Today, it’s not that difficult to secure one of those new iPhones that came out last week. Car buying, comparatively speaking, is still a chore, especially given how supply chain constraints have thrusted prices upwards considerably for new and used vehicles alike.
But if you ask an executive at a chip manufacturer who’s to blame for what automakers are going through — like, say, Intel CEO Pat Gelsinger — the answer might not be an especially sympathetic one. From Fortune:
“I’ll make them as many Intel 16 [nanometer] chips as they want,” Intel chief executive Pat Gelsinger told Fortune last week during his visit to an auto industry trade show in Germany.
Carmakers have bombarded him with requests to invest in brand-new production capacity for semiconductors featuring designs that, at best, were state of the art when the first Apple iPhone launched.
“It just makes no economic or strategic sense,” said Gelsinger, who came to the auto show to convince carmakers they need to let go of the distant past. “Rather than spending billions on new ‘old’ fabs, let’s spend millions to help migrate designs to modern ones.”
The implication here is that the auto industry is far too reliant on archaic tech that isn’t applicable to other consumer tech fields. It’s now finally reckoning with its reluctance to change, and only a fool would invest in shops to pump out the outdated silicon cars require.
But is that a fair assessment? As Fortune notes in its own piece, there are reasons why carmakers — some of the largest corporations in the world — choose the chips they do. The comparison to smartphones is moot. If your iPhone crashes, you press volume up, then volume down and hold the side button until the screen turns off and turns back on again and voila, you can lose yet more hours to Instagram like you never missed a beat. The potential ramifications of a glitch in a metal box traveling at many miles per hour are a little more severe. That’s especially true if you’re talking about modern vehicles with driver-assist functions.
However, Gelsinger’s argument was one I hadn’t heard before, and I figure the guy heading up a company like Intel probably knows something I don’t. So I asked some auto industry veterans to weigh in on the credibility of his claim.
“Automobiles have a long operating life, compared to many other consumer devices,” Thomas Coughlin, Fellow at the Institute for Electronic and Electrical Engineers, told me. “People expect their cars to last for more than 10 years, whereas many five-year-old consumer products are often considered nearly obsolete. Thus, parts for automobiles, including some chips, are often built on older, proven technology, rather than the latest available technology.”
What automakers require is somewhat at odds with what chipmakers prefer and are tooled to produce: smaller, more densely packed chips, that can be manufactured at lower cost and yield more units.
“The semiconductor companies continue to operate older fabs using bigger lithographic processes (like 90 nanometers),” Coughlin added. “But they don’t invest in new plants making these parts, because the costs of manufacturing the chips are less for newer fabs with finer lithographic features. This is because the costs of manufacturing a given size wafer (with devices for a given application) from which chips are made is relatively stable. However, if smaller lithographic features are used, more of a given chip die can be made from the wafers, or more functionalities can be made for the same price in a fixed die size. This provides either more working die (chips) from a wafer at lower cost or die with more functionality for the same cost.”
For reference, the system-on-a-chip that serves as the proverbial brain of the new iPhone is built on a 5nm lithographic process. The “process technology” of a semiconductor is defined in nanometers, and its value refers to the shortest transistor gate length — the distance between two transistors — on the chip. Here’s how PCMag’s glossary contextualizes size and process technology:
The size of the features (the elements that make up the structures on a chip) are measured in nanometers. A 22 nm process technology refers to features 22 nm or 0.022 µm in size. Also called a “technology node” and “process node,” early chips were measured in micrometers.
Historically, the feature size referred to the length of the silicon channel between source and drain in field effect transistors (see FET). Today, the feature size is typically the smallest element in the transistor.
Car companies use a varied selection of chips that tend to be many times larger than what’s employed in most consumer electronics — perhaps as big as 45 or 90 nanometers. These are often used for simple tasks like raising and lowering windows and climate control.
The dilemma is that if those chips are tiny rather than huge, a given size wafer will yield many more of them. Miniaturization thus makes supply easier to maintain, and allows chip manufacturers to reap a more lucrative return on their investments.
“So, chip fabrication for older (bigger) lithographic features tends to be retired or at least new fabs using these older technologies won’t be built to meet an increase in demand,” Coughlin sums up. “Overall, the chip companies do have a valid point.”
