Automotive IC Shortage Drags On
May 20, 2021
The current automotive semiconductor shortages won’t end anytime soon.
When the COVID-19 pandemic hit in early 2020, it wreaked havoc on the worldwide supply chain, but it especially caught automakers flat-footed. When the auto OEMs canceled chip orders during a roughly eight-week period of plant shutdowns, they later found their supplies of critical ICs had evaporated.
To make it an even more perfect storm, the pandemic was exacerbated by several other events, including the massive Texas winter ice storm that shut down chipmakers’ fabs in Austin, a major fire at Renesas Electronics’ wafer fab in Naka, Japan, and an ongoing drought in Taiwan. All of those events severely cut into automotive IC inventories.
Predictions for automotive chip availability are now all over the map. The most optimistic range from being alleviated by later this year, perhaps the third quarter, or early next year. But the auto IC shortage most likely will extend well into 2022, or even into 2023, according to industry analysts, when more chip foundry capacity is expected to come online from TMSC, Intel, and others.
The major pullback on auto IC orders by the major automotive OEMs already has proven to be extremely costly to that industry, pushing orders to the end of the line in the long, complex semiconductor supply chain, and forcing rolling plant shutdowns at virtually every vehicle manufacturer worldwide. The chip crunch has affected nearly every automotive manufacturer, from the Detroit Three of General Motors, Ford, and Stellantis (formerly Fiat Chrysler), to most of the major European, Japanese and Korean automotive builders.
Almost all major automakers have announced periodic plant shutdowns or slowdowns, with some taking extreme measures. Among the hardest hit, Ford had to resort to building its F-150 pickups — the company’s popular and by far most profitable vehicle — without the necessary chips to sell them, parking the trucks in lots until the ICs arrive to later make the units fully operational. Ford earlier announced that if the semiconductor shortage extends through the first half of 2021, it would impact Ford’s adjusted earnings before interest and taxes by somewhere between $1.0 billion and $2.5 billion. In Ford’s Q1 2021 earnings presentation on April 28, the company reported $3.3 billion in net income on $36.2 billion in revenue, noting it faced $2.5 billion year-over-year higher costs due to the global semiconductor shortage and higher commodity costs. Ford also said it has lost about 200,000 units in Q1, or 17% of production, and that it expects the semiconductor shortage may not be fully resolved until 2022.
What caused the shortage?
The auto industry’s chip woes can be traced primarily to OEMs’ pullback on IC orders during the roughly eight-week period of the pandemic, when plants were completely shuttered. An explosion in demand came from all corners of the electronics sector, as suddenly home-based workers unleased an unforeseen demand for chips in electronic devices of all types, including PCs, laptops, tablets, routers, and other networking gear. That ate into some of the markets previously catering to the auto sector.
“To best understand the challenges in addressing the supply/demand imbalance, it’s important to take a step back to take a look at how we got here, which has been 15 years in the making since the first smartphones came out,” said Kamal Khouri, vice president and general manager of GlobalFoundries’ Automotive Business Line. “Tiny compute-centric chips took one path, while feature-rich chips — which improve the user experiences of devices like touchscreen, blind spot detection, communication and smart pay — took another. R&D and investments focused on smaller chips, while feature-rich chips quietly became drivers of technological and economic growth.”
Today’s unprecedented demand for semiconductors is driven by the confluence of acceleration of the digital transformation, the surge in work from home/learn from home, the launch of the 5G super cycle, and global trade impact. In the early months of the COVID pandemic, auto sales softened, and the effects cascaded through the supply chain.
“Simultaneously, other customers — primarily those in the laptop, tablet, WIFI, and streaming supply chains — experienced a surge in demand at the time due to the global transition to working and studying from home,” Khouri said. “A few months later, unexpectedly, the auto industry saw a rebound and growth in demand. When the automakers attempted to re-engage their supply chains, much of the finite semiconductor manufacturing capacity was already reserved by other customers.”
