Mobile-Menu

Software-Defined Vehicle From SDV to AIDV

From Sven Prawitz | Translated by AI 4 min Reading Time

Related Vendors

FAULHABER Drive Systems
Taiwan Machine Tool & Accessory Builders' Association (TMBA)
SIMTOS (Seoul International Machine Tool Show)

Xpeng calls its P7+ model the "world's first AI-defined car". The term has been driving the industry ever since. The concept is more than just marketing. The technology is there, but it is lacking elsewhere.

VW is developing a new E/E architecture for its future Chinese models. Experts say that the technology in this area is developing faster than the organizations.(Image: Volkswagen AG)
VW is developing a new E/E architecture for its future Chinese models. Experts say that the technology in this area is developing faster than the organizations.
(Image: Volkswagen AG)

The world's first AI-defined car - this was the claim with which Xpeng presented its P7+ model in November 2024. At the time, founder He Xiaopeng believed that artificial intelligence would fundamentally change mobility. That is why Xpeng wanted to be a pioneer for the new category of AI-defined vehicle (AIDV).

Specifically, the OEM promised: a level 2 assistance system improved by AI, particularly natural voice control, a braking distance of 35 meters (approx. 115 ft) and energy consumption of ten kilowatt hours per 100 kilometers (approx. 62 mi)—made possible by AI-optimized energy efficiency. The special feature: The AI allows the car to "learn" every two days, according to Xpeng.

Difference Between AIDV and SDV

Since then, the term AIDV has become increasingly popular in the industry. But what is the difference between an AI-defined vehicle and a software-defined vehicle? The market research company Omdia—commissioned by Here Technologies—surveyed almost 650 industry experts to find out. Based on the results, the authors of the study defined four phases of the software-defined vehicle (SDV)—from the connected vehicle (phase 1) to the "agentic vehicle" (phase 4).

For Omdia, AI is an integral part of this evolution. From phase 2 ("Augmented"), individual AI and machine-learning applications will be integrated (36% agreement in the survey). In phase 3, OEMs will become software-first companies. 39% of respondents expect AI and ML to be used here. The final phase 4 ("Agentic") describes intelligent agents as the culmination of the SDV concept.

First OEMs Scale Up AIDV Approach

Augustin Friedel, Associated Partner at MHP, is already observing the first AIDV architectures on the market. "Most AIDVs are SDVs, but not all SDVs are AIDVs," is his thesis, which he visualized in a graphic on Linkedin. There he names the manufacturers Geely, Xpeng and Nio as well as the supplier Huawei as AI-first companies that are already scaling the technology.

"AIDV: this is not a trend that requires a new platform, but is based on the new E/E architectures. This will gradually become more common over the next few years," observes Friedel.

The difference: with SDV, software defines what the car does—rule-based, deterministic, controllable through regular updates. With AIDV, on the other hand, AI models and data define how the vehicle behaves: probabilistic, context-dependent, self-learning across all domains.

AIDV Expands E/E Architectures

Omdia describes a scenario for these cross-functional E/E architectures: the car analyses data patterns, recognizes that the driver is heading to the airport and uses a camera to identify the driver's tiredness. The vehicle therefore increases the sensitivity of its monitoring systems and displays high-contrast warnings.

In May 2025, Suraj Gajendra, Vice President at ARM, also gave a striking example: "AI models process camera, radar and lidar data in real time and anticipate not only objects, but also driving behavior, road conditions and trajectories of other vehicles—before functions such as automatic emergency braking take effect."

Not All OEMs Have Access to Data

Geely, says Friedel, uses a kind of "AI master" as the central enabler for the AI-defined vehicle - a powerful base model plus smaller, specialized AI agents. This approach fundamentally distinguishes AIDV from SDV: instead of software updates with new or improved functions, there are continuous improvements to the AI models using data from the fleet.

This is where Friedel, Gajendra and the Omdia Report see a central problem in the current business model of the automotive industry when it comes to data. "Many OEMs have millions of vehicles with cameras in the field, but not all of them have access to the camera data because they belong to a supplier," says Friedel. The Omdia report defines data access in phase 3 as a core competence. Only from this point onwards would SDV investments bring significant returns. "In this phase, OEMs think about total cost of ownership instead of bills of materials," write the authors of the study. "Processors, sensors and memory create opportunities instead of costs."

The biggest hurdle is not technological, but organizational. "The technology is ready, but the organizations are not," is the key finding of the Omdia report. Organizational maturity is the primary bottleneck. Regardless of this, Gajendra stated in May 2025: "The software architecture always reflects the organizational structure."

Subscribe to the newsletter now

Don't Miss out on Our Best Content

By clicking on „Subscribe to Newsletter“ I agree to the processing and use of my data according to the consent form (please expand for details) and accept the Terms of Use. For more information, please see our Privacy Policy. The consent declaration relates, among other things, to the sending of editorial newsletters by email and to data matching for marketing purposes with selected advertising partners (e.g., LinkedIn, Google, Meta)

Unfold for details of your consent

Value Chains are Becoming More Fragmented

"The whole issue of orchestration, partnership management, recognizing early signals in the market and reacting to them—designing the architecture so flexibly that changes are possible. That is the main challenge," Friedel is convinced. This fundamentally changes the relationship between OEMs and suppliers. "Hardware-software separation will make the value chains more fragmented." The focus of this collaboration must be on the performance of overall systems— not on individual sub-areas.

In recent months, chip suppliers in particular have come to the fore. They now negotiate directly with the manufacturers. Won't this make tier 1 suppliers superfluous themselves? Friedel's answer: "Nvidia and Qualcomm see themselves as technology platforms, but not as solution providers. That is not their core competence." When it comes to implementation, scaling and volume, Tier 1s are needed "who know exactly how an OEM program works". The key question for European OEMs: can they manage the organizational transformation quickly enough to keep up technologically?

"The vehicle really interacts and engages in continuous communication—less command—oriented like Alexa, but more dialog-oriented," says Friedel, describing how the Xpeng AIDV works.

Related content