Will software define the winners?

There was an interesting article in the Financial Times this week describing the challenges that established car manufacturers are experiencing as they try to adapt to being suppliers of software as much as hardware.  Generically this is described as the software defined vehicle, but even though widely-used, I don’t believe that this is an accurate description of either the role of the software or the business objectives.  The software doesn’t ‘define’ the vehicle, it more effectively and flexibly enables the core functions of what we have always expected in a vehicle to operate individually and in harmony, and it enables some new functions that are largely or wholly enabled by software.

As examples of the former, brake by wire and steer by wire do not introduce new functions – they still allow the driver to stop and steer without the need for hydraulic or mechanical connections between the driver position and the actuators.  That brings a hierarchy of benefits – parts count reduction, easier assembly, more flexible packaging, easier configuration (with some caveats that I will come to), and the opportunity to eliminate some other systems by taking advantage of the improved and more flexible controllability of these primary systems.  In the steering and braking areas, this would include more adjustable steering effort and weight, not requiring a separate parking brake system and eliminating the need for complex limited slip differentials to maintain traction in challenging conditions.

New software-controlled functions include the growing availability of drift modes, crab steering allowing on-the-spot manoeuvring of the car during parking and increasingly capable autonomous driving systems.  Taking the latter as an example, ICDP members had the privilege of hearing at our Summer Meeting from friend of ICDP, Max Warburton, now the CFO at Wayve, a UK-based start-up pioneering a new approach to AI-enabled autonomous driving.  Unlike others such as Tesla or Waymo, they are not relying on more accurate mapping, even more sophisticated sensors and pre-defined rules to enable high level autonomy.  Their product is entirely software based (albeit with some very sophisticated AI) and is sensor-agnostic.  A Wayve-equipped car can be taken to a new city and perform well, straight off the transporter, with no need for extensive preparation and testing.  The software is receiving inputs from and sending instructions to the other systems in the vehicle.

There is therefore no doubt in my mind that software and the integration of different software-controlled systems will become increasingly important, and that anyone who fails to deliver this reliably and quickly will fall behind the competition.  The disadvantage they will face is not that they lack some ‘tech’ features that others have, but that their cars have a cost, performance or reliability disadvantage because their software is not as capable or reliable as others.  Going back to my earlier caveat about configuration challenges, the recent Autocar road test of the Lotus Evija (0-200mph/322 kph in 13 seconds if you’re in the market) made the point that it was the management of the massive amount of power through software that was the most impressive point.  The car could be undriveable, but the power delivery and how it was metered out in a way that made the car manageable at all speeds, is perhaps the competitive differentiator.  A less skilled systems integration could have resulted in a product that was positively dangerous.  Always recognised as having a magic touch in delivering superior driving qualities with traditional technologies, they appear to have replicated these skills for the software age.

The advantage of a brand with no previous heritage like Tesla, or one coming from a software background like Xiaomi, is that they have not had to ‘unlearn’ old habits, or retrain large numbers of engineers to think differently.  For the established players, their thinking and culture starts with the ‘oily’ bits (even if electric motors have replaced V8s), and I believe that affects the way they think about product development, timescales and risk.  By the time they get a product to the standard they deem fit for market, their new competitors will already have launched one product and be working on the successor.  I’m also not sure that buying in an architecture helps.  VW Group invested in both Rivian and XPeng in 2023 and 2024, but the first product benefiting from this will not come to market until next year – and we still have to see whether they avoid the many issues faced by products based on their own Cariad architecture over the past few years.

That is then also reflected in the attitudes of some manufacturers to who ‘owns’ the software in their vehicles, and the commercial value they associate with that.  There are still some manufacturers who cling onto the idea that they are going to get new high margin revenue streams if they control the software in their vehicles.  There are others who embrace the idea that the best software will come from a software specialist like Google with Auto Android or building in Amazon’s Alexa.  Apple CarPlay including the latest Ultra version that allows car functions to be controlled through an iPhone interface is no longer included in GM products in the US, yet included in the latest Aston Martins.

I don’t pretend to know what the right approach is, or even whether there is a single ‘right’ approach.  However, what I can see is that getting the software right will increasingly define how quickly a car can come to market, and how good the user experience is in terms of the functionality they enjoy and how reliably it performs.