How to Think About the Automotive Chip Shortage
And why everyone should be watching Malaysia's monthly trade statistics
Welcome to SemiLiterate, a guide to the chip industry through the lens of public policy. Feel free to share with others you think may be interested.
BLUF: This post summarizes information about the ongoing worldwide chip shortage: 1. Why it started 2. How it got worse 3. Who is hurting the most 4. What to do about it and 5. Where to look for signs of a recovery. I argue there is little that government can do to solve what is, fundamentally, an industry-wide problem that only industry can solve. I also argue that we should interpret Malaysia’s exports of integrated circuits as a bellwether for the chip supply chain crunch easing more generally.
Why it Started
In mid-2020 the automotive industry (logically) assumed that consumers would buy fewer cars during a global pandemic. As a result, automotive companies reduced their orders for chips with their contract chip manufacturers (“foundries”). In turn, these foundries, which now had unexpected capacity they needed to fill, sought new business and orders to fill their factories’ excess capacity. They re-tooled their automotive chip factory lines so they could take orders from consumer electronics chipmakers like Nvidia, which makes the graphics processing unit (GPU) chips found in the sorts of video game systems that saw a tremendous spike in demand while everyone was home during the pandemic.
However, the automotive industry’s assumption proved incorrect, and demand for cars shot up in the second half of 2020. The automotive manufacturers, caught off-guard by this increase in demand, went back to their foundries and attempted to place orders for automotive chips, only to be told the factories were at capacity.
Careful observers of this industry noticed as early as last fall that foundry quotes were going through the roof. If you are anywhere in the foundry business, you’re going to have a fantastic 2021:
Pure-play foundries including Globalfoundries, United Microelectronics (UMC) and Vanguard International Semiconductor (VIS) have raised 8-inch foundry quotes by 10-15% in the fourth quarter, with the quotes set to rise by another 20-40%.
How it Got Worse
Note that the “8-inch foundry quotes” referred to in the previous paragraph are for factories that operate 8-inch (or 200mm) wafer manufacturing processes. 200mm wafers are a smaller/older size of wafer frequently used to manufacture chips for automotive end uses that, while advanced, are not state of the art. State of the art chips are manufactured on 300mm (12-inch) wafers to take advantage of economies of scale. Chips made on 300mm wafers are mainly high-margin chips destined for consumer electronics end uses (think Apple Watches and Playstation 5s).
Put yourselves in the shoes of a foundry that makes automotive chips: you have a 200mm wafer line that usually makes low-margin automotive chips. Its a good, consistent business. And then a global pandemic happens and your main automotive customers cancel their orders. Major problem. But, in a stunning turn of events, you get approached soon thereafter by a consumer electronics chipmaker who is desperate to make more chips for Playstation 5s in 2020 and they are willing to pay a premium to compensate you for the trouble of re-tooling your factory line to make their chips (a costly process that can take months). The worst year ever has just turned in to the best year ever. But then, your old automotive customers come knocking asking if they can re-place their orders, just as you’ve finished the expensive task of re-tooling your factory for Playstation 5 chips. What do you do? You make the Playstation 5 chips, of course.
So the automotive industry was faced with a shortage of available factory capacity to make the chips they need. This shortage was exacerbated by several incredibly unlucky events:
a February 2020 power grid failure in Texas that shut down SamsungFoundry, NXP, and Infineon chip factories. SamsungFoundry makes Tesla’s chips, while NXP and Infineon are the #1 and #2 automotive chip makers in the world.
a March 2020 fire at a Renesas factory in Japan which partially destroyed its clean room, stopping chip production there until June. This halt is costing Renesas ~$156 million/month. Renesas is the #3 automotive chipmaker in the world.
I feel like this point bears highlighting: the #1, #2, and #3 largest suppliers of automotive chips in the world all had black swan events shut down factories in early 2021 at a time when their production was already maxed out.
Who is Hurting
The largest end use markets for semiconductors are computers and consumer, industrial, and communications equipment, which collectively consume over 85 percent of all semiconductors. However, industry analysts expect that the automotive semiconductor market will be the fastest growing end use market for chips in the coming decade.
