Why I generally don’t buy semiconductor stocks and where my argument is most likely to fail

Disclaimer: I welcome all counter-arguments and genuinely prefer to be proven wrong because it can serve as a valuable learning opportunity.

Dominance in value-added services and end products is one big part of why the United States has become such an economic powerhouse. Generally it tends to be harder to earn outsized profits from extracting raw materials than from manufacturing; it tends to be harder to earn outsized profits from manufacturing than from designing and selling end products and providing value-added services. There are plenty of exceptions, although typically businesses closer to the end of the value chain take a bigger cut out of the value pie, which enables them to produce larger returns on invested capital. There are many different arguments for why this happens. Intuitively businesses closer to the end of the value chain are also closer to the source of the money and thereby typically have a more advantageous position in influencing prices and designing products and services that eat a bigger chunk of the value pie.

I think it would be unfair to argue that “semiconductor companies” are low on the value chain. Nvidia designs graphics cards containing chips fabricated in Taiwan, using machines from the Netherlands – not to mention that Nvidia graphics cards effectively sell like a consumer product in the PC gaming community. Designing chips requires a lot of expertise and adds a lot of value. Even fabs are clearly closer to the end of the value chain than, say, oil companies.

Because of opportunity costs, it's useful to evaluate investment opportunities against a universe of alternatives. For example, let's suppose that I compare a hypothetical “pure software company” with a hypothetical “semiconductor manufacturing company”. All things equal, I would tend to demand a higher price for a dominant pure software company than for a dominant semiconductor manufacturing company because the software company closer to the end of the value chain. For example, one of Google's valuable early insights was that buying cheap, generic, commodity hardware in bulk and distributing massive workloads across thousands of machines enabled it to scale more efficiently, thereby putting itself into a position to add way a bigger cut of end user value than the hardware manufacturer. If I'm a software company, I also probably have more options to navigate around supply constraints and cyclicality than even, say, Nvidia, over the long term.

However, nowadays, aside from a few exceptions, such as Micron and Intel, companies in the semiconductor stocks generally trade at similar multiples as dominant, fast-growing, high ROIC pure software companies, such as Google and Microsoft, despite being farther down the value chain and being relatively cyclical.