Comparison of Domestic and Foreign Storage Chips

Source: ittbank

Why Is Our Country Developing Storage Chips?

The following content is sourced from:Everbright Securities Electronic Research Team

The necessity of developing storage chips lies in their significance and scale. Their significance is reflected in the fact that storage chips are the granaries of electronic systems, carriers of data, and related to data security; their scale is reflected in the market size, which is large enough, accounting for about one-third of the overall semiconductor market. To use a military analogy, developing storage chips is akin to ensuring that supplies are ready before the troops march.

Significant: The Granary of Electronic Systems

A basic electronic system mainly includes the following parts: sensors, processors, storage chips, and actuators. Sensors are responsible for data acquisition, processors are responsible for data processing, storage chips are responsible for data storage, and actuators are responsible for executing the results of the processor.

We generally prefer electronic products that run faster and can operate larger software applications. The core components that determine the performance of these electronic devices, apart from the processors we are familiar with, also include the reading speed capabilities provided by storage chips, which are an important influencing factor.

Storage chips can be simply divided into flash memory and RAM, where flash memory includes NAND FLASH and NOR FLASH, while RAM is primarily DRAM. To better understand the role of storage chips, if executing a complete program is likened to manufacturing a product, then storage chips are equivalent to warehouses, and processors are akin to processing workshops. To increase the speed of product manufacturing, one method is to enhance the efficiency of the processing workshop, which means improving processor performance; another method is to shorten the time it takes for raw materials to move from the warehouse to the processing workshop, by setting up a temporary small warehouse to store raw materials for products currently being manufactured, which can significantly reduce manufacturing time. The large warehouse corresponds to the flash memory in storage chips, while the small warehouse corresponds to the RAM in storage chips. Both are indispensable for the operation of electronic products, thus they have a high degree of overlap in their application range.

RAM differs from flash memory. Although both are carriers of data required by processors, RAM provides a space for processing currently needed data. Its capacity is smaller than that of flash memory, but its data reading speed is faster, akin to a VIP channel. It provides a fast passage for the data that needs to be processed currently, allowing the processor to quickly access and execute this data.

The difference between NAND FLASH and NOR FLASH lies mainly in their application fields. NAND FLASH is mainly used in high-end, large-capacity products such as smartphones, SSDs, and SD cards, while NOR FLASH is mainly used in low-end products such as feature phones, MP3 players, USB keys, and DVDs. Additionally, NOR FLASH is also used in automotive electronics and in technologies such as TDDI and AMOLED in smartphones.

As various applications become increasingly complex and new emerging scenarios continue to be applied, such as artificial intelligence, the Internet of Things, big data, and 5G, the amount of data that needs to be stored is also growing. Currently, information data is no longer just numbers, but an asset. The application of big data has turned internet companies into data processing centers, and competition among internet companies is a competition for traffic; whoever controls the data controls the world. For example, Tencent relies on its traffic-attracting ace: WeChat, to create a complete industrial ecosystem. Through the analysis and processing of big data, it can achieve precise profiling of billions of users, thus enabling targeted marketing.

Massive: A Barometer of the Semiconductor Industry

From the perspective of year-on-year growth in global semiconductor sales, the global semiconductor industry generally operates on a 4-6 year cycle, closely related to macroeconomic conditions, downstream application demands, and its own production capacity and inventory levels. Since 2010, storage chips have gradually become the main driving force behind the development of the semiconductor industry.

According to WSTS data, the global semiconductor market size in 2017 was $408.691 billion, a year-on-year increase of 20.6%, breaking the $400 billion mark for the first time and reaching a new high in seven years (2010 saw a year-on-year increase of 31.8%).

Among them, the market sales of integrated circuit products were $340.189 billion, a year-on-year increase of 22.9%, which far exceeded industry expectations, accounting for 83.2% of the total value of the global semiconductor market. The market sales of memory products were $122.918 billion, a year-on-year increase of 60.1%, accounting for 30.1% of the total value of the global semiconductor market, surpassing the largest historical share of logic circuits ($101.413 billion), which also confirms the industry’s assertion that memory chips are a thermometer and barometer of the integrated circuit industry.

