Trend: The dawn of memory semiconductors

Source: This article is translated from "eetimes.jp " by the public account Semiconductor Industry Observer (ID: icbank), thank you.

Since the advent of "Big Data" in 2016, the memory semiconductor market has experienced explosive growth (Figure 1). The buzzword "Super Cycle" was born in 2017. How long can the explosive growth of memory semiconductors continue? - In the second half of 2018, the memory semiconductor bubble burst and fell into a slump.

The author believes that the main reasons for the sluggish storage semiconductor market are: the continued failure of Intel's 10nm process has led to a shortage of processors for PCs and servers. However, storage semiconductors that depend on PCs and servers are flooding the market, ultimately causing prices to plummet.

Figure 1: Quarterly shipments of each semiconductor. (Image source: Table created by the author based on WSTS data.)

As a result, the so-called "Super Cycle" can be said to be a cycle of "super" "up and down" in the memory semiconductor market. In other words, as a part of the "Silicon Cycle" that has always been well known, the memory semiconductor market has experienced ups and downs in recent years.

In this way, there will be a boom one day, and a recession will come sooner or later. I believe that the reason for the poor memory semiconductor market is Intel, whose head once said, "So far, the shipment of our company's 10nm processors has been delayed. We will deliver them as scheduled in June 2019 according to the plan announced in 2018."

Then, the storage semiconductors in Figure 1 are mainly DRAM and NAND flash memory (hereinafter referred to as "NAND"). The author made Figure 1 based on the data released by the World Semiconductor Trade Statistics Organization (WSTS). From Figure 1, it can be seen that the trends of the two are very different. This article will explain the difference and analyze the reasons. It will also look forward to the future storage semiconductor market.

Figure 2 shows us the shipment amounts of DRAM and NAND in each quarter. When I made this table, I fell into deep thought because the market trends of the two memory semiconductors were very different.

Figure 2: Changes in DRAM and NAND shipments per quarter. (Image source: Table created by the author based on WSTS data.)

DRAM sales peaked in the third quarter of 1995, the third quarter of 2000, the fourth quarter of 2006, the third quarter of 2010, the fourth quarter of 2014, and the third quarter of 2018. A peak occurs almost every 4-5 years, which truly reflects the "Silicon Cycle".

In addition, except for the third quarter of 2018, NAND has hardly seen any high peaks. From the second quarter of 2016 to the rapid growth period, the market has expanded almost in a straight line. In other words, NAND does not have the "Silicon Cycle" phenomenon like DRAM.

Why are the market trends for DRAM and NAND so different?

The following is an explanation of the "Silicon Cycle" of DRAM. For example, with the release of Windows 95 in 1995, the emergence of the "IT bubble" in 2000, and the popularization of "Big Data" from 2016 to 2017, the demand for servers has increased dramatically. At this time, the demand for DRAM exceeded the supply, and the price of DRAM continued to rise. As a result, the shipment amount of DRAM continued to increase, ushering in a boom period.

At this time, although memory semiconductor manufacturers increased DRAM production, once supply exceeded demand, prices would fall. From 2016 to 2017, although DRAM production did not increase in large quantities, due to insufficient supply of processors, there was a phenomenon of DRAM surplus and plummeting prices.

In other words, although the situation may vary, if supply exceeds demand, prices will fall and memory semiconductors will fall into depression. This "Silicon Cycle" phenomenon exists in the DRAM market.

Figure 3 is a chart showing the annual changes in DRAM shipment value and shipment quantity. In terms of DRAM shipment quantity, it can be divided into the following three major periods.

(1) 1991-2003: A period of slow growth in shipment volume.

(2) 2003-2010: A period of rapid growth in shipment volume.

(3) After 2010: Annual shipments remained almost stable at 15 billion units.

Figure 3: Annual changes in DRAM shipment value and shipment volume. (Image source: Table created by the author based on WSTS data.)

The growth rate of DRAM shipments has shown a great difference since 2003. The main reason for this is the economic rise of Asian countries, led by China, since the beginning of the 21st century. In other words, in Asian markets such as China, people generally buy PCs, mobile phones, digital home appliances and other products, so it is estimated that the market for DRAM used in these products has also expanded rapidly. We can also draw the same conclusion from the fact that the Asian semiconductor market has expanded rapidly since 2001 (Figure 4).

Figure 4: Quarterly changes in semiconductor shipments by region. (Image source: table created by the author based on WSTS data.)

