Online information source for semiconductor professionals

The three major memory markets, DRAM, NAND and NOR, have suffered terribly in 2007. Rather than poor end-demand or lack of killer apps for consumers to consume, the problem is self-inflicted as demand has been incredibly strong but supply has been even stronger. 

Massive capacity increases by all the major players, particularly in both DRAM and NAND, has seen ASPs for both fall drastically in 2007. DRAM ASPs declined nearly 40 percent in 2007 and many market researchers believe than no manufacturer made money on DRAM. NAND, on the other hand, has now seen declines of 60 percent two years in a row!

Greater economies of scale via 300mm mega-fabs, higher yields, shorter cycle times and obsessive fabrication cost reduction strategies have failed to keep pace with the rate of ASP decline, forcing the majority of memory manufacturers to start reporting losses for 2007.

Importantly, the writing was clearly on the wall in 2006 as prices fell faster than most memory manufacturers had projected, but fab ramps continued unabated.

In the case of DRAM, clear evidence from independent market research firms that the Vista OS adoption was not going to be better than that of XP and that the adoption rate, especially in the important corporate space, would take much longer than the hype from Microsoft was implying, should have made DRAM manufacturers curtail expansions to more realistic levels - but it didn’t!

If that wasn’t enough cause for DRAM manufacturers to take a reality pill, then the announcement by the likes of Texas Instruments that it would be a late corporate adopter of Vista because there was no compelling reasons to change from XP (for about 2 years), should have. If a tech company can’t be bothered then something is fundamentally wrong!

Of course, we did see an average bit growth increase in PCs as Vista launched, but PC manufacturers only met DRAM requirements for the full Vista OS in top-of-the-range systems with 2G, while budget systems continued to increase memory content at a much slower rate.

Ironically, the massive price drops in DRAM through 2007 had little effect on the budget end of the market, though the memory content as we start 2008 is only now coming close to the Vista optimum requirements.

What seems to be happening is that the PC manufacturers are simply not passing on the significant DRAM cost savings they have seen throughout 2007 when a massive oversupply raged.

Listen to recent quarterly conference calls from HP and Dell and you will realise that executives clearly noted that business margins were better because of lower component costs. Read DRAM!

The hyped demand expected from Vista didn’t materialise to the degree expected from the memory manufacturers even when overall PC sales in 2007 showed strong double-digit growth. The situation would have been worse without that core demand, especially in notebooks.

The critical mass expected from Vista has been slower to arrive but bit growth has been double that of 2006. DRAM is therefore in a mess.

Some ray of light is that Micron, Qimonda, Nanya and Inotera have lowered capital spending plans for 2008 and even though die shrinks are a top priority, the actual bit growth from these companies should be less than last year’s!

Samsung and Hynix haven’t broadcast any real intentions to reduce spending in 2008, while profits have been cut so severely that funding CapEx plans in 2008 will be tough.

There has already been the push-out of Qimonda’s latest 300mm fab in Singapore and judging by the drop in equipment orders from many DRAM manufacturers, the emphasis is on technology buys for faster node migrations rather than capacity buys.

Indeed, there seems to be no new 300mm fab announcements expected from DRAM manufacturers in 2008, also indicating that the mess needs fixing - and fast.

The problem still seems to be in the NAND space as Toshiba, Samsung and Hynix continue to spend and ramp aggressively in 2008. Toshiba is currently ramping Fab 4 all through 2008 and the days of switching from DRAM production to NAND in the same fab seem over due to the faster-paced NAND node migrations.

When mixed memory producers give CapEx projections, just think NAND rather than DRAM for 2008.

With respect to DRAM, I would expect a return to a better supply/demand scenario in the very latter half of this year, way too late to make decent profits for the big league players and little if any hope for the rest to return to full-year profitability.

However, NAND rapidly needs new markets to take off in 2008, which looks unlikely. The PC market is still a few years away from high unit demand and the rise in storage in mobiles isn’t enough to get the NAND market back in balance.

Both NAND and DRAM have critical mass but the mess will take another year to sort out, if we are lucky!