Some time ago, Samsung’s flash memory chip factory had an accident. In the event of an unexpected power outage for half an hour, wafers in production were damaged. The number of damaged wafers differs from the number of wafers. It is said to be 6,000. It is 60,000 pieces, and the most exaggerated is the loss of 600,000 wafers. If it is the latter two, this will have a low impact on Samsung's NAND capacity, which will inevitably lead to NAND flash memory price fluctuations.
Associating with blackouts, fires, and other accidents that occurred before the Samsung and SK Hynix companies, many netizens confessed that the accident gave Samsung an excuse to raise prices.
Joke is a joke that Samsung, as a manufacturer, is naturally reluctant to report accidents, but this is still a hot topic. Especially in the Chinese consumer community, one issue that people are concerned about is who can balance Samsung, SK Hynix, in the future. Toshiba, Micron and other companies? We all look forward to domestic companies to counterattack, but how to counter it?
For domestic memory and flash memory issues, we have recently done a lot of articles on domestically produced chips. In particular, reports on purple DDR3 and DDR4 memory have had a wide range of impacts. I have paid attention to comments from many stations and users outside the station. Some readers were cynics about domestically produced products. Some readers thought that domestic memory was good, DDR3 memory had come, and DDR4 memory was not far. It seemed that Samsung could step on the foot in a year or two, allowing the memory market to quickly cut prices and even restore. Before the price of cabbage.
For these two diametrically opposed evaluations, we must rationally look at it. Taunting is wrong, but blind optimism is even more undesirable. Memory chips are high-tech manufacturing industries. Chinese manufacturers have been absent from this field for decades. Localization can be said It is a blank space. We must be sure to come from behind. However, this process may take a long time. It cannot be reversed in a few years.
Global memory industry structure: High monopoly, Samsung alone, High threshold
According to the plan made in China 2025, the only breakthrough in domestic production is the entire semiconductor industry chain. From semiconductor equipment to chip production, testing and application are all involved. The memory chip is now the focus, but it is only part of the semiconductor chip. The memory chip is part of the memory chip, but it is an important part of it.
In 2017, the world’s DRAM memory output value was US$72 billion, which was mainly due to the fact that the DRAM chip prices increased by 74% within one year, and the NAND output value was approximately US$49.8 billion. Of course, the flash memory market price has also risen greatly in the past one or two years, but the price has no memory. So exaggerated.
In the global $72 billion memory industry, Samsung, SK Hynix, and Micron all accounted for about 95% of the shares, of which the Samsung family accounted for about 46% of the share, and the share of SK Hynix was about 29%.
Micron acquired the Inacore Corp., which made memory chips for it two years ago, with a market share of about 21%, and the remaining share was obtained by Taiwan South Asia, Powerchip and other companies, but they have no influence on the market. The two South Korean companies controlled 75% of the global memory market share, showing a highly monopolistic situation. The size of the Samsung family is almost the sum of the other two, and the production capacity is far ahead. The entire situation is similar to that of the Three Kingdoms period. Dan Dae, Wu Guo SK Hynix, Shu Guoguang's strength can not be compared.
These three companies not only monopolized the market, but also created a great barrier to entry for the market's latecomers. In other countries or regions, there is no DRAM industry chain, so there is no talk about personnel training. The memory factory is also the same as the processor factory. The need to continuously upgrade the technology, the required technology and capital investment is also a problem, SK Hynix, Micron is significantly behind Samsung because of the progress in the 20nm node upgrade is not smooth, and Samsung is not only the first mass production 20nm process, or The first mass-produced 18nm process, no accident, then the future 17nm, 16nm will be Samsung's first production.
In order to activate the memory industry, technology, talents, and funds, it is not impossible for China to save money. However, talents and technologies cannot be solved by themselves. This is also a trick for domestic development of the memory industry. Industry, Many domestic companies have grabbed talents from companies in Taiwan, South Korea, and Japan, and even scrambled current employees, which is easy to fall into legal disputes.
Not long ago, there were reports that Micron chose to initiate a lawsuit in California, and accused UMC and Fujian Jinhua of the mainland infringing their DRAM patent rights and stealing trade secrets. This is the reason why domestic companies took talents from companies such as Taiwan Asia Branch. One of the unfavorable consequences, it is easy to provoke a legal case.
In short, although the huge output value of the memory market has made many companies excited, the current pattern has led to a very high barrier to entry in the memory market. Japanese companies were once the dominant power in the memory market, but Elpida, Renesas, and other companies either withdrew from the market. In the memory chip market, either the bankruptcy went bankrupt and the market was handed over to the United States and South Korea.
Japanese companies can easily obtain technical licenses from US companies in the early march of the military DRAM market, which is still the case. Not to mention that it is difficult for Chinese companies to build the memory industry on the basis of zero.
The status quo of China's memory industry: DDR3 has just started, and the process is seriously lagging behind
What is the status of China's own memory industry?
If there is a sense of presence in the NAND flash memory and NOR flash memory markets, then the DRAM memory is really poor.
