Behind the skyrocketing memory prices is an endless war you can’t see

On February 3, 2012, Steve Appleton, CEO of Micron Technology, a world-renowned memory manufacturer, was flying a Lancair IV-PT propeller plane at an air show in Boise, Idaho, U.S. Shortly after takeoff, the plane lost control, failed to make an emergency landing, and crashed directly to the ground. The CEO was 51 years old.

The CEO who loved to play with heartbeats died, and the stock price of Micron did not fall much. However, a Japanese named Yukio Sakamoto on the other side of the ocean was as anxious as an ant on a hot pot. Before boarding the plane, Appleton had just negotiated a secret agreement with him. For Elpida, a Japanese memory giant where Yukio Sakamoto was the president, this agreement was a life-saving book.

Elpida is the "national team" of Japan's chip industry. It was formed by the merger of the memory manufacturing businesses of Hitachi, NEC and Mitsubishi. It was once prosperous. However, after the financial crisis in 2008, the memory industry was oversupplied, product prices continued to fall, and coupled with the squeeze from Korean manufacturers, Elpida's business continued to deteriorate. Despite repeated rescue efforts by the Japanese government, it still could not be saved. By the end of 2011, Elpida had accumulated a huge amount of debt and losses.

Banks all over the world like to lend umbrellas when it is sunny and take them back when it is rainy. Elpida had a lot of creditors, and they all withdrew their loans. The government-backed Policy Investment Bank of Japan came forward to coordinate and put forward the only condition for renewing loans and injecting capital, which was to introduce a large memory manufacturer as a strategic shareholder before the end of February 2012, otherwise they would have to abandon it. At that time, there were only a few companies that were qualified to take over. Apart from the mortal enemy Korean, Micron was the only one left.

Yukio Sakamoto had been negotiating with Steve Appleton for several months and had reached an agreement on the key terms. However, with only 20 days left before the deadline, Appleton, who likes racing, skydiving, flying airplanes, and scuba diving, got hammered and the acquisition was postponed indefinitely. Yukio Sakamoto, who had no plan B at all, was in tears. On February 27, 2012, Elpida declared bankruptcy.

The fortune of an industry, like the so-called fortune of a country, often carries some strange randomness.

01

The year 2012, when Elpida went bankrupt, was an unlucky year for Japanese manufacturing. That year, Japan's electronics industry collapsed across the board. First, in addition to the bankruptcy of Elpida, another semiconductor giant, Renesas, also fell into crisis; second, the total losses of the three giants, Panasonic, Sony, and Sharp, reached a record high of 1.6 trillion yen; finally, the output value of the entire electronics industry was only about 12 trillion yen, less than half of the 26 trillion yen in 2000.

Japan's "lost decade" usually refers to the decade after the bubble burst in 1991. However, since 1991, two industries in Japan have continued to rise against the trend and are the key to supporting the yen exchange rate and foreign exchange reserves after the bubble burst. These two industries are automobiles and electronics. In 1991, NHK specially made a program called "Electronics Nation - Japan's Autobiography", which put the electronics and automobile industries on the same level and listed Sony, Panasonic and other companies, and the pride was beyond words.

2000 was a watershed year. In that year, the output value of automobiles and electronics was still evenly matched, but then the two went their separate ways. The automobile industry continued to expand overseas, and Toyota, Honda, and Nissan's global status steadily rose. However, the electronics industry went from bad to worse. Not only did its output value drop by half in 10 years, but in 2013, it even had a trade deficit. It was unimaginable that the once powerful Japanese electronics industry had exports less than imports.

However, the bankruptcy of Elpida, in addition to being related to the overall environment of Japanese manufacturing, was mainly caused by the roller coaster-like fluctuations in memory prices. Electronic products usually give people the impression that they will be reduced in price every year, and you can buy better things with the same amount of money every year. But the memory industry is fresh and refined, and the price trend is similar to that of chemical products, with strong cyclicality and big ups and downs. When the price goes up, you will tremble all over when counting the money, and when the price goes down, you will be so crazy that you will even cut yourself to death.

