Skyworks Weighs Up Capacity Dilemma
KB: The replacement market for cellphones is increasingly driven by more complex, high-end handsets that have multimode capabilities such as W-CDMA and GPRS, or W-CDMA and EDGE. This increases not only the unit volume but also the GaAs content of PAs, as well as the volume of PHEMT switches.
Occurring simultaneously towards the lower end, the entire market is expanding at a more rapid pace than we ve seen in previous years. This is the result of addressing new geographies, such as in developing countries, where cellphone penetration has been relatively low but the infrastructure is already in place. So, we have many underserved markets in the world where if the phone price can be driven to a low enough point, the entire market can expand. That is something that we ve begun to see over the past year - particularly with GPRS, but also some low-end CDMA2000.
MH: How has this affected Skyworks GaAs fabs?
KB: Utilization has remained at nearly full capacity for quite some time, while we have been increasing that capacity. For HBT products, we use a wafer foundry that we have had a relationship with for a number of years now. We have introduced a copy-exact process that gives us both second-sourcing capability as well as flexible capacity, and that has been part of our ability to rapidly expand supply to meet demand.
We have ongoing investment [and more equipment] at both of our factories, and we are exploring other forms of capacity expansion beyond that. We have been strategically evaluating all of the alternatives for some time, including 6 inch conversions at our own fabs, as well as broader and deeper foundry relationships, and the possibility of acquiring a fab. But none of those decisions is something that we have ready to make public at this point.
KB: What are the key issues involved in a switch to 6 inch production?
MH: There are three key parameters. One is the cost of the investment versus return. Second is the time element - none of these projects occurs quickly. Third is continuity of supply - how do we ensure that a project like that satisfies our ongoing requirements? A 6 inch conversion in a wafer fab has the potential to affect our existing production. In the end, it s simply a question of finances, customer satisfaction and timeliness.
MH: Does the use of 4 inch processing put you at a disadvantage?
KB: No, I think that it is an advantage - we have a fully written-off 4 inch wafer fab. Any investment we were to make in 6 inch would come with a significant increase in our depreciation costs, without necessarily a slam-dunk benefit in terms of lower wafer costs. We have a very advantageous wafer-foundry relationship regarding cost - so we ve been able to satisfy a tremendous increase in demand with an outsourced manufacturing strategy, which is balanced with high utilization of our existing low-cost wafer fab. It s hard to beat that cost profile. That s not a disadvantage - it is quite the opposite, and I think that we have a cost-structural advantage over some of our competitors.
MH: Does the ramp-up of BiFETs strengthen the case for a conversion?
KB: We make our BiFETs in California and production is ramping just now. It s still a minority of our production today, but it will increase over time. The increase in die area is not significant. The idea of larger wafers is really about needing more capacity and what is the most cost-effective and timely path to achieving that - there are many drivers.
MH: Are you seeing strong demand for products used in ultra-low-cost ($20-30) handsets?
KB: Absolutely, we have for some time now. We have a 100% share of a Taiwanese ODM for exactly that market today and we ll continue to enjoy that business. Today, a low-end phone is a dual-band GPRS product. Our designs take out costs that are not valued by that segment of the market. Specific products for that market are optimized for dual-band-only applications, which is quite a departure from where we were a year ago, when we really saw no value other than to ship only quad-band GPRS PAs because that s what the vast majority of the market valued. But in this very low tier, cost is everything and those additional two bands are less important.
MH: What market-share target do you have for the PA business?
KB: Our market share is 40-45% currently, and our target is well over 50% in the next two years. Overall success will be greatly determined by our success with the top-tier vendors. Our primary focus is on them, and our ability to satisfy demand for the PA, switch and filter in complex FEMs that really address W-CDMA and EDGE. Concurrently, we have to satisfy the low-end of the market with very cost-optimized solutions.
MH: Do you expect to see CMOS PAs taking share away from GaAs?
KB: Certainly there s been lots of noise in the industry about CMOS PAs over these past couple of years. But there has been no market-share penetration because of performance and reliability, combined with no demonstrable cost advantage with CMOS. We ve proved with our low-end PAs that we can deliver price-points at acceptable margins where we compete very nicely with silicon CMOS PAs.
We do see certain applications where silicon-based switches will play a role, and we are addressing that where required. But we will also continue to see PHEMTs have a big role.
MH: What will drive your GaAs business aside from handsets?
KB: I think you ll see us increasingly participate in WiMAX and wireless LAN, using GaAs technology for each of those applications. WiMAX is exciting for what could eventually be a mobile 4G standard, and you ll see us have more to say about that in the future. Cellular infrastructure has been pretty flat, and although we re hoping to see a recovery it is probably too early to say whether there is one - we ll just have to wait and see.