Is It Time To Decentralize The Supply Chain?

One of the key requirements in any engineered system is a backup plan. A single point of failure in safety-critical or mission-critical applications can lead to disaster, whether that involves a smart phone, a car, a bridge, an airplane, or a design, manufacturing or business process. So why has this been largely ignored across the semiconductor manufacturing supply chain?

The answer is complicated, often convoluted, and not the same for all of the companies involved. Moreover, it varies greatly depending upon market segments and the risk/reward models associated with each of them. So a trade war that lasts two months may have little actual impact, while a trade war that last six months could have a significant effect. It’s the same for a virus that burns itself out in one region in two months versus one that lasts for 18 months.

And this is where things become really complicated for the semiconductor supply chain. A stoppage or slowdown that lasts for several months will vary greatly by application, by process node, and by location. For example, a problem in Taiwan or South Korea would have a devastating impact on advanced-node manufacturing and memory chips. A problem in China or Japan, meanwhile, would have a big impact on manufacturing of devices that utilize some of those chips.

Leading-edge equipment development cycles typically are long enough to span an entire node or two of development, which makes them somewhat impervious to short-term disruptions but highly sensitive to long-term slowdowns in end-user demand. On the flip side, there are only a few leading-edge fab customers for advanced processes, so any delay in equipment manufacturing would have minimal impact on overall fab business. EUV was delayed for multiple nodes and the semiconductor industry thrived. Likewise, a delay of a few months in rolling out a new manufacturing process would be a minor hiccup, while a shutdown of one of the key foundries would be a disaster.

The big problem is that it’s not always possible to separate out all of the pieces. So while it makes sense on one level to spread manufacturing across multiple geographies, that’s not feasible for everything in the supply chain. There are thousands of steps involved in semiconductor manufacturing, and hundreds of different and often complex materials.

While most people tend to associate labor arbitrage as the key factor for manufacturing in any particular location, that is far less of an issue than available expertise, the cost of shipping to end markets, or the price of environmental compliance for mining and processing rare-earths and certain metals. There also are tax breaks and subsidies, which are opaque to the outside world, but which have a significant impact on where companies decide to locate. And there is an ongoing management headache of running businesses in multiple locations, and an almost perpetual disruption of keeping different operations in sync.

Some of these factors are completely hidden from view. And while regional hedging may make sense on paper, adding redundancy across the entire supply chain is almost impossible. The challenge is figuring out where it does make sense, and that’s a complex calculation that is unlikely to be affected by a single disruption, no matter how severe it becomes. But that doesn’t mean that companies don’t need a good backup plan. Every company and every process needs a backup plan, and while those plans might never be utilized, they look awfully good in times of need.