Automobiles increasingly consume semiconductors associated with on-board navigation systems, infotainment systems, autonomous systems, and other safety features. As levels of automation increase, industry forecasts suggest that the average semiconductor content per car could increase from $160 to $970 and total 3,500 chips per vehicle. Auto companies are more intensively consuming chips and incorporating them throughout the vehicle:
Europe (33.4 percent) and Japan (24.8 percent) lead the Americas (8.4 percent) in consumption of automotive chips, though many U.S.-headquartered firms have announced plans to enter or expand in this market. To put all this in plain English: automakers in Europe (VW, BMW, Mercedes, Fiat) and Japan (Toyota, Suzuki, Mitsubishi) lead automakers in the U.S. (GM, Ford, Chrysler) when it comes to building cars that are hybrid, electric, and/or rich in navigation/infotainment systems. This is a good thing so long as the chip supply chains on which all of these features rely is healthy. Consumers love all the new features coming from Japanese and European cars. But when a supply chain crunch happens, these European and Japanese automakers are all the sudden even more exposed than their U.S. competitors.
Analysts are saying this shortage will linger through the first half of 2021 and result in 700,000 fewer new vehicles in Q1 2021. One report estimates it will cost the auto industry $61 billion. However, these supply chain constraints will disproportionately affect Japanese and European automakers and their production delays will likely extend beyond those of American automakers.
What to Do About It
The news that the White House is hosting a summit this week with chipmakers is welcome, but no one should expect any breakthroughs. In addition to the basic fact that it takes 60-90 days to make a chip and that building a new factory takes 12-18 months, there are several other specific factors that mean there is simply no way the government (any government) can be reasonably expected to help resolve the problem in the short term:
Automotive Chips Take Time: we’ve all had the experience of our smartphone unexpectedly shutting down or freezing. That’s a hardware hiccup. Imagine if the hardware that hiccupped was your power braking system while you’re taking a turn at 55 mph on a highway during rush hour. For that reason, automotive chips have much more unique performance requirements than consumer electronics chips.
Source: Another USITC report These unique performance requirements mean:
Supply Cannot Increase Quickly: It takes 12-18 months to build a chip factory, so you cannot increase supply overnight even if the government subsidized the heck out of it. The president of the Semiconductor Industry Association said things should resolve in half a year and that the government should stay out of it.
Qualification Times Cannot Shrink: Even if supply could increase quickly, because of the unique performance requirements detailed above, it takes a long time for automotive chipmakers to have their products qualified for use in cars. One U.S.-headquartered chip manufacturer reported that it can take up to six months before their semiconductors are qualified by an auto OEM for use in vehicles. During this six-month qualification time, the chips are subject to stress tests of 2,000 hours, where a chip is exposed to higher voltages and temperatures than its rated specifications. Even after meeting these stringent performance requirements, the chips will be further tested by the OEM at the system (automobile) level, making the entire process take well over one year.
Government Solutions:
The hard truth is there are no government (policy) solutions to this supply chain crunch in the short term. The German government can encourage TSMC to produce more automotive chips all it wants, but the whole chip industry is basically operating at capacity. Unfortunately, the best the government can do in the short term is make sure that workers getting furloughed at automotive factories like GM and Ford are eligible for unemployment.
Industry Solutions:
Shut down: Auto factories can and will shut down until the chip supply chain stabilizes. Ford, GM, Suzuki, Subaru, Fiat Chrysler are all doing this.
De-Content: Automakers are shipping cars without some essential electrical components.
Pay More: Foundries can re-prioritize their orders, but it comes at a cost to the buyer. So if an automotive company really wants to get their chip supply chain shored up, they can pay extra to “jump the line.”
Where to Look for Signs of Recovery
I would argue there are two major signals to watch in the midst of this chip shortage: (1) chip companies that have significant exposure to the automotive end-use market and (2) Malaysian trade statistics.
Watch Auto Chip Companies
An automotive chip supply shortage is terrible news if you’re in the business of making cars. But if you are in the business of making automotive chips, you’re going to have a great 2021. Many U.S. chip firms are engaged in designing and fabricating chips for automotive end uses. Thus far, this post has focused mainly on foundries and how they are set up for a good 2021. But more conventional U.S. chip companies are likely to have a great 2021 too.