Players: High Market Concentration

In the entire storage chip market, DRAM products account for about 53%, NAND Flash products account for about 42%, and NOR Flash accounts for only about 3%. DRAM is mainly divided into: standard (PC), server (Server), mobile (mobile), graphics (Graphic), and consumer electronics (Consumer) based on downstream demand. NAND Flash is mainly divided into: storage cards/UFD, SSD, embedded storage, and others based on downstream demand.

The storage chip market exhibits a high degree of concentration, whether it is DRAM, NAND Flash, or NOR Flash, all showing an oligopolistic pattern.

According to DRAMeXchange data, the DRAM market is mainly occupied by three manufacturers: Samsung, Hynix, and Micron, with Samsung holding about 48% market share; the combined market share of Samsung, Hynix, and Micron exceeds 90%.

According to DRAMeXchange data, the main players in the NAND Flash market include Samsung, Toshiba, SanDisk, Micron, Hynix, and Intel, with Samsung holding about 36% market share.

After a major reshuffle in recent years, the NOR Flash market saw Cypress holding about 25% market share, Macronix about 24%, Micron about 18%, Winbond about 17%, and domestic manufacturer Gigadevice Innovation holding the fifth position.

Supplement: Storage Chips and Storage Controller Chips Are Two Different Chips

This article primarily focuses on storage chips. Considering that storage chips and storage controller chips are easily confused, a brief introduction to storage controller chips is provided here for readers to distinguish between the two.

Typically, when we refer to storage chips, we mainly mean storage chips, while in reality, the chips within storage also include storage controller chips. Taking SSD products as an example, SSDs generally include PCB (including power supply circuits), storage chips (NAND flash), main control chips, interfaces, and an optional but still very important cache chip, which is the RAM chip. If the storage chip is the warehouse, then the storage controller chip is the key to the warehouse, managing the security of the granary. The storage controller chip controls the speed at which the processor reads and writes information from the storage chip; thus, the storage controller chip encompasses computer interface technologies and storage chip management technologies, playing a crucial role in protecting the information security of storage chips.

In terms of cost, NAND flash in storage chips can account for 70% or more of the material cost of SSDs. RAM chips are not essential in SSDs, depending on the type of main control, but equipping a cache can significantly enhance SSD performance, especially write performance. The cost of the main control chip occupies 10-15% of the SSD’s cost, making it not the most expensive component, but still very important.

In the early days, SSD storage controller chips were mainly produced by flash manufacturers, presenting a triopoly pattern among Intel, Samsung, and Micron. With the development of independent storage controller chip manufacturers, the market has gradually evolved into a competitive landscape, with major manufacturers including Marvell, SMI, Phison, SandForce, Realtek, etc. In recent years, many SSD controller manufacturers have emerged in the domestic SSD controller chip field, with well-known ones including JHICC, Goke Microelectronics, Yixin, Hualan Microelectronics, and companies focused on military and enterprise markets such as Zhongbo and Yifang Information. Additionally, there are Taiwanese manufacturers that have established subsidiaries in mainland China, such as Phison establishing Zhaoxin Electronics in Hefei, and Hangzhou Lianyun Technology having Taiwanese investment participation.

Why Can Mainland China Develop Storage Chips?

Storage chips are a technology, capital, and talent-intensive industry. The sufficiency for developing storage chips lies in the right timing, geographical advantages, and human resources. Timing: ① Low brand recognition; ② Slowing of Moore’s Law; ③ Emphasis on IP and manufacturing. Geographical advantages: ① Manufacturing is shifting to domestic; ② Strong national support. Human resources: ① Yangtze Memory, ② Hefei Changxin, ③ Fujian Jin Hua’s three major storage projects are progressing smoothly. With the right timing, geographical advantages, and human resources, the development of storage chips in mainland China is entering an accelerated stage, making domestic production just around the corner.

Timing: The Attributes of Storage Chips

1. Low Brand Recognition

Storage chip products differ from most consumer products, possessing typical bulk commodity attributes, with minimal differentiation in competition. The technical specifications of products produced by different companies are generally similar, and the standardization level is high, resulting in weak brand recognition and low user stickiness.

