March 11, 2021
It’s been a topsy-turvy period for the memory market, and it’s not over.
So far in 2020, demand has been slightly better than expected for the two main memory types — 3D NAND and DRAM. But now there is some uncertainty in the market amid a slowdown, inventory issues and an ongoing trade war.
In addition, the 3D NAND market is moving toward a new technology generation, but some are encountering yield issues. And suppliers of both 3D NAND and DRAM are getting new competition from China.
After a slowdown in 2019, the memory market was supposed to rebound this year. Then, the COVID-19 pandemic struck. Suddenly, a large percentage of countries implemented various measures to mitigate the outbreak, such as stay-at-home orders and business closures, among others. Economic turmoil and job losses soon followed.
As it turned out, however, the work-at-home economy drove unforeseen demand for PCs, tablets and other products. There also was booming demand for servers in data centers. All of this drove demand for memory, logic and other chip types.
The ongoing U.S.-China trade war continues to create uncertainty in the market, but it also caused a wave of panic chip buying. Basically, the U.S. has launched an assortment of trade restrictions for China’s Huawei. So, for some time, Huawei has been hoarding chips, driving up demand.
That’s coming to an end. To do business with Huawei, U.S. companies and others will require new licenses from the U.S. government after Sept. 14. Many vendors are cutting ties with Huawei, which will impact chip demand.
All told, the overall memory market is complicated, and there are several unknowns. To help the industry gain some insights into what’s ahead, Semiconductor Engineering examined the markets for DRAM, 3D NAND, and next-generation memory.
Today’s systems integrate processors, graphics, as well as memory and storage, are often referred to as the memory/storage hierarchy. In the first tier of that hierarchy, SRAM is integrated into the processor for fast data access. DRAM, the next tier, is separate and used for main memory. Disk drives and NAND-based solid-state storage drives (SSDs) are used for storage.
2019 was a tough period for DRAM, punctuated by lackluster demand and falling prices. Competition has been fierce among the three top DRAM makers. In the DRAM market, Samsung is the leader with a 43.5% share in Q2 of 2020, followed by SK Hynix (30.1%) and Micron (21%), according to TrendForce.
The competition is expected to intensify with a new entrant from China. China’s ChangXin Memory Technology (CXMT) is shipping its first 19nm DRAM line, with 17nm products in the works, according to Cowen & Co.
It remains to be seen how CXMT will impact the market. In 2020, meanwhile, the DRAM market is a mixed picture. In total, the DRAM market is expected to reach $62.0 billion, roughly flat from $61.99 billion in 2019, according to IBS.
The stay-at-home economy, coupled with the datacenter server boom, propelled strong DRAM demand for the first half and the third quarter of 2020. “Key drivers for growth in Q1 through Q3 2020 were datacenters and PCs,” said Handel Jones, CEO of IBS.
Today, DRAM vendors are shipping devices based on the 1xnm node. “We are seeing stronger DRAM demand in Q3 as DRAM suppliers begin to ramp up the ‘1nmy’ and ‘1nmz’ nodes,” said Amy Leong, senior vice president at FormFactor, a supplier of probe cards for chip testing applications.
Now, however, there are fears of a slowdown in the latter part of 2020. “In Q4 2020, there is some softness because of slowing demand in datacenters, but it’s not a deep drop,” IBS’ Jones said.
So far, meanwhile, it’s been a lackluster year for memory demand in smartphones, but that could soon change. On the mobile DRAM front, vendors are ramping up products based on the new LPDDR5 interface standard. The data transfer rate for a 16GB LPDDR5 device is 5,500Mb/s, approximately 1.3 times faster than the previous mobile memory standard (LPDDR4X, 4266Mb/s), according to Samsung.
“We expect increasing mobile DRAM and NAND demand into calendar 2020 on higher production of flagship 5G smartphone devices that carry higher DRAM content,” said Karl Ackerman, an analyst at Cowen, in a research note.
5G, a next-generation wireless technology, is expected to drive DRAM demand in 2021. The DRAM market is projected to reach $68.1 billion in 2021, according to IBS. “In 2021, the key driver for growth will be smartphones and 5G smartphones,” IBS’ Jones said. “Also, data center growth will be relatively strong.”
After a period of sluggish growth, suppliers of NAND flash memory also hope for a rebound in 2020. “We are optimistic about the long-term demand for NAND flash memory,” FormFactor’s Leong said.
In total, the NAND flash memory market is expected to reach $47.9 billion in 2020, up 9% from $43.9 billion in 2019, according to IBS. “Key application drivers in Q1 through Q3 2020 were smartphones, PCs and datacenters,” IBS’ Jones said. “We’ve seen some softness in demand in Q4 2020, but it’s not significant.”
In 2021, the NAND market is expected to reach $53.3 billion, according to IBS. “Key drivers in 2021 will be smartphones,” Jones said. “We see an increase in volumes and well as increasing NAND content per smartphone.”
