Doron Simon, Tower Semiconductor USA President

Doron Simon (Source: Tower Semiconductor)

Doron Simon has been president of Tower Semiconductor USA since April 2001, and has served in various capacities, including director of customer service, director of planning and turnkey operations, and director of worldwide sales operations, since 1993. Prior to then, Simon was production control manager for National Semiconductor in Migdal Haemek, Israel. He holds a B.S. in industrial engineering from the Technion, and an M.B.A. from Hariot-Watt University.
 
Tower Semiconductor Ltd. (Migdal Haemek, Israel) is a pure-play independent foundry that manufactures ICs with geometries ranging from 1.0 to 0.13 µm. It also provides complementary technical service and design support.

SI: How did Tower Semiconductor get its start?

Simon: Tower was acquired from a National Semiconductor facility that was destined to manufacture CPUs. National didn't make it in that market, and in the early '90s, the company made the decision to sell the factory. This was right around the time the foundry model was getting its start, and a group of financial investors and key people in the management got interested in this facility. It was not a manufacturing fab, but more of a lab running prototypes and limited production with one-of-a-kind tools and a lot of manual operation. Tower was formed in 1993, and in the early days, we focused on building the infrastructure of the facility and set our sights on being a specialized foundry.

SI: Which markets did you decide to target?

Simon: We selected two markets for two totally different reasons; CMOS image sensors and embedded memory. In 1996 and 1997, we were pretty much ahead of the market in CMOS sensors, and with our lead customers, we realized this technology could be used in so many other applications in the consumer marketplace. The first of these products were for medical sensors, high-performance cameras and, later, low-cost PC cameras, which put the technology in front of people. Today, in the mobile applications space, these images are more easily downloadable and transmittable and the quality is improving.

SI: So, as a foundry, you were involved in market creation. Isn't that unusual?

Simon: Yes, and I'll give you another example of that. We believe that CMOS is replacing CCD technology as a long-term trend; the CCDs are the more expensive technology. It's starting to replace CCDs in low-end products, but also in the very high end such as 14 megapixel cameras. I believe this technology is going to slowly become mainstream CMOS, and CCD will be the niche technology.

SI: Tower is a much smaller foundry than, say, TSMC or even IBM. What are the advantages and disadvantages?

Simon: Because we don't have the number of customers that TSMC would have, for instance, we don't have the same business methodology where the foundry puts out a standard and the customers have to align to that. We provide value added because of the way we work with the customer, understand their product application and product needs. The disadvantage is that we have to be very selective about the businesses we target.

SI: How did Tower get into the embedded memory space?

Simon: We came across a patent, a theoretical exercise at the time, and we figured out it could be a very good way of creating embedded flash memory technology. An Israeli company, Saifun Semiconductor, created in 1997, developed the core patents and the architecture of the memory cells while we developed the manufacturing process. Many companies have licensed this technology, like AMD/Fujitsu, Infineon, Micronix, and others. One interesting aspect of this business is that many of our customers and partners are coming from Taiwan. So it is interesting that they find it worthwhile to work with an Israeli company rather than working with foundries in their own backyard.

SI: Because of your involvement on the design side, what can you tell me about IP issues today?

Simon: It seems the semiconductor equipment industry may be less concerned about intellectual property protection than people are on the design side. We work in specialty areas — CMOS sensors and embedded memory — and we believe some customers are coming to us rather than going to China because of IP protection concerns. It gives us nice positioning for the time being, but in the long run, I believe that since all the downstream manufacturing, back-end assembly and testing is ending up in China, most of the products will have to go through that region of the world regardless. Eventually, the Chinese will have to step up and face those IP issues, otherwise it will impede progress.

SI: How long do you expect such a process to take — 10 years?

Simon: It might take 10 years. But I think it takes two things. You need to show some financial successes working by the law in order to promote it. I think this is what really drove Taiwan. Because, in Taiwan, the business model was proven that you could be running an honest business and be successful. But China is more diversified; there is a lot of division between different companies in different territories and governments, and everyone wants their fair share, so in that respect it is tougher.

SI: Moving to another topic, some people are talking about the increasing importance of cost of ownership today. What is your opinion on that?

Simon: I think the concept has been around for many years, but what has really changed is the longevity of using the equipment is much longer now. You are looking at 10-15 years of operation as opposed to 3-5 years. As an industry, we were used to absorbing everything via the brute force shrink — the shrink element was compensating for all the other deficiencies in the supply chain. It was compensating for buying expensive tools, not using them as much or efficiently; or compensating for lack of design tools or design processes or fine-tuned design technology. You did not put as much effort into streamlining operations because before you finished you were working on another tool generation. What is changing across the board is you live with a technology for a long lifespan, putting a lot of effort into fine-tuning design and equipment performance.

SI: Would you say that your company's success has less to do with design shrinks than other aspects of the business?

Simon: Absolutely. We have a lot of interesting applications today and something like an X-ray sensor still lives very nicely at 1 µm. In some of these applications, there's no point to shrinking, and the challenges are totally different. Instead, we're dealing with noise separation, creating a nuclear image, processing the data, but not so much with shrinking geometries.

SI: Most fabs perform fab benchmarking, usually on a regional basis. Is this a challenge with the limited amount of IC fabrication going on in Israel?

Simon: Yes, it is done. We have open sources as well as unofficial company relationships. I believe that by benchmarking and developing those techniques, everybody benefits. Fab efficiencies are progressing. If you look at today's environment, take Tower, on a very local basis we have Intel fabs, and although the business models are very different, you can still compare certain metrics, modes of operation and other general rules of operating a fab. If you just look at the people on our board of directors, we have someone like Eli Harari from SanDisk, who owns joint-venture fabs in Japan and was involved in the creation of UMC; or Dan Reddy, who was involved in Chartered and UMC. So we are exposed to many people and many benchmark points at the high-level view of how a fab should be built and operated.

Now, if you talk about topical benchmarking, metrics such as cycle time and yields, we do that mainly with specific customers in different locations, or other foundries. In China, we have a good relationship with SMIC. Although we are competitors, we know some of the people there and compare, for instance, staffing levels for R&D, support people, based on production capacity. We have similar relationships with other fabs around the world.

We also talk to IDMs, which is another component, and some of our base technologies are licensed from IDM fabs. For instance, we licensed our 0.13 µm CMOS technology from Freescale. The training and the transfer involved sending people to Austin for close to a year. There are mutual benefits from this exchange.