Strong cyclical industries usually have the following characteristics: 1. Products are highly standardized, and user stickiness is weak. They buy from whoever is cheaper; 2. The industry has economies of scale, and large-scale production can effectively spread costs; 3. They are asset-heavy, and depreciation is huge. Once production starts, it cannot be stopped. Even if it is at a loss, production must be carried out to at least have cash flow; 4. The industry structure is not stable, and there is no price alliance. When prices rise, manufacturers want to expand production crazily to kill their competitors. When prices are low, they can reduce production capacity through bankruptcy and mergers. Memory meets all of the above characteristics.

The official name of memory is "memory", which is one of the three pillars of the semiconductor industry. In 2016, the global semiconductor market size was 340 billion US dollars, and memory accounted for 76.8 billion US dollars. For all the electronic products around you, such as mobile phones, tablets, PCs, notebooks, etc., memory is similar to steel in modern industry, and it is a veritable "raw material" of the electronics industry. If the memory is further subdivided, it can be divided into DRAM, NAND Flash and Nor Flash. Among them, DRAM is mainly used for PC memory (such as DDR) and mobile phone memory (such as LPDDR), each accounting for 30%. Elpida makes DRAM.

After decades of cycles in the DRAM field, the number of players has gradually decreased from 40 to 50 in the 1980s to five before the financial crisis in 2008: Samsung (South Korea), SK Hynix (South Korea), Qimonda (Germany), Micron (USA) and Elpida (Japan). These five companies basically control the global DRAM supply. End product manufacturers such as Kingston have almost no DRAM production capacity and have to purchase raw materials from them. According to common sense, the situation has stabilized and the price war should have died down. Unfortunately, the Koreans did not agree, especially Samsung.

Samsung took full advantage of the strong cyclical characteristics of the memory industry and relied on government transfusions to expand production against the trend when prices fell, production was oversupplied, and other companies cut investment. It further cut product prices through large-scale production, thus forcing competitors to exit the market or even go bankrupt. People call this the "counter-cyclical law." In the field of memory, Samsung has used the "counter-cyclical law" three times. The first two times occurred in the mid-1980s and early 1990s, allowing Samsung to start from scratch and become the leader in memory. But Samsung obviously felt that it was not big enough, so around the 2008 financial crisis, it raised the "counter-cyclical" knife for the third time.

In early 2007, Microsoft launched the memory-hungry Vista operating system. DRAM manufacturers judged that memory demand would increase significantly, so they increased production capacity. As a result, Vista sales fell short of expectations, DRAM oversupply and prices plummeted. The 2008 financial crisis made matters worse, and DRAM chip prices plummeted from US$2.25 to US$0.31. At this time, Samsung made a jaw-dropping move: it invested 118% of Samsung Electronics' total profit in 2007 in DRAM expansion business, deliberately exacerbating industry losses and adding the last straw to its struggling competitors.

The effect was remarkable. DRAM prices plummeted, falling below cash costs in mid-2008 and below material costs at the end of 2008. In early 2009, the third-ranked German manufacturer Qimonda was the first to collapse and declared bankruptcy, and the memory players on the European continent disappeared. In early 2012, the fifth-ranked Elpida declared bankruptcy, and Japan, which once occupied more than 50% of the DRAM market, also lost its last card. On the night of Elpida's bankruptcy, Samsung's headquarters in Gyeonggi Province was lit all night, and the stock price soared the next day. The whole world knew that the Koreans had won again this time.

At this point, there are only three players left in the DRAM field: Samsung, Hynix and Micron. The mess after Elpida's bankruptcy was bought by Micron, which had a new CEO in 2013, for more than 2 billion US dollars. 2 billion US dollars is really a bargain price. Five years later, Micron's market value rose from less than 10 billion US dollars to 46 billion US dollars. 2 billion US dollars is almost the amplitude of its market value in one day.

The whole country is working hard to support a field, from being weak to overtaking others. Chinese sports can only be considered entry-level, and Korean talents are the masters.

02

After Elpida went bankrupt in early 2012, DRAM chip prices did not rise immediately, but continued to consolidate until the second half of the year, and then prices began to soar. By October 2013, DRAM prices were already twice as high as when Elpida went bankrupt. The three oligarchs had a good two years in 2013-14, and began to expand production again in 2015, causing a short-term oversupply. DRAM prices began to fall again, and fell until mid-2016, but this time, the decline was far from a serious blow to the three oligarchs.

