On September 12, 1958, Jack Kilby from Texas Instruments successfully integrated five components, including germanium transistors, to create the world’s first germanium integrated circuit.
The following July, Robert Norton Noyce from Fairchild Semiconductor invented the world’s first silicon integrated circuit based on silicon planar technology.
Jack Kilby (left) and Robert Noyce (right)
As we all know, the inventions of these two giants hold significant importance. The emergence of integrated circuits has effectively promoted the miniaturization of electronic devices and laid the foundation for the comprehensive arrival of the chip era.
█ The Birth of DRAM
Entering the 1960s, with the development of computer technology, the electronics industry began to attempt to apply integrated circuit technology in the field of computer storage.
At that time, semiconductor storage technology was divided into two directions: ROM and RAM. ROM is read-only memory, which retains data even when the power is off, also known as external storage. RAM is random access memory, used for storing operational data, which will be lost when the power is off, also known as memory.
Today, we will focus on the RAM field.
In 1966, Robert H. Dennard from IBM Thomas J. Watson Research Center first invented DRAM memory (Dynamic Random Access Memory).
Robert H. Dennard
This type of memory is based on a “MOS transistor + capacitor structure,” featuring low energy consumption, fast read/write speed, and high integration. Even today, our computer memory, mobile phone memory, graphics card memory, etc., are all based on DRAM technology.
In June 1968, IBM registered a patent for transistor DRAM. However, just as they were preparing for the industrialization of DRAM, the U.S. Justice Department launched an antitrust investigation against them.
These investigations delayed IBM’s progress in DRAM industrialization, thus providing opportunities for other companies.
Shortly after, in 1969, Advanced Memory System in California, USA, took the lead and successfully produced the world’s first DRAM chip (with a capacity of only 1KB), selling it to the computer manufacturer Honeywell.
Upon receiving this batch of DRAM chips, Honeywell discovered some issues with the process. Therefore, they sought help from a newly established company.
This company was Intel (Intel), co-founded in 1968 by Robert Noyce (the inventor of silicon integrated circuits mentioned earlier) and Gordon Moore (the proposer of Moore’s Law).
Robert Noyce (left) and Gordon Moore (right)
After Intel was established, its main business was the research and development of transistor semiconductor storage chips.
At that time, semiconductor technology primarily had two research directions: bipolar transistors and field-effect (MOS) transistors. Intel itself did not know which direction was correct, so it established two research teams to follow up on both technological paths.
In April 1969, the bipolar team made a breakthrough first, launching the 64-bit static random access memory (SRAM) chip—C3101. This chip was Intel’s first product, with Honeywell as its main customer.
Intel C3101
The field-effect transistor team was not to be outdone; in July 1969, they launched the 256-bit static random access memory chip—C1101. This was the world’s first large-capacity SRAM.
In October 1970, the field-effect transistor team continued to perform well, successfully launching their first DRAM chip (also considered the world’s first commercially mature DRAM chip)—C1103.
Intel C1103, with 18 pins, a capacity of 1Kbit, priced at 10 USD.
After the launch of C1103, it achieved great success and quickly became the best-selling semiconductor memory globally, serving important customers such as HP and DEC.
With the help of C1103, Intel rapidly grew. By 1972, Intel’s employee count exceeded 1,000, and annual revenue surpassed 23 million USD. By 1974, Intel’s global market share for DRAM products reached an astonishing 82.9%.
Intel’s early team
While Intel was profiting immensely in the DRAM field, its competitors were also rising rapidly.
In 1973, companies such as Texas Instruments (TI) and Mostek entered the DRAM market one after another.
After Intel launched C1103, Texas Instruments disassembled and replicated it, studying the architecture and processes of DRAM through reverse engineering. Later, in 1971 and 1973, they successively launched 2K and 4K DRAMs, becoming strong competitors to Intel.
Texas Instruments, Intel’s old rival
Mostek was founded by L.J. Sevin, a former chief engineer at Texas Instruments Semiconductor Center (1969), and also had strong technical capabilities.
