Software-Defined Vehicle Magazine – August / September 2025

Automotive industry buzz around advanced driver assistance systems (ADAS) is intensifying, particularly since Chinese player BYD made the technology standard across its model line-up in February 2025. While countries and companies hold divergent views on how to regulate and release automation systems, few doubt their importance to the future of mobility.

“The rate of innovation we’re seeing is unprecedented,” states Dipti Vachani, Senior Vice President and General Manager of Automotive at Arm. “With all the compute power coming into cars, I think now is the coolest time to be in the market.” She labels ADAS the key to safer transport, and research by the US National Safety Council estimates it could help prevent up to 21,000 road-based deaths annually, or 62% of the country’s total.

As five-star NCAP ratings become increasingly dependent on reliable ADAS, the technology’s importance to customers will likely grow. However, a study by S&P Global found that 40% of new cars on the road today do not have even relatively simple active safety features that could make a crucial difference, such as driver condition monitoring. That figure is expected to drop to around 30% over the next five years, but why not faster and sooner?

The problem is complexity

Vachani tells Automotive World that active safety feature uptake is hindered by both hardware and software limitations. Hardware in most vehicle models is often fragmented, the consequence of different silicon architectures sporadically bolted together by OEMs primarily focused on power and cost considerations. Subsequently, bespoke software solutions are required to run applications, resulting in overly complex systems that cause long lead times to ensure safety and performance standards.

These issues multiply as brands embrace the shift to software-defined vehicles (SDVs) and consider transitioning entire portfolios. “All this complexity means OEMs can’t bring new features to market very quickly, and that presents a competition problem,” notes Vachani. While incumbents’ operational excellence at delivering mechanical products remains, automotive’s evolution is working against them. “As new automakers with a software-first mentality enter the market, legacy players will struggle to keep up.”

Many incumbent OEMs have built up overly complex and fragmented vehicle architectures, which require bespoke software to run applications

At the same time, many tech-forward brands currently lack the scale and production capacity to compete on sheer numbers. This provides a window of opportunity for incumbents to catch up. Since more than 90% of automakers already use Arm hardware, the company is uniquely positioned to help level the playing field. Subsequently, it has been working with the industry for several years to overcome the challenges holding back the refinement and expansion of ADAS.

Building an SDV ecosystem

As the SDV concept started to gain momentum, Arm realised early on that standardisation lay at the heart of accelerating active safety feature implementation. In 2021, the company began gathering members for a new consortium: Scalable Open Architecture for Embedded Edge (SOAFEE), which has around 150 members at the time of writing. Its ongoing aim is to build parity, cost efficiency and portability in the auto software space.

Arm kickstarted the endeavour by creating a standardised firmware layer and making it available to the open-source developer community. Since then, SOAFEE’s progress has included a standard spec for how to write software in the cloud, blueprints for specific applications, and deployments of sample software. “We’re constantly looking at the ecosystem and asking how to ensure software works across every silicon solution out there,” says Vachani.

Through 2024 and early 2025, Arm also released and expanded Kleidi for software-level AI in automotive. By integrating the KleidiAI open-source library kernel with the latest machine learning frameworks, developers can more easily achieve optimum performance on Arm-based compute hardware. Another offering is the Neoverse portfolio of processor cores—originally designed for data centres, these provide high-performance computing at low power to help meet in-vehicle AI requirements. Without those characteristics, she emphasises, next-gen ADAS capabilities will “never see the light of day.”

Now with access to standardised architectures and the ability to write software and deploy AI consistently with safety and security baked in, is the industry satisfied? “Oh no; it’s never enough,” Vachani laughs. “Automakers have all the pieces but must still put them together.” In June 2025, Arm announced Zena Compute Subsystems, a new platform that brings together separate IP strands and validates them with SOAFEE. The company claims this can accelerate development of new, AI-defined vehicle models by at least a year, compared to standard product timelines.

Notably, Zena CSS this mirrors the ‘chiplet’ trend that other semiconductor manufacturers are exploring. By replacing monolithic chip architectures with modular blocks, adherents believe they can unlock more powerful, customisable, and energy and cost efficient in-vehicle computing.

A bright future?

There’s still a lot to do before advanced life-saving features can become standard across all car models. Some automakers have terminated their in-house SDV architectures due to complexity and expense, while others are relying on partnerships with newer players to fill in skill and knowledge gaps. However, Vachani believes an uptick in scale will solve many of the vehicle software issues seen today: “When there’s demand, the ecosystem figures it out.” Through its efforts to make vehicle software implementation easier, Arm is convinced it can nurture that surge.

We’re on the technological journey to put [full autonomy] functionality on the road

In fact, the near future might look very bright indeed. Despite the intermittent progress of vehicle automation in recent years, Vachani is sufficiently optimistic to believe SDVs are providing the foundation for full autonomy. “Obviously there are regulatory and consumer confidence issues that prohibit timeline predictions. However, we’re on the technological journey to put that functionality on the road, and Arm is creating the capabilities necessary for it to happen.”