Agile Development Of Software-Defined Vehicles Using Cloud-Based Virtual Prototypes

By Gunnar Braun and Stewart Williams

Automotive software is becoming more expensive and central to a car’s identity. Infotainment, advanced driver-assistance systems (ADAS), traction control, and even powertrain management are all shaped by lines of code. Vehicular codebases can now exceed those of commercial aircraft! The growing adoption of electric vehicles (EVs) and the push toward fully autonomous transport have further increased the role and importance of automotive software.

Automotive manufacturers (OEMs) and their suppliers need to shorten development cycles to compete and differentiate in a fast-paced global market. They need the ability to deliver frequent software updates throughout the vehicle’s lifecycle. Quality, safety, security, and reliability must be assured. And cost pressures are enormous.

All these challenges demand faster, more rigorous software development, testing, and validation. To enable earlier and more efficient software testing and validation, the industry is adopting cloud-based development practices and the use of virtual prototypes.

Cloud-native development reaches the automotive industry

The shift from hardware-centric to software-centric vehicles means OEMs and suppliers must rethink their development models entirely. OEMs are transforming their development processes from hardware- and component-centric to software-centric approaches. While traditional hardware-in-the-loop (HiL) testing rigs are still a mainstay for system validation, their expense, latency, and inflexibility are increasingly at odds with the agility required by today’s automotive software development practices.

The IT industry pioneered cloud-native development approaches, leading to web applications and Software-as-a-Service (SaaS) solutions that are quickly delivered and continuously updated. These approaches involve incremental development and a strong emphasis on automation. DevOps practices further bridged the gap between the software development process and its operational deployment environment. Thanks to technologies such as containerization, testing cycles are now performed and automated in a production-simulated environment.

This is one of the key challenges for OEMs and suppliers adopting cloud-native methods for SDV development: the replication of the operating environment — the vehicle — when developing and testing software.

While the cloud provides virtually unlimited compute and collaboration resources, physical hardware — the ultimate destination for automotive software — often lags behind in availability. Teams often wait months for silicon or electronic control units (ECUs) to be ready for initial testing and validation. If hardware revisions are required, it extends the waiting period before OEMs and their suppliers can fully test their software at scale.

This is where virtual prototypes enter the equation.

Virtual prototypes accelerate automotive software development

Virtual prototypes are models of target hardware that are used for software development, testing, and validation — before the hardware is available. Increasingly referred to as electronics digital twins (eDTs), these virtual prototypes:

• Simulate the behavior of automotive compute platforms with varying levels of fidelity.
• Enable continuous software testing, integration, and delivery (CI/CD).
• Minimize reliance on hardware and eliminate associated delays.

Virtual platforms have existed for decades and are used throughout the embedded software industry for “shifting left” software development. That is, starting software development before hardware is available to gain precious time in an otherwise sequential development process. This results in a significant time-to-market advantage.

A notable example is the open-source community’s use of QEMU, an emulator that allows open-source software stacks to leverage cutting-edge Arm CPU features well before the corresponding hardware is available.

But traditional emulation-based techniques and tools were never designed for the complexity of modern automotive processors, such as Arm’s Cortex-A720AE, or the runtime demands of continuous, cloud-based software development.

Advancing Arm embedded software development

With the introduction of Synopsys Virtualizer Native Execution, Synopsys is addressing these limitations and enabling Arm embedded software to be executed directly on Arm server CPUs — in the cloud, with no emulation or ECU hardware required. This means developers can run workloads at the speed of the eventual hardware, roughly 100x faster than traditional instruction set simulators, while retaining the benefits of virtual prototypes and full compatibility with the existing ecosystem of models, tools, and workflows.

The technical and operational impacts are profound:

• Software teams can validate complex system-level behaviors early and often, significantly reducing the dependency on HiL rigs during the development cycle.
• With architecture parity between cloud and vehicle CPU (via Arm’s instruction set), developers can use the same development tools to significantly reduce workflow complexity.
• Native execution on Arm-based cloud instances, such as AWS Graviton servers, enables continuous integration and DevOps pipelines, with virtual prototypes providing the critical link between development and operational environments.

Leveraging the SOAFEE reference architecture

An essential part of this puzzle is the SOAFEE (Scalable Open Architecture for Embedded Edge) initiative. Led by Arm and other industry stakeholders, SOAFEE provides a standardized framework and reference architecture — based on the principles of modularity and orchestration — for SDV workloads. By adhering to common standards like SOAFEE, the ecosystem of automotive OEMs, suppliers, and technology leaders can increase interoperability and reduce vendor lock-in.

In collaboration with Arm, Synopsys demonstrated at Embedded World 2025 how virtual prototypes, SOAFEE reference architecture, and cloud-to-edge software development come together:

• Leveraging Synopsys Virtualizer Native Execution in an AWS cloud environment, the demonstration featured an open-source autonomous driving workload running on top of the SOAFEE edge reference stack.
• Executed at real-time speed, the virtual prototype was approximately 100x faster than a comparable QEMU-based simulation.
• The demonstration showed how virtual prototypes replicate the structure of an edge device with sufficient granularity to validate real-world functionality and software behavior, while remaining scalable and shareable across teams and geographies.

Accelerating validation cycles and feature deployment

While the technical underpinnings are compelling, the business value of virtual prototypes and cloud-native development is just as important. Cloud-based workflows are becoming a prerequisite for meeting time-to-market expectations in a world where vehicle features are continuously updated via over-the-air (OTA) software pushes.

Virtual prototypes minimize the reliance on HiL rigs, reduce CapEx and OpEx, and allow distributed teams to develop and test in parallel. By enabling early and scalable validation, they help mitigate the risk of schedule slippage due to delayed hardware availability. And by aligning development and target architectures through Arm-native cloud compute, they eliminate costly architectural mismatches late in the cycle.

Together, these shifts can materially reduce validation cycles, accelerate feature deployment, and improve team productivity.

Stewart Williams is a segment management director for the Solutions Group at Synopsys, where he is the lead on the automotive software ecosystem. He has a Ph.D in electrical and computer engineering and applied physics from North Carolina State University.