Sydney, 16 May 2001 Nonsenso
A year ago, in February 2000, I shorted a number of stocks and Rambus was one of them. At the time Rambus was trading at a little over 100 USD, and had a price-to-sales ratio that made Rambus come up on one of my screens. Just last week Rambus came up on another of my screens; that one looks for quality companies at their 52-week low.
I decided to have a deeper look at Rambus, and as I do every time before I analyze a company I did a couple of rough valuations to see if my research would be a waste of time. The valuations turned out great; an unmodified valuation on Bloomberg professional gave a value greater than 200 USD (with an 80% growth projection). Ford investment services gives a fair value of around 50 USD. Valuepro.net gives a fair value of around 30 USD with a growth estimate of 35%. My own valuation spreadsheet came up with values between 30 and 50 USD, depending on the (moderate) growth estimate I used. So far, so good.
After this, I read the last few annual reports about Rambus, printed out all the news articles for the last three months, and created a PDF hardcopy of their website using Adobe Acrobat for easy reading. By then it became clear that something special was going on; this is not just any company. I spent a couple of days reading up about the various types of RAM, and the position of RDRAM in the market today and tomorrow. I found production estimates for 2001 and 2002 from Toshiba and Samsung. I found strategic RDRAM position papers from DELL, Infineon, IBM and Intel. And I found the price history of RDRAM for the last 12 months or so, and noticed how the price had halved in the last six months.
What I saw in Rambus is a tremendous opportunity, unlike any company I've researched so far.
RAMBUS is an IP (Intellectual Property) company; it licenses its advanced chip connection technology to third parties. One of the inventions Rambus licenses is its RDRAM, a revolutionary type of fast DRAM with superior bandwidth.
It is likely that in 2001 and 2002 there will be a DRAM war between RDRAM and DDR DRAM. DDR DRAM is used in the high performance AMD machines. RDRAM is the only memory that is shipped with the Pentium 4. Intel has announced support for DDR DRAM, but a chipset will not be available until the end of 2001.
Intel chose RDRAM because it is the highest performance memory on the market today. It is true that in some cases DDR DRAM has similar bandwidth performance as RDRAM, but with an efficiency of only 65%, compared to RDRAM efficiency of 95%. RDRAM does have higher latency than DDR DRAM, and this has turned up in a couple of benchmarks. Low latency is good for some applications. High bandwidth is good for some other applications. Companies like DELL expect that high bandwidth memory will be imperative to future performance growth in the PC market. DELL is a strong supporter of RDRAM.
Among the companies that want to make RDRAM the next memory standard are: DELL, INTEL, SONY, SAMSUNG, TOSHIBA and KINGSTON. These are some of the biggest names in the industry, and their collective effort will surely have a positive result on RDRAM positioning. Among RDRAM enemies are Micron technologies, Infineon (?), AMD and Hyundai. There is also a large crowd of techies that do not like RDRAM, because Rambus is a legally aggressive company, and because RDRAM has higher latency than DDR DRAM. If the standardization fight is the result of collective efforts of the above mentioned companies, I have no doubt that the RDRAM supporters will have a good chance of winning, simply because of their much larger size, and much deeper pockets.
RDRAM had a hard time in the market in 2000, because it was prohibitively expensive. But prices are dropping fast; 64 MB of RDRAM in Feb. 00 cost 110 USD; today it is listed at 35 USD. 128MB of RDRAM cost 160 USD in Feb. 00, and 75 USD today. And prices are still dropping. Dropping prices for RDRAM are good news for Rambus because it will result in a larger share of the total market. And Rambus gets royalties of ~2% of every unit sale of RDRAM.
Production of RDRAM is also being increased. Samsung projects a fourfold increase for RDRAM unit sales in 2001, and expects a total market for RDRAM of 600 million units in 2002. Samsung has also simplified the architecture for RDRAM, resulting in a projected 20% lower overall cost. Toshiba will triple its RDRAM output in 2001. RDRAM will make up 60% of all Toshiba DRAM sales by the end of 2001, up from 20% currently. Kingston has also announced an increase of RDRAM production capacity.
Because Rambus is an IP company, and not a semiconductor company, it will not suffer the boom-bust cycle of traditional RAM manufacturers. Rambus does not need to invest in new fab capacity. It does not need to compete on price with fiercely aggressive competitors. It just takes a 2% cut out of every RDRAM sale, by anyone. Rambus revenue will go up and down with the DRAM pricing cycle, but it will not have to endure losses as a consequence of that cycle, because its income is limited to royalties.
Today RDRAM is used in all Pentium 4 computers, in the Sony Playstation 2, in some Sony high definition digital TVs, and in some network devices. RDRAM is projected by Intel to capture a 50% market share of the total DRAM memory market by the end of 2003. The global DRAM market currently has a size of about 30 billion USD. When the RDRAM market share works out the way Intel projects, Rambus will have more than 300 million USD in revenue in 2004, with a gross profit margin of around 90% (it has virtually no capital expenses, no fabs to maintain, and a relatively small workforce). By that time Rambus also has pricing power because of its significant market share. Currently the royalties are a low 2%, but once they capture a large market share they have the ability to raise their royalty percentage to say 4%, which would double expected revenue.
Rambus is involved in a number of court cases. These are related to the fact that Rambus claims royalty payments over DRAM and DDR DRAM technology. They just lost their first round against Infineon, but will appeal. But regardless of the outcome of these cases, their RDRAM patents are not disputed, and they will be able to claim royalty payments over every RDRAM product shipped worldwide. And this is projected to be the most important DRAM product in the market three years from now.
I'm projecting a price of 30 USD for Rambus shares at the end of 2001, and more than 60 USD at the end of 2002. These projections are based on discounted cash flow analysis, with relatively moderate growth projections (compared to the analyst consensus growth estimate).
Sydney, 16 May 2001