You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I don't think we would even be talking about this without the chip shortage narrative. Would there even be a CHIPS ACT and on-shoring? The truth is there was never really a wafer shortage. Companies cancelled orders then had to get back in line. This was a total supply chain management debacle to the highest order. The foundries were very happy with the shortage narrative because it allowed them to raise prices and spend CAPEX with impunity. And if/when over capacity comes it will be back to price and no one makes wafers cheaper than TSMC. They have nothing to lose either way. I never thought I would see TSMC build fabs in the US but here they are. Brilliant move, absolutely.
Look at the TSMC automotive numbers: In Q2 2019 it was 4%. In Q2 2020 it was 4%. In Q2 2021 it was 4%. In Q2 2022 automotive was 5%. No additional wafer capacity needed here.
Take a look at the latest automotive numbers. In 2020 automotive dropped -15%, in 2021 it rose +2.4% and 2022 looks like +6%. 2023 is expected to drop again a percent or two. So help me understand why automotive is still experiencing chip shortages? I just don't see it.
Agreed, nothing like a lively debate (especially if it challenges my views and enriches my mind).
For sure shortages were in no small part because they bungled their ordering and until recently hoarding was driving up demand. However as more automation and other chips enter vehicles enter the market and as the auto industry reaches production levels that are approaching pre-pandemic levels I can't see any way in which auto demand doesn't turn into a large growth sector for the semiconductor industry (as opposed to the mostly stagnate sector it is has been for the better part of a decade). One important thing with automotive is they like local suppliers. Just look at the Japanese automakers and their deeply integrated local suppliers that they literally own stake in. With this comedy of errors that happened with the pandemic I wouldn't be surprised if auto manufactures want some degree of local capacity just to shore up their supply chain/mitigate trade disruptions. Because when you are buying chips for a few dollars in you tens of thousands of dollars car, you would easily pay 5x just to have guaranteed dedicated capacity just a truck drive away from my factory so you never have to shut down the line.
As for the shortages, I thought there was a substrate shortage at some point due to the demand spike/fab over-utilization and that required a bunch of fabs to bring in additional suppliers?
As for the shortages, I thought there was a substrate shortage at some point due to the demand spike/fab over-utilization and that required a bunch of fabs to bring in additional suppliers?
Un forecasted spikes definitely caused temporary shortages because as you said the semiconductor industry requires advanced planning and preparation. The pandemic was a black swan event but it was also a learning experience. The semiconductor supply chain is definitely stronger as a result. Multi sourcing is back. So many changes are being made. You will never meet a smarter group of people than semiconductor professionals. We solve very complex problems on a daily basis. Unfortunately we are surrounded by chicken littles and click starved media that misinform to get attention. Being the smart people we are we use that misinformation to our advantage and that is what is happening here. I feel sorry for the people who are duped but Darwinism is a fact of life.
I mostly agree. However I think there is a big market for more embedded/automotive/telecom chip onshoring, as the things these chips go into are actually made in the USA. For these products the chip is a relatively insignificant part of the BOM or the margins are already huge, and I think manufacturers would be wiling to pay extra for extra resilience. To say nothing of the massive savings in shipping or potentially having to deal with rolling lockdowns and power outages in the PRC.
For consumer products, yeah I don't expect iphones to be assembled in the US anytime soon. But I could see why fabless companies might want some foundry capacity outside of Taiwan in case there is a natural or manmade disaster/trade disruption. Yes I am well aware that this would screw up the whole market regardless. However, some diversification wouldn't cause any harm and might speed up a bounce back.
Additionally the whole supply chain wouldn't need to be from the US, and I don't think the US govt or most US companies care to have total autarchy. As it stands much of the semiconductor ecosystem is already "out of the line of fire". Chems are already mostly from Japan, tools from the US/EU/JP, wafers are already partially/could further be sourced from the US/JP, a great deal of assembly takes place in Malaysia/Vietnam/SK/(in the future the EU for intel). To my understanding the only big things that are basically monopolized by the Chinas are PCBs and consumer electronics assembly (which is partially shared with SK and Malaysia with India trying to break into this field). So if US policy makers want to reduce US dependence on the Chinas for critical electronics then I think this should at least build a basic level of supply chain resilience (enough for national security purposes and a bit extra for the niceties of modern life) which could grow as the PRC continues to become less business friendly/more companies continue to pull out to cheaper and less uncertain nations. After all the same thing happened to Japan and they were a stable democracy with a relatively open door to trade/less bureaucratic barriers. I see no reason why China should be immune to the middle income trap that so far has struck every developing nation since the beginning of the individual revolution.
