Led by NAND flash memory, worldwide memory revenue reached $33.7 billion in 2003, an increase of 18 percent over the 2002 total of $28.4 billion. Sales of NAND flash soared by 75 percent from $2.4 billion in 2002 to $4.1 billion last year. That huge growth was fueled by explosive demand for flash cards and USB flash drives, which led to a worldwide shortage of components as well as stable pricing.

NOR flash, DRAM and SRAM sales all increased by modest amounts in 2003 as megabyte growth was partially offset by continued downward price pressure. Last year, DRAM revenues were up 13 percent at $15.5 billion, SRAM sales increased by 16 percent to $3.4 billion and NOR flash memory revenues increased 16 percent to $6.8 billion.

In 2003, the markets for EPROM, E2PROM, mask ROM and other memory continued to diminish in significance, totaling $1.9 billion in sales and accounting for about 5 percent of the overall market.

For 2004 and 2005, memory market revenues are forecast to increase as an industry up cycle plays out. But an oversupply-induced down cycle will hit the market in 2006 before the next up cycle begins in 2007.

Memory revenues are forecast to reach $63.5 billion by 2008, representing a five-year compound annual growth rate (CAGR) of 14 percent.

The highest growth market over the five-year forecast horizon is expected to be that for NAND flash memory with a CAGR of 22 percent. At that rate, by 2008 NAND flash memory revenues will account for 18 percent of the memory total, up from 5 percent in 2001, 8 percent in 2002 and 12 percent in 2003. At the same time, NOR flash memory revenues will grow by a CAGR of 17 percent, increasing their share of the memory market from 20 percent in 2003 to 23 percent in 2008. DRAM, meanwhile, at a CAGR of 12 percent will be lower than that of the memory market average of 14 percent. However, the DRAM market will still be the largest memory market by 2008 with 48 percent of revenues compared to 40 percent for NOR and NAND flash combined.

In the past three years the NAND flash market has tripled in size, growing from $1.4 billion in 2001 to $4.1 billion last year. An explosion in demand for flash cards and USB flash drives, which combined account for about 80 percent of the market, has fueled megabyte growth in excess of 100 percent per year.

Seemingly insatiable consumer appetite for digital data storage capacity is providing the semiconductor industry with its purest commodity product yet. Although the market for NAND flash memory is still in its infancy, demand appears highly elastic. Unlike the PC and DRAM main memory, there is no ceiling on megabyte content per system for NAND flash, suggesting that as long as the price per megabyte continues to decline, megabyte consumption will continue to increase.

The growth for NAND flash will be supplemented longer term by the need for increased data storage flash in high-end digital cellular handsets. But here adoption will be resisted to some extent by the incumbent NOR flash technology.

The NOR flash memory market returned to growth overall in 2003 but it was really a tale of two markets. Supply of low-density devices--that is, 16 Mbits and below--became constrained as electronic equipment unit production rates improved while the leading NOR flash vendors concentrated production on higher-density devices — 32 Mbits and higher--for the cell phone sector. For lower-density devices, lead times lengthened and prices firmed up while downward price pressure was still a feature of the market for higher-density devices.

However, continued strong demand in the cell phone market, especially for higher-functionality handsets, will mean improving market conditions for higher-density devices through 2004 as excess capacity is burned off.

The most important trend in the NOR flash market is the proliferation of multichip packaged (MCP) volatile/nonvolatile memory components. An estimated 200 million MCP components were sold in 2003 for use in cell phones. That number could easily exceed 500 million by 2008 as higher-performance 2.5G and 3G handsets become mainstream.

Not only will more MCP components be shipped, but they will become far more complex. The simple combination of a NOR flash memory die and a low-power SRAM die is being superseded rapidly by MCP components containing any combination of NOR flash memory, NAND flash memory, low-power SRAM, pseudo-SRAM and low-power DRAM dice.

The requirement for such sophisticated cell phone memory solutions is being driven primarily by the need for increased data storage capability, which has lead some to predict the demise of NOR flash memory in favor of NAND in the cell phone. This may occur, but it is more likely, for performance reasons, that a combination of memory types will continue to be used.

Long-term megabyte growth in the DRAM market has slowed to an annual average of about 50 percent compared to the historic norm of about 70 percent. Saturation is affecting the PC sector, and the low-cost PC business model championed by Dell, not to mention the lack of megabyte-hungry software applications, will continue to keep a cap on DRAM megabyte content per PC.

In the broader SRAM market prices declined month to month through 2003 as demand remained weak. In the important telecommunications and networking segments of the market, although there was significant design-in activity, a recovery in demand proved elusive. However, as the global economy improves and corporate spending on IT infrastructure gathers momentum beginning in 2004, the SRAM market--particularly for high-speed, high-density synchronous devices--will recover.

Longer term, SRAM market growth will be limited because of further integration of SRAM arrays by logic devices and by replacement of low-power and pseudo-SRAM by low-power DRAM.

Richard Gordon is research vice president and research area leader for emerging technologies and semiconductors at Gartner Dataquest (Lowell, Mass.).