Asp flash form

EDO DRAM, Flash Memory to be big players - 1996 industry forecast for semiconductor memory industry - Industry Trend or Event

Mountain View, Calif.--Extended data out (EDO) DRAM and flash memory will capture headlines in 1996, according to industry observers--the former by dominating the main memory market, the latter by promoting the advancement of handheld consumer electronics and wireless communications.

A majority of senior executives agree that supply will continue to chase demand on both fronts, but the capacity bottleneck should loosen as the year progresses and more fabs are built or converted from other uses. EDO DRAM should finish the year strong, perhaps losing some ground on the back end of 1996 to the emerging synchronous DRAM market.

"It's very clear as we move into the first half of next year that EDO is replacing fast page," said Hitachi's memory marketing manager Dan Mahoney. "It's the predominant solution for PC main memory. Our forecast is that even though synchronous will take off and ramp throughout the year, it will not surpass EDO in terms of volume in '96. Considering EDO came from virtually nowhere in 3Q95, it's quite a remarkable transition."

"The bulk of PCs are in the 75-100MHz range, and EDO is satisfying that requirement," said Avo Kanadjian, Toshiba's director of memory marketing. "In 1996 we'll see the market moving to 120-133MHz and EDO plus cache will be used. But in '97, at 150MHz and above, we will start using synchronous DRAM."

Market research houses Dataquest and Integrated Circuit Engineering (ICE) both foresee the MOS memory market growing to about $65 billion next year, a 38 percent increase by Dataquest's estimation. Of that market, DRAM is expected to account for about $50 billion, 23 percent more than 1995 but far less than the 75 percent leaps it experienced over each of the last two years, according to ICE. Bipolar memory, which ICE described as a dying technology, should generate only about $150 million in 1996.

"The DRAM market is definitely cooling off, but at the same time any other industry segment growing at 23 percent is going to be happy," said ICE senior market analyst Brian Matas.

Having made its projection, which is about half of what Dataquest is predicting, ICE recognized that last year's phenomenal growth far surpassed even the most optimistic industry estimates.

DRAM manufacturers will continue moving to 16-megabit densities, phasing out the tenacious 4M which will still make a sizable showing. However, Dataquest says difficulty overcoming noise spikes inherent in the shallow configuration of the much sought after 1M x 16-bit DRAM organization will continue to haunt vendors. The technology problem, which apparently has been overcome by vendors including Toshiba, NEC and Samsung Semiconductor, limits other manufacturers to x1 and x4 16M DRAM configurations, and, because of bus requirements, prevents them from filling the granularity needs of OEMs.

"With the 1M x 16 configured 16M part, PC manufacturers can make an 8-megabyte Pentium PC or a 4MB 486, and the increments of granularity are in 8MB chunks at a cost per increment of $200," said Dataquest memory analyst Jim Handy. "If everybody started out with the x4, we'd need 32MB of main memory at a cost of $800."

The combination of new capacity and improved x16 organization could conspire to drive DRAM prices down significantly in 4Q96 and early 1997, according to Dataquest's Mr. Handy.

"DRAM prices are twice now what they would be in a supply/demand balance," he said. But ICE's Mr. Matas warned that a lower ASP could unleash a pent-up rush on DRAMs and again upend the delicate balance.

"If the ASP lowers in '96 due to added capacity, we sense there may be a lot of people waiting to upgrade their main memory who have been waiting on the sidelines," he said. "If enough of them come forward, this could shift supply/demand out of whack again."

"Flash will continue to be an extremely strong technology, if not the strongest," said Bill Howe, Intel's VP/GM memory components division."We're doing a pretty good job going out and talking with customers, letting them know how much they can expect next year. Everyone will get a lot more, but there will still be constraints. The rapid growth flash has experienced this year will continue next."

Having proven itself in more pedestrian applications such as PC BIOS and mass storage, flash is poised for swift penetration into the digital cellular, digital camera and digital voice recording markets next year, with its showing in the PCMCIA and small form factor PC card arena improving steadily. The non-volatility of flash, combined with its low power consumption, should invigorate a 3.3V market that was just awakening to the technology in 1995. Overall, observers expect the flash market to grow by 35-50 percent in 1996, with automotive applications increasing by as much as 70 percent as flash begins to replace EPROM.

Even so, flash will account for only about four percent of the MOS memory market, with sales of $2.3 billion projected for next year.

"Some of those applications are pretty nifty, but even with those, we don't see flash getting bigger than a sizable niche market," said ICE's Mr. Matas. "But flash is the only other memory segment other than DRAM which will either grow or hold steady through the rest of the decade."

"Flash is not meeting the aggressive expectations of mass storage," said Toshiba's Mr. Kanadjian. "We still think mass storage will employ flash at densities of 64M and above. That's what it will need for cost/bit (parity)."

Some vendors believe the supply-constrained flash arena will open up to fill existing markets once capacity ramp-ups planned for 1996-97 make more chips available.

"The last two years the market has been deprived of flash memory," said Walid Maghribi, AMD vice-president, non-volatile memory division, "Look at cellular phones. Not every phone is shipped today with flash because vendors can't get all they need. As capacity grows, we will begin filling more demand in the same application."

While the emergence of flash has prompted a growing list of companies to explore the technology, nothing on the horizon is likely to unseat flash veterans Intel and Advanced Micro Devices (AMD) from their grip on roughly 80 percent of the market. There are, however, flash technologies such as AND, NAND and DINOR which are aiming to dominate.

"At higher densities we expect to give (flash vendors) a serious challenge," said Mr. Kanadjian. "Our NAND technology ultimately will replace disk drives for solid state memory. We'll see 64M in '97, with mass production volumes of solid state memory conservatively by the end of the decade."

Others disagree: "In early '95 we were getting approached once a month by someone with a new flash technology," said Mr. Maghribi. "NOR and AND are major players and will stay. AND will be used mostly for mass storage applications like memory cards. But alternatives to NOR, like DINOR and NAND are not as reliable. These companies are way behind in flash."