In the wake of the fab-lite strategies, possibly moving to fabless strategies, of NXP, Infineon, Texas Instruments and STMicroelectronics, the pros and cons of having a fab were listed during IFS2007 in London this week organised by analysts Future Horizons.
The first argument is the cost argment. Some say that, at $3bn, fabs are too expensive. But if you look at their cost in relation to the market size, their cost is the same.
A fab cost $40m in 1970 when the industry TAM was $2.4bn; they cost $330m in 1985 when the TAM was $21.5bn; they cost $3bn in 2005 when the TAM was $245bn. That’s a 14.1 per cent CAGR over 25 years fro both the market and the fab-cost. So there’s no change in cost.
“It’s just that you have to bet the company when you build a fab, and people have lost the stomach for betting the company”, said Future Horizons CEO, Malcolm Penn.
The second argument is that some say you don’t need a fab to build state of the art product - look at Broadcom, Qualcomm, NVidia and Xilinx.
But if you want differentiation in performance or cost, or both, you need a fab, and, if the end customer doesn’t care, then you’re already in a commodity market with no chance for competitive differentiation.
The third argument is that you don’t need your own fab to build jelly bean parts. But owning a fab is the only way to get to the lowest possible manufacturing cost i.e. a fully depreciated fab which was built a few years before for leading-edge product.
The fourth argument is that state of the art suppliers should not be making jelly bean parts. But it’s inevitable. What is state of the art today, is jelly bean tomorrow. And getting into state of the art products early, which is what having a fab allows you to do, gives you the best profits.
Furthermore there’s almost as much value in the jelly bean products as in the main chips, and those without a full portfolio of chips can be shut out.
“The industry has been very good at freezing out suppliers who haven’t got the full list of parts. Most killer products have almost as much value in the jelly beans as in the main chips”, said Penn.
However there could be a benefit of the big companies giving up their fabs. It gives a more level competitive playing field for fabless companies. This could boost spin-outs and start-ups which will compete with the large companies on the basis of superior design. The 20 year decline in the market share of the top ten companies could be accelerated.
Leave a comment