Like a train leaving the station, it now seems inevitable that US companies are moving to reduce, or entirely eliminate, their reliance on China. It took a long time to get started, companies had been complaining about changing conditions in China for a decade. The 2018 Trade War was the spark that really got them moving, and their progress has only been gaining momentum since then. This process will take years, maybe decades, but at this point is probably unstoppable.

In our newsletter last week, we linked to a story about Texas Instruments opening a packaging plant in Chengdu, and we mused that this is probably the last semis plant TI, or any US company, is ever going to open in China.

Speaking of our newsletter – you can sign up for it here:

What will decoupling entail for semis? There are two sides to this – China as a supplier and China as a customer.

As a customer, China is an important market, but not as important for other products like luxury goods. In 2021, China imported about $433 billion of semis, which is a huge slice of the industry, but crucially, the large majority of those imports were packaged into other goods and then exported to the rest of the world as PCs and phones and TVs. In terms of domestic consumption, China’s demand for semis and for products containing semis is somewhere closer to 10%-15% of the industry. This is still a big number, but a large portion of that is under threat from domestic alternatives, especially around trailing edge processes, so it is going to fall no matter what US companies do in China. Barring serious geo-political strife, we think it is unlikely that US semis companies will be entirely blocked from China, so we tend to think of this as less losing a customer and more of a protracted period of sales headwinds. Less growth, not necessarily declines.

The supplier side is more complicated. Today, China really offers US semis companies two things – trailing edge fabs and Outsourced Test and Assembly (OSAT or packaging and testing).

Trailing edge fabs are somewhat constrained now, but in general the world has sufficient capacity to pick up much of the slack. Every manufacturing economy is currently looking to add trailing edge semis capacity today. Much of that will not end up getting built, but enough of it will that we think US companies will have many options to choose from. True, there are certain corners of the industry that are more dependent on China, memory for instance, but for everyone else decoupling should be easier.

OSAT is more problematic. China’s packaging and testing capacity has grown significantly over the past decade. This is high-skilled, labor-intensive work that fits nicely with China’s overall manufacturing strengths. It will take considerable work to reduce reliance on this. Other countries have some packaging capacity, notably Malaysia, but China’s position here is fairly strong. On the other hand, packaging is becoming a more important part of the semis ecosystem, it is one part of the solution to the slowing of Moore’s Law. As a result, it has attracted considerable attention lately. The leading foundries, like TSMC and Intel (not exactly a foundry yet), have been investing heavily in their packaging flows. Moreover, the typical OSAT flow is already fairly distributed. We regularly work with semis companies who contract OSAT work to a single vendor, but that vendor will ship parts around Asia to balance capacity and capabilities.

Extending this a bit further, we think it is also important to look at China’s broader manufacturing capability. This is likely to prove the hardest part of the supply chain to reduce reliance on China. China now has 40 years of compounding improvements in its manufacturing skills. Whole regions of the country have webs of inter-connected manufacturing facilities from hundreds of special-purpose vendors to training schools. In parts of China, ISO certifications are so ingrained to the local culture that we have seen bars, restaurants and associated entertainment venues proudly proclaim their adherence to certifiable process expertise. Getting things manufactured in China is just easier (not easy, but definitely easier). Ask anyone who has tried to develop a low-volume electronic gadget in recent years – finding reliable vendors in the US for design, sourcing, assembly and QA can be painful, while in China there are hundreds of firms that can do all of that in-house, usually for a lot less.

All of this is to say that decoupling semis, and electronics manufacture, from China need not be a traumatic process, but it will take many years. We have heard rumors that Apple is planning to entirely relocate the majority of its production outside China over the next seven years. And that is Apple, probably the world’s best at electronics supply chain management. For everyone else, decoupling will come, but not quickly.