Taking smartphones as an example, when users choose smartphones, they generally consider whether the processor brand and model are Qualcomm or MediaTek, but very few consider whether the screen is from LG or BOE; they only consider parameters such as screen size and resolution. This aspect is similar between storage chips and panels. Users generally only consider the storage chip’s capacity, whether it is 64G or 128G, and very few consider whether the storage chip is produced by Hynix or Toshiba. Therefore, in the storage chip industry, as long as the technical parameters meet requirements, the substitutability of products from different brands is very high, providing latecomers with the possibility of overtaking in the curve. The success of BOE in the panel industry also confirms this point.

2. Slowing of Moore’s Law

Moore’s Law states that, under constant prices for integrated circuits, the number of transistors that can be accommodated doubles every 18-24 months, resulting in a doubling of performance. However, as the manufacturing process of integrated circuits approaches physical limits, the speed of developing advanced process chips is gradually slowing, and Moore’s Law is facing obsolescence.

Currently, international giants have entered the mass production stage of 7nm processes, and Tech Insights predicts that by 2020 it will reach 5nm, but the application of such cutting-edge processes is mainly concentrated in the manufacturing of logic circuit processor chips.

The process route for storage chips, while differing from that of logic circuits, also faces bottlenecks as Moore’s Law approaches its limits, potentially even earlier than logic circuits. Currently, the manufacturing process of storage chips has developed to the 1x, 1y, and 1z (between 20nm and 10nm) stages, where further miniaturization is challenging. As the manufacturing process advances, stability declines after reaching a certain level, and 10nm is generally considered the critical point. DRAM is currently at the 1x and 1y levels, with hopes of entering the 1z stage by 2020. NAND has essentially reached its limit, shifting from 2D to 3D development. With the slowing of Moore’s Law, the gap between domestic technology and international giants is expected to gradually narrow. Currently, the common international standard for 3D NAND is 64 layers, while Yangtze Memory has already achieved 32 layers, with only a generational gap remaining.

3. Emphasis on IP and Manufacturing

The semiconductor industry has transitioned from an integrated IDM model to today’s design-manufacturing-testing outsourcing model. However, for storage chips, the mainstream manufacturers still predominantly follow the integrated IDM model, mainly due to the characteristics of storage chips emphasizing IP and manufacturing.

The difficulty with analog chips lies in design, as analog chips cannot be verified through simulation like digital chips; they can only be tested through repeated wafer productions, with results fed back for improvement. Thus, the R&D cycle for analog chips is long, and costs are high, making experience accumulation crucial, with major manufacturers like TI and ADI having decades of history.

The difficulty with processors lies in architecture IP, ecosystem, and manufacturing, each of which has a very high concentration. The architectures mainly include X86 for computers and ARM for mobile devices, corresponding to Windows and Android systems, respectively. The manufacturing of processors is also the most advanced in semiconductor manufacturing. For CPUs and analog chips, the entry barriers are very high, and latecomers face significant disadvantages.

The difficulty with storage chips lies in IP and manufacturing. In reality, any chip design must break through the IP blockade. The concentration of IP for storage chips is somewhat lower than that for processors, making it slightly easier to obtain storage chip IP through a combination of cooperative licensing and independent R&D.

Regarding DRAM IP, domestic manufacturers started relatively late, resulting in weaker patent accumulation. However, since the DRAM field is relatively mature, international capital investment has decreased, providing an opportunity for domestic manufacturers to increase capital investment for domestic substitution. Domestic manufacturers must accelerate technological iteration and achieve breakthroughs in higher technological fields as soon as possible to gain stronger bargaining power over downstream manufacturers and improve product gross margins.

Regarding NAND IP, 3D NAND stacking technology is an upgrade from 2D planar technology. The gap between China’s 3D NAND stacking technology and that of major international manufacturers is relatively small compared to the DRAM field. This is because DRAM is already relatively mature, while 3D NAND stacking technology has emerged as a new technology in recent years. Therefore, the technological gap between China and world-leading technology is not too large. However, in terms of IP reserves, domestic manufacturers still remain at a disadvantage, as storage chip giants still hold a dominant advantage in patent reserves.

Combining the aforementioned points, storage chips have low brand recognition and are standardized products, with low dependency on the upper ecosystem. The main work of storage chip manufacturers lies in the manufacturing stage, where the advantages of scale are very obvious. With the slowing of Moore’s Law, the gap between domestic and international technologies has narrowed, providing domestic manufacturers with an opportunity to catch up.