In 1973, they launched a 16-pin DRAM product—MK4096, which also posed a challenge to Intel’s market position (other companies had 22 pins; fewer pins meant lower manufacturing costs).
In 1976, Mostek launched MK4116, which adopted POLY-II (double-layer polycrystalline silicon gate) technology, reaching a capacity of 16K. This product achieved great success, dramatically reversing the market competition pattern and raising its DRAM market share to 75%.
MK4116
Unfortunately, not long after, due to malicious acquisitions from the capital market, Mostek’s equity structure underwent significant changes, resulting in severe management turmoil, and technical personnel rapidly left, causing the company to quickly decline.
In 1979, the company was acquired by United Technologies Corporation (UTC) and later sold to STMicroelectronics.
In October 1978, four technicians from Mostek left and founded a new memory technology company in the basement of a dental clinic in Idaho.
This company later became the storage giant—Micron.
The founding team of Micron
█ The Rise and Fall of Japanese Semiconductors
Besides domestic competitors, Intel faced a greater threat from abroad, specifically from Japan.
In the 1970s, Japan’s economy rapidly rose. To secure a favorable position in the global technology industry chain, they carefully laid out their semiconductor technology.
In 1976, Japan established the VLSI Consortium through a national system (VLSI: THE VERY LARGE SCALE INTEGRATED).
The consortium had six laboratories dedicated to research in high-precision processing technology, silicon crystallization technology, process handling technology, monitoring and evaluation technology, and device design technology.
Not long after, this consortium successfully overcame core semiconductor processing equipment such as electron beam lithography machines and dry etching devices, as well as leading process technologies and semiconductor design capabilities, laying the foundation for the rise of the Japanese semiconductor industry.
In 1977, with the help of the VLSI project, Japan successfully developed the 64K DRAM, catching up with the R&D progress of American companies.
By the 1980s, Japanese manufacturers (Fujitsu, Hitachi, Mitsubishi, NEC, Toshiba, etc.) continued to gain momentum, leveraging quality and price advantages to surpass American companies.
In 1986, Japan’s market share of memory products rose to 65%, while that of the U.S. dropped to 30%.
In the fierce market competition, American Intel directly announced its withdrawal from the DRAM market (in 1985). The only company that could survive among the Japanese manufacturers was Motorola.
Global semiconductor company rankings (1987)
As the Japanese semiconductor manufacturers were about to dominate the market, subtle changes began to occur in the external political environment.
In 1985, the atmosphere of the U.S.-Soviet Cold War weakened, and U.S.-Japan trade frictions increased. Under the immense pressure of fiscal deficits, the Reagan administration began to focus on suppressing the Japanese economy.
That year, the U.S. led the famous Plaza Accord, forcing the yen to appreciate. Meanwhile, the American Semiconductor Association initiated anti-dumping lawsuits against Japanese semiconductor products. Later, the two countries reached a price supervision agreement on Japanese semiconductor products.
Amidst a series of setbacks, the market share of Japanese semiconductor products plummeted, and they quickly lost their dominance.
█ The Rise of Korean Semiconductors
So, did the market share given up by Japanese manufacturers go to American manufacturers?
No.
As the saying goes, “The mantis stalks the cicada, unaware of the oriole behind.” While Japanese manufacturers quickly fell from grace, another competitor from the U.S. emerged, namely—Korea.
When Japan launched the VLSI project, the Korean government was also busy. They established the Korea Electronics Technology Institute (KIET) in the Gumi industrial area of North Gyeongsang Province, offering high salaries to attract American semiconductor talent and focusing on developing key integrated circuit technologies.
In addition to KIET, Korean conglomerates such as Samsung, LG, Hyundai, and Daewoo also recognized the market prospects of semiconductor technology. They acquired, introduced technology patents and processing equipment, and digested and absorbed them to build up technological strength.
In 1984, Samsung Semiconductor established its first memory factory, mass-producing 64K DRAM. No one expected that this relatively unknown Korean company would become the future industry “giant.”
