(Image: iStock)
By Junko Yoshida
What’s at stake:
Lifecycle management is a mandatory conversation on the topic of aging products. It’s time for automakers — eager to roll out software-defined vehicles — to acknowledge they won’t be able to support software updates forever for a 10, or 15-year-old SDV. If so, what’s the plan?
Manufacturers must plan for the end of life for their products. For example, can they continue to get replacement parts and components used in their systems?
Recycling old products is also a big deal.
Can a product’s components be safely and sustainably disposed of? Is there a built-in used market? Will used systems be shipped to developing countries, or do they get dumped into a landfill?
These are the principal questions asked in Supply Chain 101.
And yet, before launching new products, how many system designers and manufacturers think through the lifecycle? Whether it’s an AI-driven cloud connected device or a software-defined vehicle, do vendors plan for the day when the product gets bricked?
On this issue, consumers have been already fooled once, in 2022. Even reputable OEMs such as BMW declined to upgrade a 3G cellular modem with a 4G cellular modem in consumers’ vehicles, after all major US cellular operators shut down legacy network support.
Given that the software-defined vehicle (SDV) is today’s buzz among OEMs, what’s the end-of-life plan?
Fallacy of SDVs
Last week in a session, entitled “The AI Lifecycle: A Pragmatic Discussion of the Costs of AI Implementation,” at Business of Semiconductor Summit 2024 (BOSS 2024) organized by The Ojo-Yoshida Report, Bryan Reimer, research scientist at MIT Age Lab, and Phil Koopman, professor at Carnegie Mellon University, discussed the dilemma baked into carmakers’ aspiration to keep updating software for 10 or even up to 20 years. They noted that the hardware OEMs already installed in vehicles is unlikely to keep up with software progress. (Note: Video playback is available. Register free here.)
OEMs have yet to seriously discuss this issue with their technology suppliers or customers.
Are you really going to do a software update for a 15-year-old CPU and it’s going to run the same software as your new car? I don’t think so.
Phil Koopman
Carmakers are counting on SDV for new revenue streams by adding features and capabilities via software updates. Yet, as software advances, OEMs might well find that escalating software updates (and replacing the hardware to support them) can prove costlier than they imagined.
Putting it simply, Koopman said, “We keep hearing companies say that ‘we’re going to do software updates,’… but I start saying, ‘You’re really going to do a software update for a 15-year-old CPU and it’s going to run the same software as your new car? I don’t think so.”
Reimer noted, “When automakers shift their business model to a subscription-based service for up to eight to 10 years after vehicle launch, they are trying to sandbox the risk of their vehicle features [getting outdated]” for consumers. But what they are not sandboxing off is “all the fundamental control and active safety systems [that could suffer] … from a lot of the same reasons that we say when the assistive driving features won’t work anymore.”
Tech suppliers going along
Some technology suppliers are amenable to the auto industry’s SDV dream, in hopes of enabling good choices by OEMs.
Robert Day, director, Autonomous Vehicles at Arm, told us, “We are offering the automakers a plan to futureproof the SDV.” However, that bold statement is based on many assumptions on Arm’s part, including “if there is the right amount of compute available to process data,” and “the possibility to offer hardware upgrades for the central compute system in the future,” and “the automotive industry learning from software development methodologies and tooling from another heavily software defined industry – the cloud.”
In contrast, Brian Carlson, global marketing director at NXP Semiconductors, is skeptical. Over time, he said an older vehicle “won’t have the same level of capability as a new SDV … The question for OEMs is how long they plan to support upgrades to their vehicles.”
At NXP, he said, “We believe most likely in the range of 5-10 actively, but with the ability to update safety and security longer.”
Jeffry Chou, CEO of Sonatus, a middleware company, acknowledged, “I haven’t heard anything from OEMs” about their plans. He noted, however, that carmakers are [still] on Day 1,” dealing with SDV’s initial hurdles.
Chou agreed with Reimer and Koopman that SDVs will be “like your older PC not being able to run the latest high performance-requirement applications.” He added, “And, unlike PCs, the life cycle of your vehicles is much longer.”
In Chou’s view, vehicle software providers [like Sonatus] must rise to this occasion. It is their responsibility “to add value in software with older hardware.” He said, “This isn’t impossible. [But] it takes planning, skill, and experience.”
Moritz Neukirchner, senior director, strategic product management at Elektrobit, an Erlagen, Germany software company, agreed that the lifecycle mismatch in SDVs is a problem. For passenger vehicles, he noted, “We are looking at 10 years of cybersecurity maintenance after ‘the end of production’ mandated by the R155 regulation… So not even looking at functional upgrades may be a challenge from a hardware perspective.”
He noted, “My strong assumption is that we will see a staged maintenance/upgrade timeline for many vehicles, as we already have for smartphones today, i.e. functional upgrades for a few years (smartphones are currently on 3-5 years today) and then a few years cybersecurity patches.”
What’s the plan?
Neither Reimer nor Koopman is saying that the industry should jettison SDVs. But the industry needs a plan.
Reimer said, “Just because it’s hard to maintain a lifecycle doesn’t mean we shouldn’t do it.” The industry must outgrow the fallacy of a perfect-world scenario, in which carmakers leverage AI and roll out — for the next 25 years — improved features via software updates that make everyone happy.
There will be a D-Day, when carmakers can no longer support their software updates. The question to OEMs is whether they are prepared for D-Day, said Reimer.
