July 1, 2005

View from Silicon Valley- Denali Memory Report- June 2005

Passing along the latest from Denali.  Their take on the market is refreshing for its inside knowledge with lack of a slant.  (Underlines and comments added.)

* * * * *

Memory Market Outlook

Though there is a more optimistic mood for the overall chip business today than just a month or two ago, significant uncertainties still surround the outlook for the memory segment— ahead of memory makers’ 2Q05 reports, which almost certainly, and in almost every case, will show further reduced profitability from the earlier 1Q05 retreat.

The upturn of 2003-04 is over.

Whether price stability and improved margins take root in the second half of 2005 remains to be seen. Whether seasonal strength and product transitions mitigate against an always-tough competitive market also remains an open question that will not be decided until we move further into summer and fall. But for sure, the market is not falling over itself, topsy-turvy, to lower and lower, and unsustainable price domains.

There are some positives. Capital spending has slowed, even as earlier spent money still is yet to become productive in fabs worldwide. Demand remains reasonably strong almost across the board, from Top Dollar Markets of cell phones and PCs, to smaller memory market niches in myriad networking and automotive markets. Meanwhile, the larger Digital Consumer Electronics (DCE) transition, generally felt to be the driver of broad demand for the next decade, steadily gathers steam. As more of the portfolio is filled out and its various component markets gain critical mass, they can drive large amount of demand for DRAM, flash, and other assorted silicon. It is no small feat to replace the maturing and rather homogeneous PC market as a chip-industry driver. But that is what the diverse and chaotic array of ‘toys’, games and game consoles, communications devices, next generation HDTVs, music machines and their wired and wireless linkages hope to do. They are rollicking forward amid technology that is changing in unforeseen and unpredictable ways: hard to size, harder to play in, harder to grasp, harder to gain sustainable market dominance.

At the top level, pundits revised their overall industry forecasts within the past 30-60 days—all upward. Low-ball forecasts were moved up incrementally, with some resizing of the 2005-6-7 market trajectory. A very nice analysis, by a “Bull for 2005”, Future Horizons, can be read in their most recent Semiconductor Monthly Update (see www.futurehorizons.com), which pretty well states the case for a higher forecast. While their expectations for 2005 of “+15 percent growth over 2004” still seems ‘over the top’, it is not as unlikely as it was when they put their stake in the ground late last year. We shall see.

Down in the memory trenches, things are going many ways at once. DRAM prices, as low as they should go before too much red ink appears, are more stable. And after a 40 percent price drop in just six months, which no one can keep up with insofar as cost reductions, it is about time, if not long overdue. NAND flash pricing has been relatively stable, even after a large 2004 price drop-off of more than 60 percent/MB. [Editor's note: some NAND prices have fallen significantly in the last few weeks]  But unlike DRAM, NAND flash is still mostly profitable, and the suppliers cannot keep up with demand, especially for the lower density NAND flash that has been EOLed as the industry charges on to more and more dense devices. Memory makers are hoping that today’s apparent stability in pricing will prove to be the plateau from which to drive (1) major revenue and output growth in the hoped for buildup to end-of year sales; (2) PC DRAM product transition to second-generation and higher performance DDR2-667 DRAMs; (3) production transitions to high volumes and high yields in 100nm and 90nm processes; and (4) steady mastery of 300mm manufacturing processes and economics. For many companies, the high-depreciation period of their earlier 300mm investments is now winding down, and that Rubicon is crossed, too. 300mm production costs are lower than 200mm costs for an increasing percentage of vendors and products.


* * * * *
Company Financials for 1Q05
Revenues fall back, profits squeezed

Revenues in 1Q05 remained about flat for memory makers, but strong price pressures, especially in DRAMs and NOR flash, pushed memory industry profits down for all memory makers combined. Overall, sales for the 22 companies tracked by Denali totaled about $12.3B, compared with 4Q04’s $13.4B; this represents a drop of about 9 percent, from what is now seen to have been a seasonably weak 4Q04. Profits were harder hit, dropping nearly 50 percent (Both estimates take into account estimates for Intel’s flash and Communications operations).

Table 1 shows the WSTS data for the memory product segments for the past three quarters. The sales drop-off from 4Q04 to 1Q05, is shown in Table 2, below, and reflects the memory companies Denali tracks. It is somewhat greater than the decline shown in the WSTS memory data.

Prepared using a different methodology, WSTS data showed 1Q05 memory sales of $12.0B, off about 2 percent from 4Q’s $12.2B. WSTS gathers memory-only monthly sales data from most memory makers under a confidentiality agreement, aggregates this company data, and publishes the totals; Denali uses publicly available quarterly financial reports for the 22 companies it covers, which sometimes are ‘memory-only’ and sometimes include non-memory production by companies that are mostly memory.  Denali then publishes those data in tables in the DMR. Denali data includes some Denali estimates of profits, and for sales for some companies, and reports only ‘operating profits’ from companies that are diversified into several divisions.