Several U.S. semiconductor firms have made substantial investments targeting the automotive semiconductor market and are among leading firms worldwide in supplying this end use market. Texas Instruments, ON Semiconductor and Microchip Technology are all in the top 10 firms globally, with other leading firms concentrated in Europe and Japan:
The companies listed above are going to have a good 2021 for their auto chip business. Because of the long qualification time and high quality expected, demand for automotive chips are more “sticky” than standard consumer electronic chips. In general, auto OEMs maintain the same sourcing throughout the model run of a vehicle (approximately 5 years) and consider new suppliers only when they are launching new models. Auto OEMs are likely to be repeat customers of companies and brands that they have previously qualified both for use in existing vehicles as well as future models, given the expectation that the supplier supports their product for the lifetime of a given vehicle. U.S. headquartered semiconductor firms that are qualified suppliers for automakers will benefit from this stickiness, customer loyalty, and reputation for quality.
Follow the Silicon: All Eyes on Malaysia
If you’ve made it this far in the post, well done. Bear with me here because this is about to get even more geeky: I am going to argue that we should interpret Malaysia’s exports of integrated circuits as a bellwether for the chip supply chain crunch easing.
So, the production of semiconductors occurs in three distinct stages: design, manufacturing, and assembly, test, & packaging (ATP). Though ATP providers only account for 10% of a chip’s value, because they are the final step in semiconductor production, the export of the finished good registers the full value of the good against the balance of trade with a given country, even in countries where the design and/or manufacture of that chip (and thus 45-90% of its value) may have originated. Because of the high concentration of ATP facilities in Southeast Asia (due to low labor costs), U.S. trade data reports a significant bilateral trade deficit with select SE Asian countries for trade in semiconductors:
U.S. imports from SE Asia of semiconductors are comprised of products that already contain significant U.S. value. U.S. semiconductor firms often import their own products after the ATP process for warehousing or sales purposes. U.S. manufacturers of products that incorporate semiconductors (such as computers or automobiles) also import a significant portion of products, which include U.S. added value, directly from overseas.
For example, a company like Texas Instruments might design and manufacture a $10 automotive chip in Richardson, TX but then send that chip to a test facility in Malaysia, when $9 of the value is “built in,” for ATP. Then when that chip is exported from Malaysia back to the U.S. so Ford can put it on a printed circuit board behind the dashboard of an F-150, the value of that export is registered as a $10 chip, even though Malaysia only added $1 of the value.
This part of the semiconductor global value chain leads to some funny trade statistics, especially when you consider that Malaysia doesn’t have many semiconductor factories but it has many many many more semiconductor ATP facilities.
Malaysia: Semiconductor Production vs. Semiconductor ATP
There are different kinds of “exports” recorded by international trade statistics.
Domestic Exports refer to commodities grown, produced or manufactured in a country.
Malaysia’s Domestic Exports of chips was ~$29 billion in 2020.
Total Exports refer to the total physical movement of merchandise out of a country to another country.
Malaysia’s Total Exports of chips was ~$50 billion in 2020.
So there was roughly a $21 billion gap between the value of chips domestically produced-and-then exported vs. the value of all chips exported from Malaysia in 2020. That $21 billion gap is the value of all the chips that were sent to Malaysia for ATP before being re-exported for end use (automotive, communications, etc).
The vast majority of chips that are made in semiconductor factories in Malaysia are made by the big automotive chip companies: ON Semiconductor and Infineon both have semiconductor factories in-country. But, in addition to this domestic production (which again, when exported, would count towards the Domestic Exports total), Malaysia also is home to ATP facilities used by nearly every automotive chip maker AND the foundries that are being asked to ramp up production in response to the chip shortage.
So what we should expect is that Malaysia’s exports of chips (both Domestic Exports and Total Exports) are spiking right now as auto chipmakers and foundries increase production and send those chips to Malaysia for ATP. But we’re not seeing that.
In fact, when one looks at the monthly trade data provided by the Malaysian government on its (surprisingly user-friendly) trade data website we find that Malaysian Total Exports of chips sharply declined in January and February of 2021 (the light blue line in the chart below).
Conclusion
I wont put up another chart, but suffice it to say that the story looks similar when one runs the numbers on Malaysia’s Domestic Exports of chips too. That is not good news if you are an automotive company, but note that it reflects the trend line in previous years of declines in exports during January/February. Once that trend line starts to tick up we will know we’re starting to see an actual easing of this chip supply chain crunch. Until then, there is not much policymakers can do except wait for Malaysia’s March 2021 trade data release.
Note: the views here are my own and drawn solely from the documents I cite here.