Recently, our article on testing DDR memory has attracted the attention of many peers and players. The overall performance of Violet's DDR3 particles is not bad, and some media have written very exaggerated. It seems that Violet DDR3 memory has to be whistled by Kingston, Zhi Qi, and Samsung. However, the actual situation is that Ziguang's DDR3 memory is just starting. The DDR3 memory that we currently buy online is mainly distributed by dealers. The number is not large, and the store display is only more than 66 pieces.
Even DDR3 memory is not independently developed by Ziguang. The production unit Xi'an Ziguang Guoxin is actually bought. Initially, it was Infineon’s memory chip division established in Xi’an in 2003. In 2006, Infineon’s semiconductor business was split into Qimonda Technologies. In 2009, they were acquired by China’s Inspur Corporation and restructured into Xi’an Huaxin Semiconductor. In 2015, Ziguang acquired a 36% stake in the Tongfang Guoxin from another Tsinghua-based company's Tongfang, and established Ziguang Guoxin. Xi'an Ziguang Guoxin bought them from Inspur and now is a subsidiary. Theirs DDR3 memory is actually a legacy of Qimonda.
DDR3 memory has not yet been eliminated. However, it is not currently the mainstream of desktops, servers and smart devices. Therefore, many people are looking forward to the use of violet-made DDR4 memory. However, Violet has previously refuted this issue. They have not mass-produced DDR4 yet. Memory, Later we got news that Ziguang is expected to launch DDR4 memory chips in the second half of this year, but the specific situation is unknown, and the key information such as capacity and price is still unknown.
Not only is the mainstream memory production scale lagging behind, domestically produced memory is still at risk of technological obsolescence. To date, Ziguang’s DDR3 particle process technology has not been announced yet. Considering Qimonda was more than 10 years ago, Xi’an Ziguang Guoxin’s The DRAM process is at least 30nm and above, and now the mainstream memory technology is already at the 20nm level. Samsung will mass-produce the 18nm process in 2016. The next step is the 17nm process.
The exact manufacturing process of domestic memory has always been a mystery, but we can confirm with other information. Last year, Zhaoyi Innovation announced its entry into DRAM memory. They signed a cooperation agreement with the Hefei Municipal Government Holding Investment Company with a project budget of RMB 18 billion. It was at the end of 2018 that a memory chip with a 19nm process was developed -
If successful, their memory technology is still quite advanced. Although it is still Samsung, it is at least mainstream. However, the goal of Siu Yi Innovation is to achieve a yield of no less than 10%. This standard is not mass production. The horizontal, 10% yield rate can only be a technical experiment, which means that there is a long way to go from mass production to market.
Not to mention DRAM memory, the domestic NAND flash memory technology must be quite backward. The Yangtze River storage company in Wuhan now researches and develops only 32-layer stacks of flash memory, but the core capacity is only 64Gb. This is in line with the current mainstream 64-layer stack, with a core capacity of 256Gb. Even 512Gb 3D NAND will have to fall behind for two years.
However, Zhaoyi's innovative NAND flash memory production level is the 38nm process. The 24nm class is still under development. Note that they are still 2D NAND flash memory. Samsung, Toshiba has already mass-produced 2D NAND flash memory with 15, 16nm process, and 2D. Flash memory is the technology they are eliminating.
Domestic Memory Opportunities: Two Steps: Independent R&D, Technology Licensing
The domestic memory base is so poor, and the technological gap is so big. Do we have to abandon this field? No, not just memory chips. China has lagging behind in the past. We used to envy Japan, Europe's high-speed rail, and now China's high-speed rail mileage. Far more than other countries, are we not catching up with it?
The same is true for the memory industry. It is not feasible for China’s such a large market to rely entirely on imports. It is an excellent tradition for Chinese companies to know the difficulties. On the issue of domestic memory, China has the largest market in the world, and there are large investments. Our direction is correct, what is lacking now is rational development strategy.
China has a tradition of self-reliance, so domestic companies will not stop on the road of independent research and development. However, in this era of economic globalization, Chinese companies are behind closed doors, and the practice of recreating wheels is not desirable. It is also necessary to find partners as much as possible to unite the forces that can be used. In this regard, there is an urgent need for partners in the Chinese market and capacity.
Intel: China's Best Partner in Developing Storage Industry?
The memory chips that have been determined to be developed in China are primarily flash memory. The representative of the national team is the Yangtze River storage technology of Wuhan in Wuhan. The memory industry is more difficult because of this, so it was originally intended to seek the cooperation of foreign companies.
Many people may remember the original rumor that Ziguang’s US$23 billion acquisition of Micron had no increase in storage chip prices, the US$23 billion offer was not low, and Micron’s stock price fell, and its revenue declined, if not by US government agencies. Rejected, there was still cooperation in the game. Then there were rumors that the company would cooperate with SK Hynix, seeking their memory technology license, the two sides set up a joint venture company in the country to produce chips, but eventually lost.