Geographical Advantages: Opportunities for Domestic Development

1. Manufacturing Shifting to Domestic

With the trend of semiconductor manufacturing shifting to domestic, international giants are increasingly setting up factories or expanding their existing factories in mainland China. According to SEMI data, it is estimated that between 2017 and 2020, approximately 62 wafer fabs will be put into operation globally, with 26 located in mainland China, accounting for 42% of the total.

With the establishment of numerous wafer fabs in mainland China, this will help promote the development of domestic manufacturing, while also driving the development of the entire industry chain, including design, testing, materials, and equipment, thereby facilitating the establishment of a domestic semiconductor industry ecosystem. Manufacturing is the most crucial link in storage chips, and thus, the shift of manufacturing to domestic will also favor the development of the domestic storage industry.

2. Strong National Support

In June 2014, the State Council issued the “National Integrated Circuit Industry Development Promotion Outline,” proposing the establishment of a national integrated circuit industry fund (referred to as the “Big Fund”), elevating the R&D of new semiconductor technologies to a national strategic level. It also explicitly stated that by 2020, the gap between the integrated circuit industry and international advanced levels would gradually narrow, with an average annual growth rate of over 20% in the industry’s sales revenue and a significant enhancement in the sustainable development capacity of enterprises; by 2030, the main links in the integrated circuit industry chain would reach international advanced levels, with a number of enterprises entering the international first tier, achieving leapfrog development.

According to statistics from TrendForce, as of November 30, 2017, the Big Fund had made effective decisions on 62 projects involving 46 enterprises, with a total effective commitment of 106.3 billion yuan and actual investment of 79.4 billion yuan, accounting for 77% and 57% of the first phase’s total scale, respectively. The Big Fund has made investments across the entire industry chain, including manufacturing, design, testing, and equipment materials, with the proportion of commitments in each link being 63%, 20%, 10%, and 7% respectively.

In addition to the national integrated circuit industry investment fund, many provinces and cities have also established or are preparing to establish integrated circuit industry investment funds. Currently, more than a dozen provinces and cities, including Beijing, Shanghai, and Guangdong, have established local government funds specifically to support the development of the semiconductor industry. According to statistics from the national integrated circuit industry fund, as of June 2017, the local integrated circuit industry investment funds (including those under preparation) leveraged by the “Big Fund” reached 514.5 billion yuan.

The second phase of the Big Fund has already been launched, with the amount expected to exceed that of the first phase, which had a scale of 138.7 billion yuan. Ding Wenwu, the general manager of the Big Fund, revealed that the fund will increase its investment ratio in the design industry and will plan investments around national strategies and emerging industries, such as smart cars, smart grids, artificial intelligence, the Internet of Things, and 5G, while also providing support to equipment and materials to accelerate their development. Additionally, we expect the second phase of the Big Fund to focus on storage chips, integrated circuit design, and compound semiconductors.

3. Talent Concentration: Smooth Progress of Three Major Projects

Under the conditions of timely opportunities, slowing Moore’s Law, and gradually catching up in technology by mainland manufacturers; under geographical advantages, with strong national support for the storage industry, capital is no longer a bottleneck; and in terms of human resources, mainland projects have also made promising progress.

With the right timing, geographical advantages, and human resources, the three major storage projects in mainland China, Yangtze Memory (NAND), Hefei Changxin (DRAM), and Fujian Jin Hua (DRAM), are progressing smoothly, with mass production expected in the second half of 2018. The year 2018 is also anticipated to be a milestone year for the mainstream development of domestic storage chips.

1. Yangtze Memory

Yangtze Memory is a national storage chip base project established in cooperation with Tsinghua Unigroup and Wuhan Xinxin, focusing on the R&D and manufacturing of 12-inch 3D NAND flash.

Human Resources: The global executive vice president of Unigroup and acting chairman of Yangtze Memory is the “father of DRAM” in Taiwan, former chairman of Winbond, Gao Qichuan. In April 2017, Gao Qichuan stated that 500 R&D personnel had been gathered in Wuhan for 3D NAND development, and considerations were being made for R&D on 20/18nm DRAM. In May 2018, former director of the Electronic Information Department of the Ministry of Industry and Information Technology, Diao Shijing, joined Tsinghua Unigroup as co-president. Prior to this, former deputy director of the Electronic Information Department, Peng Hongbing, had already assumed the position of vice president of the Big Fund and chairman of the supervisory board of Yangtze Memory.