In the second half of 2016, the newly released production capacity has been almost fully utilized, while market demand is still growing rapidly. Previously, 30% of DRAM production capacity was supplied to mobile phones. As mobile phone manufacturers started an "arms race" in memory, nearly 60% of DRAM production capacity was consumed by mobile phones. Especially in the third quarter, the peak season for mobile phone stocking, DRAM was in short supply and prices continued to rise. The price of memory sticks for PCs also rose accordingly, turning them into "financial products."

Recently, there was a joke like this: "In 2016, I opened an Internet cafe and bought more than 400 DDR4 8g memory sticks, each costing nearly 180 yuan. In 2017, the Internet cafe lost more than 100,000 yuan. Yesterday, I sold all the computers in the Internet cafe and sold the second-hand memory sticks for 500 yuan each. I actually made back the money I spent on opening the Internet cafe."

I don't know if the joke is true or not, but the price of memory sticks has been rising like crazy. Currently, a DDR4 8G desktop memory stick has risen to 900 yuan, while in mid-2016, the same product was only sold for about 200 yuan. A Weibo user calculated that currently an entry-level DDR4 16G memory stick = 28-inch 4K monitor = i5 processor motherboard set = 512GB 960EVO solid-state drive = Zotac overclocked GTX1060 graphics card. If the price continues to rise like this, I'm afraid I won't even be able to play "PlayerUnknown's Battlegrounds".

In addition to DRAM, NAND Flash, another field of memory, is also facing a similar situation. Players in the NAND Flash market include Samsung, Toshiba/SanDisk, Micron, and SK Hynix, which together account for 99% of the market share. Compared with the DRAM market, there is one more Toshiba/SanDisk. NAND Flash is mainly used in two fields, one is the flash memory of mobile phones, and the other is solid-state drives (SSDs). These two fields are both rapidly growing fields, driving NAND prices to soar.

The Nor Flash market is quite special. Although the price is also rising, the logic is different. In the era of feature phones, mobile phones did not have high requirements for memory. Nor Flash was widely used with the XiP architecture of NOR+PSRAM. But in the era of smart phones, a large number of memory-intensive apps emerged, and the shortcomings of NOR's small capacity and high cost were undoubtedly exposed. It was gradually replaced by NAND, and the market continued to shrink. Companies such as Samsung, Micron, and Cypress gradually withdrew from the NOR market.

But while major manufacturers shut down Nor production lines, Nor has enjoyed a second spring. The main reason is that AMOLED screens need a Nor Flash for electrical compensation. The penetration rate of AMOLED displays is accelerating, especially after Apple adopted it, so the demand for Nor has been boosted. However, companies such as Micron have already taken one step forward and are too lazy to reopen production lines, resulting in limited supply of Nor in the market. A group of small manufacturers have benefited from this, such as Taiwan's Macronix and Winbond, as well as the A-share star GigaDevice.

It should be said that the Nor market is too small, and the main battlefield of memory is still on DRAM and NAND, and the pattern of these two fields is already very stable, with Samsung, Micron, Hynix, and Toshiba, which is selling itself everywhere. It is already difficult for Samsung to eliminate its opponents through the "counter-cyclical law". Since it cannot eliminate its opponents, the suicidal charge of self-destruction will not happen again. The biggest possibility is: memory prices will fluctuate in short cycles with changes in supply/demand, but the industry will maintain a state of huge profits for a long time.

The problem is, how can we Chinese be missing from such a profitable and fun game?

03

In the early spring of 2012, when Elpida went bankrupt, Zhao Weiguo of Tsinghua Unigroup was working on a real estate project called Hagu Villa in Putian, Fujian. The huge changes in the memory field seemed to have nothing to do with him. I'm afraid he himself would never have thought that five years later, he would shout out the slogan "become one of the top five in the global memory industry within 10 years."

But the situation changed in the second year. In 2013, Tsinghua Unigroup acquired Spreadtrum, entering the semiconductor field for the first time. In 2014, without the "small road permit" from the National Development and Reform Commission, it snatched RDA Microelectronics from Pudong Science and Technology Investment. Zhao Weiguo put the two companies into a holding company, named "Unigroup Spreadtrum", and successfully attracted Intel to invest $1.5 billion, with a post-investment valuation of $7.5 billion. Compared with the $2.7 billion spent on the acquisition of the two companies, Zhao Weiguo made the biggest pot of gold since he became rich.