Re "For these products the chip is a relatively insignificant part of the BOM or the margins are already huge, and I think manufacturers would be wiling to pay extra for extra resilience."
I suggest the last 2 years tells us the reverse - the auto makers got into huge trouble because they would not pay anything extra. And the margins on automotive chips have never been that great. Which is why they ran supply so lean. The total BOM for all chips in a car is also probably quite significant these days.
Agreed, nothing like a lively debate (especially if it challenges my views and enriches my mind).
For sure shortages were in no small part because they bungled their ordering and until recently hoarding was driving up demand. However as more automation and other chips enter vehicles enter the market and as the auto industry reaches production levels that are approaching pre-pandemic levels I can't see any way in which auto demand doesn't turn into a large growth sector for the semiconductor industry (as opposed to the mostly stagnate sector it is has been for the better part of a decade). One important thing with automotive is they like local suppliers. Just look at the Japanese automakers and their deeply integrated local suppliers that they literally own stake in. With this comedy of errors that happened with the pandemic I wouldn't be surprised if auto manufactures want some degree of local capacity just to shore up their supply chain/mitigate trade disruptions. Because when you are buying chips for a few dollars in you tens of thousands of dollars car, you would easily pay 5x just to have guaranteed dedicated capacity just a truck drive away from my factory so you never have to shut down the line.
As for the shortages, I thought there was a substrate shortage at some point due to the demand spike/fab over-utilization and that required a bunch of fabs to bring in additional suppliers?
"Because when you are buying chips for a few dollars in you tens of thousands of dollars car, you would easily pay 5x just to have guaranteed dedicated capacity just a truck drive away from my factory so you never have to shut down the line."
Did it really happen that automakers paid 5 times or 10 times of pre-pandemic price in order to get sufficient supply of chips?
Global fears of semiconductor scarcity have ramped up pressure on Taiwan’s chip production to meet overseas orders, while fears of a potential military clash with mainland China are also weighing on the industry.
On the topic of corruption: There is a specific flavor to Federal government United States corruption. Corruption with US characteristics if you will. If you drive around the countryside surrounding Washington DC you see the evidence of wealth distribution to people with connections; lots of expensive houses.
The top of this food chain is lawyers. New laws require $5000/hour legal advice, and help handling the logistics of dealing with the Commerce Department.
The second layer is consultants. The new law will require their services, not sure exactly how, but it's certain. Maybe the law mandates "Lean Six Sigma" as a requirement to operate the supply chain efficiently at the new fab, and a consultant is needed to prove you fulfilled this diligently.
You should expect about 60% of the money to go to these layers. It feeds the Feds, via a revolving door between the law firms and the consultants. So $52B is actually $21B, net of corruption. The 60% figure is pretty consistent across federal programs. Most universities that engage in science research receive federal money for that support. How much actually goes to researchers? 40%. 60% on average goes to administration.
Not true, that is not how wafer agreements work. They are legally binding contracts for price and delivery. New wafer agreements or mutually renegotiated agreements are fair game though. If a customer wants more or less wafers that will include a price negotiation.
Let's say in 2019 fabs were at 80% utilization and 2021 it was 100% utilization. Additional capacity will bring that number up quite a bit assuming that utilization stays at 100%. So yes, it is easy to believe output is up 20%.
Briefly, because the automotive chip content is skyrocketing (double-digit growth over the past 5 years I think, and projected to continue the next 5 years as we transition more to electric vehicles? I'll have to look around again to see which sources I read about this. It was reputable ones like ICInsights or McKinsey.) and because mature nodes aren't seeing an investment in increase in capacity from the foundries because nobody wants to pay for "old" technology.
The top 9 semi companies on the revenue list you posted don't interest me much, honestly, since they don't really operate in this space. Look at the next 5 and Microchip and Renesas and ON Semi.
Briefly, because the automotive chip content is skyrocketing (double-digit growth over the past 5 years I think, and projected to continue the next 5 years as we transition more to electric vehicles? I'll have to look around again to see which sources I read about this. It was reputable ones like ICInsights or McKinsey.) and because mature nodes aren't seeing an investment in increase in capacity from the foundries because nobody wants to pay for "old" technology.
