by Pete Singer, editorial director
May 18, 2010 - The flash memory market is young and growing, according to Eli Harari, founder, chairman and CEO of SanDisk Corp (Milpitas, CA). Speaking at The ConFab in Las Vegas, Harari said to expect a lot of growth in legacy computer products (digital cameras, digital camcorders, GPS, MP3 players, gaming, USB flash drive, and e-books), mobile applications and various computing applications, including netbooks, notebooks, tablets and mobile internet devices. He said this growth will be measured by million of gigabytes of data: about 10M in 2010, 20M in 2011, 40M in 2012, and 75M in 2013.
Harari, who has served as CEO of SanDisk since the company's founding in 1988, described how flash memory storage evolved from the early days in the '90s, where it was mostly used for industrial and military applications, early development of digital film and early PDAs. From 2000-2009, the digital consumer revolution found flash used as a ubiquitous, strategic enabler for portable personal content, and for the early days of the mobile internet, including 3G networks, iPhone and mobile apps and social networking.
|Eli Harari, Sandisk|
"What really drove flash to become such an important technology was very much the fact that we were able to drive the cost down relentlessly and very successfully,” Harari said.
NAND saw significant cost reductions from 2005-2009, on the order of 45%-65%, due to the move from 200mm to 300mm wafers, the introduction of mega-fabs, increased automation and immersion lithography. There was also a move from single-level cell (SLC), which offers higher performance for applications that are less cost-sensitive, to multi-level cell (MLC) which offers high performance at an attractive cost-per-bit. All of these rapid technology transitions resulted in unprecedented productivity, he pointed out.
From 2010-2013, Harari expects continued cost reductions, but at a slower pace of ~25% annually. He sees a transition from MLC to three-bit technology (called X3) for about 50% of NAND bits. SanDisk put X3 into production in the first part of 2008 using the company's standard 56nm flash technology. X3 provides more than 20% more die per wafer compared to standard MLC memory on the same technology node.
Looking down the road, Harari sees potential in 3D read/write memory in the form of a scalable cross-point diode array. Joint development with Toshiba is now underway in Yokkaichi, Japan, and although they have made good progress with the read/write layer, it is not yet ready for production. "At x8 equivalent, and given our know-how in 3D diode arrays, we believe 3D R/W is the most likely successor for NAND in the coming decade,” he said. "It could usher the second solid-state drive wave."
How will NAND and post-NAND fabs look in the coming decade? Harari said existing NAND mega-fabs at 1Xnm will be highly depreciated, generating attractive margins and "throwing off a long tail of cash." New NAND mega-fabs will be pricey (~$8B for 200,000 wafers/month), will need to support EUV, and will need to adapt to post-NAND technologies production, Harari predicts. While the NAND transition from SLC to MLC took three years (2002-2005), the transition to post-NAND will be more complex, expect for an extended period of overlapping NAND and post-NAND designs in production, he believes.
Like many semiconductor manufacturers, SanDisk's capital expenditures dropped dramatically in 2009. Harari said the company spent 48% of revenue on capex in 2007 and 2008 ($1.86B and $1.62B, respectively), but only 10% ($370M) in 2009. In 2010, he expects that $700M-$900M will be spent, or about 20% of revenue. "We believe that growing demand should drive a healthy industry-wide supply/demand balance in 2010, likely to continue into 2011," he said. The Toshiba-Sandisk JV fabs (Fab 3 and Fab 4) are rapidly transitioning to 32nm X3 and X4 technology, he said. "In 2010/2011 we are planning to acquire our 50% of remaining unused capacity in Fab4." Harari also noted that the industry has not built a new NAND fab in the past 2 years. "Investment in new NAND fabs will require healthy demand/supply balance, which is a prerequisite to a reasonable ROI on the massive capex required for a new megafab." -- P.S.