In 2014, when Tsinghua Unigroup completed its acquisition of RDA Microelectronics, the atmosphere in the industry suddenly became hot and agitated, but few people realized that 2014 would be called the "first year" of China's semiconductor industry. Prior to this, more than a dozen academicians and experts wrote to the central government, asking the country to fully support the development of semiconductors, and received a positive response from the top level. In September, a national fund with a scale of 100 billion yuan was established. In the next three years, it will completely change the ecology of China's and even the world's semiconductor industry.

The memory industry, which imports tens of billions of dollars every year and has almost zero localization, naturally first entered the vision of various capitals. But even a domineering person like Zhao Weiguo did not dare to directly challenge Samsung Hynix, which was standing on countless corpses. Instead, he took a roundabout way and launched a takeover offer for the weaker Micron, offering $23 billion, a premium of 19%. If the acquisition is successful, according to Micron's closing price yesterday, Micron's market value will exceed $46 billion, and the floating profit of this transaction will more than double.

This is not the first time that Chinese capital has attempted to acquire overseas semiconductor companies. As early as 2012, Lenovo's Hony Capital once joined TPG to bid for Elpida after it went bankrupt. If the acquisition was successful, it would have been of great significance to Lenovo, which lacked core technology and components, but unfortunately it failed. Since then, Hony has never tried to make industrial chain mergers and acquisitions around Lenovo's main business, and the investment cases that often jump out of the eye are Pizza Express and Xishaoye Roujiamo.

Tsinghua Unigroup's tender offer for Micron was unsurprisingly blocked by the US government, and Zhao Weiguo did not make a floating profit of 20 billion US dollars. Of course, Micron also went to great lengths to assure the outside world that it would "not build a wafer fab in the mainland within three years", which undoubtedly angered Zhao Weiguo and others who had huge funds. Since shortcuts were not feasible, they could only start from scratch. As a result, the three major memory projects of Yangtze Memory, Hefei Changxin, and Fujian Jinhua broke ground, and the north and south of the Yangtze River were working hard to catch up with the UK and the US.

Samsung's "counter-cyclical law" may be difficult to replicate in other countries, but in mainland China, it will only be played more skillfully. For example, Samsung and LG successfully defeated the Japanese panel industry through counter-cyclical investment. The mainland government followed suit and finally created BOE through years of transfusions and subsidies. Therefore, in the field of memory, mainland China also has road confidence, but compared with the panel industry, the technical threshold of memory projects is more difficult and the opponents are more powerful.

In response to the hot market and the entry of mainland capital, Samsung, Hynix and Micron have started a new round of expansion. In this round of surge, these companies have also stockpiled enough food and ammunition to welcome the entry of new players. It is foreseeable that the day when the mainland memory project reaches full production will be the day when memory will once again fall to cash cost or even material cost, which may be around 2019. By then, those who are now complaining that memory sticks are too expensive and cannot play PUBG smoothly may wish to buy a few more sticks and store them.

Someone once revealed a rule: many industries were first invented by Americans. After the Americans made a lot of money, the industries basically circled around Japan, South Korea, mainland China, and Taiwan. These four players are like playing a mahjong game, fighting each other until they bleed, but they take turns to win, and in the end they make money from people all over the world outside the mahjong circle. This rule is probably most applicable to the electronics industry.

Back to the beginning of this article, Yukio Sakamoto, the last president of Elpida, did two things after Elpida went bankrupt. One was to write a book called "Unwilling Defeat", the title of which reflects the content. The second was to come to China and try to make a comeback with Chinese funds. Yukio Sakamoto founded a DRAM design and development company called Sino King Technology. Sino = China, King = King. The first partner SKT found in China was the Hefei Changxin Memory Project, which cost US$7.2 billion. But obviously, local governments with capital have passed the era of listening to foreign experts. Against the backdrop of the rise of local talents and the influx of overseas talents, the 70-year-old Yukio Sakamoto, representing the last generation of semiconductor talents in Japan, may not be their first choice.

Therefore, whether it is the fate of the industry or the fate of the country, it is both accidental and inevitable, cyclical and reincarnated. The industrial grievances among China, Japan and South Korea have not yet come to an end.

As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity.