more on semiconductor design vs. manufacture

Broadly speaking, a semiconductor firm can make its products more powerful by creating more efficient designs, or cramming more semiconductor pathways onto a given chip, or making bigger chips. For commercial use, the first and second are the ways to go. (Firms can also use larger silicon wafers, tp get more chips/wafer, but that’s a different issue.)

Today’s state of the art semiconductor manufacturing plant costs $10 billion+. It’s highly automated, runs non-stop and spews out around $25 billion – $30 billion worth of output. These very high numbers express the problem for any firm that wants to both design and manufacture semiconductors–who has the money to build a new fab? and who has the markets for that much worth of product?

The answer–almost nobody except Intel and Samsung.

Enter Arm Holdings of the UK, which gives chip designers a framework in which to operate independently, and TSMC of Taiwan, which provides cutting-edge contract manufacturing services to designers.

A decade ago, TSMC was a year or two behind Intel in the sophistication of its manufacturing plants. Today, it’s a year or two ahead, with no signs that I can see of Intel being able/willing to regain lost ground.

Hence, the crucial political importance of Taiwan, and the potential vulnerability of the US tech industry were Taiwan to become a part of the PRC.

Exit mobile version