Even with billions on the line, semiconductor shortages can take years to resolve.
Consider the 2020–2023 chip shortage, which hit automakers especially hard. When COVID19 struck in 2020, car sales plunged by 30%–90%, inducing many automakers to cancel future chip orders. Then when demand rebounded unexpectedly quickly, automakers found that there were not enough chips to buy, nor enough available manufacturing capacity to make them.
This shortage of chips was severe. Every major automaker (including Toyota,Volkswagen, Daimler, Ford, Honda, and GM) were forced to idle manufacturing plants. In 2021 alone, the chip shortage reduced auto production by ~11 million vehicles, costing an estimated $210B in auto revenue(!). In response, the White House convened 3 summits and signed the CHIPS and Science Act of 2022.
Building extra semiconductor manufacturing tools takes years. Even in late 2022, lead times of various tools were as high as 24-30 months, and lead times of some chips were still over 100 weeks.
Literally hundreds of billions of dollars of automotive revenue (tens of billions of profit) were sitting as a prize to anyone who could make some extra chips for automakers. And this is a relatively “easy” task - automotive is only a small ~10% slice of the semiconductor market, and doesn’t even need the leading edge technology nodes, where production is even more bottlenecked. Yet even with this massive prize of billions of dollars, the industry couldn’t match this demand in under 3 years. It’s not for lack of skill or effort — these companies have hundreds of thousands of employees, a large fraction with advanced degrees — but simply because the modern semiconductor supply chain is fantastically complex.
When designing a supply chain, cost trades off against resilience. If you minimize cost (e.g., single sourcing crucibles from the lowest cost provider), you lose resilience (e.g., it may take you months to recover from that single source going offline unexpectedly).