[ IEEE Spectrum Article ]
The June 1997 issue of Spectrum has an article on thin-film SOI, written by Michael Alles, who is the director of applications engineering at Ibis. As a follow-up to reading the article, I decided to research Ibis as an investment opportunity. My conclusion is that Ibis is extremely overvalued. It seems that a couple of recent Ibis press releases have really pumped up the stock price. I can only conclude that non-technical people are misinterpreting Ibis's competitive position in SOI. All of the huge multi-national electronics companies are involved, and will win.
The following is a summary of the Spectrum article, with some comments on the recent press releases.
There is a small two page sub-article on thin-film SOI manufacturing processes. Alles divides implanters for SIMOX SOI wafer manufacture into three generations. The oldest equipment is made by Eaton, and is being phased out. The current generation is made by Ibis and Hitachi, and are currently being used. Looking at Ibis's trailing sales, it seems that mostly Hitachi implanters are being used. Alles sees the third generation of equipment taking advantage of the ITOX and Advantox low-dose SIMOX developments. ITOX was developed by NTT (Nippon Telegraph & Telephone), and is already being used by Nippon Steel and Komatsu Electronic Materials. Ibis is developing Advantox, but Alles does not mention any customers. From the main article, I get the feeling that NTT has a major long-term SOI development effort, and has sold many commercial SOI circuits. I wouldn't bet on Ibis' third generation technology winning. Finally, Alles looks to the future, and picks as winners two of the four technologies under development. He picks the Smartcut process, being developed by France's Laboratory of Electronics, Technology, and Instrumentation. This process is already being used by Soitec SA. Alles also picks the Eltran process, developed by Canon, with no customers mentioned. Alles, a top employee at Ibis, does not see any Ibis technology in the future. Do the investors know something that he doesn't know?
The bulk of the Spectrum article is on development of thin-film SOI applications. There is a nice table of commercial CMOS products, with entries from Fujitsu, IBM, Samsung, Mitsubishi, NTT, Motorola, and Toshiba. Since Ibis has a joint venture with Mitsubishi, it is worth taking a closer look at Mitsubishi's three product categories listed in the 5/21/97 press release.
In the gate array category, NTT had the first commercial product in 1996, a 300k gate array operating at 1.2-2 V. Mitsubishi came in second in 1997, with a 220k gate array operating at 2 V. I doubt Mitsubishi sells much of this product, since it a worse product introduced later. The text of the article states the NTT 300k gate array consumes 30 mW. The 5/21/97 press release states that Mitsubishi will soon have a new 560k gate array operating at 1-2V, 50-150 MHz. The power consumption is listed as 0.05 microW / MHz / gate, which works out to 1,400 - 4,200 mW. Either the press release contains an error, or Mitsubishi will not sell much of this product. The advantages of SOI circuits are high speed and low power. People using a SOI gate array would want to operate at 150 MHz, which would consume 4.2 W! In terms of power per gate, that is 75 times as much power consumption at NTT's product. Further, Mitsubishi sample quantities will not be available until November 1997, whereas the NTT product is available now.
In the SRAM category, things are just as bad. IBM came out with a 512kb SRAM in 1993, which operated at 1 V. IBM topped that in 1996, with a 1Mb SRAM operating at <1V. The press release announces that Mitsubishi has prototyped a 192kb SRAM operating at 2 V. IBM's 1996 commercial product has 5 times the memory and 1/2 the operating voltage of Mitsubishi's 1997 prototype! When Mitsubishi finally gets a commercial product, they will not have any sales.
In the third category, NTT was first in 1997 with a 40 Gb/s 8 x 8 bit ATM switch, operating at 2 V. Mitsubishi now has a prototype operating at 2 V. There are no more technical details.
The trailing P/S is 6, indicating investors believe sales may double or triple soon. It appears that people got excited about Ibis after reading the press release that said the Ibis/Mitsubishi joint venture will sell SOI wafers to Mitsubishi. The problem is that Mitsubishi will not buy many wafers since they won't have a viable commercial product. Will the Ibis investors wait a few years until the sales don't materialize? |
