The Real Reason Behind the Automotive Industry IC Shortage — A Step-Function Surge in Demand!

[hkwint]

Both on this very website, but also semi engineering.com, it has been described how 'trailing node' 200mm-foundries, like TowerJazz, SMIC, VanGuard and Global Foundries were barely profitable the last few years.

Therefore they didn't want to expand 200mm-capacity between 2016-2020, already leading to a constant near-shortage of 200mm wafers starting in 2016.

The reasoning was, the one who would build extra capacity would have lower fab utilization and therefore would not be profitable anymore.

So it became a custom to associate extra capacity (more fabs) with losses. That's a hard mantra to change, only if customers are willing to pay up there will be extra capacity.

If those foundries made better demand forecasting in 2018 or so, they put up extra capacity up the next few years and they would be laughing all the way to the bank now. So it seems automotive isn't the only one who failed at planning.

 

[hkwint]

I'm looking for quotes from automotive executives on investor calls that say "sorry, we made a mistake, we cancelled orders and we shouldn't have"

Very true! They do tell exactly this on the phone when calling their _suppliers_ though, all the time; still Q4 '21.

And then comes December and they're sorry for uncancelling in November which they shouldn't have.

But that stuff is indeed never repeated on conference calls. It makes planning impossible these days, to such a great extent the thing actually has a name now: 'VUCA'. These days, internally, it's accepted we live in a 'VUCA' world.

 

[hist78]

Lots of insights here, I liked this one, then I thought for a second. Lead time is a symptom not a cause.

Cause is a lack of capa. Specifically, the capacity to meet demand of the part desired at the price and lead time desired.

So you have to pick one:

- Do I substitute another part?
- Do I pay more?
- Do I wait longer?

I get the sense though that after substituting, paying more, and waiting longer, there still is no light at the end of the tunnel. Something else is afoot.

Regardless of what that something is, and it can be hard to figure out, even if you have total visibility into all aspects of the supply chain, adding some excess capacity is the only thing that will bring relief. For sure. There is no other cure. More fabs are needed, right now.

"adding some excess capacity is the only thing that will bring relief. "

The problem is who is willing to pay for it? Those 4 new Intel fabs are in the leading edge game and probably won't help automobile industry chip shortage too much. And who is willing to start or expand the legacy fabs knowing the market and margins are limited?

Probably "Do I pay more" needs to change to "Do I pay more and do I pay enough?" The price must reach to a level to have balanced supply and demand.

 

[jms_embedded]

I agree, it's personal, it's not logic, it's emotion. Especially when the narrative is "We must build ICs in the USA" like that is going to solve the problem of automotive companies cancelling orders and coin miners buying up all the GPUs. But the US is going to throw billions of tax dollars at it anyway because cars are sitting unfinished due to the lack of semiconductors.

Yes, let's all take a breath here and try to be detached. I've enjoyed this thread, even the differences in perspective.

And my take on it is that any executive who is talking about events outside his/her company's own control is just engaging in public theater, and we shouldn't give them much if any weight as an expert.

I'm looking for quotes from automotive executives on investor calls that say "sorry, we made a mistake, we cancelled orders and we shouldn't have" and I can't find one, it's always "we can't get chips, supply chain problems, etc..." which is finger pointing.

Interesting, could you cite a couple of examples of finger-pointing? (I was thinking on checking some myself, mostly Tier1s I've heard of like BorgWarner / Magna / Lear / Continental / etc... but if you've already looked then maybe you can save me some trouble )

Daniel: I have listened to a couple of episodes of your podcast, and enjoy the dialogue. I would challenge you to extend an invitation to a few executives in the automotive industry who are willing to discuss what happened, from their point of view, and how they are looking forward.

 

[jms_embedded]

"The real root cause, is there's no level playing field between the buyers of semiconductors."

If automobile industry can't afford to pay more and/or doesn't want to pay more for the semiconductor products they need, what options do they have?

Or they are doomed to extinction?

They will have to adapt; they say they are (e.g. talks between Ford/GloFo for some kind of partnership), but we shall see how.

Also, be aware that "can't afford" is not a constant state. In March 2020 they likely could not afford (or had to make hard decisions), but now there's money, and if they learn from the situation like Toyota did after Fukushima in 2011, then they'll do better in the future when there are disruptions.

 

[jms_embedded]

Lots of insights here, I liked this one, then I thought for a second. Lead time is a symptom not a cause.

Cause is a lack of capa. Specifically, the capacity to meet demand of the part desired at the price and lead time desired.

So you have to pick one:

- Do I substitute another part?
- Do I pay more?
- Do I wait longer?

I get the sense though that after substituting, paying more, and waiting longer, there still is no light at the end of the tunnel. Something else is afoot.

Regardless of what that something is, and it can be hard to figure out, even if you have total visibility into all aspects of the supply chain, adding some excess capacity is the only thing that will bring relief. For sure. There is no other cure. More fabs are needed, right now.

