Shockwaves continue in the auto industry following global shortages of chips and battery-grade lithium and cobalt to support EV supply chains. Both forced automakers to fundamentally rethink who was on their supplier lists and how much lead time they required.
But now a larger, albeit less dramatic, force is in play. OEMs are scrambling to restructure their software and hardware assets – and their org charts – to keep pace with the production of AI-driven SDVs, software-defined vehicles that are increasingly referred to as AIDVs.
AIDVs have features, safety systems, and performance that are driven by AI for the lifetime of the vehicle, not just to the point at which it leaves the factory.
It's a challenge resulting from decades of legacy architecture that was built for a pre-software era
OEMs are now racing to reconstruct.
Pain Point #1: Decades of outsourced, siloed code
For the last three decades, OEMs didn't build their own vehicle electronics. According to MarketsandMarkets, OEMs outsourced much of their E/E architecture, ECU development, and embedded software to Tier-1 suppliers. Each ECU had a dedicated function, leaving OEMs to contend with fragmented systems, limited software ownership, and unknown update cycles.
The result was a codebase challenge: software written by dozens of suppliers on different toolchains, with no shared architecture or a go-to team that understands how it all fits together. It's not dissimilar to the chip shortage, during which a critical input was sourced from too many places, with too little visibility, for too long.
Pain Point #2: Too many boxes, not enough platform
Imagine ECUs as individual black boxes. S&P Global Mobility estimates that early vehicle architecture comprised 70 to 100 ECUs, with each one independently sourced, wired, and updated. The industry has spent the last several years consolidating that sprawl of individual ECUs into "zonal" architecture — fewer, more powerful compute nodes to replace dozens of single-purpose boxes. IoT Analytics' 2026 adoption report said 80% of OEMs and suppliers are now doing this. In fact, 45% of the industry reports that implementing zonal architecture is its number one strategic priority. In a separate study, Deloitte found that 90% of surveyed OEM executives are actively executing the strategy, with 81% of them projected to complete the effort by 2030.
Enter vehicle AI and the consolidation is beyond cosmetic. It's physically requisite for AI to effectively navigate a car's safety, perception, and comfort systems. At CES 2026, Qualcomm demonstrated how its cross-domain partnership with Leapmotor, as well as its Snapdragon Digital Chassis deal for the 2026 Toyota RAV4, were framed to specifically reduce ECU count so the resulting platform could support AI workloads.
Pain Point #3: Automakers aren't software companies
Perhaps the biggest challenge for OEMs is on the inside. Automakers aren't software companies, and retrofitting their mechanical engineering is difficult even for well-resourced OEMs.
Volkswagen's software unit, Cariad, is the industry's most-cited cautionary tale. Cariad struggled for years trying to build and deliver a consolidated in-house software platform. Ultimately, VW restructured the relationship entirely by licensing Rivian's zonal software architecture across VW, Audi, and Porsche in a deal worth up to $5.8 billion rather than continuing to build in isolation.
General Motors is amidst a similar circumstance. Over the past several months, GM has combined its vehicle software engineering and global product units into a single organization under new Chief Product Officer Sterling Anderson, a former Tesla Autopilot leader and co-founder of Aurora Innovation. GM's AI team has also been folded into the manufacturing engineering organization. Going forward, “software and product must be thought of as one and the same," Anderson told CNBC reporters.
The moral of the story is that building AIDVs isn't just an engineering challenge. It's an organizational one that automakers are still working out. After all, refabbing a century-old manufacturer as a software-and-AI-first company is no easy task. But the financial incentive to do so is significant. Forbes' Business Development Council estimates the AIDV-adjacent software and connectivity market will reach $470 billion in 2026 and $1.19 trillion by 2036.
The Missing Link: Unified data for life
AI is only as good as the data available to train and validate it. Like the ECUs that generate it, vehicle data has been scattered across suppliers, formats, and systems that were never built to feed a single pipeline. But solving that problem isn't a single event. It has to perform through every stage of a vehicle's life, from the first pre-production prototype to the last service appointment years after it drives off the dealership lot.
Testing and validation is where it starts. Pre-production vehicles run on systems that are incomplete and bugged, and engineers typically send instrumented test vehicles to the track to capture data on a specific subsystem. The data are ingested and reviewed separately, which is a slow, intensive process made noisier by failures in unrelated (also unfinished) systems bleeding into the results. AI-driven analysis is directly changing that math. Nissan Technical Centre Europe has recorded a 90% reduction in system debugging time – from roughly two weeks to two days – by using AI to assess vehicle development data. Pre-production vehicles also cost as much as 10 times more to build than mass-production units, so cutting the number of physical test vehicles needed — and thereby using engineering time more efficiently — translates into real savings well before a vehicle reaches Start of Production.
Production has been similarly fragmented. Data generated on the assembly line and through end-of-line testing have traditionally lived apart from both the pre-production data that preceded it and the field data that follows after a vehicle ships. This makes it difficult for OEMs to trace a defect back to its origin or apply lessons from one vehicle program to the next, but the introduction of zonal architecture is changing that equation.
After-sales diagnostics are where fragmented data most impacts an OEM’s bottom line. Modern vehicles contain as many as 1,400 computer chips, far more than what traditional diagnostics were built to handle. Conventional methods rely on basic code readers, technician "gut feel," and reactive maintenance that only kicks in once a warning light appears. Such an approach drives up warranty costs when technicians replace parts that were never actually faulty.
AI-assisted, or AI automotive diagnostics, work differently. They cross-reference diagnostic trouble codes against sensor readings, technical manuals, and fleet-wide repair history to rank likely root causes, which improves first-time fix rates and helps prevent technicians from chasing the wrong problem. Because the analysis can draw on data from an entire connected fleet, a fix identified in one car can proactively flag a similar issue in another before its warning light ever turns on. Some commercial fleet operators applying these tools have pushed vehicle availability to as high as 93%.
Over-the-air updates add a top-layer of complexity. OTA delivery is now critical to AIDVs’ currency, but failed updates, intermittent faults, and unexplained "black screen" incidents have become a growing source of warranty claims and service center confusion — issues that are software in nature but often get diagnosed, at real cost, as if they were hardware.
That's what companies like Sonatus are addressing, by delivering automotive AI and data platforms that give OEMs a coherent, unified foundation across testing, validation, and production productions, as well as after-sales service. That way, AI initiatives at any one stage of a vehicle lifecycle aren't left to rebuild the foundation from scratch. It's a backstage act relative to the chip-and-cockpit that’s taken center stage at CES, but it's the role that ensures the front-end restructuring — software, hardware, and org chart alike — generates gangbuster ROIs with smart vehicles that get even smarter over time.
