NewsCompany newsIndustry News



Cai Yimao: Our country's catcher has infinite opportunities in the field of semiconductor memory


On June 22, the 2009 National Entrepreneur Day and the Annual Conference of Chinese Entrepreneurs were solemnly held today in Beihai City, Guangxi. Cai Yimao, Vice President of the School of Information Science and Technology of Peking University, attended and delivered a speech at the Forum of the Third Conference on Industrial and Information Technology Achievements and the One-stop Development of Strong Industrial Foundation Industries.

Cai Yimao said that some entrepreneurs once said that many enterprises in China need special support when they start. In fact, our country has provided great help to the relevant enterprises, so enterprises need to think about how to make good use of policies when they get support.

Cai Yimao believes that a hundred-year-old shop will not fail, once the route is wrong, it may lead enterprises into difficulties. But on the other hand, this is an infinite opportunity for our country's pursuers and the beginners of the industry.

Following is the transcript of the speech:

Cai Yimao:

Leaders, friends of entrepreneurs and experts are delighted to be invited to this forum. In the past, I mostly gave reports to my academic colleagues. Today, I give reports to my friends, who are mainly entrepreneurs. I may publish some views that are not mature enough. If they are inappropriate, please criticize and correct them.

The topic of my report is the development of semiconductor memory technology and industry, which is related to today's one-stop licensing of industrial strong-base memory enterprises. Just now, a friend of entrepreneurs said that enterprises need special support when they start. I agree that in the current trading environment, how to make good use of our policies and give support to enterprises will be of great help to enterprises. I am also vice president of the National Memory Industry Alliance. Just now, I think you all noticed that the market of electronic information industry is very large, but our local enterprises still have a small share in the emerging information industry market. I know that leaders and entrepreneurs are working hard, so how do we do this? Everyone needs to think. Memory industry is facing a lot of competition, and the technology threshold is too high. What should we do? Do we have a chance? Just now, a friend of an entrepreneur said that the bigger the enterprise is, the more stable it will be, the more old stores will not fail for a hundred years. This is particularly good because, taking semiconductor memory technology as an example, information technology is developing rapidly. If we lose innovation or take the wrong route of technological innovation, it may bring great risks to formerly powerful enterprises. On the other hand, the rapid development of technology and the demand for innovation also have unlimited opportunities for industry catchers.

Memory is very important. It is not only used in consumer electronics of current terminals, but also used in many other devices and instruments. Especially in the era when data is king, many software companies and Internet companies pay more attention to memory technology, because the application of memory is indispensable to data centers.

Semiconductor memory is mainly divided into two categories: one is volatile memory devices, which will lose the information stored after power failure. But because it needs to be maintained by electricity all the time, its power consumption is a challenge. Especially now a major application market of memory is terminal, power consumption is more concerned. Therefore, it is a very important trend to develop lower power memory or other types of memory to replace it. Second, non-volatile memory devices, such as tape disks, here we talk about semiconductor memory, mainly flash, which has better reliability and speed than tape disks and is widely used in mobile phones, notebooks and other terminals. In fact, Flash is not so perfect, we have to use it, because we have not found something to replace it. Flash has many disadvantages, such as the need for 10-20V voltage, but our logic circuit has adopted 0.8-1V voltage, 0.8-10V has a large span, need to consume a large chip area and power consumption to power Flash, and large voltage will make it difficult to continue to reduce and improve Flash integration, which is a great limitation. Some special application scenarios, such as aerospace and automotive electronics, also bring opportunities to other memory, such as FRAM. Although the market is small, there are products on the market very early. In some segments of the market, some large companies begin to intervene first, and later may find it very hard to do, because the market is not big, so they are unwilling to increase investment, so the related technology has not been well iterated and promoted. On the other hand, large companies can not focus on the market, it may also bring opportunities to our local enterprises. Just now I saw a published data about the market share of emerging information technology of Chinese local enterprises. Zhaoyi Innovation has a large share in Flash market. Zhaoyi Innovation is our main board listed company. It can get the opportunity of technological iteration in the market, because when it invests in NOR Flash, Samsung and other companies just want to give up this piece. Zhaoyi Innovation makes good products through its own technology. So I have a little thinking, because I was also doing Flash research and development in Samsung, Korea. Samsung is a very advanced technology, but later did not, not because there is no opportunity or market for this technology, but for some large companies, the cost maintenance or profit generated by this market is less than its main products, so these opportunities can be local enterprises. Career assurance.

Most of the new memory is non-volatile, mainly our mobile, terminal applications, the most important thing is how to save data in case of power failure, to meet customer needs with faster speed and higher reliability. New types of memory include phase change memory, MRAM, and resistive RRAM.

So now DRAM + NAND is equal to 95% of the total output value of the storage industry. As we all know, the main players are enterprises outside the country, so the technology threshold is very high. Now, under the guidance of our country's big gene or policy, some enterprises have begun to do mainstream DRAM and NAND storage, including Yangtze River Storage, Changxin and so on. It is necessary for them to do these things. Even if the cost is high, we need substitutes, because this is a strategic thing. We can't give up strategic products. For entrepreneurs, making money may be the most important, but for countries, strategic needs are always the most important.