Project Progress: The initial investment exceeds $24 billion, with an expected additional investment of $30 billion in the future. In September 2017, the Yangtze Memory national storage chip base project (phase one) saw the completion of the main production and power plant ahead of schedule. In February 2018, clean room decoration and installation of air-conditioning, fire protection, and other systems began within the factory. On April 11, 2018, Yangtze Memory officially launched its equipment installation ceremony, with equipment from Northern Huachuang successfully entering the production line of Yangtze Memory. In May 2018, a lithography machine ordered from ASML in the Netherlands, valued at $72 million, arrived in Wuhan. Over the next two years, Yangtze Memory’s storage chip base will import nearly 30,000 tons of precision instruments from around the world to Wuhan.

In February 2017, Yangtze Memory announced that its 32-layer 3D NAND flash chip successfully passed testing, with mass production expected by the end of 2018 and a projected monthly capacity of 300,000 wafers by 2020. At the same time, Yangtze Memory is also advancing the 64-layer stacked 3D flash, aiming to achieve large-scale production by the end of 2019, thereby narrowing the gap with world-leading levels to within two years.

Unigroup also plans to invest in two 12-inch production lines in Chengdu and Nanjing, with a total capacity of 500,000 wafers/month, while also advancing the development of 20/18nm DRAM, with DRAM progress being slower than that of NAND FLASH, and expected to achieve mass production by 2020 at the earliest.

2. Hefei Changxin

Hefei Changxin Storage was established by Gigadevice Innovation, former CEO of SMIC Wang Ningguo, and Hefei Industrial Investment, with a project budget of 18 billion yuan.

Human Resources: According to Wang Ningguo, chairman of Hefei Changxin Storage Technology Co., Ltd. and CEO of Ruili Integrated Circuit Co., during the “National Integrated Circuit Major Special Project Entering Anhui Activity” in April 2018, one of the implementation entities of the Hefei storage project, Ruili Integrated, was established on July 13, 2016, with only one person at the time. After 21 months and 630 days, the current employee count has reached 1,539, which equates to an impressive growth rate of 2.5 new hires each day over less than two years. Among the employees, there are 447 from Taiwan and 1,013 from mainland China, accounting for two-thirds of the workforce.

Project Progress: Gigadevice Innovation is responsible for developing 19nm process 12-inch wafer mobile DRAM, with a target to achieve success by the end of 2018, ensuring a product yield of no less than 10%. By then, Hefei Changxin will become the first large-scale DRAM factory in China and the fourth company in the world to break through the production technology of DRAM below 20nm. Hefei Changxin completed the construction of its first factory in January 2018 and began equipment installation, with engineering samples of the 19nm product expected to be released by the end of the year.

The company’s current technology level is about five years behind the world leading level. If the company can achieve the development of 17nm technology according to schedule by 2021, the technology gap will narrow to about three years.

3. Fujian Jin Hua

Fujian Jin Hua mainly engages in the R&D and production of niche DRAM, primarily applied in consumer electronics. Although these industries have entered a stage of stock game, the market scale is still massive.

Human Resources: The leader of Fujian Jin Hua, Chen Zhengkun, was originally the head of the joint venture between Elpida and Taiwan’s Powerchip, and after Elpida’s bankruptcy and acquisition by Micron, he transitioned from the Japanese semiconductor industry to the American system. Jin Hua Integrated Circuit adopts a combined approach of recruiting and training talent both domestically and internationally. By 2018, the talent pool is expected to reach 1,200 people, with over 800 already recruited. Due to the relatively weak foundation of the Jinjiang integrated circuit industry, Jinjiang City is using the Jin Hua project as a leading project to build a “three parks and one area” industrial development space carrier, creating a complete integrated circuit industry chain ecosystem covering design, manufacturing, packaging testing, equipment and materials, and terminal applications, with the goal of forming an industry scale of 100 billion yuan by 2025. It is reported that over 20 industry chain projects, including Taiwanese companies like Siliconware and G.Skill, as well as American Air Products, have landed in Jinjiang, with a total investment of nearly 60 billion yuan, gradually forming an industry chain ecosystem.

