRAM prices for both companies for another six weeks, confidence is high the QoQ ASP change will be negative. For example, ASPs dropped from 2.5% in Q2 2018 to 0% in Q3 for Samsung, and from 2.7% in Q2 to 1.2% in Q3 for SK Hynix.

I was off by three quarters for DRAMs because I neglected to take into consideration the fact that the DRAM industry is an oligopoly. That is, the three major DRAM companies Samsung, SK Hynix and Micron Technology (MU) could adjust supply to stall the drop in ASPs to negative growth QoQ.

In this article, I wanted to update data as well as forecast ASPs through 2019. I'm also including an analysis of NAND and DRAM from Micron Technology.

In Chart 1, I present data for Samsung, forecasting ASPs through CY 2019.

Chart 1

In Chart 2, I present data for SK Hynix, forecasting ASPs through CY 2019.

Chart 2

In Chart 3, I present data for Micron Technology, forecasting ASPs through FY 2019.

Chart 3

In all three charts, we note that the trend is similar - NAND ASP changes were more pronounced than DRAM. Also, differences among the companies are due to customer base and product mix.

Chart 4 presents the same data from Chart 1 for Samsung, giving a different prospective on the QoQ changes. Based on this analysis, NAND ASP changes already reached their lowest point in Q3 2018 and we will witness an upturn beginning in Q4 2018. ASPs changes will continue to be negative through 2019, but QoQ changes will be less significant.

DRAM ASP changes will reach their lowest point in Q1 2019 for Samsung, and an upturn will begin in Q2. In Q3, I forecast that DRAM ASP changes will reach 0% QoQ in Q3.

Chart 4

Chart 5 presents the same data from Chart 2 for SK Hynix, giving a different prospective on the QoQ changes. NAND ASP changes will reach its lowest point in Q4 2018 and we will witness an upturn beginning in Q2 2019. ASPs changes will continue to be negative through 2019, but QoQ changes will be less significant.

DRAM ASP changes will reach their lowest point in Q1 2019 and an upturn will begin in Q2. In Q3, I forecast that DRAM ASP changes will reach 0% QoQ in Q4.

Chart 5

Chart 6 presents the same data from Chart 3 for Micron Technology, giving a different prospective on the QoQ changes. NAND ASP changes will reach its lowest point in Q1 FY2019 ending November 2018, and we will witness an upturn beginning in Q2 FY 2019. ASPs changes will continue to be negative through 2019, but QoQ changes will be less significant.

DRAM ASP changes will reach their lowest point in Q2 FY2019 (ending February 2019 and an upturn will begin in Q3.

Chart 6

Basis for Analysis

The above analysis for ASP changes is based on several trends and forecasts that point to continued strong demand for memory devices, which deviates from a traditional cyclical nature of the semiconductor industry. These are individually detailed below.

Memory Capex Spend is Slowing

I discussed in an October 17, 2018, Seeking Alpha article entitled “The Disconnect Between Memory Chip Output And Processing Equipment Sales: Part 2, Equipment,” that despite a huge amount of capex spend by the memory manufacturers, there was little effect on the number of NAND and DRAM chips shipped, as shown in Chart 7.

Chart 7

Semiconductor equipment capex spend is slowing down because of pushouts and cancellations of equipment by the memory companies, as illustrated in Table 2. DRAM capex spend is forecast to decrease 33% YoY in 2019. Remember the three major DRAM companies are an oligopoly, and by adjusting capex, they can adjust the supply of chips. Reduced capex spend also increases the company’s bottom line. NAND spend is forecast to increase just 4%.

This slowdown in equipment spend will result in a slowdown in bit growth of DRAMs and NAND, and have the effect of decreasing supply while demand continues to increase. It also will have a negative impact on equipment companies with a high exposure to memory processing, such as Lam Research (LRCX) and Applied Materials (AMAT).

Bit Growth

Although capex spend did not effectively increase the number of memory chips shipped, capex spend on processing equipment and cleanrooms did have the effect of increasing bit growth, as shown in Chart 8. Data are for Micron Technology. DRAM bit growth increased 52% in FY 2017 ending August and 20% in FY 2018. NAND bit growth increased 66% in FY 2017 and 44% in 2018.

Chart 8

For CY 2018, bit growth for Samsung and SK Hynix are essentially the same as MU. Table 3 tabulates forecasts of bit growth, ASP change, and margin.

