The semiconductor story is a bit more complicated.

background

Semiconductor development has always been all about making chips that are smaller, faster and use less energy (or generate less heat).

In the early days, a given company’s engineers designed chips that the firm made in its own fab (factory). By the mid-1990s, the two processes, design and manufacture, began to separate. Two reasons:

–the cost of a fab making state-of-the-art chips has been continually rising as manufacturing technology has become more complex. Today, for example, a cutting edge fab costs about $15 billion. Just as important, it will spew out at least $40 – $50 billion worth of chips a year. Few firms can afford to build one; fewer still have the sales volume to justify doing so alone.

–these factors, the desire of designers to work for themselves rather than in a bureaucratic behemoth, the development of third-party design tools (ARM is the prime example) and the startup of third-party fabs (foundries, TSMC by far the leader) have led to the development of specialized design and manufacturing firms.

–five years ago, Intel was the undisputed chip manufacturing leader, making its own designs, primarily for personal computers and for servers. Today, however, its manufacturing capabilities have fallen behind those of TSMC and its designs behind those of TSMC customer AMD.

–in somewhat similar fashion, there’s no American manufacturing leader in 5G, the latest generation of telecom equipment. The baton has passed to Huawei, which also makes better/cheaper conventional telecom equipment, as well. Huawei does. however, use American-designed chips.

why supply disruption?

The pandemic is the biggest culprit, in two ways. The virus caused the shutdown of chip manufacturing and distribution early last year. At the same time, demand for stay-at-home devices, from PCs to video game consoles to intelligent autos, began to boom.

The collapse of the makeshift Texas electricity grid caused another halt to the manufacture of chips there (a surprisingly large amount).

Also, facing the loss of the next generation telecom market to China, the only counter the Trump administration could come up with was to undermine Huawei by deny Chinese firms access to semiconductors made by US firms or made using US intellectual property, or made on machines that employ US intellectual property.

This action has implications. It reduces today’s effective capacity, since US-linked sources can’t supply China. It also makes the question of where to locate new capacity more problematic–except for TSMC, which has announced a massive expansion program, centered on Taiwan. It may also have kickstarted China’s languishing program to develop its own semiconductor industry. This presumably means eventual global overcapacity, but probably not during any time frame important to stock market investors.