Project Progress: The manufacturing technology work of Fujian Jin Hua is primarily carried out by UMC, with a process technology starting from 32nm and a planned capacity of 60,000 wafers per month, with trial production expected to start in September 2018. The company’s ultimate goal is to launch 20nm products, with plans to complete four phases of production capacity reaching 240,000 wafers per month by 2025.

What Impact Will the Development of Storage Chips in Mainland China Have?

Price: A Downward Trend May Be Inevitable

Driven by downstream applications such as smartphones, AI, data centers, automotive, and the Internet of Things, the storage chip market is expected to continue maintaining high growth. Micron predicts that from 2017 to 2021, the compound annual growth rate for DRAM demand will reach 20%, while NAND demand will see a compound annual growth rate of 40-45%.

The price increase of storage chips began with supply not meeting demand; whether it continues will depend on supply and demand. Demand is growing slowly, while supply will suddenly increase. With international manufacturers releasing capacity and domestic storage projects progressing steadily, a decline in storage chip prices may become inevitable.

DRAM

Demand Side: The continuous upgrades of mobile DRAM from 1G to 2G, 3G, and 4G in downstream smartphones have led to rapid growth in demand for mobile DRAM, while the rapid development of data centers has promoted the growth of server memory demand.

Supply Side: DRAM is mainly controlled by a few companies like Samsung, Hynix, and Micron, presenting an oligopolistic structure, with Samsung holding about 45% market share. Before Q3 2016, DRAM prices were on a downward trend, and all DRAM manufacturers were hesitant to expand production.

Price: The supply-demand imbalance caused DRAM prices to soar from Q2/Q3 2016, with the DXI index rising from 6,000 points to a peak of 30,000 points. The DXI index, created by TrendForce in 2013, reflects mainstream DRAM prices and remains at a high level. From the perspective of spot prices, the price of 4G products began to decline in Q1 2018, but the price of 2G products remains strong.

In the short term, under the influence of weakened smartphone shipments in mainland China, demand for mobile memory has declined. At the same time, the three major DRAM manufacturers will increase production capacity by 5-7% in 2018, which will ease the supply-demand tension, leading to a decline in mobile memory prices. However, with the global development of data centers, demand for server memory remains strong, and we expect the prices of server memory to continue to rise in 2018. According to predictions from the Korean Industry Technology Promotion Institute, not considering mainland manufacturers, the price of 4Gb DRAM is expected to drop from $4.71 in 2017 to $4.39 in 2018, a decline of 6.8%.

In the long term, as the three major manufacturers release capacity and the DRAM projects in Hefei Changxin and Fujian Jin Hua progress steadily, the supply will increase rapidly while demand grows slowly. The resulting oversupply may trigger a price war, leading to a significant drop in DRAM prices. According to predictions from the Korean Industry Technology Promotion Institute, the price of 4Gb DRAM is expected to drop significantly to $3.53 by 2019; due to the impact of mainland manufacturers, the estimated reduction in DRAM revenue for Korea over the next five years is $6.7 billion.

NAND Flash

Demand Side: The continuous upgrades of embedded storage in downstream smartphones from 16G to 32G, 64G, 128G, and even 256G have led to rapid growth in demand for embedded storage, while the increasing penetration of SSDs in PCs and the rising number of data center servers have also driven rapid growth in SSD demand.

Supply Side: Major NAND manufacturers include Samsung, Toshiba, Micron, and Hynix, with Samsung also being the industry leader, holding about 37% market share. The years 2016 and 2017 were the transition years for NAND Flash from 2D to 3D processes, and there has been a gradual release of capacity, leading to a slow increase in supply.

Price: The supply-demand imbalance caused NAND prices to soar from Q2/Q3 2016, with the highest increase exceeding 50%. As supply gradually increases with the release of capacity, NAND prices have decreased by about 20% from the second half of 2017 to the present.

In the short term, with the weak sales growth of smartphones, NAND demand is expected to fall short of expectations in the first half of 2018. As 3D capacity continues to be released, the market is likely to shift to oversupply, increasing the probability of NAND Flash prices continuing to decline. According to predictions from the Korean Industry Technology Promotion Institute, not considering mainland manufacturers, the price of 32Gb NAND is expected to drop from $2.8 in 2017 to $1.93 in 2018, a decrease of 31%.