Prior to the pandemic, there was no chip shortage. In fact, semiconductor sales declined slightly due to price softness in the DRAM market. “In 2019, there wasn’t really a chip shortage. The overall industry for semiconductors actually declined 7% in units,” noted Jim Feldhan, president of Semico Research. “It was sort of a correction year. Of course automotive does use memory. It’s not as extensive as some of the other industries, like servers, but we saw tremendous price increases. In 2019, there was a little bit of a slowdown and it was a correction year, and prices really fell in memory, but units were down.”
Once the pandemic hit, automotive companies reined in production. “They were closed for a number of weeks, and GDP fell by 30-some percent,” Feldhan said. “Things looked pretty bleak, and there was a lot of pullback from the automotive industry. There were a ton of cars they were just going to be sitting on and financing. What wasn’t realized was how strong the car market was going to be.”
Meanwhile, newly home-based users were buying all sorts of electronics gear — laptops, tablets, larger monitors, faster desktops and routers — to work more effectively full-time at home while the pandemic shut everything down. “For years, we’ve been saying people have been slow updating, upgrading their PCs, including on the business side,” Feldhan said. “But nobody thought that it was going to be a pandemic that caused it.”
In 2019 and 2020, there were some spot shortages in the automotive market, particularly for EVs from Tesla and a few other automakers, Feldhan said. “Tesla was the original adopter of silicon carbide MOSFETs, and they were in short supply,” he said. “There were two manufacturers (Tesla and Toyota) out there, and they consumed most of it. In fact, Toyota was looking at silicon carbide MOSFETs vs. standard silicon MOSFETs because the carbide is much more efficient, but they were afraid of the shortage. And they opted not to use that technology in their plug-in and hybrid vehicles because they were so concerned about the limited supply. What was miscalculated or not realized is when you have a pandemic, how that can affect the purchasing of electronics, so all the auto guys cut back. The thinking was probably was, ‘We can turn it on when we want.’ But then a lot of the capacity got swept up by other things.”
Some of these parts are interchangeable. DRAM, for example, can be used across a wide swath of applications. But automotive chips require special processes and a wider temperature range, and it’s expensive when they fail in the field. Automotive OEMs are looking for long lifetimes under extreme conditions with extremely low defectivity.
What’s missing and who’s making what
The majority of automotive ICs in scarce supply are microcontrollers (MCUs), which typically are produced on more mature process nodes in older 200mm and 300mm fabs. Most estimates call for auto IC lead times of 26 weeks, or even significantly longer, depending on the type of MCUs involved and the semiconductor process nodes involved in manufacturing the chips.
IHS Markit’s Supply Chain and Technology team, which has tracked the chip shortage since April 2020, recently released a white paper on the topic. “Because the cause of these constraints is the result of increasing demand from OEMs and limited supply of semiconductors, it will not be resolved until both forces are aligned,” wrote Phil Amsrud, senior principal analyst for ADAS, semiconductors and components at IHS Markit. “If the cause was a natural disaster, then the supply chain would respond with the appropriate recovery plans, and while that would still take months or quarters to implement, plans already exist. This is a case of balancing supply and demand, and with microcontroller unit (MCU) lead times being 26 weeks or longer, the supply chain constraints will likely persist until at least the third quarter of this year.”
What makes this particularly problematic is growing amount of semiconductor content in vehicles. Virtually every new vehicle today comes chock-full of automotive electronics, and that’s true even for those models that aren’t loaded with the fanciest options. While this trend has continued for the past couple of decades, it wasn’t until relatively recently — sparked by the advent of advanced driver-assistance systems (ADAS), the focus on electric and hybrid vehicles, and more connectivity functions in traditional vehicles — that foundries made automotive a top priority.
In the automotive industry, ICs fall into one of five domains — body, connectivity, fusion/safety, infotainment, and power train. The current shortages affect every one of those domains.
