August 25, 2005
View from Silicon Valley- Denali Memory Market Outlook
Here is the August, 2005 memory report from Denali:
* * * * * *
Memory Market Outlook
RAM price slide stops; NAND price slide resumes
What was observed to be a recent pause in price drops for DRAMs in the last issue
of DMR has continued, and in fact, brought with it some price increases, however modest. As it usually does, this has convinced
some DRAM makers that they will get a breather from ever-lower prices, and will be able to reverse their financial losses
(to be discussed further in next month’s DMR), and again start making money. As you can see in the Micron Technology
Fiscal 3Q05 results, the last several months have proven to be very tough for DRAM makers, after only about 4-5 quarters of
modest profits. Along with Micron Technology, Infineon, Nanya, ProMOS and Elpida all lost money in DRAMs in 2Q05.
The causes of the current DRAM price stability are several, not the least of which
are the movement of many thousands of DRAM wafer starts per month over to make more profitable NAND flash, at Samsung, Hynix,
Infineon and Micron. In addition, after a quite strong CapExp investment climate in 2003, continuing on throughout most of
2004, spending has tapered off and capacity is not being added as fast as formerly.
Historically, DRAM makers have consistently added production capacity -- too much
and too early -- thereby swamping their own boats so they have to bail water again sooner than they might otherwise. Alternatively,
they could run a bit behind demand with better, lower cost production that does not add excess DRAMs to markets that are ill-equipped
to absorb them—and continue to shift wafer starts to more profitable products.
As we have noted before, one die shrink a year for all DRAM portfolio products (typical
in the DRAM business), plus the natural transition to denser DRAMs and more cost-efficient designs, automatically generates
about 40-50 percent bit growth by itself, without adding any DRAM wafer starts. If you add in today’s very productive
transition to 300mm wafer, and it is easy to see how this supply easily exceeds recent history’s DRAM MB demand growth
of 40-55 percent. (You can read a good analysis of the industry’s Cap Exp issue in Future Horizon’s ‘The
Semiconductor Monthly Update Report, July 2005’ on pages 11-12, at
, along with some of FH’s analysis which we mentioned in our previous
DMR 4-3.)
Investment restraint can do no harm to DRAM makers, except to perhaps allow some
DRAM makers to cede some DRAM market share to other DRAM makers. But less DRAM capacity is more profitable for everyone: net,
net and net. And besides, who wants ‘DRAM market share’, when it correlates so strongly with financial losses,
as it has for so long? Why not focus more on cost reduction strategies, develop ‘burst capacity expansion’ and
‘quality of portfolio’ strategies that do not overwhelm demand in total production capacity and supply, in the
necessary pursuit of improved financial returns?
On the other hand, NAND flash, which constitutes an apparently insatiable market
and has been something of a safe haven for some DRAM makers for many months, is now recently again competitive, which means
lower prices for consumers, and lower profits for vendors. Still, the reduction in revenues and profits in the NAND flash
market is far less than what the industry stands to gain with improved DRAM pricing The DRAM market was $2.01B in June; NAND
flash was $758M, a more than 2.65x multiple.
In addition to these top line fundamentals, the market transitions from DDR1 to DDR2
is moving ahead fast these days, as DDR2 prices are within 5-10 percent of DDR1—with lower power and higher performance,
as a consequence, virtually for free.
And, as low as prices are for mainstay 256M DDR DRAMs, prices per MB for 512M’s
are even lower, with 512M low-water mark within the past several months coming in sometimes in the low $4.00 range, against
prices for 256M’s in the mid-$2.00 range. And in addition to the DDR1-2 transition, and the 256M to 512M transition,
there is a steady and broad improvement in the performance of both the DRAMs and the systems that use them.
News and New Products
Infineon Stands Pat in DRAMs…so says Manager Magazine
Putting to rest the latest rumor that it would vacate its DRAM market position by
spinning off the DRAM unit with a Fall 2005 IPO, Infineon’s CEO Wolfgang Ziebart has told Manager Magazine that he has
no such plans, but cautiously being quoted as, "the measures that outsiders recommend are not appropriate." This leaves him
an opening to suggest any ‘alternative solution.’ to Infineon’s DRAM financial problems, and many options
are being bandied about the industry, among them a possible merger of DRAM operations with Nanya, with whom Infineon already
shares common DRAM technology roadmap and a Joint Venture fab in Taiwan.
Of course, Infineon is not alone. NOR flash prices and investment returns threaten
also to throttle AMD (through Spansion), and even Intel, both of which are under some pressure to either right their ‘memory
ships’, or cut them loose. As we discussed in earlier DMRs (See further discussion of "History and Rationale for DRAM
Market Exits" in DMR 3-5 and DMR 3-6), it is no mean feat to jettison such a large revenue base, and headcount, even while
it is draining the larger corporation of funds that might better be used to defend or strengthen existing market positions,
or develop new opportunities. Even though DRAM prices are seeming to stabilize today, and maybe even move up, the fundamentals
for the DRAM business, especially, are still grim, with only a few suppliers making money (See also, Micron Reports 3Q05 Financial
results, below).
The same net that is cast upon Intel and AMD/Spansion’s flash operations is
also covering ST Micro, which is increasingly mentioned as one company that ‘should’ unload its flash memory business.
Totaling about $50B in revenues and consistently 20-25 percent of the total semiconductor
industry revenue base (at times), but sometimes more than 100 percent of its losses, these massive markets (1) command immense
financial and manufacturing resources, (2) are essentially non-exitable once the commitment is made, and (3) have killed more
than one career and billions of investment dollars. As we said earlier, probably no one has ever regretted leaving DRAMs,
and maybe no one ever regretted leaving flash, either. What about SRAMs or EPROMs? The largest chip company for the last decade,
Intel itself, was once the market leader and technology foundation stone of DRAMs, Slow SRAMs, and EPROMs, but eventually
left them all. In addition Intel abandoned a #4 position in bipolar PROMs in the early 1980s, joined the mass exodus of magnetic
bubble memories in 1982, abandoned a market leadership position in fast SRAMs (The first MOS SRAMs to compete with bipolar
SRAM speeds), and left a modest mask ROMs position. For Intel, the potential for making sustainable money in hard-to-differentiate
commodity memories was just not there, and depended on the vagaries of the marketplace, almost regardless of one’s technical
skills and commitment" win the week, lose the month-year-eternity.
Samsung alone is making memory money, more money than anyone, in each and every product
class, showing that it CAN be done, with the right formula and execution. (See Samsung articles in DMR 1-7 and 2-1, plus several
interviews with Samsung Management in various other DMRs.)
Whether the memory market is big and diverse enough for more than one ‘Samsung’
is impossible to tell, though their uniform market shares of about 35 percent or greater in each memory market likely precludes
anyone else playing with the force they do.
Visit
