The Covid effect came in combination with secular trends in home appliances and cars becoming smarter, equipped with more chips than in the past (e.g., an electric vehicle uses roughly double the semiconductor content of a regular car). The digitalization of the economy in general (e.g., the shift to e-commerce and digital work) is also fueling a constant increase in the demand for chips. Unfortunately, even geopolitical actions had their impact: US imposed sanctions on Chinese technology companies (e.g., Huawei) in 2019 led to Chinese manufacturers hoarding chips to ensure supply.⁵

The chip industry has historically suffered from “boom-and-bust” cycles as well as “double ordering,” where clients buy more chips than needed. As supply and demand normalized and tightness in the chip market abated, demand dramatically declined. Based on past experience, most players in the semiconductor market were understandably reluctant to invest into more capacity to satisfy temporarily high demand without strong, long-term customer commitments. Even though TSMC, Intel and Samsung did eventually increase their investment in new fabs, these modern semiconductor plants cost more than USD 10bn each and take years to build as chip complexity increases.⁶ As a result, most invested capacity will only be available in 2023 or even 2024 – as most equipment makers have seen lead times stretch to more than 52 weeks (double the usual).

Capacity is generally only added at the leading edge, meaning fabs are built to produce the most modern semiconductors, as only these chips carry the profit margins required to amortize the high fixed capital costs. Over time, as chip technology gets commoditized, older chip designs (like those for less sophisticated automotive applications) also benefit from moving into this higher-end capacity. However, this is a long process and provides no immediate solution. For the time being, we currently face the unusual situation of legacy chips (40nm to 90nm nodes) used for automotive and white goods, feeling the squeeze, whereas capacity is mainly expanding for bleeding-edge nodes (5nm and 7nm).

Accordingly, chip manufacturers have reacted with selective price increases that reflect the tightness of the market at different node levels. Figure 1 illustrates price hiking mechanism of leading fabs across different nodes for 2021 and 2022.⁷

Several risks also loom in the highly consolidated global semiconductor supply chain. A Corona-driven lockdown in Xi’an, a leading manufacturing hub in China that is home to many memory chip component factories, meant supplies for semiconductor leaders, Samsung and Micron, were temporarily disrupted. Moreover, a fire at ASML’s optical production site in Berlin over New Year’s Eve may also impact its customers. Incidents like these could further exacerbate supply imbalances and inflate prices.

Several large technology companies are working hard to make virtual and artificial intelligence (VR & AR) mainstream, leading to potentially huge demand for smart glasses equipped with high-powered semiconductor content. The recent stronger push for a more carbon-neutral world often means a shift from “dumb energy” such as burning fossil fuels to “smart energy” like batteries, solar cells and wind turbines, all of which make heavy use of semiconductors to control and optimize energy production, storage and distribution. However, the history of the semiconductor industry indicates that once tension is released, the bust cycle will be, once again, at play.

The Smart Energy Strategy is overweight in analogue and power semiconductors driven by strong secular trends in electric vehicles and smart grids as well as the robust economic recovery. These should help increase utilization, tighten supply, raise prices and lead to continued higher profits even given inflationary pressure. However, as these factors are priced in, the Strategy has become more selective of its semiconductor holdings.