Voltaire
I know Rambus is your darling plus many other porchers. Thought everyone would enjoy this article, loved the summary and the price target in relation to the PE ratio at the end.
Please forgive me if this has already be posted.
English Karen
p.s Is it etiquette on Silicon Investor to mention I pinched it from another thread???
To: mishedlo who wrote (62701)
From: Estephen
Saturday, Dec 9, 2000 3:52 PM ET
Reply # of 62812
Author: ptnewell
Number: of 23554
Subject: Valuing Rambus
Date: 12/9/00 2:59 PM
Post New • Post Reply • Reply Later • Create Poll • Post News
Problem Post • Recommend it!
Recommendations: 10
The DRAM industry has the classic boom and bust cycles of most other commodity industries.
However unlike, say, wheat or steel, there is a long term upward trend in the total dollar value of
the DRAM industry, albeit with frequent and often sharp downturns. The market usually awards
cyclic industries a lower P/E than those with less fluctuating fortunes. Thus Applied Materials
(AMAT) grew throughout the 90s with an annualized rate of 40%, but is never afforded the P/E of,
say, Microsoft, which experienced the same 10 year growth rate. (I think AMAT will have the
better growth rate over the next 10 years, but that is another story).
Rambus is partially immunized against these cycles in several ways. First, as a potential monopoly,
earnings are far more secure. Second, its IP, royalty-based model ensures that it does not
periodically slip into actual losses, as say the DRAM makers themselves do. Finally, probably
more than half its income will actually come from the logic controllers that interface to memory
rather than from the DRAM itself. Scores of different companies make their own memory
interfaces adapted to their own purposes. A plunge in commodity DRAM prices will not normally
affect the much more specialized and individualized memory controllers as severely. Nonetheless, a
fair valuation of Rambus must take into account the sharp fluctuations in the DRAM industry.
DataQuest's latest prediction (Nov 2000) of the size of the DRAM market in 2002 is $76 billion,
up from a 2002 target of $70 billion predicted in 1999. The bad news is that DataQuest now
predicts the dollar value of the DRAM market will fall in 2003, whereas previously they had
forecast continued growth, possibly to as much as $100 billion. The reality is that because DRAM
prices depend so sensitively on the balance between supply and demand, these estimates, although
the best available, are really guesstimates. (For reference, the dollar value of the DRAM industry in
1999 was $20 billion; in 2000 about $30-35 billion). DataQuest is predicting a 29% share for
RDRAM in 2001 (I assume by dollar value, not megabits), rising to 40-50% in 2002.
Let us assume a 2002 DRAM industry with a $76 billion value, divided as shown below (all
market shares are dollar values, not megabit shares). The royalty rates are the best that can be
inferred from past Rambus results, and various guarded statements by management. The previous
memory standard, EDO, is slowly fading. Rambus gets no royalties on EDO.
Share/type Dollar value royalty rate Rambus revenue
50% RDRAM $38 billion 1.5% $0.57 billion
40% SDRAM $30.4 billion 0.95% $0.29 billion
5% DDR $3.8 billion 1.75% $0.19 billion
5% EDO $3.8 billion 0% $0
Total 2002 DRAM revenue $1.05 billion
Memory controllers are even harder to accurately pin down. Announced royalty rates run 3-5%,
but the size of the market is harder to quantify. Rambus believes they will earn as much or more on
memory controllers as DRAM. Currently, Rambus only earns royalties on Rambus memory
controllers (RAC), such as are used in the Sony PlayStation2 and Pentium 4, and on the SDRAM
controllers made by Hitachi and NEC. In principle, even after winning their cases against the
DRAM makers, Rambus may face additional fights against AMD, Intel, Transmeta, Motorola,
Nvida and scores of other companies that use DRAM. I believe most of these companies will
settle peacefully, particularly if Rambus wins against the remaining DRAM makers. The DRAM
makers (particularly Hyundai, Micron, and Infineon) have a track record of fighting and losing
hopeless battles against royalties, e.g., against TXN. Most other technology companies better
respect intellectual property rights. Few also would be willing to take the high-stakes risks Micron,
Infineon, and Hyundai are incurring. I will punt on the memory controller question, and simply
assume Rambus is correct that this market will at least equal the DRAM royalties.
Estimated 2002 Memory controller revenue: $1.05 billion.
In principle Rambus will collect on royalties back to April 2000 from scores of companies it
notified of patent infringement then. The first several deals did clearly include such back payments,
which Rambus has decided to account for by amortizing over a 5 year period. Already Rambus
should have at least $11 million in 2001 income just from amortizing past settlements. Future deals
could add greatly to this income from back royalties.
However the Samsung deal, struck in November, which Rambus management described as
"difficult, but ultimately win-win", only backdates to July 2000. Obviously one can only guess what
Rambus will ultimately collect in back royalties. I will assume $100 million/year by 2002, which
seems fairly modest given the billions potentially owed.
