Let’s say you ran out of milk. You like fancy, organic milk made to very specific standards. You were in the grocery store, but you knew you still had two inches of milk in the carton back home. Buying another gallon would be inefficient with all that milk at home. So you pass it by. Then a few mornings later and someone has finished off the milk the night before. You have no milk for your morning coffee. Frustrating. But it gets worse, you go to the grocery store and they are out of milk – maybe the cows are on strike. You could just go without milk for a few days. You could make a mental note that next time availability is worth a little bit of inefficiency. Or you could drive way out into the countryside and buy a cow. Your ancestors raised cows, so how hard could it be.
This is a story about the auto industry and semiconductors.
For years, decades, auto companies have been gradually increasing the semiconductor content in cars. But for chip companies auto makers are among the worst customers (the US government is the worst). Auto makers have ridiculously (by semis standards) long product cycles. It takes years to get a design win and then years more to get into production. Often, the chip companies are end-of-lifing support for the chips before the auto company gets to volume production. The auto makers also have stringent design qualifications for their chips. The chips have to be rugged, resistant to shocks, and heat, and humidity and snow. Most chips can withstand these tolerances, but the auto makers have a years long, expensive process to certify these standards. And for most chip companies, these most demanding customers are not big customers, often buying modest volumes of the cheapest chips. And to top it all off, the auto companies have developed ‘sophisticated’ supply chain strategies based on “just in time” (JIT) deliveries. While the business school case studies extol JIT as optimizing efficiency, another way to describe JIT is “let our vendors hold inventory so we can boost our cash metrics”.
Of course, this all came crashing down during the current supply crunch. A combination of Covid, labor shortages and surprise demand surges wreaked havoc on the auto makers supply chains. The auto makers are large entities, and command a lot of attention, so the press was full of stories of $60,000 cars going unfinished for lack of proverbial $2 chips. And since the auto makers are relatively small customers for each individual chip company, they often went to the end of the queue for what parts were available.
Most chip executives we know just shook their heads at this, chalking it up to chickens coming home to roost. The chip makers had tried for a long time to get the auto companies to speed things up and be more flexible with their inventory policies. To no avail.
We were reminded of all this by a story in the Journal a few weeks back, Ford is partnering with Global Foundries to potentially build its own chips. In fairness, there is some sense to this partnership. Ford appears to be investing in its relationship with the semis supply chain. GloFlo is the right partner, as it will be a long time before Ford could even hope to be a significant client for leading foundries TSMC and Samsung. Much of Ford’s chips use older processes, and GloFlo has been very aggressive in providing customer service to every segment they can support.
That being said, we get the sense that Ford may be driving to the countryside to buy that cow. In their press release, Ford’s CEO said:
“This agreement is just the beginning, and a key part of our plan to vertically integrate key technologies and capabilities that will differentiate Ford far into the future.”
Ford is planning to vertically integrate. Admittedly, we are not experts in the automotive industry, but we seem to remember that Ford was once entirely vertically integrated, even making their own steel. They had some good reasons for abandoning that model (less good reasons for abandoning the workforce and towns that went with the model). Has the auto industry changed so much that vertical integration now makes sense?
Even in normal times, we imagine that Ford would struggle with semis. First, they need a design team, maybe a few hundred people, all of whom are expensive, some of whom are very expensive. Then they need to be careful what products they design themselves. Do they need to design their own micro-controllers? Or power management modules? Recall our thesis that it only makes sense for companies to design chips which convey strategic advantage, building a chip to save a bit of margin does not really work out mathematically. But having that big team, and knowing what we do about auto industry culture, there will likely be huge pressure to design everything themselves.
And these are not normal times. The rest of the auto supply chain looks poised to move in a radically different direction – away from vertical integration towards discrete systems, i.e. the model used by the electronics industry. Apple alone is likely to pry this open and pave the way for other, newer entrants (read China) to take advantage of radically altered industry economics. If the auto industry moves towards this kind of model, it will mean the vertically integrated companies will struggle with unworkable cost structures burdened by excess R&D teams, duplicating the work done better and cheaper by product-specific companies.
It may make sense for Ford to design their own ADAS and autonomous processors, but building up a complete semis team, with direct ties to foundries, is going to be a very expensive way to get a glass of milk.