In the long term, as international manufacturers release capacity and the NAND projects of Yangtze Memory come online, the supply will increase rapidly while demand grows slowly, potentially leading to oversupply and triggering price wars that cause NAND prices to decline significantly. According to predictions from the Korean Industry Technology Promotion Institute, the price of 32Gb NAND is expected to drop significantly to $0.65 by 2020; due to the impact of mainland manufacturers, the estimated reduction in NAND revenue for Korea over the next five years is $1.1 billion.

NOR Flash

Demand Side: Although the market share of NOR FLASH is relatively small, it is still often used to store boot codes and device drivers due to its capability to execute code within the chip. As IoT, smart applications (smart homes, smart cities, smart cars), and drones increasingly adopt NOR Flash as storage devices and microcontrollers, and with smartphones needing external NOR Flash to store program codes for OLED panels, the demand for NOR Flash continues to grow.

Supply Side: On the one hand, the upstream silicon wafer raw materials are in short supply and prices are rising; on the other hand, major players like Micron and Cypress have announced their exit from the market, shutting down some production lines, leading to a supply gap and price increases.

Price: In 2017, due to the supply-demand imbalance in the NOR Flash market, prices surged. In Q1 2018, as the smartphone production chain entered an inventory adjustment phase, the NOR Flash market reached a balance between supply and demand, with prices stabilizing. Only some lower-capacity NOR Flash products saw price declines due to the emergence of domestic capacity.

In the short term, in Q2 2018, as the Android smartphone camp began entering the component stocking season, and with the increasing penetration of AMOLED panels in the market, demand for NOR Flash has shown signs of recovery. Intel raised the NOR Flash capacity for storing BIOS in the eighth generation Core processor platform from 64Mb/128Mb to 256Mb, tightening the supply of mid-to-high capacity NOR Flash. However, since 2017, international giants have not made large-scale capacity expansion actions. As market demand transitions from the off-season to the peak season, supply shortages may resurface, leading to potential price increases.

In the long term, high-end NOR Flash demand is expected to surge with the development of automotive intelligence and electrification, while low-end NOR Flash demand will also remain strong due to the growth of new applications such as IoT, smart speakers, and AMOLED.

Security: Gradually Achieving Autonomy and Control

The overall market share of the mainland industry chain is low, making domestic production urgent. In core chips such as CPU/MPU in computer systems, FPGA/EPLD and DSP in general electronics, embedded MPU and DSP in communication equipment, DRAM and NAND Flash in storage devices, and Display Driver in display and video systems, the market share of domestic chips is almost zero. In the manufacturing sector, although mainland China has established a foothold in mature processes above 28nm, the market share of advanced processes below 28nm and compound semiconductors is still very low. The market share of high-end equipment, materials, EDA tools, and core IP is also extremely low.

This situation is unfavorable for both the country and enterprises, whether from the perspective of national security or the development of the electronic industry. Fully promoting the semiconductor industry has now become a consensus across the country, providing ample motivation for the industry’s development.

With the steady progress of the three major storage projects, the self-sufficiency rate of mainland storage chips is expected to gradually increase, thereby achieving autonomy and control.

Investment Recommendations

Storage chips are the barometer of the semiconductor industry, and the global storage chip industry is thriving. The three major storage projects in mainland China are progressing rapidly, and 2018 is expected to be a milestone year for the development of the mainland storage industry. We recommend the domestic storage chip design leader Gigadevice Innovation, initiating coverage with a “Buy” rating. The construction of storage factories will boost demand for upstream equipment, and we recommend the domestic semiconductor manufacturing equipment leader Northern Huachuang, maintaining a “Buy” rating.

Initial Coverage of Gigadevice Innovation: Domestic Storage Chip Design Leader

1. High-Growth Storage Chip Design Scarce Target

The company is the leading enterprise in the domestic storage chip design field, operating under a typical fabless model. Its main products include storage chips and MCUs, with storage chips accounting for about 85% of revenue.

The largest shareholder is the actual controller Zhu Yiming, holding a 13.58% stake; the second largest shareholder is the Big Fund, holding an 11.0% stake.