In the NAND market, Samsung is the leader with a 31.4% share in the second quarter of 2020, followed by Kioxia (17.2%), Western Digital (15.5%), SK Hynix (11.7%) and then Micron (11.5%) and Intel (11.5%), according to TrendForce.
If that’s not enough competition, China’s Yangtze Memory Technologies (YMTC) recently entered the 3D NAND market with a 64-layer device. “YMTC will have relatively strong growth in 2021, but its market share is very low,” Jones said.
Meanwhile, for some time, suppliers have been ramping up 3D NAND, the successor to planar NAND flash memory. Unlike planar NAND, which is a 2D structure, 3D NAND resembles a vertical skyscraper in which horizontal layers of memory cells are stacked and then connected using tiny vertical channels.
3D NAND is quantified by the number of layers stacked in a device. As more layers are added, the bit density increases in systems. But the manufacturing challenges escalate as you add more layers.
3D NAND also requires some difficult deposition and etch steps. “You are using different chemistries. You are also after certain etch profiles, especially for high-aspect ratio etching or what they call HAR. For 3D NAND, that’s become extremely critical,” said Ben Rathsack, vice president and deputy general manager at TEL America, during a recent presentation.
Last year, suppliers were shipping 64-layer 3D NAND products. “Today, 92- and 96-layer 3D NAND devices are common,” said Jeongdong Choe, senior technical fellow at TechInsights. “These devices are common in mobile, SSDs and the enterprise market.”
128-layer 3D NAND is the next technology generation. Reports have surfaced that there are some delays here amid yield issues. “128L has just been released. 128L SSDs have just been released on the market,” Choe said. “It’s a little delayed. Yield issues are still there, though.”
It’s unclear how long the problem will last. Nonetheless, suppliers are taking different routes to scale 3D NAND. Some are using the so-called string stacking approach. For example, some are developing two 64-layer devices and stacking them, forming a 128-layer device.
Others are taking another route. “Samsung kept the single stack approach for 128L, which involves very high aspect ratio vertical channel etching,” Choe said.
The industry will continue to scale 3D NAND. By the end of 2021, Choe expects 176- to 192-layer 3D NAND parts will be in risk production.
There are some challenges here. “We are optimistic about 3D NAND scaling,” said Rick Gottscho, CTO of Lam Research. “There are two big challenges in scaling 3D NAND. One is the stress in the films that builds up as you deposit more and more layers, which can warp the wafer and distort the patterns, so when you go double deck or triple deck, alignment becomes a bigger challenge.”
It’s unclear how far 3D NAND will scale, but there is always demand for more bits. “There is strong demand long-term,” Gottscho said. “There is explosive growth in data, and data generation and storage. All of these applications for mining the data are going to feed new applications for more data, so there is an insatiable demand for data and to store the data forever.”
For some time, the industry has been developing several next-generation memory types, such as phase-change memory (PCM), STT-MRAM, ReRAM, and others.
These memory types are attractive because they combine the speed of SRAM and the non-volatility of flash with unlimited endurance. But the new memories have taken longer to develop because they use complex materials and switching schemes to store data.
Of the new memory types, PCM has been most successful. For some time, Intel has been shipping 3D XPoint, which is a PCM. Micron also is shipping PCM. A nonvolatile memory, PCM stores data by changing the state of a material. It’s faster than flash, with better endurance.
STT-MRAM also is shipping. It features the speed of SRAM and the non-volatility of flash with unlimited endurance. It uses the magnetism of electron spin to provide non-volatile properties in chips.
STT-MRAM is offered in standalone and embedded applications. In embedded, it’s targeted to replace NOR (eFlash) at 22nm and beyond in microcontrollers and other chips.
ReRAM has lower read latencies and faster write performance than flash. In ReRAM, a voltage is applied to a material stack, creating a change in the resistance that records data in the memory.
“ReRAM and to some extent MRAM have been affected by the lack of successful volume use cases,” said David Uriu, technical director of product management at UMC. “Each technology from PCM to MRAM to ReRAM have had their strong and weak points. We have seen exciting predictions on many of these technologies, but the fact is that they are still in the works.”
While PCM is gaining steam, the other technologies are just taking root. “The issue of maturity in product adoption is what needs to be demonstrated over time to gain confidence in the solutions capabilities,” Uriu said. “Issues of costs, analog performance and usage cases in general have been brought forward and only a few are meeting the challenges. The majority are simply too risky to bet production and total cost-of-ownership exposures.”
This is not to say that MRAM and ReRAM have limited potential. “We do see future potential in MRAM and ReRAM. PCM, while comparatively expensive, has been shown to work and has begun to mature,” he said. “Our industry continuously improves the materials and use cases associated with developing the maturity acceptance of these newer memory designs and these will be brought to market for advanced applications like artificial intelligence, machine learning and processing-in-memory or computing-in-memory applications. They will expand into the many machines we use today for consumer, smart IoT, communications, 3D sensing, medical, transportation and infotainment applications.”(From Mark LaPedus)