I don't think there is a simple answer; I've been asking myself this question for the past 6-8 months and trying to learn. It's complex. (shameless plug for my short take on the subject: https://www.embeddedrelated.com/showarticle/1440.php)

There's a long-term issue and a short-term issue. The long term issue appears to be increasing semi content in cars, along with a mismatch in supply capacity and demand at the process nodes used in automotive electronics. The question that should be asked by the fabless/fab-lite companies is "We project X demand in the next five years; is this something you will have capacity for?" and if the answer is "no" then someone has to come to the right kind of solution, e.g. redesigning on smaller nodes (which may be more costly both per chip and for one-time-design costs) or purchasing capacity themselves or through the foundries, or persuading the foundries it's in their best interests to expand capacity, or giving up.

The short-term issue (as far as I can tell) is the interaction between JIT supply-chain strategy and the bullwhip effect and being close to this long-term capacity that the long-term problem got accelerated... along with the automotive companies making decisions and having bad luck in the long term that forced them into a corner in spring 2020 in a very uncertain future.

 

[Daniel Nenni]

Interesting to note TSMC revenue history:

2019 Automotive -7%
2020 Automotive -7%
2021 Automotive +51%

Just a slight correction... Even at a 51% upswing automotive is only 4% of TSMC's revenue.

 

[tooLongInEDA]

"The real root cause, is there's no level playing field between the buyers of semiconductors."

If automobile industry can't afford to pay more and/or doesn't want to pay more for the semiconductor products they need, what options do they have?

Or they are doomed to extinction?

The really interesting result of the auto chip shortage - at least in the UK and I suspect everywhere - is that both new and used car prices have risen very sharply. Apparently, the auto companies are making *more revenue* on the reduced volumes because they can choose to produce more higher spec cars now there is under-supply and buyers have to take whatever they produce or wait a very long time. So it's not totally clear the legacy auto companies are losers in this. Their customers certainly are though !

 

[Xebec]

Slightly OT - but related - Tesla mentioned during their year end earnings call 2 days ago that their 2022 capacity would be limited by "the chip shortage" and *not* battery capacity. They said they expected next year to shift back to battery capacity being the limiting factor. Interesting as they were one of the companies better able to navigate the shortage in 2021 though of course they're still smaller than the bigger legacy automakers (about Subaru size in terms of revenue).

 

[Paul2]

Both on this very website, but also semi engineering.com, it has been described how 'trailing node' 200mm-foundries, like TowerJazz, SMIC, VanGuard and Global Foundries were barely profitable the last few years.

Certainly not the case for SMIC — the owner of the biggest general purpose CMOS 200mm fab in the world.

 

[hkwint]

Slightly OT - but related - Tesla mentioned during their year end earnings call 2 days ago that their 2022 capacity would be limited by "the chip shortage" and *not* battery capacity. They said they expected next year to shift back to battery capacity being the limiting factor. Interesting as they were one of the companies better able to navigate the shortage in 2021 though of course they're still smaller than the bigger legacy automakers (about Subaru size in terms of revenue).

It's in line with Infineon, claiming the automotive chip semiconductor shortage:
-will have ended in 2023 for the parts they procure at other foundries,
-and H2 '22 already for the chips they manufacture "in house".

 

[hkwint]

Interesting to note TSMC revenue history:

2019 Automotive -7%
2020 Automotive -7%
2021 Automotive +51%

Just a slight correction... Even at a 51% upswing automotive is only 4% of TSMC's revenue.

The numbers from IC Insights (millions of chip units delivered to automotive) can be divided by the number of cars sold, to reach a "chips per car" number.
For 2021 it will probably be off, because of the "filling of the stock channel".

But from 2011 to 2019, it went from 250 to nearly 500; and is estimated to rise to 1000 for fully connected / AI driven / electric cars. So I think it's the biggest growth market for TSMC, even if starting from a low point of only 4% of TSMC revenue.

 

[jms_embedded]

I'm looking for quotes from automotive executives on investor calls that say "sorry, we made a mistake, we cancelled orders and we shouldn't have" and I can't find one, it's always "we can't get chips, supply chain problems, etc..." which is finger pointing.

And there may also be a shortage of car customers as well but we will never know because there are no cars on the lots to find out. It's the perfect crime...

Daniel, I've heard you mention this ("It's the perfect crime") several times, and although I don't trust car dealers, somehow I doubt there's a grand conspiracy among the auto industry to hide low sales behind the chip shortage.

I recently ran across this dataset on the St. Louis Fed, titled Auto Inventory / Sales Ratio, which I think might shed some light on the situation, as it uses the same kinds of terms we use in the semiconductor industry, e.g. days of inventory. (I'm not sure how they measure sales... per month? so a long-term average of 2.5 means that 40% of inventory gets sold each month?)

Note how low the ratio got in 2021. Looks like it's coming back up, slowly.

 

[count]

The automotive industry is used to being able to squeeze suppliers, and are now finding out there are limits to how hard they can squeeze. The semi industry now has much more supplier power and the auto OEMs are highly fragmented.

 

[cliff]

Penny-wise, dollar-foolish. Earthquakes happen. Toyota keeps extra inventory of chips. The American and other bean counters didn't.