From the 1980s to the present, the DRAM industry has undergone nearly forty years of development. If one word were to describe these forty years, it would be—“bloody and fierce”.
The reason is simple: the biggest characteristic of the DRAM semiconductor industry is its cyclical nature. Industry insiders have summarized: DRAM semiconductor storage makes a profit for one year and loses for two years, hence the saying “earn one, lose two.”
Under such strong cyclical rules, it is very difficult to survive in the long term. DRAM manufacturers need strong cash flow and financing capabilities, capable of maintaining high-intensity R&D expenditures and stabilizing their teams.
During loss periods, DRAM manufacturers need more money to survive. During prosperous periods, they must also be cautious. Manufacturers need to be very careful when choosing the timing to expand production capacity. Otherwise, it could lead to oversupply, turning profits into losses.
Forty years ago, there were about 40-50 DRAM manufacturers globally. Today, only three remain, highlighting the brutal competition.
Over these forty years, one company not only survived but also eliminated countless competitors, maintaining a dominant position. This company is the aforementioned Samsung.
Samsung Electronics
Many may have heard of Samsung’s story. They adopted a “killer strategy” that has been documented in numerous business school textbooks—counter-cyclical investment.
In simple terms, counter-cyclical investment means utilizing the cyclical development characteristics of the industry. When the industry enters a downturn and competitors are shrinking, one goes against the trend, increasing investment, expanding production capacity, and further driving down prices, thereby intensifying competitors’ losses and even leading to their bankruptcy.
In other words, while everyone is facing destruction, I have more money, and if I burn you to death, I can continue to survive.
Samsung, with the backing of the Korean government, repeatedly adopted the “counter-cyclical investment” strategy, eliminating countless competitors and becoming the leader in the semiconductor storage field.
Next, let’s take a detailed look at what has happened over these decades.
-
First “Counter-Cyclical Investment”
Samsung’s first “counter-cyclical investment” occurred in the mid-1980s mentioned earlier.
At that time, the U.S. and Japan were in fierce competition, and the DRAM market was generally sluggish, with prices plummeting. The price of DRAM chips fell from 4 USD per chip (in 1984) to 0.3 USD per chip (in 1985).
When Samsung built its factory and launched 64K DRAM, its production cost was 1.3 USD/chip. In the face of an industry winter, Samsung not only did not reduce investment but instead began to invest in reverse, expanding production capacity.By the end of 1986, Samsung Semiconductor had accumulated losses of 300 million USD, and its equity capital was completely depleted, nearing bankruptcy.At this critical moment, the Korean government intervened to “rescue the market,” investing nearly 350 million USD in total and backing Samsung, bringing in 2 billion USD in private fundraising.Later, Japanese semiconductors were defeated by the U.S., and with the booming demand brought by the PC market, Samsung successfully turned the tide and welcomed significant growth.Shortly after, Korean DRAM manufacturers represented by Samsung gradually seized the market share vacated by Japanese semiconductor companies, dominating the market.
-
Second “Counter-Cyclical Investment”
In 1992, an explosion occurred at Japan’s Sumitomo Chemical Plant, leading to a tight supply of raw materials and a surge in memory prices. That year, Samsung was the first to launch the world’s first 64M DRAM.
In 1993, the global semiconductor market began to weaken again. At this time, Samsung resorted to its old trick and implemented its second “counter-cyclical investment.” They invested in building an 8-inch silicon wafer production line for DRAM production.
In 1995, Microsoft released the Windows 95 operating system, which greatly stimulated memory demand, driving memory prices up significantly, and Samsung’s investments paid off. Major manufacturers were caught off guard and scrambled to invest in expanding production capacity.
However, good times didn’t last long. By the end of 1995, as manufacturers’ 8-inch wafer factories came online, production capacity surged, leading to oversupply. Thus, the seller’s market turned into a buyer’s market, and prices began to fall again.
In this situation, manufacturers were forced to cut production and reduce investment scale.
Samsung continued to expand its investments. In 1996, Samsung launched the world’s first 1GB DRAM, establishing its industry-leading position.
