“A scalable software platform starts with hardware agnosticism”

As software-defined vehicles become more complex, the pressure on E/E architectures is no longer limited to connectivity and compute alone. Automakers now have to decide how edge intelligence can be integrated in a way that supports scalability, maintainability and faster software-driven adaptation in the field.

Alexandre Corjon, Senior Vice President and Technical Fellow at Sonatus, addressed these questions in his presentation “Integrating Edge AI into Modern E/E Architectures” at the Automobil-Elektronik Kongress 2026. Drawing on more than 30 years of experience across aerospace, automotive and software, he outlined where current SDV approaches still fall short and what a scalable software platform must deliver in practice.

US Automotive Wire Harness & EDS Conference

20-21 October 2026 | The Henry Hotel, Dearborn, Michigan

After the event, we spoke with him about the architectural bottlenecks, software foundations and edge intelligence strategies that will shape the next phase of software-defined mobility.

ADT: Looking ahead three to five years, what will be the biggest bottleneck in turning SDV and AI strategies into scalable, industrialized vehicle platforms?

In the near term, the primary bottleneck will shift from hardware availability to the challenge of bridging legacy systems with dynamic software. While hardware availability is no longer the crisis it was during the chip shortage, automakers still face real hurdles in deploying AI at the edge – especially when it comes to updating, monitoring, and orchestrating different AI models across varying vehicle configurations without overrunning network bandwidth or compute budgets. Adding to this is a dependence on physical testing. Even today, many validation processes still rely on physical fleets, real-world road exposure, and long test cycles to verify even minor software changes. The industry needs to move toward fully virtualized automated testing and AI-driven diagnostics, from pre-SOP to post-sales, if it is going to keep pace with software’s rapid innovation cycles.

Vector Informatik simplifies SOVD diagnostics

Vector Informatik has launched an end-to-end SOVD toolchain for service-oriented vehicle diagnostics. The solution helps OEMs and suppliers evaluate, test and integrate the new diagnostic standard into modern SDV architectures.

Nicole Ahner

Which decision being made today will most strongly determine where value is created in the future automotive ecosystem?

The OEMs that will lead the future automotive ecosystem will be the ones investing today in a consistent, hardware-agnostic software foundation. Right now, most vehicles still rely on fixed, hard-coded data logging, meaning engineers have to decide months in advance which signals the car will collect, and changing that later requires rewriting software and pushing an update. That is too rigid for a world in which issues emerge in real time and AI models constantly evolve. Moving to a dynamic, AI-driven data pipeline flips that model. You can instrument the fleet dynamically, pull only the signals required for a given analysis, and adapt as conditions change. That flexibility is what will enable the ecosystem to unlock value, from predicting maintenance needs to understanding driver behavior and delivering targeted services.

Where do current approaches to SDVs and next-generation E/E architectures still fall short in real-world programs?

Many of today’s SDV implementations fall short because they use a “sledgehammer” approach to OTA updates. You have to flash huge portions of the vehicle’s software just to adjust a small feature or change the data you are collecting. On top of that, many next-generation E/E architectures keep domains like powertrain and infotainment in separate silos, which makes it almost impossible to gain true cross-domain insights. The result is an architecture that is still too rigid. Instead of targeted, containerized updates that isolate changes to only what is needed, it often lacks the flexible network fabric, compute resources, or shared storage needed for a vehicle to adapt independently in the field without forcing it back into a factory-level workflow.

How Central Compute, Chiplets, and Open Toolchains Enable the SDV

Software-defined vehicles are increasing demands on automotive computing. Central compute, chiplet-based scalability, and open toolchains provide a practical foundation for performance growth, software reuse, and long-term platform development.

Peter Bechberger

What defines a scalable software platform for SDVs in practice?

A scalable software platform starts with hardware agnosticism and a rigorously defined service-oriented architecture. You need a unified abstraction layer – a common operating system and vehicle API – that works across whatever chips or Tier 1 hardware you are using. Vehicle functions need to be isolated, interoperable components – effectively microservices – so features can move across vehicle lines without rewriting code. But that foundation only works if you pair it with fine-grained orchestration. Automakers need the ability to push edge AI models, adjust network behavior, or reconfigure data-collection rules on the fly without spinning a new software build or performing a full firmware update.

Where do current software approaches still struggle to deliver reuse and long-term maintainability?

Current software approaches struggle because they are still tightly coupled to hardware and lack the abstraction and lifecycle visibility needed for long-term maintainability. Reuse is limited because much of the code is written for specific ECUs instead of as portable, cross-platform services, so hardware changes trigger major rewrites. And without AI-driven diagnostics to catch issues early and automate regression checks, teams remain trapped in reactive maintenance cycles, continually chasing regressions instead of isolating root causes early.

Save the date: 31st AUTOMOBIL-ELEKTRONIK Kongress

The 31st International AUTOMOBIL-ELEKTRONIK Kongress (AEK) will take place on 22 and 23 June 2027. For many years, the networking conference has brought together leading decision-makers from the automotive electronics sector and senior executives from the technology industry to discuss the integrated customer experience required for the vehicles of the future.

Despite its increasingly international profile, participants still describe the AUTOMOBIL-ELEKTRONIK Kongress as an “automotive family gathering”.

Secure your conference ticket for the 31st AUTOMOBIL-ELEKTRONIK Kongress in 2027. You can also follow the AEK LinkedIn channel and #AEK_live.

How does edge intelligence change the role of traditional E/E architectures in vehicles?

In-vehicle AI at the edge changes the game. It turns the vehicle’s E/E architecture from a simple set of data pipes into a local processing hub. These systems can process high-volume sensor and telemetry data directly in the car, which results in faster responses, lower cloud and transmission costs, and improved data privacy and sovereignty. These architectures also need to be able to move heavy AI workloads and high-bandwidth sensor data in real time, sending them to the compute resources available at that moment.