Both Reimer and Koopman, as well as tech suppliers we talked to, shared some options.
Quoting an OTA presentation at a recent SDV event where an ex-Telsa employee (early Model S days) talked, NXP’s Carlson explained that some companies are already front-loading vehicles with a lot more hardware (including memory/storage) than current models need. This was Tesla’s strategy. The ex-Tesla speaker said it has paid off with the ability to support a decade of upgrades.
Elektrobit’s Neuikirchner observed, “I have seen approaches, e.g. with blade-architectures to look into the upgradeability of hardware.”
Other options include designing electronics to deal with limitations in software complexity, or modular designs to swap older computational systems for new ones after four or five years.
Think about a submarine, Koopman said. The military cut a giant hole in one to forklift out an old computer to install a new one.
He said, “Maybe carmakers can create a slot where you just slide in a new tablet every five years?” Whatever the choices, doing it so upfront is essential when designing a system. “But I haven’t heard people talking about this as a plan,” said Koopman.
Add the issue of chips for AI.
“Your chips 10 years from now can be different,” said Koopman. “Scaling in five-year time frames is different than scaling in 20-year time frames. So, these are the kind of discussions we need to have.”
Computing is a big part of the SDV, Reimer reminded. “Vehicles will not be made with the same materials 20 or 30 years from now … Radars and vision systems are going to have to be modified. It’s likely all the sensing hardware and even the network components will change.”
Reimer added, “You may be defining the vehicle based upon software, but it’s hardware componentry that’s going to be harder to change.”
Data centers: A model for SDVs?
SDVs are often described as “Data Centers on Wheels.” Data centers embody SDV promoters’ aspirations for collecting and processing data to improve their vehicles.
Arm’s Day said, “The more data the car collects from real-world scenarios, the smarter the car becomes.”
Sonatus’ Chou noted, “Data centers are heterogenous in their hardware and software, which means they have old hardware and new hardware. They also continuously replace and upgrade their hardware.”
Data center operators, in fact, constantly rip their data centers apart, discarding old hardware and putting in new ones every few years.
“But are you going to throw away your cars every four years? I don’t think so,” said Koopman.
Resale value
Another unspoken hitch is the resale value of an old SDV, after its OEM stops software updates because it’s too costly to replace hardware.
“At the very least, the car companies need to promise that they are not going to brick our car after so many years,” said Koopman.
Why? Whether you are leasing or planning to resell your vehicle, you assume it’s worth something. Maybe not, warned Koopman. “You’re going to see faster depreciation … your vehicle near zero worth after 10 years, either because it won’t be supported, or it has to be refitted.”
Neither consumers nor the used car market has been forewarned of this. Koopman suspects that the current used car market “is not really factoring the cliff that’s going to come when these manufacturers can’t support their vehicles anymore.”
For an end-of-life of your product, the best approach is ‘to design for it up front, so that you can deal with some of the difficult issues a lot better.’
Bryan Reimer
Electronic waste
Going a step further, Reimer worries about used vehicles that will be inevitably shipped to developing countries. When safety features updated via software can no longer match a vehicle’s old hardware, what then? “Are we going to turn off those safety features and ship them to countries where the risks of auto fatality are magnitudes greater than they’re here in the United States,” he asked.
Reimer worked on “an end-of-life, recycling and logistics” for his master’s thesis 25 years ago. He cites electronic waste as a big issue. Again, the best approach is “to design for it up front, so that you can deal with some of the difficult issues a lot better,” said Reimer. Consider EV batteries. Had EV companies originally designed batteries that are easier to recycle, they would not now be scrambling for technology solutions for recycling.
There is a way out
So, how to solve the SDV software-hardware mismatch riddle? Koopman sees “a way out of the box.”
Given that carmakers are counting on updatable AI to be critical, it must be acknowledged that OEMs who make money off software updates can’t build a business case for supporting SDVs after they’ve been sold.
Think about a 20-year-old used SDV, said Koopman, bought by consumers who can’t afford to pay carmakers to maintain obsolete technologies.
How will carmakers make money?
The way out of the box, as Koopman noted, is to pick a number of years during which carmakers must provide 100-percent support for whatever they’ve sold. After that, carmakers would turn off all computer-based functionality, because they can’t afford to maintain it and it’s not new revenue.
Instead of dumping the older car (“you’re not building disposable cars”), carmakers should open a new business model for others in the secondary market to step in.
“Maybe you make the computer-based functionality into a module that you unbolt. There’s this great secondary market for aftermarket electronics you bolt in.” The result would be a basic, inexpensive vehicle with no self-driving, but that still works, he explained.
In Reimer’s view, this solution is similar to today’s after market for components like engines and transmissions. It poses an opportunity “for replacing componentry with something better, something newer a decade or two down the line. The more we plan for that, the more successful we’re going to be.”
Bottom line:
It’s time for automakers to recognize the fallacy of the SDV. There is no such thing as a vehicle that can keep getting better forever simply via software updates, unless hardware is refitted. As Reimer noted, “The key is collaborative discussion within the industry. If manufacturer A is not going to handle this, be honest and transparent about opening a business opportunity for aftermarket companies. Work with them now, enable them so that they can build the off-market parts to solve this problem a decade from now.”
Junko Yoshida is the editor in chief of The Ojo-Yoshida Report. She can be reached at [email protected].
Copyright permission/reprint service of a full Ojo-Yoshida Report story is available for promotional use on your website, marketing materials and social media promotions. Please send us an email at [email protected] for details.