From 1996 to 1998, DRAM remained in a downward cycle.
In 1999, the downward trend in DRAM prices eased slightly due to the emergence of the internet bubble, leading the DRAM industry into a brief period of prosperity.
That year, amid intense competition, several significant changes occurred in the memory industry:
In the Korean sector, Hyundai Memory merged with LG Semiconductors to form Hyundai Semiconductor, which later split from the Hyundai Group (in 2001) and was renamed Hynix.
In the American sector, Micron acquired Texas Instruments’ memory division.
In the Japanese sector, Hitachi, NEC, and Mitsubishi Electric’s DRAM businesses consolidated to form Elpida.
In the European sector, Siemens’ semiconductor division became independent to form Infineon Technologies. A few years later, in 2002, it was renamed Qimonda. Later, in 2006, Infineon Technologies’ memory division became independent, transforming into Qimonda.
In 2000, among the top five global DRAM market shares, two were Korean manufacturers: Samsung (23.00%) ranked first, and Hyundai (19.36%) ranked third.
Not long after, the internet bubble burst, leading to a global economic crisis. The PC market suffered a severe blow, and the market demand for DRAM rapidly declined, causing prices to plummet again.
In 2001, the DRAM market size halved from 28.8 billion USD to 11 billion USD.
From 2002 to 2006, the DRAM market gradually recovered from its low point, showing good overall growth.
In 2006, Samsung developed the world’s first 1GB DRAM using 50nm technology. Hynix developed the highest-speed 200MHz 512MB Mobile DRAM at that time.
During that period, the DRAM market gradually formed a strong five-player structure: Samsung (Korea), SK Hynix (Korea), Qimonda (Germany), Micron (U.S.), and Elpida (Japan).
-
Third “Counter-Cyclical Investment”
In 2007, Microsoft launched the Vista system, which consumed a lot of memory. DRAM manufacturers anticipated a significant increase in memory demand and began to increase production capacity.
However, in reality, Vista had poor sales and did not boost the memory market, leading to oversupply again.
Worse still, in 2008, the financial crisis broke out, exacerbating the situation in the DRAM market. Memory prices plummeted, even falling below material costs.
At this critical moment, Samsung implemented its third “counter-cyclical investment,” further expanding production capacity and exacerbating industry losses.
In the spring of 2009, the third-ranked German manufacturer Qimonda declared bankruptcy, marking the official exit of European manufacturers from the DRAM market.
Qimonda
In 2011, DRAM supply again exceeded actual demand, causing prices to plummet. This time, Elpida could not survive and declared bankruptcy, marking the complete exit of Japanese manufacturers from the DRAM industry.
Elpida chip
Thus, the strong five became three, with only Samsung (Korea), Micron (U.S.), and Hynix (Korea) remaining in the DRAM field. The combined market share of these three companies exceeds 93%.
█ The Current State of DRAM Technology
Since 2011, there have been no significant changes in the market structure of DRAM memory. However, the user demand and market environment for DRAM have changed dramatically.
In addition to traditional PCs, the rapid development of mobile internet and IoT has led to the rise of smartphones, wearable devices, and IoT devices (such as cameras), greatly driving the demand for DRAM.
The development of cloud computing, big data, and AI has also spurred the increase in the number of data centers, resulting in a rapid increase in servers and network devices, further stimulating DRAM sales growth.
These demands have gradually segmented DRAM into categories such as standard DRAM, mobile DRAM, graphics DRAM, and niche DRAM.
Standard DRAM is mainly used in PCs and servers. Mobile DRAM, mainly LPDDR, is applied in smartphones and tablets. Graphics DDR is used for graphics card memory (GDDR). Niche DRAM is mainly used in products such as LCD TVs, digital set-top boxes, and network players.
LPDDR
The strong demand across multiple product scenarios has driven up DRAM prices. Around 2018, the surge in demand for cryptocurrencies such as Bitcoin brought about a rare “golden period” for the DRAM market.