After 2000, DRAM shipments remained at almost 15 billion units per year, due to the "centralization" of multiple DRAM manufacturers. In particular, in 2012, Elpida went bankrupt and was acquired by Micron Technology of the United States. In the end, there were only three DRAM manufacturers left: Samsung Electronics, SK Hynix, and Micron. As a result, DRAM shipments remained roughly the same because the three companies had "tacit consultations" and adjusted production.

Now let's go back to Figure 2 to look at the market trend of NAND. Before 2016, NAND did not have a "Silicon Cycle" like "DRAM". Here, like DRAM, an annual chart of NAND shipment value and shipment quantity was made, and NAND shipment quantity grew almost linearly (Figure 5). However, since 2016, when it entered the "Super Cycle", its growth seems to be slowing down.

Figure 5: Annual changes in NAND shipment value and shipment volume. (Image source: Table created by the author based on WSTS data.)

First, let's think about the reasons for the continuous increase in shipment value and shipment volume from 2000 to 2016. NAND is a non-volatile memory semiconductor invented in 1987 by Mr. Fujio Masuoka, who was formerly employed by Toshiba, but NAND was not used in Toshiba's business reform in the 1990s.

However, by 2000, large-scale application markets such as digital cameras, Apple's music player "iPod", and mobile phones emerged. In addition, the iPhone was launched in 2007, opening the door to smartphones, and the HDDs of PCs and servers began to be replaced with SSDs, further expanding the market size.

That is to say, as NAND continues to develop applications, the market size continues to expand, so its demand has been rising, so there has not been a price plunge like DRAM. In addition, even if the price falls, new applications will drive demand growth, so the shipment amount and shipment quantity are showing a linear growth.

After 2017, the growth of NAND shipments began to slow down, which may be due to the transition of NAND from 2D to 3D. Since the production technology of 3D NAND is extremely difficult and the yield rate is difficult to improve, 3D NAND is likely to fail to ship as planned.

At the same time, like DRAM, NAND shipments also experienced a leap in value after 2016, but entered a recession in the second half of 2018. The reason was the same as DRAM, which was due to insufficient supply of Intel processors. In other words, because the supply of NAND exceeded its demand, prices plummeted and shipments dropped sharply. As a result, NAND experienced its first "Silicon Cycle" in history.

Regarding DRAM, the market is actually dominated by three memory semiconductor manufacturers. In the future, they will probably conduct "tacit consultations" and will not increase the number of DRAM shipments too much, and will slowly advance the DRAM business. However, "accidents" like this "Super Cycle" will probably appear again in the future. Therefore, it is expected that the "Silicon Cycle" will appear repeatedly in the DRAM market in the future.

In addition, China is also trending towards joining the DRAM market. The United States has included JHICC (Fujian Jinhua) in the Entity List (EL), and American companies such as Applied Materials, Lam Research, and KLA are prohibited from exporting equipment produced in the United States, which has led to setbacks in the development and production of DRAM.

In other words, the United States is "defending" by setting up a "breakwater" to "destroy the Chinese DRAM market." However, there is no guarantee that this "breakwater" will have a permanent effect. China is promoting the localization of production equipment and materials. Once successful, Chinese-made DRAM is likely to sweep the world.

On the other hand, NAND, which is experiencing the "Silicon Cycle" for the first time, has become a storage semiconductor with drastic market fluctuations, just like DRAM, depending on supply and demand. It can even be said to be a "grown-up storage semiconductor." Therefore, it is expected that the shipment volume of NAND will increase significantly in the future, and the shipment amount will fluctuate periodically.

In short, DRAM and NAND-related equipment and material manufacturers are paying more attention to marketing in order to predict the "Silicon Cycle" in advance.

Yunokami TakashiDirector of Microfabrication Research Institute

Born in Shizuoka Prefecture, Japan in 1961. After graduating from Kyoto University (major: nuclear industry), he joined Hitachi, Ltd. In the following 16 years, he worked in the Central Research Institute, the Semiconductor Division, Elpida Storage Semiconductor (external assignment), and Semiconductor Advanced Technology (external assignment) to engage in the research and development of semiconductor micro-processing technology. In 2000, he obtained a Bosch degree from Kyoto University. Currently, he is the director of the Micro-Processing Institute and works as a consultant and editor for companies related to the semiconductor and electrical industries. His works include "The Defeat of Japan's "Semiconductors", "Lessons from the Failure of Electrical and Semiconductors", and "The Defeat of Japanese-style Manufacturing, Zero Carrier-based Fighters, Semiconductors, and Televisions".

*Click on the end of the article to read the original text.