EDN Asia: Semindia is building a fab in Hyderabad, the first fab in the country. Please comment.
Jain: Having a fab does not mean cost effectiveness or competitiveness or even better, cheaper chips for local consumption. In fact, building a fab is a huge capital investment. Latest estimates put the cost at $3 billion or more and will continue to go up with each new technology process. It is much more than start-up costs, as a fab needs regular maintenance, not to mention the cost of tracking new process technologies and constant retooling.
EDN Asia: What are the direct and indirect benefits Semlndia Fab is expected to generate by 2015?
Jain: While it could create 5,000 to 10,000 jobs, we need to consider the $3 billion investment for a single fab and compare it to the returns it will provide. The market leader in the foundry business is Taiwan Semiconductor Manufacturing Company (TSMC). It has built its business on long-term technology and business partnerships with key captive customers. Intel’s announcement that it intends to build a fab in China reinforces the point that a captive fab could drive development and provide a ready-made customer base. Captive fabs such as Intel’s can bring the much needed semiconductor ecosystem and help develop the infrastructure.
EDN Asia: Should India focus on the design of the product or its manufacture?
Jain: The government should not only focus on fab, but also on design. The real value is in owning the design. Traditional Indian companies are focused on the services model, which does not allow having ownership of the IP. Qualcore is one of the leading independent companies developing semiconductors IP in analog and digital areas.
EDN Asia: Should India continue to invest in the design segment and offer incentives to build products instead of just providing services?
Jain: India should continue to invest in design segment and develop its own product for the local and the world market. The value goes up as you move up the chain. Indian system companies should focus on developing products. India is planning to develop its own automobile, and I think it can do similar in the consumer electronics space and chip design companies can help in that area.
EDN Asia: Would it be wiser for India to remain fabless and focus on its core competency?
Jain: For strategic and security reasons a fab may be required, but otherwise it may be more advantageous to focus on design and remain fabless. As compared to a fab business, the software and very large scale integration (VLSI) design engineering industry is clean, can be distributed over various geographic locations, and does not need a huge investment. The only requirement is access to engineering talent and Internet connectivity. Conversely, the fab is one production facility in one location.
EDN Asia: India has the largest pool of talented software engineers. Should India concentrate solely on the designs?
Jain: India’s talent base is steeped in software development. The designers are not process or manufacturing experts. That means that the country would need to create university-level training programs, including bringing expert talent from other countries to get a fab operational and working efficiently. Traditionally, Indian software and chip design companies have focused on service, not on developing products. India should continue to invest in this talent and offer incentives to build products instead of only providing services. Perhaps it is time we learned to design, build and market products, though without incentives, this will not happen.
Biogarphy of Mahendra Jain
Mahendra Jain is the President of QualCore Logic Inc. Previously the Vice President of operations of TeraBurst Networks Inc., his earlier roles include, Vice President of Worldwide Marketing for Sagantec, Executive Director of VHDL International, Director of ASIC vendor relationships at Compass Design Automation, Co-Founder and Editor-in-Chief of ASIC Technology and News. He also served as Director of ASIC customer engineering at National Semiconductor and Fairchild Semiconductor.