We took a cost accounting class in business school. We transferred out before we failed. It was among the most boring experiences we have ever encountered. But that was the ignorance of youth, today we think Cost Accounting could be the most important skills required to compete in semiconductors.
To be clear, we are not saying that semiconductor companies need to become 100% cost-conscious and slash the few perks they offer. We are definitely not suggesting all CFOs start to play Dr. No. Instead, we think a key competitive advantage in the coming years for semi companies will be the way in which they balance the demand from customers for custom and semi-custom chips against the very real costs that each new chip bring. We have touched on this subject in the past where we noted that many semis companies have been operating under these constraints for a very long time, but we think it is a lesson that needs repeating.
Here is the root of the problem. Each new chip carries a host of fixed costs. These is more than just the cost of cranking out a new design, which is something that most semis companies are now fairly proficient at. These other costs not readily apparent to outsiders. First, there are the obvious costs of extra payments to the foundries. Like any manufacturer, the foundries do not like to retool for small runs of different products because that takes their machines off-line. This is an old dance, and by now the ecosystem has developed many processes for working through this. For instance, the fabless companies can design a chip whose functions can be activated or deactivated after coming back from the foundry. There are some costs here, but they are tiny compared to the penalty incurred at the foundry level.
But there are other costs as well. Notably, every chip company has a host of requirements for New Production Introductions (NPI). For instance, they need fully document each chip’s features and specifications. They also may have to run each product through a complex testing regime. This is incredibly labor-intensive and not in a fun way. Most big companies have serious new chip processes in place to make sure all the required work for each product line or SKU is fully compliant. There is negative scale involved here and so the chip designers spend a surprising amount of time debating just how few SKUs they have to produce.
Admittedly, there are ways to automate much of this, and definitely room for chip companies to improve and modernize their new product processes. And we think those are going to become very important. We frequently point to Indie Semiconductor as a good example of this. They make an assortment of chips for car companies. They have built a reputation for catering to each customer in what appears a highly custom way while simultaneously maintaining a cost structure that looks a lot more like a purely catalog business. They were able to do this because they started small and built a new approach from scratch. They had to do that or starve.
The larger companies are going to have to carefully rethink their approach to New Product Introduction and find ways to significantly increase their capacity without blowing up costs.
So we apologize to our accounting professor for not being able to stay awake better, it turns out he was right about a lot of things,