WW NAND Revenue to Grow at 2018-2024 CAGR of Only 4% – Yole Developpement

Explosive memory bit demand from data-centric systems and applications
will drive long-term market growth
Despite some cyclicality and seasonality, the stand-alone memory market has experienced extraordinary growth over the past decade. This has been driven by important megatrends, such as mobility, cloud computing, AI, and the IoT, according The Status of the Memory Industry (€5,990) report from Yole Developpement, providing a 360° overview of the stand-alone memory market and its competitive landscape, detailing the technology and market trends in the elds of NAND, DRAM, NOR, eNVM, and other stand-alone devices.

NAND and DRAM account together for around 97% of the overall stand-alone memory market. Their revenues hit a registering high of around $160 billion in 2018, recording an impressive CAGR of 32% between 2016 and 2018.

At the end of 2018, both NAND and DRAM markets started experiencing oversupply caused by unseasonably weak demand, including lower-than-expected smartphone sales and a slowdown in datacenter demand. DRAM prices are projected to decline by around 40% this year and likely will not increase again until 2020. For NAND, the outlook is more positive with potential for tight market conditions in 2H19.

In the long-term, NAND and DRAM revenues are forecast to grow with CAGR 2018-2024 of 4% and 1% respectively. This is thanks to ever-growing bit demand fueled by novel AI/IoT applications and systems, such as smart cities, connected homes and intelligent factories, smartphones and Echo-like personal assistants, virtual and augmented reality and autonomous vehicles. All these rely on a massive amount of data and on the networks that connect them all. Thus, the coming rollout of fifth generation (5G) wireless technology will be critical for their future market expansion.

Notably, emerging non-volatile memory (eNVM) started making inroads into the storage-class memory (SCM) market with the introduction in 2017 of Intel’s Optane phase-change memory (PCM) products. As the new generations of Xeon server CPUs are conceived to be compatible with the new Optane persistent memory modules (NVDIMM), Intel might be able to gain significant business at the expense of Samsung and SK Hynix, who are now getting ready with their own PCM products.

Complex technology-scaling challenges require increasingly higher capital expenditures to achieve significant bit density growth

Continuous advancements in semiconductor memory technologies are critical for the technical and economic viability of newly emerging data-centric applications.

NAND and DRAM scalability was supposed to peak in 2020, but memory manufacturers and equipment players have found new solutions – just like in the past – to exceed this limit.

New manufacturing techniques include self-aligned multiple patterning, extreme ultra-violet (EUV) lithography and metrology, as well as high aspect ratio (HAR) etching to produce contact holes in 3D NAND devices. Novel advanced packaging methods for heterogeneous integration include 2.5D silicon interposer and 3D stacking. Together, these technologies create new opportunities for the semiconductor industry to increase the bit density in next-generation memory devices, to improve their bandwidth and reduce their power consumption and cost-per-bit.

However, with each technology generation bit growth is also becoming significantly more expensive. For instance, multi-patterning requirements are increasing with each node migration, resulting in additional processing steps and therefore more cleanroom space per wafer produced. Huge annual capital expenditure (CAPEX) and R&D investments, reaching several billion US dollars, are still being made in existing DRAM and NAND technologies. However, due to the degradation of the pricing environment initiated in late 2018, CAPEX investments are now being reduced in order to quickly achieve a balanced market condition. In the long-term, the CAPEX required per 1% of bit growth is expected to continue rising, as illustrated in the gure for the case of DRAM.

Entrance of Chinese players might trigger new phase of consolidation of memory market
Market concentration has accelerated dramatically in the last decade and is now very high, with three dominant NAND and DRAM players, namely Samsung, Micron, and SK Hynix, and two pure NAND players, namely Toshiba and SanDisk/Western Digital, holding a combined 95% market share.

Nowadays, it is unlikely that the memory market could consolidate further. No significant mergers or acquisitions were recorded in 2018 in the DRAM industry. The major event in the NAND business was the sale of Toshiba’s memory unit to an investment group led by Bain Capital and including Apple, SK Hynix, Dell, Kingston and Seagate.

Meanwhile, Chinese players could become a threat and might trigger a new phase of market consolidation in the long term. In China, central and local governments, in partnership with a number of private players, are investing billions of dollars to develop a local semiconductor memory ecosystem. The objectives are: to bridge the gap between domestic production and consumption, reduce dependency on the supply of global memory companies, and full huge memory chip demand in strong growth segments like mobile/ wireless, consumer, servers, AI, IoT, and automotive.

In the NAND business, Yangtze Memory Technologies Co. (YMTC) is the most likely to succeed out of all the publicly-known Chinese memory players, thanks to significant financial backing from government investment funds and a head start on R&D and manufacturing. The company’s volume production of 64-layer 3D NAND devices based on the newly-developed Xtackingtechnology is expected to take off in the second half of 2019.

On the other hand, DRAM in China is still in the technology-development phase. DRAM manufacturing is incredibly difficult and it will likely take a while longer for China to achieve competitive parity with the rest of the industry.

We expect that significant output from Chinese vendors could reach the market in 2020 for NAND, likely later for DRAM. Meanwhile, stand-alone NOR will remain the most solid memory business in China thanks to a well-developed local supply-chain system. In coming years, a possible relaxation of the China-US trade tension might open up new opportunities for China to acquire key companies across the memory supply chain, which could reinforce China’s position in the semiconductor memory industry.

Companies cited in the report include:
4DS, Adata, Adesto, Advantest, Alliance Memory, AP Memory, Apacer, Apple, Applied Materials, ASE Tech. Hold., ASML, Avalanche, Buffalo, Canon, Centon, CXMT, Cisco, CNE, Crocus, Crossbar, Cypress, Dell, Dosilicon, Etron, ESMT, Everspin, Facebook, Ferroelectric Memory company, Fidelix, Freescale, Fudan Microelectronics, Fujitsu, Fusion IO, GigaDevice, GlobalFoundries, Google, GSI Technology, H-Grace, Hikstor, Hitachi, HLMC, Honeywell, HP, Huawei, IBM, IDT, Imec, In neon, Intel, ISSI, JHICC, Kingston, KLA Tencor, Lam Research, Lapis, Lenovo, Longsys, Liteon, Lyontek, Macronix, Marvell, Maxio, Maxim, Materion, Mediatek, Microchip, Micron, Montage Technology, Nantero, Nanya, Naura, Nikon, NEC, NetApp, NetList, Numonyx, NXP, ON Semiconductors, Panasonic, Phison, Powerchip, Powertech, ProMOS Technologies, Qualcomm, Rambus, Reliance, Realtek, Renesas, Rohm, Samsung, Sandisk, Seagate, Semtech, Silicon Motion, SK Hynix, SK Materials, Smart Modular Technologies, SMIC, Sony, SPIL, SST, Spin Memory, STMicroelectronics, STEC, Swissbit, TDK, TX Instruments, Tezzaron, TEL, Teledyne e2v, Toshiba, Towerjazz, TPSCo, Transcend, Tsinghua Unigroup, TSMC, UMC, UniIC Semiconductors, Unimos Microelectronics, Viking, Violin Memory, Weebit, Western Digital, Winbond, XFab, XMC, YMTC, new Chinese memory players.