Estimated 2002 amortized income from back royalties: $100 million
Rambus has introduced a serializer/deserializer (Ser/des) which is targeted at a reported $19 billion
market. So far only one customer has signed on, PMC-Sierra. The PMCS product is supposed to
hit the market in early 2001. If Rambus get a 5% market share, and charges a 4% royalty
(admittedly wild guesses), income would be:
Ser/des 2002 income (guess): $38 million
I have no idea how successful the Rambus Ser/des will really be.
Rambus additionally hopes to collect royalties from other chip-to-chip applications that use
synchronous interfaces, but are not memory specific. One example would be Intel's 400 MHz
FSB, except Intel alone specifically has a waiver to use Rambus IP for non-memory devices.
Rambus is also marketing their RACs (Rambus ASIC), which can control RDRAM, for other
high-speed chip-chip data transfer purposes. Thus far LSI Logic and quite recently Agilent, offer
these in their ASIC solution "tool kits" to their customers. It is not clear that there are any takers
yet. Rambus thinks this market could be lucrative, but in the absence of any proven customers, I
discount this for 2002 earnings.
Chip-to-chip (non-memory) RAC and other: $0
When Toshiba signed a license with Rambus for QSRL (Quad Rambus Signal Logic), they implied
they already had a customer in mind. QSRL may be implemented as Static RAM (SRAM) rather
than DRAM. Regardless, QSRL is targeted toward communications devices and applications
where DRAM is little used. Hence I believe QSRL should count as distinct income from the
current DRAM market. Obviously any assigned value is a wild guess. But hints from Toshiba and
elsewhere make it appear that real products with QSRL will be selling already in 2001. I can't
count this as $0. Therefore a conservative guess is made. (Note: it is quite unlikely QSRL will be
used in PCs anytime soon. RDRAM covers the needs of current computers, and has plenty of
room for upgrades through at least 5 GHz CPUs or so.)
QSRL 2002 income (wild guess): $75 million
Rambus has been accumulating cash at a even faster rate than their earnings imply. This is partly
because of back payments, and partly for other reasons. For example, they don't really have to
pay the income taxes their pro forma earnings assume. The reason is the large write-off for future
stock options taken in April, 2000. That write-off, which included employee stock options which
will not vest, let alone be exercised, for several years, provides a long term tax shield. The upshot
is that Rambus already has $123 million in cash and marketable securities (half in cash, half in
securities). As further payments, especially back payments materialize, income on investments will
grow. I will assume that by 2002 they have $500 million in cash and back payments, which
generates investment income at an average rate of 6%.
Investment income in 2002: $30 million
Contract revenue in 2000 was about $40 million. Let us cautiously drop this to $30 million, as
fewer companies need tutoring on the use of RDRAM. It is possible that MORE companies will
want such aid by 2002. But note that LSI Logic, for example, has been teaching companies to use
RDRAM. Both Vitesse and Sun are introducing RDRAM products with assistance from LSI
Logic, rather than Rambus directly. Likewise, the Japanese company, Zuken, has recently
announced a sophisticated software package to let other companies design PCB layouts with
RDRAM. Companies like Zuken and LSI may take much of the "Rambus consulting" business.
Contract revenue in 2002: $30 million
Total 2002 income: $2,.34 billion
The great news about a pure IP company is the low cost structure. For example, between 1999
and 2000, total costs went from $33 million to $44 million, even as total revenue and royalties
soared. They do not have the usual marketing, shipping, manufacturing, and other costs. There
appears to be no good reason why costs should soar now, unless revenue producing ventures are
invested in. Let us assume 2002 costs of $80 million. This is much faster than costs actually have
been rising. The numbers are so small that breaking out details seems pointless.
Total 2002 costs: $80 million
2002 income: Revenue – costs = $2.34 billion - $0.08 billion = $2.26 billion.
Pro forma income (assumes 35% taxes, which won't really be paid): $1.47 billion.
As of September 2000, there were 98 million shares outstanding. However, including warrants and
options which will vest over the next few years brings the total to 108 million, fully diluted. To be
cautious, let us assume 120 million fully diluted shares by 2002.
Fully diluted pro forma EPS in 2002: $12.24
I suggest that the P/E on these earnings will be lower than many readers hope, given the volatility of
the DRAM field. A P/E of 50 or 60 seems unjustified to me. It will depend on many factors, such
as Rambus's traction in some of the new areas (non-memory chip-chip interfaces, Ser/des, QSRL,
etc.), and what the DRAM market looks like in 2002. Provisionally I suggest a P/E of 20 is
reasonable, in view of the strong cash flow and market dominance.
2002 price RMBS price target on $12.24/share earnings and 20 P/E: $245/share |
