Market outlook of memory in 2018
DRAM's growth in 2017 comes mainly from the growth of server content, which has not slowed down.
At the end of 2016, a couple of memory market annual reviews and prospect reports, see, "only the 3Q's back memory market may begin to enter the supply of the memory market long cycle of the long cycle of the long cycle of memory." These words unconsciously smile, quite a feeling of the night.
Memory market economic cycle is in almost perfect competition market, Moore's law and the demand for the center with computers is shaped by three big pillars of market a rough model, now many conditions don't exist, then continue to use this model was dangerous.
Some of the predictions for 2018 are surprisingly consistent: DRAM continues to be strong, while NAND is slightly weaker in some applications, and my expectations are similar, but slightly more optimistic.
DRAM's growth in 2017 comes mainly from the growth of server content, which has not slowed down. Moreover, 5G will be implemented in China and South Korea in 2019, and there will be an early demand for Internet construction in 2018, which is one of the highlights of the 2018 winter Olympics in Korea. The high frequency is more buffer, and the server is not consumer product, it's less sensitive to price. In the experiment field demand for DRAM is big, the launching of the AI knew from the second generation of TPU systolic design data from AI processor DRAM to NAND repeatedly passed between caused much burden, need special designed to ease the burden. These are persistent hard requirements.
In terms of supply side, the sales of semiconductor equipment grew last year. The Korean market jumped to the top of the world, and the investment of DRAM equipment was far less than the peak of 2015. Why is the market so attractive that companies are so stingy with their investments? It is easy to understand the position of the company. The investment in machine equipment has not been paid for by Moore's law, and the fee amortization is done in five years, and this is a totally different set of financial rules. As for the production of the mainland, it is very different from the environment before Taiwan. As the market is oligopolistic, almost no mergers and acquisitions, joint development, may authorize the transfer, every step has infringement, so I'm afraid to enter mass production and some time.
The price of NAND has fallen somewhat in 4Q, mainly because it is now in the crossover period between SSD and HDD, and the price of NAND increases, and HDD naturally returns to the market, and the market share of SSD application in the data center is contained. This factor is still there for a short time, but when the process continues to evolve, and the HDD costs a lot of money and the cost of each unit goes away, it disappears. Another is so difficult to optimism is memory company capital expenditure growth last year was a 3 d NAND process equipment, and QLC (4 bits) and 9 x L bits of space to grow up, the benefit of the kind of Moore's law, which is why memory company equipment investment so generous, and 3 d technology and possibly subsequent benefits, Pi if use of MRAM.
But those who make DRAM to buoyant demand reason - content, bandwidth, Internet of things, and machine learning to NAND none shall not apply, and multiplier is bigger, this is the reason I report optimistic than normal!
At the end of 2016, a couple of memory market annual reviews and prospect reports, see, "only the 3Q's back memory market may begin to enter the supply of the memory market long cycle of the long cycle of the long cycle of memory." These words unconsciously smile, quite a feeling of the night.
Memory market economic cycle is in almost perfect competition market, Moore's law and the demand for the center with computers is shaped by three big pillars of market a rough model, now many conditions don't exist, then continue to use this model was dangerous.
Some of the predictions for 2018 are surprisingly consistent: DRAM continues to be strong, while NAND is slightly weaker in some applications, and my expectations are similar, but slightly more optimistic.
DRAM's growth in 2017 comes mainly from the growth of server content, which has not slowed down. Moreover, 5G will be implemented in China and South Korea in 2019, and there will be an early demand for Internet construction in 2018, which is one of the highlights of the 2018 winter Olympics in Korea. The high frequency is more buffer, and the server is not consumer product, it's less sensitive to price. In the experiment field demand for DRAM is big, the launching of the AI knew from the second generation of TPU systolic design data from AI processor DRAM to NAND repeatedly passed between caused much burden, need special designed to ease the burden. These are persistent hard requirements.
In terms of supply side, the sales of semiconductor equipment grew last year. The Korean market jumped to the top of the world, and the investment of DRAM equipment was far less than the peak of 2015. Why is the market so attractive that companies are so stingy with their investments? It is easy to understand the position of the company. The investment in machine equipment has not been paid for by Moore's law, and the fee amortization is done in five years, and this is a totally different set of financial rules. As for the production of the mainland, it is very different from the environment before Taiwan. As the market is oligopolistic, almost no mergers and acquisitions, joint development, may authorize the transfer, every step has infringement, so I'm afraid to enter mass production and some time.
The price of NAND has fallen somewhat in 4Q, mainly because it is now in the crossover period between SSD and HDD, and the price of NAND increases, and HDD naturally returns to the market, and the market share of SSD application in the data center is contained. This factor is still there for a short time, but when the process continues to evolve, and the HDD costs a lot of money and the cost of each unit goes away, it disappears. Another is so difficult to optimism is memory company capital expenditure growth last year was a 3 d NAND process equipment, and QLC (4 bits) and 9 x L bits of space to grow up, the benefit of the kind of Moore's law, which is why memory company equipment investment so generous, and 3 d technology and possibly subsequent benefits, Pi if use of MRAM.
But those who make DRAM to buoyant demand reason - content, bandwidth, Internet of things, and machine learning to NAND none shall not apply, and multiplier is bigger, this is the reason I report optimistic than normal!