However, to suggest as Gelsinger did that the burden to adapt should fall squarely on automakers simplifies the issue. General purpose chipmakers don’t seem to grasp the unique challenges of the automotive sector — something that became clear to me after chatting with Jon M. Quigley, Society of Automotive Engineers member and columnist at Automotive Industries.
“Qualifying a product, specifically testing activities, are costly and requires time, talent, and equipment,” Quigley said. “Some of the test equipment requirements are expensive and often not on hand at the OEM but will require an external lab, and booking time at this lab can be a long lead time activity, and is necessary for certain product certifications. Depending upon the vehicle system commonality, this testing might have to be performed on multiple vehicle platforms.
“Making changes to an existing product, changing an integrated circuit that only has the difference in the manufacturing processes would still require this sort of testing. Unless there are some compelling associated cost improvements to recoup the investment, this is not very plausible.”
It’s easy for those of us on the outside to miss the many steps of validation automotive components are required to go through before they end up in what we drive. Ultimately, carmakers don’t care how small or new a chip is; all that matters is that it works for its intended purpose and is properly vetted, Quigley said.
“When a product is being developed, engineering either is already aware of the integrated circuits that can be part of achieving the customer and the OEM’s needs. This is bound by cost, design constraints, performance, and what already exists. There are no OEMs that put integrated circuit (IC) manufacturing requirements in their product documentation. We expect the component to perform as documented as we base our designs on the technical specifications of the IC.
Quigley added that trying to design new chips and vehicles that will use them in parallel often introduces yet more headaches.
“The few times I have been a part of product development that was also connected to the development of a new integrated circuit ended poorly, and never connected to a manufacturing process. That is, the integrated circuit identified was under development and the product was going to use this integrated circuit as the production date was in line with the production date of the OEM product. The integrated circuit was subsequently not able to be produced, the yield was too low for it to be an economically viable product for the IC manufacturer. This failure delayed the product under development by the OEM required retooling, retesting and the project extended for months longer, with increased costs.”
Chipmakers want as much miniaturization as possible to maximize production efficiency, automakers need significant lead time to make sure a chip will work for them. Each industry has reasons for operating the way it does. That doesn’t change the fact that someone’s going to have to budge to address this shortfall. I asked Judy Curran, Head of Global Automotive Strategy at Ansys, if there’s any hope the two sides could meet in the middle.
“If semiconductor companies were to invest in more capacity, they will need to increase capacity of newer tech, especially as vehicles will continue to grow into more ADAS and electrified technologies which cannot function on the older process nodes,” Curran told me. “However, the automotive industry will still need a variety of chips for some time. Over the years, electrical architectures will change to focus more on larger ‘supercomputer’ chips and less on smaller microcontrollers, but there is a long way to go.
“That said, there are certainly opportunities to merge functionality from discrete electronics and controllers into a smaller number of chips and electronic systems. System-on-chip (SoC), for example, has been successful in data centers and mobile devices, and 3D IC (three-dimensional integrated circuit) brings higher integration. But the challenges to adoption, including difficult and expensive designs, persist.”
The rate of miniaturization will pick up for cars, but it’s going to take time. I wondered if an automaker acquiring and vertically integrating a fab into its pipeline might present one way for carmakers to prioritize production of the silicon they require. If automakers need these chips that semiconductor companies don’t want to make, why not make them themselves? Curran said not to bet on that — at least, not in the most direct sense.
“I do not anticipate automakers getting into the fab business, but I do suspect automakers will engage in strategic partnerships with chip suppliers (big and small) to ensure their needs are met both in the short and long term; automakers always insource and outsource engineering and manufacturing depending on the technology and business constraints.”
You can expect these to play out like the many battery partnerships that have cropped up in the automotive industry every other week, like BlueOvalSK, or that joint venture between Mercedes-Benz, Stellantis and Total.
“Because of the importance of the chips, as they are now a significant business constraint, automakers will find ways to own or control more of the chip supply base going forward by partnering with ASIC design companies who do similar design service for networking companies,” Curran said. “This is no different than the battery announcements we are seeing for electrification. While automakers likely won’t buy a fab, this demand will likely drive consolidation of the small fabs, possibly into larger companies.”
Over time, the transition to newer technology may naturally happen, but certainly not quickly enough to Band-Aid the snags of the present moment. That doesn’t give anyone a single, solitary scapegoat, and it’s not the easy answer anyone likely wants to hear — not prospective shoppers, not automakers and not the CEO of Intel. But it’s the most realistic answer nonetheless.