2. Rapid Revenue and Net Profit Growth, NOR Flash Price Increase Expected to Continue

The company’s performance has consistently maintained rapid growth, with a compound annual growth rate of 35.92% in operating revenue from 2011 to 2017, and a compound annual growth rate of 67.70% in net profit. In 2017, the company achieved operating revenue of 2.03 billion yuan, a year-on-year increase of 36.32%, and net profit of 398 million yuan, a year-on-year increase of 127.56%; in Q1 2018, the company achieved operating revenue of 542 million yuan, a year-on-year increase of 19.71%, with net profit of 90 million yuan, a year-on-year increase of 28.65%.

The main reason for the company’s rapid growth in 2017 was the supply-demand imbalance in the NOR Flash market that began in Q4 2016, with supply-side players like Micron and Cypress exiting the low-capacity market to focus on high-segment markets for automotive, industrial, and IoT applications. The demand side has seen an accelerated penetration of AMOLED screens requiring a NOR Flash for electrical compensation, particularly driven by Apple’s adoption.

In the second half of 2018, as Android smartphones enter the stocking peak season, the penetration rate of OLED panels is expected to increase further. Additionally, the demand for NOR Flash from automotive electronics and IoT is strengthening, with new models launched this year featuring LiDAR and automatic emergency braking systems, leading to a rapid surge in demand for automotive-grade NOR Flash.

Currently, the company has achieved mass production of high-capacity 256Mb NOR Flash products, with R&D on 55nm and 45nm technologies progressing rapidly. In 2016, the company ranked fifth in the global NOR Flash market, with a market share of 7%. With Micron and Cypress gradually exiting the low-end market, we expect the company’s market share to surpass 10% in 2017.

3. Continuous Increase in R&D Investment, NAND/DRAM Open Up Broad New Space

The company has significant technological and talent barriers. By the end of 2017, the number of R&D personnel reached 253, a 42.94% increase compared to the end of 2016. The main management and technical teams come from advanced industrial regions such as the United States, Canada, and Taiwan. In 2017, R&D expenses amounted to 167 million yuan, a year-on-year increase of 63.31%, with a compound annual growth rate of 51.66% from 2011 to 2017. Key technical R&D personnel are from top microelectronics institutions such as Tsinghua, Peking University, Fudan University, and the Chinese Academy of Sciences. As of the end of 2017, the company had applied for 718 patents and obtained 261 patents, covering key technology areas for NOR Flash, NAND Flash, and MCUs.

Currently, the company’s NAND Flash products can reach a capacity of 32GB, with self-developed 38nm products already in mass production and 24nm R&D progressing smoothly. Additionally, in October 2017, the company signed a joint development agreement with Hefei Industrial Investment for the DARM storage chip project, with a budget of approximately 18 billion yuan, with both parties raising funds in a 1:4 ratio. The project aims to achieve a product yield of no less than 10% by the end of 2018. NAND Flash and DRAM, as mainstream storage devices, have a global market size approximately 20 times that of NOR Flash, with China currently heavily reliant on imports, indicating a vast space for domestic substitution in the future.

4. Acquiring Shares in SMIC Forms Virtual IDM, Acquiring Silergy for Synergistic Development

For chip design companies, wafer foundries require a high degree of capital and scale, with relatively concentrated production capacity. The synergy between manufacturing processes and design takes time to accumulate, and the coordination of process nodes directly determines product quality. Often, the collaboration between design companies and foundries becomes a critical barrier to their business development. In November 2017, the company participated in the subscription of shares issued by SMIC through its wholly-owned offshore subsidiary, investing no more than $70 million to further strengthen strategic relations and form a virtual IDM, which will also help ensure the company’s production capacity.

In July 2018, the company announced plans to acquire 100% of Silergy’s shares through a combination of issuing shares and cash payments, while also planning to raise matching funds through a non-public share issuance to pay for this transaction, as well as for R&D projects related to 14nm process embedded heterogeneous AI inference signal processing chips, 30MHz active ultrasonic CMEMS process and transducer sensor R&D, and the construction of an intelligent human-computer interaction R&D center, as well as intermediary costs related to this transaction.

Silergy’s products mainly focus on next-generation smart mobile terminal sensor SoC chips such as touch chips and fingerprint chips. This transaction will supplement the company’s R&D technology in sensors, signal processing, algorithms, and human-computer interaction to enhance the productization capabilities in related technology fields, ultimately forming a complete MCU + storage + interaction system solution, injecting momentum for the company’s further rapid development.