After 2019, due to previous capacity expansion and destocking factors, memory prices dropped significantly. The collapse of cryptocurrency market prices and the maturity of the smartphone market led to weak market demand, and DRAM entered another downturn.
According to relevant institutions, from the second half of 2020 to May 2022, it was a period of improvement for the DRAM market.
Starting in June of this year, DRAM prices plummeted. In June, sales dropped by 36%, and in July, they fell by another 21%, marking a comprehensive collapse, which was quite tragic. According to institutions’ predictions, the decline will further expand in the fourth quarter.
DRAM prices plummet
Next, let’s look at the development of DRAM from a technical perspective over the years.
Historically, DRAM chips have relied on process miniaturization to increase storage density.
Every update of the DRAM process requires significant investment.
For example, transitioning from 30nm to 20nm requires a 30% increase in the number of photomasks and a doubling of the number of non-photolithography process steps. The requirements for cleanroom facilities also increase by over 80% with the rise in equipment numbers.
In the past, these costs could be offset by the higher yield of chips per wafer and the premiums from performance. However, as process technology continues to shrink, the gap between increased costs and revenues has gradually narrowed.
Around 2013, when the process technology entered 20nm, manufacturing difficulty significantly increased. After 18/16nm, continuing to reduce size in the two-dimensional direction no longer provided cost and performance advantages.
Thus, DRAM chip manufacturers began to explore new avenues, researching the Z-direction expansion capabilities. In other words, they began to push for 3D packaging.
As an industry leader, Samsung was the first to achieve 3D DRAM from a packaging perspective. They adopted TSV packaging technology to stack multiple DRAM chips, significantly enhancing the capacity and performance of a single memory stick. Later, various manufacturers followed suit, making 3D DRAM mainstream.
In terms of product standards, the industry generally adopts product standards set by the Solid State Technology Association (JEDEC), which includes the familiar DDR1 to DDR5.
Image source: Global Semiconductor Observation
The three giants of DRAM have all achieved mass production capabilities for DDR5/LPDDR5. Samsung is reportedly working on DDR6, with designs expected to be completed by 2024.
In terms of chip process technology, the terminology for DRAM has changed from previous direct references to 40nm or 20nm. Now, because the circuit structure is three-dimensional, linear measurement methods are no longer applicable, leading to terms like 1X, 1Y, 1Z, 1α, 1β, and 1γ to express the processes.
The industry believes that the 10nm to 20nm series processes include at least six generations: 1X is roughly equivalent to 19nm, 1Y to 18nm, 1Z to 16-17nm, and 1α, 1β, 1γ correspond to 12-14nm (below 15nm).
Image source: Global Semiconductor ObservationSamsung, SK Hynix, and Micron have entered the 1Xnm stage between 2016 and 2017, the 1Ynm stage in 2018-2019, and the 1Znm stage after 2020.Currently, major manufacturers continue to approach the 10nm mark. The latest 1αnm is still in the 10+nm stage.
█ The Past and Present of China’s DRAM Industry
Finally, let’s take a look at the development of the domestic DRAM industry.
China is one of the important markets for global semiconductor memory and a “must-fight” place for global semiconductor manufacturers.
However, to be realistic, our own DRAM industry development is far behind competitors.
The start of the domestic DRAM industry can be traced back to the 1990s.
At that time, Japan’s NEC established two joint ventures in mainland China to produce DRAM.
The first was Shougang NEC, established in 1991 through a joint venture between NEC and Shougang.
Shougang NEC began producing 4M DRAM (later upgraded to 16M) using a 6-inch 1.2-micron process in 1995. Later, in 1997, with a global DRAM price crash, Shougang NEC suffered heavy losses and never recovered, eventually becoming a foundry base for NEC overseas and exiting the DRAM industry.
The second was Hua Hong NEC, established in 1997 through a joint venture between NEC and Hua Hong Group.
Hua Hong NEC began producing mainstream 64M DRAM chips using an 8-inch 0.35-micron process technology in September 1999. After 2001, with NEC exiting the DRAM market, Hua Hong also left the DRAM industry.
