A Softer Cycle for the Semiconductor Industry
In late June, we saw a sell-off of the stocks of companies in the semiconductor value chain. As the market wobbled on negative tariff-talk sentiment and fears of a potential slowdown in worldwide growth, investors took a chunk of their profits and ran from sectors that are perceived as economically sensitive. Although the prices of semiconductor-related companies have rebounded a bit recently, they are still off of the highs to which they had risen – and at which they were consolidating – after two years of steady earnings performance. In our opinion, many of them are trading at attractive prices. 

This may sound surprising. After all, the semiconductor industry has traditionally been highly cyclical. Typically, demand was driven by model cycles in the dominant end-use markets. During the 1980s and 1990s, that was primarily the personal computer market. During the 2000s, the mobile and then smartphone took over as the determinant of marginal demand. More recently though, the use of chips has become more pervasive, and this is having an effect on the shape of the semiconductor demand curve.

During the 1990s and early 2000s, the semiconductor market was dominated by personal computers. After the introduction of the iPhone at the end of 2007 – yes, it really was only 10 years ago – the smartphone cycle dictated the demand pattern for chips. The introduction of the first widely adopted tablet – the iPad in January 2010 – spurred the proliferation of large-display mobile devices for reading, web searching, watching movies and TV shows on the road, and video-calling, adding to the semiconductor demand stream. Since then, the range of devices incorporating semiconductor chips, and the variety of uses to which they are put, has been quietly exploding.

Think of the range of industries and activities that require an increasing number of increasingly complex chips.

They include augmented reality and virtual reality. Tried any of the virtual reality headsets yet? They retail for $28 to $700, depending on how good you would like the experience to be. Think for a moment of the amount and complexity of chips needed by the content creator machines to render lifelike, 3D-looking images, by the screen in the headset to show parts of those images to the eye, and by the sensors in the headset to collect and transmit the data on the wearer’s movements to the screen. 

Think of the driver-assist features that have become standard on new vehicles. Each new function and refinement requires a plethora of additional chips, and more are being added each year.

What about artificial intelligence? Has anyone asked Alexa to play particular music or reorder paper towels? Imagine what we’ll be managing with such robotic assistants a mere five years from now. Particularly as we develop the Internet of Things, which is still in its earliest infancy. You can buy refrigerators that send you photographs of their interiors, so that you know whether you need to buy milk on the way home from work. But soon we should see refrigerators that can work together with a robotic assistant, auto-payment system, and sensor-tagged food to buy milk and order its delivery for us.

Basic artificial intelligence applications are being incorporated into manufacturing, power transmission, building management and maintenance, transportation, inventory control, irrigation systems, medical diagnostics … the list goes on, and we are still talking about very rudimentary levels of sensing and processing.

The point here is that chips are required for each of these applications, and more are needed as their usage and complexity rises. Furthermore, the various industries and activities in which these applications are and will be used each have their own different cycles. The demand for chips that flows from these other industries and activities impacts the aggregate demand cycle in varied and often uncorrelated rhythms, and this reduces the amplitude of the swings in the aggregate cycle, thereby producing a softer – as in gentler – semiconductor cycle.

In our opinion, the current prices of companies in the semiconductor value chain represent past history, rather than future probability. We believe that demand growth will continue and that the industry’s cycles will soften.