“I’ve had some people saying that the problem that we’re having right now is kind of a result of all of the new semiconductor applications that are on the vehicle,” said Amsrud. “But most of what we’re seeing the constraints in right now is MCUs, and that’s a lot of the legacy automotive stuff. It’s engine controls, transmission controls, steering, airbags, radios. It does touch some of the ADAS stuff — radars, front-view cameras. But the reason we’re having this problem is because the MCUs are literally everywhere in the car. It’s not like I’m just going to order a car and I’ll take some of the more modern features off, and then I’ll be able to get it. No, if you want the car to run, it’s got to have several microcontrollers in it. So it’s not the case that ADAS is the thing that’s causing this. It’s really the trend of putting more electronics in cars.”
Automotive MCUs primarily are made on 45nm and larger process nodes, Amsrud said. “There may be a few exceptions down in the 28nm range, but it is predominantly going to be in the 45, 90, 100, 130nm range. It’s kind of a worst-case scenario in that there hasn’t been a ton of expansion in those areas.”
This is compounded by a shortage of inexpensive used 8-inch equipment. Fabs now have to make almost the same investment for new 8-inch equipment as for new 12-inch equipment. “The economics suggest putting the 12-inch equipment in place, and then you tend to use that for more your leading-edge processes, as opposed to your legacy processes,” Amsrud said. “You get about 50% more out of a 12-inch wafer than you do out of an 8-inch wafer.”
With so many auto ICs now used in automotive applications, missing just one component can effectively gum up an assembly line. “We’ve seen this before in auto, where something as simple as a diode or a capacitor or some two-terminal device shuts down manufacturing lines,” Amsrud noted. “There’s not always an easy substitution, and that’s when you don’t have software involved. With MCUs, it’s affecting all the semiconductor suppliers pretty much equally. And even if you could say, ‘I can’t get an MCU from supplier A, but I can get it from supplier B,’ the problem is that the software’s not going to be compatible between the two of them. Whether we’re talking about a dollar part or a penny part, we estimate there’s about 1,200 semiconductors that go into an average car. That’s everything from a diode to a memory device, the most powerful SoC that goes into a car. So there are 1,200 opportunities on every vehicle where if something’s not there, that car can’t be built.”
No auto OEM spared
That has translated into worldwide shortages. Nearly all OEMs to one degree or another. “It’s really across the board. Even in Korea, they’ve shut some plants,” said Semico’s Feldhan. “Ford is building partial F-150s. They pull them off the line, and they’re sitting there until they get the chip, and then they’ll put them back on the line. For Ford, their most profitable vehicle is the F-150. But the engine controller is probably the same engine controller, whether it’s in an F-150 or if it’s in one of their other cars that have a V-8 engine in it.”
General Motors has cut production. So have Honda and Nissan, which collectively said they would build 250,000 fewer vehicles.
“Volkswagen said they’re seeing a computer chip crunch that’s hampering their operations, as well, particularly in the U.S., and Toyota announced they were going to have intermittent cuts in their shifts in production line,” Feldhan said. “So it really seems like nobody’s been spared the pain.”
Near-term, the shortage will be difficult, said GlobalFoundries’ Khouri. “Medium-term, we’re solving the supply-demand mismatch and will get customers the chips they need. The long-term is very bright. Auto manufacturers aren’t going to let this happen again. We’re already creating new direct partnerships with Tier 1 auto suppliers to secure capacity, and seek out closer, more strategic partnerships with foundries.”
This closer-knit supply chain is only going to grow in importance as the silicon content in autos continues to increase — both due to the shift to electric vehicles and to enhanced features including infotainment, radar and ADAS. “Computing has become the heart of the vehicle — to manage the battery, run the electric motors, brakes, lights and other critical systems,” said Khouri. “Chips reach every corner of today’s vehicles, from the microcontrollers that allow you to put your window up and down, to safety features including radar and ADAS, to sophisticated engine functionality. Vehicles are now part device, and the features we expect can’t exist without pervasive chips.”
