In 2002, the worldwide semiconductor memory market grew by 6 percent, from $26.7 billion in 2001 to $28.4 billion. Memory revenues are forecast to grow by another 13 percent, to $32.2 billion, in 2003-but that's far short of the $54.7 billion in sales recorded during the boom year of 2000.

Stronger growth is expected to return to the memory market in 2004 and 2005 as a result of improving demand in the face of fab capacity constraints. In 2006 and 2007, a traditional semiconductor industry supply-side-driven downturn is projected. That downturn is expected to be the result of increased capital spending that is forecast to occur as semiconductor vendor revenues and profits improve in 2004 and 2005.

In 2002, the star memory market was that for NAND flash, where strong bit growth, of 106 percent, fueled revenue growth of 72 percent, despite steep price declines (down 33 percent per megabyte). In DRAM the reverse was true, with a firmer pricing environment in the first half of the year driving 33 percent revenue growth despite a historically low bit growth rate, of 39 percent.

In SRAM, continued weakness in the wired communications sector in particular and downward pressure on SRAM device pricing in general held the market back, with revenues declining by 33 percent. In NOR flash, sales in 2002 were down by 15 percent compared with 2001, as steady downward price pressure throughout the year acted to inhibit revenue growth. The markets for EPROM and mask ROM declined by 35 percent and 38 percent, respectively, as demand for those legacy technologies suffered because of generally weak electronic equipment production in a soft global economy. Similarly, a weak global economy and its effect on electronic-equipment manufacturing adversely affected the market for E2PROMs, which declined by 16 percent.

We expect a modest increase, of 13 percent, in memory revenues in 2003, with the DRAM, SRAM and flash markets all expected to grow. Electronic-equipment production drives semiconductor consumption, and two applications dominate and are critical to the memory market: digital cellular handsets and PCs.

The short-term health of the memory markets depends on a three-phase recovery in end-user demand for electronic equipment. Phase one, an improvement in the digital cellular-handset market, began in the second half of this year and is continuing, fueled by the incorporation into handsets of digital cameras and color screens as well as the transition to 2.5G and third-generation phones and to packetized data capability.

Phase two, higher PC unit shipment growth, driven primarily by a corporate PC replacement cycle, is forecast to gather momentum during the second half of 2003. Phase three calls for a more broad-based recovery in electronic-equipment production, especially in wired communications, as macroeconomic conditions improve.

Continued demand from mobile applications for increased code and data storage capability will fuel bit growth in both the NOR and NAND flash markets. That said, overcapacity and intense competition will combine to keep prices under downward pressure, so revenue growth in those markets is forecast at only 25 percent and 18 percent, respectively, in 2003.

The current weakness in the PC market means continued sluggish bit growth and continued downward pressure on mainstream PC DRAM pricing, the two critical factors in DRAM revenue growth. Consequently, DRAM revenues are forecast to grow at a modest 11 percent in 2003.

The SRAM market will benefit from the continued shift to feature-rich cell-phone handsets that use higher-density volatile memory devices. The markets for workhorse asynchronous SRAMs and high-density/high-speed synchronous SRAMs (e.g., wired communications) are not expected to recover until 2004, so SRAM revenue growth in 2003 is forecast at a mere 8 percent.

We believe that the semiconductor industry in general and the memory sector in particular will remain cyclical, and therefore we have factored into our forecast a supply-side-driven downturn in 2006. This forecast reflects the expectation that strong revenues and healthy profits will drive an investment cycle in 2004 and 2005. This investment cycle will be centered on 300-mm wafer, 90-nm technology and will bring significant new capacity to the industry in the 2006 and 2007 time frame.

Andrew Norwood is an analyst with Gartner Dataquest (San Jose, Calif.).