Direct answer: AI is succeeding in the automotive industry where the software-defined vehicle (SDV) struggled, because its value shows up in the same two numbers every OEM already tracks: time and cost. Sonatus CEO Jeff Chou points to a concrete result. AI tools helped Nissan Technical Centre Europe cut vehicle testing investigation time from two weeks to two days, the kind of proof point SDV investment alone rarely produced.
Key Takeaways
- AI succeeds financially where SDV struggled because its value is measurable in time and cost, not projected infrastructure benefits.
- Nissan Technical Centre Europe cut vehicle testing investigation time from two weeks to two days using Sonatus AI tools, according to Nissan’s own announcement.
- Jeff Chou frames AI as built on top of the SDV foundation, not a replacement for it. Infrastructure comes first, applications follow.
- The industry is converging on a new term, the AI-defined vehicle (AIDV), now formalized through B2B marketplaces like SDVerse.
- Regional adoption speed varies sharply. Omdia’s 2026 SDV survey found German automakers rank predictive maintenance as a top priority while reporting the lowest AI deployment rate globally.
Jeff Chou spent close to three decades building data centers, networking equipment, and storage systems for banking, retail, and media companies, work that ran through Sun Microsystems and through startups later acquired by Cisco and by Brocade. Sonatus, founded in 2018, is his bet that the same disruption was coming to vehicles. Today the company’s technology runs in more than seven million production vehicles from Hyundai, Kia, and Genesis, according to Sonatus’s own reporting.
Why couldn’t OEMs quantify the ROI of software-defined vehicles?
The software-defined vehicle promised original equipment manufacturers (OEMs) new revenue streams and lower costs by shifting vehicle functionality into software. In practice, Chou says, that promise was hard to price.
“If I put more software in the car, how much is it going to save me, or how much faster is it going to make me, or am I going to get a new revenue stream? These are the things every OEM asks,” Chou says.
He points to a moment at a CES conference roughly a year before this conversation, where the industry openly discussed what he calls the trough of disillusionment around SDV. The technology hadn’t delivered the promised savings. If anything, it cost OEMs more, since building it meant putting thousands of engineers on the problem and debugging far more lines of code than before.
How does AI make that value measurable, according to Sonatus?
Chou says Sonatus deliberately narrowed its AI pitch to two variables OEMs already budget against.
“The way we are talking about our AI products, we’re focusing on only two values: it will save you money, and it will accelerate your time to value. That’s it,” Chou says.
That framing matters because it sidesteps the abstraction that made SDV hard to sell internally. An OEM finance team can model faster time-to-market and fewer engineering hours. It struggles to model the value of “infrastructure” on its own.
What happened when Nissan applied AI to vehicle testing?
The clearest evidence for Chou’s argument comes from outside the interview itself. Nissan Technical Centre Europe (NTCE), based in Cranfield, UK, partnered with Sonatus to apply Collector AI and AI Technician to vehicle development testing, including work on the Nissan LEAF and Juke. According to Nissan’s own announcement, early trials cut investigation time from two weeks to just two days, while reducing reliance on physical test vehicles.
In the interview, Chou frames this as exactly the kind of result SDV alone couldn’t produce on its own timeline.
“The time to value for them is getting to start of production (SOP) quicker. That’s it. Debugging their car quicker and getting into manufacturing,” Chou says.
Fewer physical test vehicles also means fewer hard drives to collect data from and fewer engineering cycles spent on repetitive debugging, which is where Chou locates the cost half of the equation.
How did a data center founder end up building AI for cars?
Sonatus didn’t start as an AI company, and it didn’t start in automotive by accident either. Before founding Sonatus, Chou built data center infrastructure for Hyundai in a previous role. It was Hyundai, he says, that asked the question that became the company’s founding thesis.
“They asked me, can you take what you did in the data center and somehow extend it to the edge in the vehicle? That was one of the first discussions we had about taking Sonatus technology and connecting it to the data center technology that I delivered in my previous lives,” Chou says.
The underlying bet, as Chou puts it, was that the vehicle was going to become a data center on wheels. Sonatus’s first product, built around in-vehicle networking, mirrored the point in data center history when universal connectivity and networking first tipped everything else forward.
Why does Chou believe infrastructure always comes before AI?
Chou’s argument for sequencing AI after SDV rather than instead of it rests on a pattern he says he’s watched repeat across multiple technology cycles.
“I think it’s always infrastructure drives application, not the other way around. The internet was there before social media. We didn’t invent networking and internet thinking that there would be Facebook,” Chou says.
He extends the same logic to automotive: hardware infrastructure first, then software infrastructure to orchestrate it, then applications, with AI as the current application layer sitting on top of the SDV foundation Sonatus and others have spent years building.
What is an AI-defined vehicle, and is it replacing the SDV?
Chou uses the term AI-defined vehicle (AIDV) in the conversation, a label that has since become part of Sonatus’s own product language. In April 2026, Sonatus joined SDVerse, a B2B software marketplace founded by General Motors, Magna, and Wipro, specifically to make its AI Technician, Collector AI, and AI Director products available to OEMs and Tier 1 suppliers building what the marketplace also calls AI-defined vehicles.
Chou is careful not to present AIDV as a break from SDV.
“I personally think SDV and AI-DV could be one in the same, because they’re both defined by software,” Chou says.
Does region change how fast OEMs adopt AI?
Chou describes China’s pace of technology adoption in the automotive sector as strikingly fast, and treats it as a signal worth watching rather than a threat to dismiss.
“It is not surprising to me at all how advanced that they have gotten, and I think the world needs to take notice on that,” Chou says.
His read on Western incumbents is candid rather than alarmist: “it’s their race to lose,” he says, pointing to brand strength, distribution, and existing customer bases as advantages that remain theirs to use or squander.
Independent survey data backs up the regional gap Chou describes anecdotally. Omdia’s 2026 SDV survey, conducted with Sonatus, found that German automakers rank predictive maintenance as a top revenue driver (47%) yet report the lowest AI deployment for it globally (just 18%), while Chinese OEMs pivoted hard toward automated driving and personalization as differentiators.
Frequently Asked Questions
Does Sonatus’s AI technology work only during vehicle development, or after cars are sold too?
Both, according to Chou. He frames the after-sales case around a simple idea: letting an owner or technician “ask my own car what’s wrong with you” instead of relying on a dashboard warning light and a dealership service department that, in his words, has a built-in “conflict of interest” in diagnosing the problem.
Is this AI technology exclusive to the OEMs Sonatus already works with?
No. Since joining SDVerse in April 2026, Sonatus’s AI Technician, Collector AI, and AI Director products have been available to any OEM or Tier 1 supplier on that marketplace, not just Sonatus’s existing customer base.
Does AI diagnostics reduce the number of physical test vehicles OEMs need to build?
Yes, in the Nissan case. Nissan’s own announcement credits the AI tooling with reducing reliance on physical test vehicles alongside the faster investigation times, which Chou connects directly to lower cost, fewer prototype vehicles, fewer hard drives, fewer engineering cycles per debugging pass.
Will “AI-defined vehicle” replace “software-defined vehicle” as the industry’s standard term?
Chou himself isn’t sure, and says so directly: “probably that term won’t stick in the auto industry. I have no idea.” His own position is that the underlying technology is converging even if the label hasn’t settled yet.
Why does Chou think Western incumbent OEMs still have an advantage over faster-moving competitors?
Brand recognition, existing distribution networks, and an installed customer base, in his view. He frames the opportunity as time-limited rather than permanent: the technology to close the gap is available now, but the advantage isn’t guaranteed to last.