In 2004, China attempted the DRAM industry for the second time. This time, the action was taken by SMIC.
At that time, SMIC invested in building the first 12-inch wafer fab (Fab4) in mainland China, and in 2006, it began mass production using 80nm technology, producing DRAM for Qimonda and Elpida.
However, this did not last long. In 2008, due to business adjustments, SMIC exited the DRAM business. The second attempt ended in failure.
In 2015, China’s DRAM procurement amount reached approximately 12 billion USD, accounting for 21.6% of global DRAM supply. The serious reliance on imports prompted domestic attempts for the third time.
The most representative of this attempt is the three major memory bases in Wuhan, Hefei, and Xiamen. These bases, leveraging national and local industrial policies, invested heavily (over 250 billion RMB) to develop semiconductor storage technology and cultivate talent.
Currently, the more representative companies in the domestic DRAM field include Hefei Changxin, Fujian Jin Hua, Unisoc, GigaDevice,Beijing Aisino, Dongxin Semiconductor,Nanya Technology (Taiwan),Winbond Electronics (Taiwan), and Powerchip (Taiwan).
Hefei Changxin is the leading company in the domestic DRAM storage chip industry. Their DRAM technology primarily comes from the now-bankrupt German DRAM manufacturer Qimonda and the Japanese manufacturer Elpida.
On September 20, 2019, Hefei Changxin announced the commissioning of the first 12-inch DRAM factory in mainland China and released the first 8G DDR4 manufactured using 19nm technology, marking a historic breakthrough.
According to institutions’ estimates, Hefei Changxin’s production capacity is expected to reach 125,000 wafers by 2022-2023.
Fujian Jin Hua may sound familiar to many. A few years ago, they were sanctioned by the U.S. government, and the news was quite sensational.
In May 2016, Fujian Jin Hua cooperated with UMC to produce niche DRAM.In December 2017, Micron accused Fujian Jin Hua and UMC of stealing its memory chip technology.In January 2018, Fujian Jin Hua also filed a lawsuit against Micron for patent infringement.In October 2018, Fujian Jin Hua was listed on the export control entity list.In November 2018, the U.S. Department of Justice sued UMC and Fujian Jin Hua for stealing Micron’s trade secrets.
After a series of upheavals, UMC could not hold on any longer. At the end of January 2019, UMC announced its withdrawal from the Fujian Jin Hua DRAM project. In November 2021, UMC and Micron reached a settlement. Currently, the review process on Fujian Jin Hua’s side has not yet yielded a complete final result.
█ Conclusion
Alright, having written so much, those who have read this far are true fans.
In summary, DRAM memory is an important component of computers, mobile phones, and other products, and an indispensable “part” of digital infrastructure.
Currently, the domestic DRAM memory has basically solved the issue of existence. The next step is to address the issues of yield improvement and capacity ramp-up. In terms of financing capabilities, industrial chain support, and talent development, we still need to strengthen continuously and proceed cautiously.
We look forward to breaking the “three strong” pattern and occupying a more important position in the DRAM field as soon as possible.
References:
1. “Who Will Break This DRAM Technology Dilemma?” Wang Kaiqi, Global Semiconductor Observation;
2. “The American Narrative of the DRAM World” by Xinguang Society;
3. “The Development Process of Storage Technology” by Xie Changsheng;
4. “The Process of Domestic DRAM Chip Localization Replacement is Tortuous, but the Future is Bright” by Xiangcai Securities, Wang Pan, Wang Wenrui;
5. “The Big Gamble of Storage Giants” by Semiconductor Industry Observation;
6. “Storage Chip Industry Research Report” by Guosen Securities;
7. “Domestic Storage is Waiting for a Revolution” by Fu Bin, Guoke;
8. “There’s No More Comprehensive Article on Semiconductor Storage than This” by Chip Master;
9. “Technology Brief 035 – Semiconductor Storage of Flash Memory” by Wu Mi Jin, Zhihu; 10. Baidu Encyclopedia, Wikipedia related entries.