Long auto lead times, complex qualifications
Depending on the auto ICs involved, lead times for chips have stretched out significantly during the crunch, going from a typical 16 to 18 weeks for some components to much longer. This is partly due to material availability, and partly to chip manufacturing equipment, which is also in short supply. The industry’s complex supply chain for auto ICs typically involves chips purchased through the automotive electronics suppliers and chip distributors, rather than direct orders from chipmakers.
“In general, overall semiconductor supply right now is very, very tight. Somehow we just cannot meet the demand,” said I-Wen Huang, vice president, sales-Americas for Winbond Electronics, a Taiwanese maker of specialty memories and ADAS systems for automotive applications. Before the pandemic hit, typical cycle time to manufacture a semiconductor device from start to finished goods and delivered to the customer was four months, Huang said. Tor many parts, lead times have stretched to five or six months.
Huang sees the shortages lasting at least until late 2021, or even into 2022. “It really depends on how big the demand will be next year,” he said. “This year for sure, if you ask any semiconductor manufacturer, they say ‘I’m sorry, this year’s capacity is already fully booked by our customers.’”
The current delays are partly due to substrate material shortages, said Huang. Winbond doesn’t make wafers, so it has had delays receiving substrates to make its automotive chips. Other factors affecting lead times are shortages of semiconductor equipment. But another key obstacle is the very long qualification of auto ICs, in which qualifying a new device can be an arduous process, as well as the automotive industry’s long-held just-in-time (JIT) business model of delivering components only as cars and trucks are assembled.
The automotive industry uses very strict, rigid process that takes a long time to get ICs qualified, Huang said. “Typically you won’t be able to qualify a device for automotive usage in shorter than a year,” he said, noting this stringent control on automotive electronics ensures safety. “When they order a product from us, we have to put in a very highly controlled process during the manufacturing, so to supply an automotive customer, it takes an even longer lead time, and we apply more quality control process during the manufacturing. Automotive customers specify every detailed step. You cannot use a replacement because we never qualified the replacement material.”
Single sourcing of auto ICs for carmakers also makes it impossible to take a chip designed for one auto application and swap it for a similar device manufactured specifically for another automaker’s application. “Every car manufacturer has their own design, their own specifications. The Renesas ADAS device may not be the same, for example, as Intel’s device,” Huang said, “So you cannot replace it, because all the hardware and software are customized.”
In the auto industry, OEMs use common platform designs within each manufacturer, but there’s no commonality between different OEMs. “A carmaker, let’s say Ford, they may have the common platform design for different models that all are a Ford car,” Huang said. “But the platform they design and develop, it cannot be used on General Motors car. So there is no commonality in this industry. Everything is customized, and this makes the supply very complicated.”
Expanding foundry capacity
As the auto IC crunch reached a crisis point earlier this year, the industry scrambled for new solutions. Intel announced plans to re-enter the foundry business. TSMC, meanwhile, said it plans to spend a total of $100 billion over the next three years. And on April 28, UMC announced plans to expand capacity for its 300mm Fab 12A Phase 6 (P6) in Taiwan’s Tainan Science Park.
In addition, the U.S. government may increase semiconductor industry support. The Biden Administration pledged to seek congressional support for authorizing new industry funding at the virtual global chip summit April 12 held with many participating automotive and semiconductor CEOs. The Biden plan would inject some $50 billion into restoring the U.S. semiconductor industry as part of the administration’s proposed infrastructure plan, which ultimately would need congressional approval on funding.
Still, recovery in automotive IC capacities may take a lot longer than expected. Any new fabs being built now likely would not come online and produce effective yields for nearly two years, but TSMC actually has some fabs in Taiwan that are coming online this year, noted Adrienne Downey, Semico’s director of research.
“We think the shortage is going to last through this year and into early next year,” Feldhan added. “It’s hard to pin an exact date, but it could go into mid-next year. The real kicker is how well the current fabs execute this year, because it seems like every year there is a bit of process and yield improvement, so they get more chips out. But if the shortages make it so that the OEMs can’t make as many products as the market is demanding, then that demand will shift. The more that happens, the longer we’ll see the shortage be with us. When push comes to shove, I would say it will go through April of next year.”