Increased Memory Content of Smartphones

Memory content in electronics has been increasing over the years. DRAM content in smartphones grew 47% in 2016 because of the oversupply in memory chips and lower ASPs, enabling end users to pack more memory at a lower cost. As memory ASPs increased in 2017 and early 2018, end-user customers continued to demand more memory. DRAM content increased 24% in 2017. DRAM content in GigaBytes per smartphone is shown in Chart 9.

Why the increase in memory despite higher ASPs? If you have a smartphone with 4GB of RAM, with an average memory usage of around 2.3GB, it can hold 47 apps in that memory. A 6GB can operate well over 60 apps in your memory at any given moment of time. Smartphones shipments are flat, growing only 1.1% HoH in 2018. Nevertheless, increased content of memory averaging 20% per year will be a strong tailwind in 2018 and 2019.

Chart 9

In Chart 10 I've plotted DRAM content of the 2018 flagship smartphones compared with DRAM content for flagship smartphones going back to 2014. Here we can see a consistent theme that DRAM content is increasing each year, with some reaching 8GB in 2018.

Chart 10

Chart 11 shows the NAND content. Here we can also see a consistent theme that NAND content is increasing each year, albeit not reaching the 512 GB level of the iPhone XS.

Chart 11

Increased Memory Content of Cloud Serverss

Cloud capex is slowing, but only after stratospheric spend by the top service providers for the past two years of nearly 50%, including 62% year-to-date 2018 over last year's Q1-Q3. Although cloud capex spend is expected to slow in 2019, an increased amount of DRAM per server will lessen the impact of lower revenues for Micron Technology. Opportunities for Micron Technology beyond 2019 exist not only with DRAMs, but with its highly-touted 3D XPoint technology due to Google's migration to the chip.

Chart 12 shows that the amount of DRAM per server is increasing, mitigating a slowdown in cloud capex. Thus, while cloud capex is forecast to increase 16.2% in 2019, GigaBytes per server is forecast to grow 27.0% in 2019.

Chart 12

Investor Takeaway

A major problem with the memory market going into 2019 is that 2017 and 2018 were too good. As a result, investors and analysts extrapolate meteoric growth in ASPs, demand, and stock prices and wrongly think that the trend will continue ad infinitum. If there are any perturbations in this straight-line growth, mass hysteria results by analysts and their underperform signals and by MU longs reading my articles.

In the server market, for example, capex spend increased a robust 87.6% growth YoY in Q1 2018 and Q2 growth increased 83.9%. However, Q3 capex grew just 46.6% YoY. For the first three quarters, YoY growth is 65.2%. Analysts are now freaking out because of lower capex guidance from server companies and thus there's only a 17% YoY consensus for cloud capex spend.

Negative news such as servers or poor iPhone sales are negatively impacting the market because analysts haven’t explained to clients that indeed server capex may be slowing, but memory demand per server is outweighing this slowdown. The same holds true for smartphone demand.

Memory chips are unique to this industry, and should not be pigeonholed along with other semiconductor devices. For example, an iPhone has one processor, the A12 system on a chip manufactured by TSMC. Yes it gets better and faster with each new iPhone, but there's only one processor. Memory content of an iPhone, on the other hand, is increasing with each new iPhone. In addition to data in Chart 9, Charts 10 and 11 point to the increased DRAM and NAND content over the past five years.

But the real key to continued demand for memory chips is that the memory companies have started reigning in capital expenditures to avoid adding more manufacturing capacity than customers need. Referring to Table 2, DRAM capex spend is projected to drop at least 33% in 2019. NAND capex spend will increase only 4%. This action, although bad for semiconductor equipment manufacturers such as Applied Materials and Lam Research, will have a positive effect on DRAM and NAND ASPs.

Stock value of Micron was intimately tied to ASPs in 2017 and 2018. As ASPs have dropped, so too has the stock. Changing the supply-demand dynamics by adjusting capex spend and bit growth will reduce supply. Unfortunately, memory companies are walking a tightrope on two fronts:

  1. Reducing output will reduce chip shipments while increasing ASPs. The balancing act comes with maximizing operating margins in the face of declining ASPs and reduced shipments.
  2. The Chinese have been investigating Samsung, SK Hynix, and Micron for DRAM price fixing for the past six months. If DRAM supply is overtly constrained, it will add to the evidence the Chinese government already has. In light of the U.S. Commerce action against China’s Fujian, and the current trade war, this could become an issue in government discussions on the trade war.

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.