Yet again, the industry must contend with uncertainty as to when supply and demand will come into balance. Gartner Dataquest predicts that, despite the modest growth expected in supply late this year, supply and demand will not come into balance until mid-2002. And by that time DRAM companies will be facing another problem-figuring out how to balance their product mix.

With DRAM technology's future road map unclear, DRAM suppliers will find it increasingly difficult to forecast demand, and therefore production, among competing SDRAM, DDR SDRAM and RDRAM technologies.

Other sectors of the memory industry will not be hit as hard as DRAM suppliers. The least affected will be those in the flash market, where revenue is expected to be flat in 2001. Bit growth remains relatively healthy, as applications continue to consume higher-density flash devices, but a slowdown in the growth in production of handsets is biting into demand for code storage. Weak demand is occurring as supply has increased, due to investments in fab capacity and continued geometry shrinks. As a result, a sustained decline in the prices of code storage flash memory has occurred since the fourth quarter of 2000.

It is this pressure on prices, more than anything else, which will stall any growth in flash revenue this year. The market for data storage flash is also suffering from price pressure. Weaker consumer demand for digital cameras and Internet audio players, combined with a desire by OEMs for high-density flash cards, is producing consumer-friendly price points. There is no doubt that the outlook for the flash market is healthy in the long term. But recovery is unlikely in 2001; monthly revenues have fallen to a two-year low.

As expected, SRAM revenue took a dive in April as new, greatly reduced contract pricing levels hit the market. For the rest of the year, the SRAM market will continue to suffer from weak demand caused by an inventory buildup that is particularly acute in the communications sector, its single most important market. Rather mysteriously, SRAM sales held up well in the first quarter of 2001, but the precipitous drop in the second quarter does not bode well for the remainder of the year. Gartner Dataquest expects that SRAM revenue will decline by 24.3 percent for the year as a whole. It is too early to attribute falling SRAM sales to anything other than the effect of excess inventory, but the longer-term outlook for SRAM is far from clear. Competition from embedded memory in system level-integration devices and high-performance, low-power DRAM will continue to threaten the SRAM market, tempering long-term growth expectations.

Overall, Gartner Dataquest expects sales in the memory market to fall from $54.1 billion in 2000 to $33.8 billion in 2001. But that will be followed by partial recovery before sustained growth brings a new peak of around $87 billion in revenue in 2004.

Industry folklore has it that 1995 was the last great boom year for memory. Then, memory revenue exceeded $54 billion, driven by a $42 billion DRAM market. However, the industry hardly noticed that memory revenue again exceeded $54 billion in 2000. One explanation could be that memory revenue had become far more fragmented in 2000, when DRAM and SRAM combined to account for just over 70 percent of total memory revenue, compared with almost 90 percent in 1995. The big difference has been the emergence of the flash memory market, which is now firmly established as the second-largest memory market behind DRAM.

The shift in composition of total memory market revenue has occurred because of a fundamental shift in the application driver for the market-in the past it was the PC but now and in the future it will be mobile devices, most notably digital cellular handsets. The requirement in portable applications for nonvolatile storage of operating code and data files is what enabled the flash market to grow to account for a fifth of memory revenue last year; by 2005, that proportion is expected to increase to more than a third.

In addition, the shift toward mobile applications is increasingly driving technology development in DRAM and SRAM as well. Instead of cost per bit being the dominant technology driver, things like low power, chip scale packaging and multichip packaged "combo" devices have taken on equal importance.

As the memory market fragments and becomes more "application specific," some in the industry are suggesting that we have seen the last of boom-and-bust cycles. Rapidly slowing growth in the flash market of 2001, following triple-digit growth in 2000, would tend to suggest otherwise.

We are not going to see an end to memory market boom-and-bust cycles. What we can expect to see, however, are memory vendors increasingly trying to insulate themselves from the worst aspects of the commodity market by developing products for specific applications. That trend is already under way, most notably in SRAMs, where fast SRAMs have been developed with the networking sector in mind, and in the code storage flash market, where the technology road map has been largely dictated by the digital cellular handset OEMs.

Andrew Norwood and Richard Gordon are Memory Market Analysts at Gartner Dataquest (San Jose, Calif.).

• Memories follow boom/bust cycle

Return to 2001 Midyear Forcast