Table 1. WSTS Memory Data, 3Q04 thru 1Q05 in $M
Revenues  3Q04   4Q04   1Q05
DRAM      7081   7230   6879
Flash     3929   3877   4046
NOR       2342   2186   1972
NAND      1587   1691   2074
SRAM       819    640    689
Other      445    490    373
Total    12274  12237  11987

Aggregate ASPs ($)
DRAM      4.37   4.27   4.05
Flash     4.60   4.54   4.50
NOR       3.45   3.31   2.92
NAND      9.12   8.73   8.71
SRAM      2.58   2.42   2.37
Other      NM     NM     NM
Total     3.36   3.38   3.32

Source: WSTS/SIA Blue Book

Due to differences in methodology, the data do not always match on a quarterly basis, but trend closely over the long term.

For 1Q05, the biggest impact on profits was felt by those making NOR flash and DRAMs. DRAM prices dropped more than 25 percent from the December quarter to the end of 1Q05, [Editor's note: DDR prices are down -40%, the average is lower mainly due to steady SDRAM prices.] making it very hard to hold profits, or even revenues, up to earlier levels. For DRAM makers, 2Q05 looks to be worse still, and some DRAM makers’ bottom line will likely turn red this quarter, after five to six profitable quarters since 3Q03’s uptick in demand and price stability.

On the NOR flash side, several things are at work. Intel, since its ‘reinvigorated’ strategy was instituted in late 2003, has been aggressively cutting prices (and costs) in the cell phone NOR flash space. In addition, Intel has been getting back into the broader flash market of lower density parts that it had deemphasized three years ago to focus on a “cell phones mostly” market strategy. Judging from the small indications Intel gives as to its flash profitability, it has reduced costs, and improved the unit’s bottom line, but is still likely very red, and making everyone else in the NOR flash business red, too. In 2004, Intel combined its two communications groups into a single group for financial reporting, but continued to call out flash revenues separately, and operating profits for the combined group. In 1Q05, Intel again realigned its operating divisions, and it is likely that ‘Intel’s flash profitability’ will be buried deeply in its Mobility Division, which itself is famously profitably, led by their Centrino product family.

That being said, we can probably tell something about Intel’s 1Q05 profits, by looking at the more visible results of its joined-at-the-hip soul mate and lifetime competitor AMD (Spansion). AMD showed a marked decline in profitability, from $-39M in 4Q04 to –$110M in 1Q05, with a growth in units shipped, in contrast to Intel’s statement that its units actually declined from 4Q04 to 1Q05.

But, 1Q05 was a grisly business for NOR flash makers, and 2Q will be more of the same.

There are other things at work on NOR flash, too. NAND makers would have you believe “The Future is all NAND”, but we believe it is not so simple. For a variety of reasons, NOR flash is locked into, or limited to, many diverse flash markets which face ceilings on how much flash they require and in how fast they are growing. Writing code for NOR flash costs something like $50 per line; one song on an MP3 player takes about 4MB. By far the most prominent NOR market is that of cell phones, in which almost every phone uses 64M- 256M of NOR flash, plus maybe some other NAND and RAM. But the flash market growth is, by and large, heavily skewed towards NAND for removable media, which constitutes the largest and fastest growing flash segment. And the more the cell phone industry moves to 3G phones, camera phones, and Internet downloading, the more NAND flash will make its presence felt even in that bastion of NOR dominance.

On the NAND front, after price declines of more than 60 percent in 2004, prices were mostly stable, and for lower density NAND (128M to 1Gb), prices actually drifted upward during the quarter and NAND vendors ‘moved on’. [Editor's note:  These figures are a few weeks out of date, during which more prices have softened.] Bit growth continued to explode, and was mostly dependent of strong 2Gb NAND sales from Toshiba and Samsung. The new NAND ‘kids on the block’, Micron, Hynix, Infineon, and ST Micro, continued to move forward rapidly.

However, coming late to market and facing the usual challenges of gaining a seat at established users against the NAND Titans who were making 100- 1000x more product than they were, their movement has not always been smooth. The established players were also way down the learning curve, especially in light of Toshiba/SanDisk and Samsung, ramping up their 70nm processes with 2Gb, 4Gb and 8Gb devices. And it will not get any easier.

In DRAMs, supplies continue to grow and grow, thus keeping downward pressure on prices despite a stably growing demand from PCs and other compute products. Micron, which shifted its strategy about 15 months ago to stress its strong presence in legacy SDR DRAMs, cellular RAMs and LP DRAMs, has found much success in an SDR market that had very stable prices in 1Q05. Meanwhile, other DRAM makers rode the roller coaster of DRAM prices down from about $4.35 (benchmark 256M) to about $2.75 in early 2005. Today, SDR commands a nice 50 percent premium for DRAM makers, and is especially nice for those vendors who continued to invest in shrinks and design improvements.