So let's take a look at the market. This is the latest data we have. In the next five years or so, NAND Flash will continue to develop, and its price will be cheaper. In the future, it will be a three-dimensional superposition, which is a visible technical solution. Then another memory is DRAM. The price has experienced rapid growth in 2017 and 2018. The follow-up forecast is that the memory market has periodic behavior. If the enterprise fails to step in this cycle and enter at the lowest point and the lowest point of the industry, many people will question the profitability of the industry. They will question whether it is right to invest in this way and how the low price breaks the cost line. What to do? But to give some suggestions, to do the memory industry, including the whole integrated circuit industry, there is no matter of quick success and instant benefit, we must have long-term planning and investment, in order to solve the problem, it is not like farmers planting land, I did not do it this year, next year the price is still here, but the integrated circuit is you do not do this year, 10 years do not necessarily do it back, this and the whole technology. Developing rapidly and correlatively.

DRAM is an important semiconductor memory, which is a very competitive industry and requires high technological innovation. Maybe some entrepreneur friends who are not doing integrated circuits or memory will ask us why we do such advanced technology? Our cars have BMW and Mercedes-Benz in the high end and market in the low end. Why must integrated circuits be aimed at the most advanced enterprises? Because memory or integrated circuits are basically standards, all standards are very strict, so for system manufacturers to come in, buy products that meet the standards pay more attention to the cost. So what is the cost associated with it? It is mainly related to technology, so strong technological innovation is needed to reduce costs.

The mainstream of NAND technology is three-dimensional. Samsung is optimistic that it can stack up to 500 layers, and Changjiang Storage says it can reach 96 layers this year. Meguiar and Toshiba have their own technological routes, but no matter what, now they have to be above 96 levels to compete in the market.

So do we have any other opportunities? Memory interface technology standards are also innovative, which is closely related to the system manufacturers, such as making controllers in addition to granules. So every technological development will bring a lot of opportunities to design companies, and possibly to our packaging manufacturers. It is very long for our country to cultivate a upstream and downstream chain of the memory industry. We can have excellent enterprises to make technological innovation in every link.

Technology and demand may lead to changes in the industry. Where is the change in the industry? In the data age, how to develop memory technology is a trend of development. This also brings a good opportunity for memory development. Now more and more development to the diversity of storage requirements, automatic driving has the demand for automatic driving, 5G has the demand for 5G, AI has a high demand for high data and high bandwidth, which also brings an opportunity. If a large company wants to maintain the needs of all customers, its cost burden is very heavy. If we can do a very sophisticated and specialized field, maybe we will. There will be opportunities.

Samsung itself pays more attention to the NAND storage scheme, because many companies want to do SSD after the cost is down, but if they want to face the data center, its advantage is cheap and large capacity, so it develops QSSD, because it has more iterations. So HBM2 is compatible with low power consumption, and its market price is in short supply. So DRAM has seized a good opportunity, sometimes the cost is not so important, it is possible to develop outside the mainstream memory, we are still exploring, when the demand is very large, if we have the opportunity to meet these technological innovations to meet the best combination of customer performance pursuit, it will bring us some opportunities.

We have entered the era of artificial intelligence from the era of PC. It may also be a great opportunity for us to bring about a variety of memory requirements.

So which part of memory is missing from the perspective of future technology? Memory technology classification is probably a triangle structure, with the highest cost, but the best performance. There is a huge gap between HDD and SDD, DRAM and SSD. This huge gap is mainly in speed. Our DRAM can do cache, but NAND Flash has a high voltage. Its device size is very large, its speed is in microseconds. There is a big gap between the class of 100 microseconds and the class of tens of nanoseconds.

Now the difficulty of the whole memory technology development is the storage capacity and the actual system's demand for storage. After so many years of development, this gap has not been tightened, so we need to research and develop technology layout other than DRAM and NAND. I think this is a good opportunity to break the gap between demand and technological innovation. Looking at the development trend of DRAM technology, our storage energy is 768GB. This storage mode is voltage reduction, cost reduction, cost and performance power consumption. But now there is no way to continue this trend confidently with visible technology and fumbled technology. Therefore, when it comes to 768GB, we need to update technology and technology. Architecture optimization. I particularly agree with the entrepreneur friend just now that technological innovation is absolutely the most core and most important element.

The three-dimensional structure of NAND is the kingdom to maintain the price advantage. To do this, we must do the strongest. Now we are talking about the new information technology, not the traditional elimination of backward technology, but the new technology must go forward.

Technological challenges, including the comprehensive demand for performance, capacity and scalability, are all from the industry. So in a dragon, we can see that there are financial and service enterprises in it, and large data e-commerce enterprises like Jingdong, which let system and enterprise users feedback back to the memory industry with time requirements. This will be a good ecological environment structure. 。

So what is the current mainstream memory problem? Just now, I also mentioned that NAND has such a large market, its life is still short, its delay is long, power consumption is large, DRAM's problem is power off, capacity is not as large as NAND, embedded memory is now mainly Flash, the mainstream is 40 nanometers, our SOC has been 14-7 nanometers, now some mainstream factories have provided some new embedded memory IP, that is to say. Ming Flash has been unable to meet the design requirements of advanced SOC.

So back to what I just said, why do we need to do research on new memory in academia and industry? That's because different needs need to be able to meet different applications. One is to make up for the missing memory, the other is that the embedded memory is still very missing, which can not meet the pre-evolution needs of SOC. Here are several major manufacturers, replacing or substituting factories, new storage or substituting factories, including DRAM, RRAM and so on. Many enterprises have their own embedded technology. There are also many American start-ups. It can be said that these enterprises are rising very fast. Which technology has the most development potential in the future? Although the academia has focused on several kinds, I think entrepreneur friends can also come in to see, the academia does not see the perspective that entrepreneurs see, there are various factors such as manufacturability, compatibility and so on need to be considered in order to decide which new type of memory has the most development opportunities.