EDN Asia: What is changing on the FPGA front and how is Xilinx responding to these changes?
Roelandts: Moving towards new generation of products is the fundamental change we are seeing. Xilinx has traditionally been a leader in introducing products at the latest nodes. Our 65nm technology is now very solid. We get 20 percent of our revenue from products at 90nm and 65nm. We hope to be the first to come out with 45nm products. We introduced 65nm node two years after we introduced 90nm products. However, 45nm product introduction will take longer because of transition difficulties. We shall introduce 45nm products in 2008-end or early 2009. Complexity in designs is increasing tremendously. In the world of digital convergence, new challenges are arising for the designer, and he is increasingly moving towards programmable logic to meet these challenges. Take, for instance, the drastically falling time-to-market. The designer can do within few days using programmable logic what he can do in eighteen months using ASICs. Besides time-to-market, rapidly changing standards are also pushing designers towards programmable. The only true solution emerging is programmable logic.
In the new cycle of technology, which will probably peak around 2010, digital convergence will reach maturity with voice, data, and video together being carried over the Internet. I believe there is a fundamental change in technology once every 10 years, and this triple-play of voice-data-video convergence on the Internet will be the next fundamental change. Designers will find programmable logic mandatory to handle technology. I believe programmable logic industry revenue will grow at a CAGR of 25-30 percent during 2008-2012, which will be much greater than the growth of ASICs and ASSPs.
EDN Asia: Do you think Xilinx is paying a higher price by aiming to be among the first to come out with a product at a new node?
Roelandts: Initially we pay a higher price. But in the long run we are benefited because of the lead we get over competition. Through early entry, we are able to get a better yield and build more sturdy technology.
EDN Asia: How do you look at the Structured ASIC market segment? Have they comfortably placed themselves between FPGAs and ASICs/ASSPs, as they were supposed to do?
Roelandts: Structured ASICs have almost gone. Even EDA companies are by and large withdrawing support to Structured ASICs.
EDN Asia: What are the next milestones for Xilinx?
Roelandts: Building Virtex at 45nm node, and Spartan at 65nm node in 2008.
EDN Asia: How does Xilinx look at Asia-Pacific, and what are your future plans for Asia?
Roelandts: We are pushing into Asia-Pacific on a broad front. We are in Asia not because of low cost, but because of the growing market and high technical skills. The level of design we do in Asia is the same as that in the US. In January last we set up an R&D team in Singapore to assist in developing 45nm FPGAs. This team is the first in Singapore to design at 45nm node. Its R&D work includes functional design, functional/holistic verification, and physical implementation to design validation, silicon characterization, and verification. In December last we set up an applications development lab in Shenzhen dedicated to solving customer design challenges in low-cost, high-volume applications using Spartan. This team develops reference designs and prototypes for customers across Asia-Pacific. In October last we opened our development center in Hyderabad, India, which develops IP cores for automotive, embedded, and high-speed serial I/O connectivity. The Hyderabad facility is one of our largest development facilities outside the US. Xilinx has set up $75 million venture fund to drive innovation and development in Asia-Pacific, and is investing in companies creating innovative applications based on programmable logic. We have already invested in India’s CG CoreEl, and are looking out for five or six more companies for investment. We’re working with universities to popularize programmable logic.