Two of the three memory companies did not have the possibility of cooperation. The remaining Samsung companies are even less likely. After all, they are the bosses, and there is no need for cooperation. Instead, they risk bringing the wolf into the room.
Because of this, domestic partners cannot be the three companies, but there may be a better choice - Intel Corporation. After Ziguang decided on a big strategy for the development of its chip business, it successively acquired Spreadtrum. Such semiconductor companies as the company, established Ziguang Zhanrui company, and later introduced Intel investment, the latter invested US$1.5 billion (approximately 9 billion yuan) to invest in Sharp company and obtained a 20% stake.
For Intel, investing in Violet is also beneficial. At that time, Intel was still subsidizing the flat-panel processor market. However, Intel’s model eventually proved to be unsuitable for the flat-panel low-cost product market. The same is true for the smart phone market. After the cooperation, Spreadtrum won the favor of Intel. Not only obtained the authorization of Intel's X86 mobile phone processor, but also 14nm process OEM opportunity, the two sides cooperated with a variety of 14nm process X86 mobile phone processors, from high-end to low-end have involved.
At this year's MWC show, Intel also signed a cooperation agreement with Spreadtrum. The two parties will continue to cooperate on the 5G network. I said these things are mainly related to the processor, 5G, and the memory chip, but it can be seen Violet's friendly and cooperative relations with Intel Corp., the cooperation between the two sides will become more in-depth.
This is not a guess, but it has gradually become a reality. Both sides have the same cooperation on the memory chip.
Not long ago, Intel had a public show with Ziguang, and Ziguang launched a series of flash memory products. The first one involved the UFS 2.1 product market, including Intel flash memory, which was the first and only domestic customer of Intel flash memory.
At the same time, the market is also rumored that the future of the purple will be authorized by Intel to produce 64-layer 3D flash memory. Behind this rumors, Intel and Micron decided to split their hands together. We all know that Intel's flash memory was developed together with Micron, and the two parties were established. IMFT co-produced the chips, but Intel decided three years ago to rebuild the chipset foundry in Dalian, China, as a 3D flash memory factory, and also reduced the number of shares in the IMFT factory, leaving only a factory share. Before Intel, Micron decided Both parties will break up completely after the 96-layer flash memory, and will probably be completely independent by early 2019.
Market Analysis Intel's move has cleared the way for cooperation with Unisplendour. In the future, it will authorize the Ziguang factory to survive 3D flash memory. Intel does not worry about any conflict between the two parties, because Intel is focusing on the high-end enterprise-class market in flash memory business. The consumer market is just playing the ticket nature. After working with Ziguang, Ziguang can be responsible for the consumer market, and the two sides can jointly increase their share.
Can Intel return in the memory market?
The purple light mentioned above, Intel cooperation is for the flash memory market, is not the same as the memory in this paper? So Intel will play a role in the development of memory in the purple light?
At present, the two sides have not announced the news in this regard, but we speculate that both parties have the same needs - Ziguang Needless to say, Intel also needs to develop memory, because this is also important for their future business transformation.
Now players know that Intel is the most powerful X86 processor company and has been ramping up the processor market for more than 40 years. However, Intel Corporation initially relied not on X86 processors but on memory chips.
They were founded in 1968. The first product was the 3101 RAM chip. In 1969, the first oxide SRAM chip was introduced. Later, due to fierce competition, the company developed a processor for the Japanese company. It inadvertently opened the door to the 8086 processor. After that, it abandoned the RAM memory business and focused on the processor.
In the past two decades, Intel has been the world's largest semiconductor company, and its technology is the best. However, in 2017 Samsung surpassed Intel to become a global semiconductor. The root behind this is memory, flash memory, Samsung. Revenue, profits soared, and Intel's PC processor business did not have such a good thing. It was also severely hit by AMD.
In 2015, Intel Corporation also announced a new strategy. Future Intel is no longer a simple PC company. Data centers, cloud services, autopilot, AI computing, and 5G markets are also Intel's priorities.
To this end, Intel has invested heavily in the acquisition of Altera's FPGA chips and Mobileye's autonomous driving company. Regardless of business, it will be necessary to provide Intel with a package solution instead of a simple supply chip. Memory chips and flash chips are also important. According to data, by 2020, the storage demand for Internet vehicles will reach 1 trillion bytes, which is 1000GB, and autopilot will need more than 300GB/s of system bandwidth. If Intel does not deploy the memory chip market as soon as possible, then the future will be Will be at a disadvantage.
Intel does not currently produce DRAM memory chips, but Intel is undoubtedly the right speaker in memory-related technologies, including the development of DDR memory, Intel has always been involved, and when Intel will support new technology standards will also affect the entire Industrial development plays an important role.
With Intel’s status, they are not suitable to intervene in large-scale memory production. If they invest in themselves, Intel will face many problems and the costs will not be as good as the purple light. Therefore, both parties have common needs for cooperation. In the future, Intel will provide convenience to Ziguang in terms of technology licensing. With the huge production capacity and capacity of the Chinese market, Ziguang will inevitably gain market share from the other three companies and change the original industrial structure.
