At CES 2023, industry leaders highlighted a major shift in automotive software development. This transformation is reshaping how vehicles are designed, built, and maintained throughout their lifecycle.
Sandy Munro and Thomas Mueller, Vice President and Chief Technology Officer (CTO) of Wipro Engineering Edge, led the discussion. They emphasized the shift from legacy, hardware-bound systems to scalable, AI-enhanced architectures.
These platforms can evolve for decades while supporting continuous upgrades and new capabilities. The transformation goes beyond satisfying consumer expectations. It redefines engineering priorities and places safety, efficiency, and long-term relevance at the forefront.
100 Million Lines of Code — and Counting
Modern cars now contain roughly 100 million lines of code, dwarfing even the software complexity of a Boeing Dreamliner. That’s ten times the aviation benchmark in terms of software volume. This sheer scale poses challenges in maintenance, security, and adaptability. Most of this code resides in hundreds of Electronic Control Units (ECUs) scattered throughout the vehicle — a design approach that limits scalability and hinders the integration of new features.
The industry’s next leap involves consolidating these ECUs into a few High-Performance Computers (HPCs) capable of hosting multiple functions, updating seamlessly, and enabling over-the-air improvements for years after production.
ECU Consolidation and Latency Reduction
Legacy vehicle designs can contain more than 150 ECUs. Each unit is typically dedicated to a single function, from HVAC control to door latches, connected by extensive wiring harnesses. This architecture increases weight, complexity, and latency — the critical time delay between sensor input and system response.
By consolidating functions into fewer HPCs, automakers can dramatically reduce latency, improve reliability, and open the door to more sophisticated AI-driven applications. For example, Tesla’s approach uses redundant chips in “shadow mode” to collect and analyze driving data without disrupting live operations. This method maximizes available compute resources while improving the company’s autonomous driving algorithms.
Edge AI and Real-Time Data Filtering
Processing data at the “far edge” — directly in the vehicle — is becoming essential as sensor resolutions climb from 1.8 megapixels to over 10 megapixels. High-resolution cameras, radar, and lidar generate enormous raw data streams. Without intelligent filtering, transmitting all of it to the cloud is inefficient and costly.
Tesla’s architecture filters data in real time, sending only the most relevant information to its cloud systems. Other manufacturers often transfer bulk sensor data for later processing, which slows insights and increases costs. Moving toward real-time edge AI aligns with long-term goals of autonomy and feature expansion without overwhelming cloud infrastructure.
Breaking Away from Legacy Code
A surprising obstacle to progress is the continued reliance on the C programming language, developed in the late 1960s. While C remains powerful, it’s less secure and slower to develop compared to modern languages like Rust, Go, Java, and WebAssembly. Transitioning to modern languages not only accelerates feature deployment but also strengthens cybersecurity.
The industry is also beginning to shift away from monolithic “spaghetti code” toward microservices architectures. This modular approach, similar to cloud-based systems, allows updates to individual components without retesting the entire software stack. The result is faster development cycles and reduced R&D overhead.
BMW, Renault, and the Android Shift
Some automakers are making bold moves. BMW, for example, is replacing its Linux-based cockpit software with Android Automotive OS. This shift promises faster route planning, a richer app ecosystem, and quicker adoption of consumer-grade features. Renault has partnered with Google to develop a “smartphone-like” architecture for cars, enabling both safety-critical and non-safety-critical apps to be updated more easily.
These decisions reflect a willingness to “cut and run” from entrenched systems — a painful but necessary step toward sustainable innovation.
Lifecycle Management and Continuous Updates
In the future, vehicles could receive software updates as frequently as cloud-based platforms like Netflix or Microsoft Teams, which deploy changes multiple times per day. This continuous improvement model could extend a vehicle’s functional lifespan well beyond its mechanical life.
For example, a safety app to detect and prevent child heatstroke deaths could be developed and deployed in weeks under a microservices architecture. Today, such a feature might require months of cross-domain coordination and full regression testing across all ECUs.
The Cost Advantage of Modern Architectures
Traditional automotive R&D budgets allocate 50–55% of spending to overhead and quality assurance when adding new features. In industries using service-oriented architectures, this overhead drops to 30–35%. For automakers, that difference translates into more features, faster deployment, and better ROI on engineering investments.
By consolidating hardware, modernizing codebases, and adopting standard middleware, automakers can redirect R&D resources from redundant work toward innovation that directly benefits drivers.
The Psychological and Organizational Barriers
Despite the technical and financial incentives, progress is slowed by compartmentalized engineering cultures. Teams often resist sharing resources across domains — for example, using surplus cooling capacity from one subsystem to eliminate redundant components in another. Overcoming these silos requires both leadership vision and a willingness to adopt off-the-shelf solutions where appropriate.
Lessons from Telecom
The telecom industry’s rapid adoption of open, software-defined networking provides a useful parallel. Faced with supplier disruptions, operators collaborated to define common software requirements, enabling the use of standard server hardware instead of proprietary, vendor-locked systems. Similar collaboration in automotive could accelerate the transition to centralized computing and software-defined vehicles.
The Road Ahead
The transformation of automotive software is not optional — it’s a competitive necessity. Automakers who embrace centralized computing, edge AI, modern programming languages, and modular architectures will be able to deliver features faster, maintain safety standards, and extend the useful life of their vehicles.
Those who cling to legacy systems risk obsolescence in a market where technology cycles move faster than ever. As CES 2023 made clear, the winners in the next decade of mobility will be those who treat software as a living, evolving asset — not a static component frozen at the time of manufacture.
Explore More Automotive Teardown and Software Insights
For detailed reports and consulting services, or more teardown reviews, expert breakdowns, and lean engineering insights, visit Munro & Associates or subscribe to Munro Live. Learn how software and hardware innovation are shaping the vehicles of tomorrow.