In our recent newsletters (drop us a line if you want a free subscription) we have been linking to a lot of press stories about the current semis cycle. We have also written quite a bit on the subject. And we have gotten a ton of questions from investors on the subject. It is now pretty clear that the semis super-cycle is tapering off. We can debate when we really start to see this reflected in public company numbers (Q4 is our guess), but another important question is who will feel the most pain when the downturn comes.
For those who have never seen a real semis down cycle, here is a sense of what is to come. At some point, demand will greatly exceed supply. Customers will start canceling orders crimping revenues. Due to Moore’s Law, semis are a fairly perishable commodity, this year’s chips are worth less than next year’s chips. So chip companies will have to write down inventories and pricing will weaken, so gross margins will shrink as well. There are far fewer semis companies today than there were twenty years ago during the last Bubble Popping, so we are not likely to see anyone go out of business, but there will be a fair amount of pain out there in the form of weak earnings. And despite the recent sell-off, a lot of chip companies are still selling at healthy multiples, so there stocks have plenty of room to fall.
Since we are talking about a massive supply chain, like squeezing a balloon, the distribution of that pain may appear in unexpected places.
In past cycles, some of the biggest losses appeared far up the supply chain with the foundries and the companies that make semiconductor manufacturing equipment (which we are just going to call semicap companies here). These companies sit at the far end of the whip, they have the least visibility of anyone in the supply chain, and no one has great visibility to begin with. So by the time they get the signal that demand is easing, they are deeply committed to their long-term production cycles. Also recall that in times of tight demand, semis customers will double order – putting in orders for more parts than they need in order to jump the queue ahead of smaller customers. This becomes a loss multiplier on the way down, exacerbating the pain of the vendors.
However, this time is likely to be different. The pandemic revealed a lot of vulnerabilities in the supply chain, and now everyone is aware how tight production capacity is. Over the past two years, the foundries in particular have been able to lock in long term customer commitments. There is an immense amount of variability in these commitments, but it seems likely these long-term contracts will shield the foundries this time around from the worst of the write-downs. So then the question becomes how hard are the commitments the foundries made to their semicap suppliers. It certainly looks like they made real commitments, ASML and the others are all reporting serious backlog. Adding that together it looks like these companies will be at least partially shielded from profit-killing writedowns. These companies will likely see some reduction in orders, but they definitely seem to have a lot of cushion.
That likely means the worst pain will be felt further down the chain by the fabless chip design companies. Here we have to segment the market a bit. The high-volume digital chip companies like Qualcomm, Nvidia and AMD have all made significant prepayments to their foundries. Most of these companies are already seeing rising inventory levels as various pools of demand dry up. So it seems likely that they will have to continue to make large payments to the foundries for capacity they do not need. This means either compression of gross or operating margins depending on the vagaries of accounting. There be some areas that are protected for other reasons. For instance, the memory companies look pretty robust at the moment (at least until the Chinese enter the market), but overall it looks like considerable vulnerability in this segment.
By contrast, the analog chip companies are in a slightly different boat. The shortages on this side of the supply chain look far more persistent, and may linger into 2023 or 2024. This does mean they are immune. Many of these companies manufacture some of their own chips, so they may have made big commitments to the semicap companies which will eventually come due. And while these companies’ products tend to have much longer shelf lives, they are also dealing with some very slow moving customers. In particular, the auto makers are still struggling to define their semis supply strategies. Most of the big automakers are in the process of boosting their inventories of semis helping demand, but if history is any guide, they will change this strategy just at the moment when overall demand is shrinking. Smaller analog players may also be up against purchase commitments to their foundries, Given the vagaries of forecasting demand in analog end-markets, we think it is likely that these companies start to feel real pain somewhere next year.
The good news in all this is that secular trends remain very positive. The world needs an ever-growing number of chips. The pain that is coming will be short-term in nature, a few quarters. But the pain will come and make life difficult for many companies (and their shareholders) in the near-term.