For decades, the value of a car was measured in horsepower, torque, and acceleration. In 2026, the industry’s most consequential battleground is measured in parameters, tokens, and inference speed.
At CES 2026 in Las Vegas, the shift was unmistakable. With electric vehicle demand softening and regulatory pressures mounting, automakers pivoted their messaging from batteries to brains. “Their focus moved to so-called physical and context-aware AI—systems that interpret real-world conditions in real time—positioning cars as software-defined vehicles rather than fixed hardware” . The era of the AI-defined car has arrived.
The Cockpit Becomes Conversational
Fawaz Sheikh says the most visible transformation is happening in the cabin, where voice assistants are evolving from rigid command-followers into genuine conversational partners.
Mercedes-Benz’s new CLA introduces MB.OS, an in-house operating system that underpins the first production car to combine AI from both Google and Microsoft. The system’s Google Maps-based navigation includes an Automotive AI Agent that interprets natural speech. Occupants can say “I’m too cold” rather than specifying a temperature, and the assistant understands intent without requiring strict phrasing. It can retain conversational context, allowing follow-up queries like checking restaurant reviews after selecting a destination.
BMW has taken a different route, becoming the first automaker to fully embed Amazon’s Alexa+ architecture. Debuting in the new iX3 in the second half of 2026, the system allows multi-part questions and linked requests. As Stephan Durach, BMW’s Senior Vice President for Digital Services, put it, “This has resulted in a product that sets new standards in the naturalness of human-vehicle interaction through the use of artificial intelligence” . The system can handle queries like “What’s the most famous painting in the world and take me there?”—identifying the Mona Lisa and setting navigation to the Louvre in a single exchange.
Volvo and Google have pushed the concept further with context-aware AI that can see and understand the vehicle’s surroundings. At Google I/O 2026, the companies demonstrated Gemini interpreting parking signs in real time, explaining restrictions and time limits rather than leaving drivers to guess. “The EX60 provides an ideal platform to explore the future of contextually aware driving experiences,” said Alwin Bakkenes, Volvo’s head of global software engineering.
From Voice Commands to Agentic Action
The distinction between a voice assistant and an AI “agent” is becoming the industry’s key differentiator. Volkswagen’s top executive in China, Ralf Brandstaetter, explained the difference bluntly: unlike a voice assistant that answers simple questions, AI agents can handle more complex tasks and decision-making.
Volkswagen plans to equip new cars built for China with onboard AI agents from the second half of 2026. The system can search for the highest-rated restaurant in an area, make and confirm a reservation, navigate there, and organize parking—all autonomously. The automaker is positioning this as part of its “In China, for China” strategy, developed with Chinese chipmaker Horizon Robotics.
At Auto China 2026 in Beijing, the shift was equally pronounced. More than 60 percent of global premieres came from Chinese brands, and AI was the centrepiece. Geely highlighted its full-domain AI 2.0 system; SAIC’s Roewe brand showcased AI-based in-car applications developed with Volcano Engine; and XPeng CEO He Xiaopeng outlined a strategy centered on “physical AI,” describing autonomous driving as “the first large-scale application of physical AI” .
The Underlying Architecture Shift
These front-end experiences rest on a profound architectural transformation. Vehicles are moving from distributed electronic control units to centralized computing platforms that consolidate previously separate domains.
Horizon Robotics introduced its Starry chip at Auto China 2026—a 5-nanometer automotive-grade processor with 650 TOPS of computing power that supports both intelligent driving and cockpit AI models on a unified platform. According to the company, integrated architectures could cut vehicle-level costs by up to 4,000 yuan ($585) and reduce development timelines from 18 months to eight months. More than 10 carmakers and suppliers, including BYD, Chery, Volkswagen, and Bosch, have shown interest.
This shift is enabling what TomTom calls the “differentiation layer” of the software-defined vehicle. As TomTom’s Chief Product Officer Leo Sei explained at the Future of the Car summit, the foundational layer—silicon, operating systems, cloud infrastructure, maps—is increasingly becoming shared industry infrastructure, while automakers focus their capital on brand-specific driving characteristics and user experience.
“Customization is where carmakers keep their brand DNA,” Sei said. “You probably don’t want a Porsche and a Škoda to have the same automated driving experience—but that doesn’t mean you need to own the entire layer all the way down” .
AI Beyond the Cockpit
The AI transformation extends well beyond voice interaction. An Omdia study conducted with Sonatus found that 34 percent of respondents worldwide name intelligent diagnostics and predictive maintenance as the most important use cases for AI in software-defined vehicles. In North America and Europe, 48 percent cite predictive maintenance as important for customer retention and after-sales revenue.
The gap between ambition and implementation remains significant in some regions. In Germany, manufacturers rate predictive maintenance at 47 percent as a revenue driver, but report only 18 percent of actual AI deployment in the field. Japan prioritizes automated driving (50 percent), while China shows a clear shift away from classic data monetization toward automated driving (54 percent) and advanced personalization (53 percent).
In manufacturing, generative AI and spatial computing are reshaping how vehicles are built. The concept of “software-defined technicians” is emerging, where expert knowledge is externalized into machine-legible reasoning chains, allowing AI systems to perform repetitive inspection tasks while human experts evolve into teachers and overseers.
The Trust Barrier
For all the technical progress, trust remains the industry’s most formidable challenge. As Cinemo’s analysis notes, “A wrong answer from AI is annoying on a laptop, but in the car it can become a safety issue” . In-car AI must cope with incomplete data, changing conditions, regional privacy rules, and the cost of running cloud-based intelligence at scale—and it must perform consistently despite all of them.
This is particularly complex in the shared cabin environment. Most AI tools are designed for one user, but a car may contain a driver, a front passenger, children in the back, and multiple screens—each with different preferences, permissions, and privacy expectations. “The system has to understand who is speaking, which actions are allowed, which services can be accessed, and what content should be shown to whom”.
SAE technical papers highlight similar concerns at the field level: data scarcity, regulatory fragmentation, sensory fusion reliability, and user trust all pose barriers to scaled AI adoption. The recommendations point toward safe, modular, and scalable integration roadmaps, emphasizing continual learning, hybrid digital twins, and legacy-system interoperability.
The Road Ahead
The automotive industry is converging on a common destination: cars that learn, evolve, and improve across generations. Giovanni Giancaspro of TomTom captured the shift succinctly: “The industry is moving towards software and AI architectures that learn, evolve and improve across multiple generations”.
But as Cinemo cautions, the cockpit is not a smartphone. “AI changes constantly, while vehicles are designed, validated, produced and used over more than ten years”. The challenge for automakers is building flexible architectures that can accommodate rapid AI advancement without requiring hardware replacement every few years.
The winners in this new era won’t be those with the most powerful engines or the longest range. They’ll be the ones who master the delicate balance between innovation and reliability, between intelligence and trust—and who understand that in an AI-defined car, the most important horsepower is measured in context, not kilowatts.
No Responses